Home Search Collections Journals About Contact us My IOPscience

NMR introscopy

This content has been downloaded from IOPscience. Please scroll down to see the full text.

1981 Sov. Phys. Usp. 24 841

(http://iopscience.iop.org/0038-5670/24/10/R03)

View the table of contents for this issue, or go to the journal homepage for more

Download details:

IP Address: 129.170.195.64
This content was downloaded on 12/07/2015 at 03:25

Please note that terms and conditions apply.

NMR introscopy
V. A. Atsarkin, G. V. Skrotskii, L. M. Soroko, and E. I. Fedin
Institute of Radiotechnology and Electronics of the Academy of Sciences of the USSR,
Moscow Physicotechnical Institute, Dolgoprudnyf (Moscow Oblast'),
Joint Institute of Nuclear Research, Dubna (Moscow Oblast'),
and Institute of Elementoorganic Compounds of the Academy of Sciences of the USSR
Usp. Fiz. Nauk 135,285-315 (October 1981)

This review presents the physical foundations of NMR introscopy, a method of obtaining images by nuclear
magnetic resonance. We describe various methods of producing NMR images and compare their
characteristics. We discuss the technical, instrumentation, and applied aspects of NMR introscopy, and in
particular the prospects for applying it in biology and medicine.

PACS numbers: 42.30. Va

CONTENTS

1. Introduction 841
2. Physical foundations of NMR introscopy 842
a) General information on nuclear magnetic resonance b) The principle of
NMR introscopy
3. Methods of producing NMR images 844
a) Reconstruction from projections b) Fourier introscopy c) Selective
methods; magnetic focusing d) The method of the sensitive point (line)
e) The method of selective excitation f) Discrete periodic structures
g) NMR introscopy in a rotating system of coordinates h) Prospects
of NMR introscopy of solid objects
4. Applied and technical aspects 853
a) The problem of sensitivity b) Instrumentation c) Applications and
prospects
5. Conclusion 857
References 857

1. INTRODUCTION latter in a deliberately inhomogeneous magnetic field.

After appropriate processing, the spectra of these sig-
The ensemble of methods that enable one to reveal
nals enable one to obtain information on the distribution
the inner structure of opaque objects without destroy-
of nuclear magnetic moments and on their dynamic
ing them is the object of introscopy (inner vision). The
characteristics throughout the volume of the specimen.
vast cognitive and applied significance of introscopy is
This idea was first realized by Lauterbur 5 in 1973, and
evident. Its various methods, which employ X-rays,
several more effective methods for its technical reali-
y-rays, neutrons, fast ions, ultrasound, etc., are zation6"8 were soon proposed. 1 '
widely applied in the most varied fields, from the de-
fectoscopy of industrial products to medical diagnostics. The path traveled by NMR introscopy during these
several years looks impressive, even against the back-
Along with the development and perfection of the tra- ground of the present-day headlong development of
ditional methods, among which x-ray reconstructive science and technology. While the cited pioneer stud-
tomography 1 has met with special success, in recent ies of the early seventies had the aim of NMR imaging
years a new field has arisen: NMR introscopy, which of very simply model objects, at present this method
rests on the well developed technology of nuclear mag- is used to obtain detailed pictures of cross-sections
netic resonance (NMR). of tissues and organs of animals and man in vivo, thus
As we know, nuclear magnetic resonance has been recognizing various pathological changes, and in par-
employed successfully for a long time for studying the ticular, malignant tumors (Sec. 4c). Naturally, all
inner structure and dynamics of matter on the atomic- this draws close attention to the new method from both
molecular level (see, e.g., Refs. 2-4). Here the speci- its potential users—physicians and biologists—and from
men to be studied is placed in a homogeneous magnetic the inventors and developers—physicists and radio en-
field. The information obtained, which consists of the gineers. This review is mainly addressed to the latter
NMR spectra and the nuclear spin relaxation times, of these categories of specialists. Its major aim is to
pertains to the entire volume of the specimen as a present as clearly as possible the physical principles
whole, without revealing its spatial macroscopic struc- and technical methods on which the best known and most
ture. promising methods of NMR introscopy are based. The
On the contrary, NMR introscopy is based on singling
out the NMR signals from individual volume elements "Similar ideas had been advanced even earlier; see, e.g. ,
of the object of study. For this purpose one places the Bef. 89.

841 Sov. Phys. Usp. 24(10), Oct. 1981 0038-5670/81/100841-19S01.10 © 1982 American Institute of Physics 841
biological and medical aspects will be touched upon only ponent does not give rise to transitions at the frequen-
on the illustrative level, while the examples presented cy u>. The classical description of magnetic resonance
of NMR images are designed mainly for comparison of is exact enough in many cases, and hence the graphic
the sensitivity, information content, and resolving pow- picture of precessing magnetic moments often proves
er of the different methods. very useful.
We assume that the reader is acquainted (if only in For the usually employed fields B 0 ~1T, the frequen-
most general outline) with the physical foundations and cy i'()=a)0/27r lies in the meter range of radio waves
technology of NMR. Nevertheless, Sec. 2 a of this re- (protons have y/2ir = 42.57 MHz/T). Thus the experi-
view briefly presents these topics in the minimum mental technique of NMR belongs to the field of radio-
space needed for introducing the fundamental concepts spectroscopy. The energy of the quanta in the radio
and notation that we shall use later on. We refer those range is considerably smaller than the energy of ther-
wishing to acquaint themselves with this material in mal motion of the atoms, and all the smaller in com-
greater detail to the monographs of Refs. 2-4. parison with the energy of a chemical bond. Therefore
This review has been based on the studies published the study of the inner structure of matter by NMR
up to September 1980. methods causes no physical and chemical changes in
specimens. Thus the NMR methods are nondestructive
2. PHYSICAL FOUNDATIONS OF NMR INTROSCOPY testing methods.
One can observe the NMR signals in two fundamen-
a) General information on nuclear magnetic resonance
tally different ways. In the first of these, which is
The phenomenon of NMR amounts to the selective in- called the stationary method, the specimen is irradia-
teraction of the nuclear magnetic moments of matter ted with a very weak rf field at the constant frequency
with a radiofrequency (rf) field at a definite (resonance) co, while the field B0 is slowly varied so as to pass
frequency. In order to observe NMR, one places the through resonance (or conversely, the field B0 is kept
specimen to be studied, which contains nuclei having a constant, while the frequency u> is slowly varied). In a
spin/#0 (e.g., *H or 19F), in a strong enough dc mag- state of thermal equilibrium, the lower magnetic sub-
netic field having the induction B 0 . The quantization of level is somewhat more populated than the upper level,
the spin moment gives rise to 21 + 1 magnetic sublevels while the probabilities of transitions between the mag-
that differ in orientation of the nuclear spin and in the netic sublevels (1 — 2) stimulated by the rf field are the
associated nuclear magnetic moment (i1 in the field B0. same. Therefore the interaction of the nuclei with the
In the simplest case of 7=1/2, there are only two such resonance field leads to absorption of energy from the
sublevels, and here the energy spacing between them latter; this is detected by using ordinary radiotechnical
is methods. The outcome of such an experiment is the
NMR spectrum—the relationship of the absorption co-
A? = 2n,B0 = fiyB0.
efficient x" to the field B 0 , or equivalently, to the fre-
Here y =n1/KI is the nuclear gyromagnetic ratio. quency it). The sensitivity of this method is small,
while it takes a rather long time to record the spec-
The interaction of the nuclear spins with the other
trum.
degrees of freedom (thermal motion of the molecules
in a liquid, crystal-lattice vibrations in solids) gives The other method, termed the pulsed method, con-
rise to longitudinal magnetic relaxation with the char- sists in subjecting the specimen to a short and suffi-
acteristic time T,. This establishes thermodynamic ciently intense pulse of an rf field at the resonance fre-
relaxation: the preferential occupation of the lower quency with the duration ts = ir/fyB[. Such a pulse syn-
magnetic sublevel and the appearance of a macroscopic chronously rotates by 90° all the nuclear magnetic mo-
nuclear magnetization M directed along the field B0. ments, which were initially oriented along the direction
B 0 1|2. Immediately after such a "ir/2-pulse" ceases,
Transitions between the nuclear magnetic sublevels
Larmor precession of the macroscopic nuclear magne-
can be excited by a rotating, circularly polarized mag-
tization M in the xy plane occurs—the so-called free
netic field (Blcosii}t, -Blsinut) or by an ac magnetic
induction (FI). It is detected by the emf induced in an
field 2.B, coswf linearly polarized in a plane perpendicu-
appropriately oriented receiver coil.
lar to BO that has the frequency o> and satisfies the
resonance condition With the passage of time, the precession of the indi-
- TB0s co 0 . (1)
vidual nuclear magnetic moments becomes dephased
M
owing to spin-spin interactions. This leads to a free-
This same phenomenon can also be described in induction decay (FID) having the characteristic trans-
classical language. In the field B0 the nuclear mag- verse spin relaxation time T 2 . The decay of the FI sig-
netic moments undergo Larmor precession around the nal also arises from the scatter in the resonance fre-
direction of B 0 . The frequency of this precession is quencies of the individual nuclear spins, which is as-
defined by the condition (1). A linearly polarized field sociated with the spatial inhomogeneity of the field B0.
can be resolved into two circularly polarized fields However, in this case the decay of the FI signal is not
rotating in opposite directions. The component of the irreversible, and the nuclear magnetic moments can
circularly polarized ac field that rotates synchronously be again phased by employing additional rf pulses. As
with the nuclear moment interacts effectively with it a result a "spin echo" signal arises that reproduces
and consequently alters its orientation. The other com- the shape of the FID.

842 Sov. Phys. Usp. 24(10), Oct. 1981 Atsarkin eta/. 842
 The observed FI signal amounts to a superposition tissues, and organs that differ only slightly in compo-
of the free-induction signals from all the nuclear spins sition.
of the specimen. Thus it contains full information on
the frequency NMR spectrum, to which it is related by Thus the study of the spatial distribution of NMR sig-
nals makes it possible to compose a picture of the
Fourier transformation. Instruments that employ this
method for obtaining spectra and are equipped with a state of aggregation and chemical structure of an ob-
ject inhomogeneous in composition, i.e., to solve fully
computer for Fourier transformation are called Four-
the fundamental problem of introscopy.
ier NMR spectrometers. While more complex and ex-
pensive than the spectrometers for observing station- In order to obtain NMR images, one must distinguish
ary NMR signals, they enable a considerably higher the NMR signals that arrive at the receiving system
sensitivity to be achieved, which is equivalent to short- from different parts of the specimen, i.e., label them
ening the time for obtaining the spectra.9 by using some parameter that depends on the coordi-
nates. One can take as such a label the frequency o)0
The information that one can obtain by using NMR
of nuclear magnetic resonance. If one makes the mag-
is very extensive. The area under the contour of an
netic field B0 spatially inhomogeneous, i.e., dependent
NMR absorption line and the initial amplitude of the
on the coordinates: B0£Bo(*,;y ,2), then the NMR fre-
FI signal are proportional to the concentration of nu-
quency, which is proportional to the coordinates, will
clei of the given type. Further, the natural line width
be related to the coordinates:
<5, which is related to the relaxation time by 6 ~1/T 2 , is
to0 E= o)0 (x, y, z) = vS» (x, y, z).
highly sensitive to the state of aggregation of the ma-
terial and to the molecular mobility. In so lids the mag- According to this relationship, each value of the fre-
nitude of 5 is usually determined by the magnetic di- quency of the resonance signal can be correlated with
pole interaction between the nuclei, and attains values the volume element from which it was obtained. This
of the order of 104 Hz. In liquids, where the dipole is the general principle of most of the methods of ob-
interactions are averaged out by the rapid motion of the taining NMR images.
molecules, the width of the NMR line declines sharply,
In principle the resolving power of these methods is
and usually amounts to from 100 to 1 Hz. In this case
restricted only by the degree of inhomogeneity of the
one can resolve the fine structure of the spectrum that
field B0 attainable in practice and by the natural width
arises from the interaction of the nuclear spins with
5 of the NMR line. Thus, the resolution Ax along the
the electron cloud of the atom or molecule (the so-call-
x axis is determined by the relationship
ed chemical shifts of the NMR frequency), and also
from the scalar component of the nuclear-nuclear in- A,*^_. (2)
teraction that does not vanish upon motion. High-reso-
Here G, is the gradient of the field B0 in the x direc-
lution NMR spectra have the greatest information con-
tion (as always in NMR, we assume the field B0 to be
tent: they enable one to decide on the subtle details of
unidirectional. That is, it is characterized by a single
the chemical bonding and structure of molecules.10
component B0,, so that, e.g., G,,= 3B0,/9x, etc.). We
The longitudinal spin relaxation time r{, which de- see from (2) that one can attain a better resolution
pends strongly on the degree of molecular mobility and only when the line width is small, i.e., in liquid media.
is especially sensitive to the presence of paramagnetic Therefore one deals in NMR introscopy preferentially
ions and free radicals, also conveys substantial infor- with liquid or "semiliquid" objects. In studying bio-
mation. logical structures, one is mainly concerned with the
spatial distribution of the physiological liquids, i.e.,
b) The principle of NMR introscopy ultimately with water, in the organ or organism under
study. The prospects of NMR introscopy of solids have
The procedure of NMR introscopy amounts to the as yet only been hinted at (see Sec. 3 h).
construction on a recording instrument (on an oscillo-
Upon adopting for example the values G.,= 0.1 T/m,
graph or chart recorder, or in the memory of a com-
y/2ff = 42.6 MHz/T (1H nucleus), and 5 = 1 Hz, we ob-
puter) of a so-called NMR image. The status of each tain from (2) the value Ax =0.2 >im. This satisfies the
of its elements, e.g., the brightness and color at the strictest requirements on image quality. Anticipating
given point on an oscilloscope screen, is determined later discussion, we not that this resolution has not yet
by the amplitude of the NMR signal from the corre- been reached in practice, though the reasons for this
sponding volume element of the object of study. This
are more of technical than fundamental nature.
amplitude is proportional to the number of resonating
nuclei. Hence the image thus obtained primarily char- We must stress that, in contrast to optics, the spatial
acterizes the spatial distribution of the nuclear spins — resolution of the NMR-introscopic methods is in no
the so-called spin density p(x,y , z ) . Moreover, with an way limited by the wavelength of the rf field employed
appropriate methodology, the observations of the NMR (\ = 1-10 m). Of course, there is no violation of the
signals provide information also on the nuclear relaxa- uncertainty principle here. The point is that the spatial
tion times. As we have noted above, the latter depend selection (singling out of the volume element) is per-
very highly on the state of aggregation and the chemi- formed in this case by employing two fields at the
cal composition of the material being studied. This same time: a dc field B0 and an rf field 2Bj cos ut,
situation is very valuable, since it enables one to which are spatially coupled by means of resonating
produce in an NMR image contrast between materials, nuclear magnetic moments. This characteristic fea-

843 Sov. Phys. Usp. 24(10), Oct. 1981 Atsarkin et al. 843
ture of the method is reflected in the name that Lauter-
bur 5 proposed for it of "zeugmatography" (from the
Greek £tvyp.<x, that which binds), which has become
rather widespread in the scientific literature.
If it were possible to assign a configuration of the
static field B9(x,y , z ) to make its value differ at each
of the points of the volume under study, then a mutual-
ly single-valued correspondence between the Larmor
frequency w0 and the coordinates x, y, and z would
arise. Then the obtaining of NMR images would con-
stitute no special problem. However, the Maxwell FIG. 2. One-dimensional projections of Hie spin-density
equations do not allow such a structure of the magnetic distribution from a specimen in the form of two homogeneous
field. For any configuration of B$(x,y,z), two-dimen- cylinders oriented along the z axis for different directions of
the gradient of the magnetic field G.
sional surfaces exist on which the value of B0 (and
hence also of u> 0 ) does not vary. The nuclear spins
lying in a thin layer of the specimen adjoining such an rection.
" isochromatic" surface unavoidably contribute to the
same NMR signal. That is, they prove indistinguish- In order to obtain information on the distribution of
able. In this regard, the construction of the final pic- nuclei in the xy plane, one repeats this experiment
ture of the distribution of nuclear spins throughout the many times, while rotating the direction of the field
volume faces difficulties. The various methods of sur- gradient or the specimen each time around the 2 axis
mounting them constitute the essence of the many by some angle. This procedure is illustrated in Fig.
methods that have been developed up to now. We shall 2 with the example of three projections for the case in
examine the most widespread and promising of them which the "specimen" consists of two parallel cylinders
thoroughly in the next section. of differing diameters.
One can construct the two-dimensional projection
3. METHODS OF PRODUCING NMR IMAGES of the object on the xy plane from the set of one-di-
a) Reconstruction from projections
mensional projections thus obtained by employing a
special computer program. The methods of solving
The first method of obtaining NMR images was pro- this type of problems are well known and are employed,
posed by Lauterbur in his pioneer study. 5 In this meth- in particular, in reconstructive x-ray tomography.1'11
od one superimposes on the spatially homogeneous We note that the very principle of reconstructing an
component of the magnetic field Bw = const a constant object from its projections seems very natural: in
grandient G I = const#0, G S =G X = 0. Thus the resultant essence we act in this very way when we turn in our
field increases linearly in the direction of the x axis, hands a transparent object of interest to us and exam-
while the condition for NMR at the frequency of the rf ine it from different sides.
field has the form
In the example given in Fig. 2, the construction of the
u> = o>0 (z')== f(Bol>+ Gxx'). two-dimensional projection exhausts the problem,
Evidently this condition is satisfied only in a plane since for specimens homogeneous along the z axis, all
layer of thickness Ax perpendicular to the x axis (Fig. cross-sections perpendicular to this axis are equiva-
lent. Yet if cylindrical symmetry is lacking, one must
1).
extend the experiment by rotating the direction of the
The NMR signal at the fixed frequency u> arises from gradient gradually about a different axis. However, in
the volume elements lying in the given layer, while the practice, one more often proceeds in a different way.
dependence of the amplitude of this signal on u> (or on One singles out a thin "working" layer in the specimen
BM), i.e., the NMR spectrum, defines the one-dimen- perpendicular to the z axis, and constructs its two-
sional projection of the function p(x,y ,z) on the x di- dimensional NMR image, or tomogram, without dis-
placing the gradient of the field B0 outside the plane of
this layer. Then one repeats the entire cycle of mea-
surements in other layers to obtain a stack of tomo-
grams, which constitutes the full volume image of the
object.
The thin layer is singled out by using a special
(plane) shape of the rf coils of the NMR spectrometer
or by one of the selective methods described below in
Sees. 3d,e.
In a discrete (numerical) processing of the data, one
FIG. 1. Singling out the layer (hatched) that gives an NMR constructs the image from a finite number of elements
signal at the frequency w in the presence of the magnetic- (cells), each of which corresponds to an element of the
field gradient Gx. Below—NMB frequency distribution along volume being studied (Fig. 3). If the number of "steps"
the x axis. along each of the orthogonal directions is n, the two-

844 Sov. Phys. Usp. 24(10), Oct. 1981 Atsarkin et al. 844
 4-

FIG. 4. NMB image of a cross-section of two parallel capil-

laries containing water.8

FIG. 3. Dissection of a two-dimensional object into elements

(above) and the one-dimensional projection of the spin density
p (x) for discrete processing of the data. This same first experiment showed the possibility of
selection in terms of nuclear relaxation times. In or-
der to do this, one of the test tubes was filled with a
dilute solution of a manganese salt instead of water.
dimensional image contains n2, and the three-dimen- As we know, this greatly shortens the spin-lattice re-
sional image w3 such elements. Evidently, in order to laxation time of the hydrogen nuclei. When the ampli-
fill all these cells, one must have the same number of tude of the resonance rf field was large enough, the
equations relating the unknown values of the spin den- NMR signal from the protons of the pure water was
sity in each of the elements to the observed amplitudes saturated, and only the image of the test tube contain-
of the NMR signals. Evidently, each one-dimensional ing the salt solution remained in the picture. Thus a
projection yields only n such equations, since the pro- contrast in the NMR image was observed here that
jections of the volume elements that lie in the same arose from the difference in chemical composition.
column are superposed on one another (Fig. 3). Thus
In the experiment of Ref. 5, the NMR spectra corre-
an unambiguous construction of a two-dimensional
sponding to the one-dimensional projections of the spin
pattern requires n independent measurements (projec-
density were recorded by the stationary method. As we
tions), while a three-dimensional pattern requires w 2 .
have already noted, this possesses a low sensitivity
Hence we see that, to obtain an image with a spatial
resolution of the order of 1 /im, a specimen of size and requires large expenditures of time to obtain an
image. One can substantially improve the situation by
about 10 cm must be divided into n~10 5 layers. In this
case the number of projections reaches an astronomi- employing FID signals that arise after Tr/2-pulses (Sec.
2 a). Just as in continuous recording, Fourier analysis
cal value, while the information thus obtained cannot
of this signal yields a one-dimensional projection of
be handled on any contemporary computer. According-
the spin density onto the defined direction. In this case
ly, in practice the number of projections does not ex-
all this information is obtained from a single FID sig-
ceed 100-200. As we have said, here one usually re-
nal that decays in the time necessary for dephasing of
stricts the treatment to two-dimensional images (tomo-
the precession of the nuclear magnetic moments of
grams). The number of image elements can also be
the whole specimen in the inhomogeneous magnetic
larger than the number of original data. In this case
field:
one employs special algorithms that solve the problem
of underdetermination. Of course, this involves losses if
in the resolving power of the method.
Here L^ is the linear dimension of the specimen in the
It is instructive to examine the result of Lauterbur's x direction. Since this dephasing is reversible, the
first experiment performed in 1973 (Fig. 4).5 This FID signals can be repeated many times by the spin
shows a transverse section of two test tubes containing echo method so as to accumulate the information dur-
water, 1 mm in diameter and lying 3 mm apart. The ing the irreversible transverse relaxation time T 2 .
image was obtained by using only four projections (the Consequently one obtains a gain in sensitivity of about
direction of the gradient of B0 was rotated by 90°) at a factor of -fn".
c = 60 MHz and | grad5 0 | «0.02 T/m. The number of
Examples of images obtained by the method of recon-
image elements is 20X20, and the attained resolution
struction from sections are given in Re£s. 13-16. One
was about 0.2 mm.
of them is shown in Fig. 5.16 This shows a cross-sec-
In order to demonstrate the analytical potentialities tion of the test object, which was made from a Teflon
of the method, the two test tubes were immersed in a washer having an N-shaped hole. The washer was
vessel containing heavy water (E^O). As we should ex- placed in an ampoule containing water. The plane of
pect, the deuterium nuclei, which have a different gy- the image is fixed by the height of the receiving coil.
romagnetic ratio from the protons, gave no contribu- The dimensions of the washer and its orientation in the
tion to the observed pattern. gap of the magnet are shown in Fig. 5a. The forms of

845 Sov. Phys. Usp. 24(10), Oct. 1981 Atsarkin et at. 845
FIG. 5. Obtaining of the NMB image of a test object in the
form of a Teflon disk having an N-shape hole filled with
water.16 a) The object and its one-dimensional projections; FIG. 6. Switching sequence of the mutually orthogonal mag-
b) the NMB image obtained by reconstruction from projec-
netic field gradients Gx, Gy, and Gt (above) and the observed
tions.
free-induction decay signal S (t) (below) in the method of Four-
ier introscopy.18

the spectra corresponding to different orientations of

the gradient of B0 are shown in the upper part of the S (0 = j J j Ax Ay dz p (x, y, z) exp ii (yG^tx + fGayta + yG,zt,)}.
diagram in 30° steps. Actually the gradient was rotat-
ted in 1° steps, and the image was constructed from (3)
(We shall neglect relaxation in order to simplify the
180 projections. The gradient amounted to 40 juT/cm,
presentation.)
and the NMR frequency was 8.13 MHz. After process-
ing on a minicomputer, the resulting image was drawn The expression (3) amounts to a three-dimensional
on an xy plotter. This made it possible to obtain an Fourier image of the spin density treated as a function
axonometric projection of the test object (see Fig. 5b). of the Larmor frequencies a) x =yG»*> w,=yG^y, and
We see that a high definition of imaging was obtained u,=yGgz, which are related unambiguously to the co-
with a spatial resolution of about 0.3 mm. ordinates. In an experiment one measures the FID
signals as a function of t, (with f x , f y = const), and then
Recently considerable progress has been attained in
repeats the measurements many times, while success-
speed and resolving power of the method. Thus, the
ively varying t% and t,. The set of values of the function
first true three-dimensional NMR images of biological
S(t) thus obtained is subjected to three-dimensional
objects have been recently described. They contained
Fourier transformation, and one finds the function
33X33X33 elements and were obtained in only 7 min. 17
p(x,y,z), i.e., the NMR image of the studied object.
b) Fourier introscopy Analysis shows that one must take 2n steps for each
of the intervals tx, ty, and t, to obtain a picture made of
While the methods of Fourier spectroscopy were
w X n X w elements.18 The total amount of information to
employed in the example given above only to improve
be stored in memory is too large in this case. Hence,
the sensitivity, in the "Fourier zeugmatography" devel-
just as in the Lauterbur method, in practice one usual-
oped by Kumar, Welti, and Ernst 18 they are employed
ly restricts the treatment to two-dimensional images.
for reasons of principle. The method was developed to
In particular, in the study18 cited above, an image of a
eliminate the inconvenience and loss of time involved
cross-section of two test tubes containing water con-
in the multiple rotations of the specimen (or of the field
sisting of 64X64 elements was constructed. The field
gradient) in obtaining the projections according to the
gradients employed in this experiment were about 2
method of Ref. 5. The authors18 proposed restricting
xiO" 3 T/m, while the time spent on the entire experi-
the procedure to only three successive orientations of
ment amounted to 15 min, of which the printer at the
the gradient of 2?0 (along the x, y, and z axes). Here
output of the computer took 7 min.
all three switchings of these directions must be per-
formed in the course of a single decay of the free-in- In addition to the decreased number of orientations
duction signal following a ir/2 pulse. of the gradient of J30 that we have already noted, a con-
venience of this method consists in the fact that a se-
The sequence of operations in this method is shown quence of Fourier transformations is performed here
in Fig. 6. Immediately after cessation of the ir/2 pulse instead of the complex algorithm for reconstruction
(at the instant t = 0), the magnetic-field gradient G,
from projections.11 These transformations are widely
is applied to the specimen. After the time t, this is applied in NMR technique, and are automatically real-
suddenly replaced by G y , and after the further time ty ized in the modern commercial spectrometers.19
by G,. The free-induction signal is observed under
these conditions as a function of the time t,. The nu- On the other hand, the fast switching of the magnetic-
clear spins lying in the volume element AxAyAz having field gradients can give rise to induction currents,
the coordinates xyz precess about the direction of B0 which are undesirable in biomedical applications.
successively with the angular velocities o)0 +fGxx, u>o c) Selective methods; magnetic focusing
+yGy , and u)0 +yG,z. Hence the envelope of the FID
signal from the entire specimen at the instant of time The two methods described above (reconstruction
t=tx + ty + t. has the form from projections and Fourier introscopy) belong to the

846 Sov. Phys. Usp. 24(10), Oct. 1981 Atsarkin et al. 846
class of integral methods in terms of the structure of spatially localized regions in the Earth's magnetic
the recorded information. The NMR signals directly field can prove to be of interest also as applied to ex-
recorded in the course of the measurements contain tended objects of nonbiological nature, e.g., in prob-
information simultaneously on many (or even on all) ing mineral resources.
volume elements of the specimen, while the informa-
tion on each of them individually becomes accessible R. Damadian and his associates (USA) have made deci-
only after complicated mathematical processing of the sive advances in the methods of magnetic focusing.
results. The advantage of the integral methods con- They employed a special configuration of both a dc and
sists in their higher sensitivity, which is attained by a high-frequency magnetic field to single out a re-
collecting the signals from the whole specimen. At the stricted region in the object being studied.8'22"24 As we
same time, these methods involve a complicated analy- have mentioned above, the topological characteristics
sis of the primary information and are highly sensitive of a dc magnetic field do not allow one to localize it
to variation in the gradients of £0 in space and time, within a restricted region of space. Nevertheless, one
and also to phase shifts and instability in the circuits can create a configuration of the magnetic field B0 in
of the NMR spectrometer. which it will be considerably more homogeneous in a
relatively small volume than in the remaining regions
These difficulties do not arise in another approach to of the specimen. The spectrum of the NMR signal
constructing NMR images, which we can call the se- from the nuclei contained in this volume will be nar-
lective approach. In these methods the signal applied rower and its amplitude larger than those from the
to the receiving device is collected from only one vol- nuclei lying in adjacent regions. If, moreover, one
ume element, whose location is then shifted throughout takes care with the special construction of the rf coil,
the specimen according to an established rule. The which also enables a certain concentration of the rf
image is constructed point by point and line by line, field in a defined region, one can attain the desired
just as is done in forming a television image. spatial selection of the NMR signal with a resolution
of the order of 1 mm. 8 By this method, Damadian and
The methods of singling out (selecting) small volume his associates obtained in 1976 the image of a cross-
elements, which we shall term below the "working" section of a living mouse that reliably distinguished the
volume elements, can differ: from the somewhat contours of the heart and lungs.8 The cover of the is-
straightforward, but not at all simple to execute tech- sue of Science in which his article was published show-
nically, idea of magnetic focusing,8 which will be ed the first NMR image of a malignant tumor implanted
treated in this section, to the more refined methods, in a living mouse. Here the contrast arises from the
including periodic modulation or pulsed switching of fact that the nuclear spin-lattice relaxation time in the
the field gradients (Sees. 3 d,e). tissues affected by the malignant tumor is considerably
As we see from its name, the principle of magnetic larger than in healthy tissues.25 The NMR images of
focusing consists in concentrating a dc and/or a reso- cross-sections of a living person at the level of the
nance high-frequency magnetic fields within the bounds chest, which have also appeared on the covers of sci-
of the volume element being examined. The simplest entific journals, 22 " 24 are even more impressive. The
scheme of this type leaves the field B0 homogeneous, experiments were performed in a magnetic field of
whereas the configuration of the rf field, which is fixed about 0.1 T (the diameter of the solenoid was 1.35 m).
by the geometry of the corresponding coils of the NMR Here the construction of the whole image with a reso-
transducer, enables one to irradiate selectively only a lution of about 6 mm took from 4.5 hours in the first
small fraction of the object of study. Thus, upon plac- experiments22 to 35 min in the following experiments.24
ing the patient in the field B 0 , one applies to different An evident advantage of the method of magnetic
parts of his body a miniature rf coil, similarly to what focusing is the possibility of selectively studying any
a physician does in listening to a patient with a stetho- part of the studied object, e.g., an organ under suspi-
scope. cion. The fundamental defect of the method is the rel-
atively low sensitivity and the complexity of focusing
The greatest advances in realizing this method have
the magnetic field. The latter difficulty even more
been attained in the laboratory of Professor Bene
heightens the need for scanning the defined small vol-
(Switzerland). 20 ' 21 Here the magnetic field of the Earth
ume (the "resonance aperture" in the terminology of
was employed as B0. Then the NMR frequency of pro-
the authors of the method) throughout the object of
tons is very low—only about 2 kHz. Therefore an ac-
study. This problem is solved only by using a compli-
ceptable signal-to-noise ratio of the order of 100 is
cated design of the magnet that was developed with a
reached only with working volumes of the order of a
computer. The basis of this magnetic system is a gi-
cubic decimeter. The resolving power of this method
gantic superconductive solenoid.26'27
is so low that here one can hardly speak seriously of
NMR images of biological objects. Nevertheless, by
employing the appreciable differences in nuclear spin d) The method of the sensitive point (line)
relaxation times in different human organs, Bene and
Another example of selective NMR introscopy is the
his colleagues have been able reliably to distinguish the
method of the sensitive point, which was proposed by
NMR signals from the blood filling the heart, from the Hinshaw in 1974,7 and then perfected and reduced to
stomach contents, from the urinary bladder, etc. biomedical applications in the laboratory of E.R.Andrew
We note that the observation of NMR signals from at Nottingham University.28"41 The essence of the method

847 Sov. Phys. Usp. 24(10), Oct. 1981 Atsarkin et al. 847
is that the entire specimen apart from one small vol-
ume element (the "sensitive point") is subjected to an
audiofrequency ac magnetic field. Under these condi-
tions one synchronously detects the NMR signal at the
frequency w0 corresponding to the value of B0 at the
sensitive point. Evidently the dc component of the out-
put signal contains only the contribution from the nuclei FIG. 8. Behavior of the nuclear magnetization M in the sys-
lying in the region of the sensitive point, since the tem of coordinates rotating along with the vector BJ in the sta-
tionary-free-induction method at the time of the first ir/2 pulse
NMR spectrum from the rest of the specimen proves (a) and of the second f/2 pulse (b).
to be modulated in frequency.
This method is realized in practice by employing
three time-variant magnetic-field gradients whose point (x 0 ,y 0 ,2 0 ). Scanning of the latter allows one to
directions are mutually orthogonal, while their mag- construct a three-dimensional image of the entire ob-
nitudes oscillate at different frequencies. In order to ject of study.
understand what happens here, let us first examine one
such gradient (e.g., along the x axis), which has the We note immediately that, in the contemporary vari-
form ant of this method,31"40 the third gradient, e.g., G, is
left unmodulated (static). The FI signal observed under
G, (t) = C, sin Q!*. these conditions arises from the sensitive line (x^^),
In this case the magnetic field in the specimen var- along which the frequency of nuclear precession varies
ies according to the law according to a linear law. Thus, each element of the
sensitive line proves to be labeled with the frequency
B, (x, t) = B00 + (x — x0) Gx sin Qtt. U0(x0,ya,z). Hence a Fourier transformation of the FI
We see from this relationship that the field B9(x,t) de- signal enables one to find directly the spin-density dis-
tribution along the line p(x 0 ,;y 0 ,2).
pends on the time for all x except x=x 0 (Fig. 7a).
Those nuclei of the specimen will contribute to the ob- After this stage is completed, the sensitive line is
served NMR signal that lie in a layer of thickness A* shifted for a small distance in the plane of the cross-
near the plane perpendicular to the x axis passing section being studied, the procedure is repeated, etc.,
through x 0 . until the required tomogram has been obtained.
In practice one often generates the oscillating gradi- The method of recording NMR signals adopted here
ent with a pair of coaxial coils (of the Helmholtz type) is based on the so-called stationary free induction.
supplied with alternating current, and with opposed The specimen is subjected to a sequence of ?r/2 pulses
magnetic fields. If the currents in the coils are the in which the phase of the rf field is in turn shifted by
same, the working plane lies exactly in the middle 180°. Figure 8 illustrates the behavior of the nuclear
between them. By increasing the amplitude of the al- magnetization M in the system of coordinates rotating
ternating current in one of the coils, one can shift this in the direction of the nuclear Larmor precession at
plane throughout the specimen (Fig. 7b). the frequency av Before the pulses begin to act, the
magnetization lies along the z axis. After the first
Analogously, a gradient Gy(t) oscillating with a fre-
pulse, during which the rf field B, lies along the x axis,
quency ^2 * nt defines a plane perpendicular to the y
the magnetization is rotated by 90° and is oriented
axis. The intersection of the two working planes yields
alongy. After the second pulse, in which B( is direct-
a "sensitive line," or bar of cross-section AxAy paral-
ed along -x, the magnetization returns back to the z
lel to the z axis. And finally, a gradient G,(t) oscillat-
axis, etc. Let the time T between the pulses be much
ing with the frequency O3 * n ( , ttj defines the sensitive
shorter than the characteristic longitudinal and trans-
verse spin relaxation times rl and r 2 . Then the de-
crease in the magnetization during the time that it is
perpendicular to the z axis amounts to Mr^/T, while
its increase owing to longitudinal relaxation in the per-
iod when it is parallel to z is (M0 -M)rjT. Here M$
is the equilibrium magnetization. In a stationary re-
gime these quantities are equal to one another. This
implies that
,f MK f A\

^-"."H-CT./TO- W
This expression determines the amplitude of the free-
FIG. 7. Singling out the working ("sensitive") plane using a induction signal measured in the sensitive-line method.
magnetic-field gradient that oscillates in time with the fre- We see from (4) that, when r, »r 2 as is characteristic
quency JJj. 7 of many liquids, the value of Mslat amounts to about
a) Magnetic-field distribution along the x axis for the times A/ 0 /2. That is, it is relatively large, whereas for
t- 0 and t= ir/fit; b) location of the working plane in the speci- solids, for which T, » T 2 , the signal approaches zero.
men (hatched) and the direction of its displacement as the cur- The sensitivity to nuclear relaxation times implied by
rent in the gradient coils is varied. (4) enables a high contrast of the images with respect

848 Sov. Phys. Usp. 24(10), Oct. 1981 Atsarkin et at. 848
FIG. 9. NMB image of a thin cross-section of the left wrist
of a person obtained in vivo by the sensitive-line method
(the darker regions correspond to larger NMB signals) (b),
and a photograph of the corresponding cross-section in a
cadaver (a) (according to Refs. 35, 38).
FIG. 10. Sequence of operations in the selective-excitation
method.43
a) Singling out the working plane in the object; b) singling out
to the state of aggregation and chemical composition the working line in the plane. Below—spectral composition
of the object being studied, which renders this method of the rf ir/2 pulses.
very attractive.
A theoretical analysis of the resolving power41 has
shown that the effective thickness of the sensitive line of a cross-section of the head of a person containing
along the x or y axis amounts to 128X128 elements took only 150 sec.39 Further in-,
crease in sensitivity and speed can be obtained by com-
3.5
= Ay; (5) bining the methods of the sensitive plane and recon-
Here £ =x or y, while the spatial resolution along the struction from projections. A corresponding example
line is will be presented in Sec. 4 c.
3.8
(6) e) The method of selective excitation
Since T« T,,T 2 , the relationships (5) and (6) imply that Another method for singling out the working region
stronger magnetic field gradients are required here has been proposed and realized by Mansfield and his
to obtain the same resolution than in the method of re- associates.43"50 In this method the stated problem is
construction from projections [see (2)]. solved by a special choice of the frequency spectrum
On the other hand, the undoubted advantages of the of the exciting rf pulses.
sensitive-line method are its relative simplicity and In one of the variants, 43 the specimen is placed in a
the less rigorous requirements on homogeneity of the magnetic field B0 having the gradient G K . Under these
values ofG,,, G,,, andB(. conditions it is subjected to an rf pulse whose spectrum
The noted practical advantages have had the result overlaps the NMR frequencies throughout the specimen,
that but little time has passed from the first experi- apart from the narrow layer Ax (Fig. lOa). This pulse
ments on model objects (still the time test tubes con- has an amplitude and duration that cause the resonating
taining water) to the attainment by the sensitive-line nuclear spins to be rotated by 90°. The precession of
method of high-quality images of biological objects. the nuclear magnetization that arises after the pulse
These images have been reproduced repeatedly in vari- has ceased decays rapidly owing to the dephasing of
ous publications,31'33"39 including the covers of scienti- the motion of the individual nuclei in the inhomogene-
fic and popular-scientific journals. 31 ' 42 One of them, ous field. Thus, after a short time the nuclear spins
which amounts to the NMR tomogram of a person's of the entire specimen, except for the defined layer,
wrist, is shown in Fig. 9. 2) The experiment was per- drop out of play in a time of the order of r{. They
formed with ^ = 30 MHz, B0 = 0.7 T, and gradB 0 subsequently no longer contribute to the overall signal
»10~2 T/m. The resolution attained was 0.4 mm, and in the receiving coil.
the duration of the experiment was 9 minutes. We see Then the direction of the field gradient is changed
in the image practically all the details of the wrist: abruptly by 90°, and a Tr/2 pulse is again applied to the
bones, inside of which one discerns an intense signal specimen. However, now its spectrum corresponds to
from the fat and bone marrow (dark regions), tendons, the resonance frequencies of only the nuclei lying in
muscles, and large blood vessels. the narrow layer Ay (Fig. lOb). Consequently a free-
induction signal arises only from the nuclear spins
Even more impressive results have been obtained
that lie inside the thin bar of cross-section A#Ay paral-
very recently by a refined technique that includes real-
time Fourier transformation of the FI signals on a fast lel to the z axis. This signal is recorded after a new
minicomputer. 39 ' 40 The construction of the tomogram switchover of the field gradient, this time along the z
axis, so that a Fourier transformation of the FID sig-
nal gives the spin-density distribution along the z axis
21 in the defined volume element Ax Ay.
The authors thank Professor E. R. Andrew and Dr. P. A.
Bottomley for permission to use this illustration. It is evident from what we have said that a successful

849 Sov. Phys. Usp. 24(10), Oct. 1981 Atsarkinef al. 849
realization of the Mansfield method requires careful
control of the spectrum of the radiofrequency pulses.
The usual square rf pulses having a constant frequency
throughout the duty cycle and duration tf prove inappro-
priate here. This is because, as we know, their spec-
trum is not only broadened by amounts of the order of
l / t f , but also it contains rather intense side lobes that
cover a considerable frequency range. Hence in Refs. #0)

43-50 "tailored pulses" were employed, i.e., they

were formed from oscillations of different frequencies,
amplitudes, and phases by using a special synthesizer
controlled by a computer. This same computer, which
was a part of the NMR introscope, was employed for
accumulating and processing the NMR signals, for
scanning the working volume element, and constructing
the image.44
Images of biological objects of small dimensions FIG. 11. " NMR diffraction" in a one-dimensional structure,
were obtained by the described method, which was a) One-dimensional grating made of parallel plane layers in the
called linear scanning by its authors: hollow stems of presence of a magnetic-field gradient; b) spectral composition
plants, thin bones, etc.44 The time for constructing of the free-induction (FI) signal; c) envelope of the observed
NMR images in such a section of diameter 1.5 cm with FI signal.
a resolution of the order of 0.5 mm amounted to about
a minute. The obtained picture was not distinguished
by high quality; this apparently involved radiotechnical (Fig. lla). The free -induction signal that arises after
problems arising in the selective excitation. The point the action of a v/2 pulse on such an object evidently
is that the correct functioning of the above-mentioned contains a discrete set of frequencies separated by
sequence of synthesized rf pulses requires very high spacings otyGJ,* (Fig. lib). The beats between the
homogeneity and phase stability of the high-frequency signals from the individual plane layers produce a
field, which is difficult to obtain over any significant series of maxima on the time pattern of the FID, the
volume. Accordingly, small specimens are apparently principal ones of which arise at the times t^ —
an inevitable feature of the given method, whose pros- lie).
pects should most likely be sought in NMR microscopy,
i.e., in studying very small structures, e.g., cell This pattern, which arises from the phase difference
structures. 3 ' An elaboration of this method, which in- of the free -induction signals from adjacent planes, is
volves simultaneous excitation of an entire set of sen- fully analogous to the maxima and minima of illumina-
sitive lines and planes, will be treated in the next sec- tion in optical diffraction by a grating, where the phase
tion. differences involve the difference in ray path from ad-
jacent slits. The noted analogy proves to be far-reach-
f) Discrete periodic structures
ing. It offers grounds for speaking of "NMR diffrac-
tion" and applying the mathematical apparatus develop-
In all the cases treated above, the distribution of nu- ed in optics51 in analyzing this phenomenon.
clear spins p ( x , y , z ) was assumed continuous. Let us
turn now to the situation in which the specimen under The theoretical predictions have been confirmed ex-
study amounts to a regular discrete structure or lat- perimentally in the same study 51 using model objects
tice. The interest in such objects arises from two fac- that amount to a series of thin layers of an organic li-
tors: first, the distinctive nature of the process of quid separated by solid partitions about 1 mm thick.
obtaining their NMR images, and second, the practical As we have noted, the NMR signals from solid speci-
importance and abundance of periodic and quasiperio- mens are strongly broadened by the dipole interactions
dic structures—from crystal lattices to the cellular and hence are practically invisible, whereas the in-
structure of living organisms. Moreover, as we shall tense signals from the liquid regions yield a distinct
see from the material below, an important role in NMR pattern of "NMR diffraction," which allows one to de-
introscopy can be played by artificial periodic objects termine the parameters of the one-dimensional lattice.
created by the method of selective excitation. On the other hand, one can also calculate from this
pattern the arrangement of the solid partitions, which
At first we shall examine, following Ref. 51, a very are not directly detectable, and which fill the remain-
simple one-dimensional lattice consisting of a set of ing fraction of the volume of the specimen (in optics
equally spaced, homogeneous plane layers of thickness this method is based on employing the Babinet prin-
Ax and period lx, and shall assume a magnetic-field ciple).
gradient G, directed perpendicular to these planes
Ingoing to two- and three-dimensional lattices, the
production of the images generally requires successive
"in Ref. 50 a tomogram of the abdominal cavity of a person repetition of the experiments with different orienta-
by the linear-scanning method was successfully obtained. tions of the field gradient. That is, it involves recon-
The experiment took 40 min. struction from projections. However, the reconstruc-

850 Sov. Phys. Usp. 24(10), Oct. 1981 Atsarkin et al. 850
 allow one to construct the image of a complete cross-
section of the human body in several seconds. How-
ever, up to now it has been possible to analyze in such
a time only objects of dimensions of the order of sev-
eral centimeters.43"49 This discrepancy between the
theoretical estimates and practice is explained by
technical difficulties mainly involving the creation in
large volumes of highly uniform rf fields that are com-
plex in composition.
FIG. 12. On the construction of the projection of a two-dimen-
sional periodic discrete structure on the direction of grad B0
g) NMR introscopy in a rotating system of coordinates
that forms an angle a with the x axis.
One of the defects of all the methods described above
is the need of modulation or rapid switching of the
tioti algorithm here can be substantially simplified as magnetic-fieId gradients. This fact complicates the
compared with Lauterbur's method.5 Moreover, in a procedure of obtaining images, and can lead to induc-
number of cases it turns out that a correct choice of tion currents in the object of study, which is especially
the field-gradient direction enables one to obtain all undesirable when this object is a living organism. One
the information from a single projection, thus shorten- of the methods of eliminating this defect has been pro-
ing substantially the duration of the experiment. 45 ' 46 posed by Hoult.52 It consists of making both magnetic
fields spatially inhomogeneous: the dc field B0 and the
What we have said is illustrated in Fig. 12, where a
hf field 2B1cosotf perpendicular to it. That is, in addi-
two-dimensional system has been chosen as the object
tion to the gradient of the field B 0 , one produces a
of study, consisting of a set of parallel strips extend-
gradient of the amplitude of the ac field B t perpendicu-
ing along they axis. Here the distribution of nuclear
lar to it. Thus, if B O HB O , and Gx = 9B^,/dx, then one
spins inside these strips is assumed to be arbitrary.
selects B 1 £(B 1 ) X , while producing the gradient gytt in
We see from Fig. 12 that, if the direction of gradB 0
it along the y or z axis.
makes a small angle a with the x axis, then a single
projection of the spin density on this direction yields Under these conditions each volume element of the
information on the distribution of nuclei throughout the specimen proves to be labeled with two independent
object, although the spatial resolution along the y axis variables, e.g., B0(#) and B t ( z ) . In principle, this en-
is diminished here by a factor of cot y. ables one to construct a two-dimensional image of the
object in the xz plane.
The method described here was the basis for develop-
ment of the "planar" method of obtaining NMR ima- The method proposed52 to realize this suggestion is
ges 45-47 The singling out of the required plane and the explained in Fig. 13. In a system of coordinates that
creation in it of a periodic structure made of parallel rotates about the direction of the z axis with the fre-
strips are conducted here in the same way as in Ref. quency a) along with the nuclear magnetization, as we
43 (see Sec. 3e). At first, one uses synthesized pulses know,2 the nuclear spins sense the time-independent
to saturate the system of nuclear spins throughout the effective field Be having the components (B,),=B0 -(CD/
specimen except for one thin layer. Then, again by y) and (BJI = B 1 . The magnetization vector M in this
using specially shaped pulses, one rotates by it/2 the rotating system of coordinates (RSC) processes about
nuclear magnetization inside the discrete strips Lying the direction of B, with the frequency n e =yB e . This
in this plane. In this case the synthesis of the rf pulses motion is called nutation. If B! » |B O - (w/y)|, then we
is technically even more complicated than in Refs. 43 haveB e = B[, so that wheng, = ( S B t / d z ) ^ Q , the frequen-
and 44. We shall not take up this problem here, but cy n, is a single-valued function of the coordinate z,
refer the reader to Ref. 46. and can serve as a marker for this coordinate.
We can easily convince ourselves that the described The frequency fte was measured in Ref. 52 in terms
procedure for shortening the number of measurements, of the angle <p of rotation of the nuclear magnetization
i.e., reducing the information on the two-dimensional during the time of action of the rf field pulse (see Fig.
object to a one-dimensional projection, can also be ex- 13): evidently the spins, which have rotated through
tended to three-dimensional periodic structures. 48 In
this case all the information on the spin-density dis-
tribution along each of the parallel bars forming the
regular structure can be obtained from a single pro-
jection along a direction that forms the oblique angles
y, (3, andy with the x, y, and 2 axes. The details of
the selective excitation of FID in such a structure (the
method of "multiplane" NMR imaging), as well as the
modification of this method using the spin-echo effect,
are described in Refs. 48 and 49. All these refine-
ments are primarily designed to shorten the time of FIG. 13. Precession of the nuclear magnetization M around
obtaining NMR images. According to the estimates the direction of the effective field in the rotating system of co-
of the authors of Fefs. 48 and 49, applying them should ordinates.

851 Sov. Phys. Usp. 24(10), Oct. 1981 Atsarkin et al. 851
different angles in the same time, give different contri- from a one -dimensional grating consisting of several
butions to the FI signal observed following the pulse. parallel plates made of solid camphor (see Sec. 3f).
These measurements allow one to distinguish the NMR The resolution that they attained amounted to about 0.5
signals from volume elements differing in the z coor- mm with | gradB 0 | = 7.7X1Q"3 T/m.
dinate. As usual, the selection with respect to x is
carried out in terms of the frequency of the Larmour As a maximum program, the same study51 discusses
precession in the field B$(x). the possible achievement of a resolution of the order
of 3 A, which would enable the employment of NMR
Two-dimensional projections of a set of simple ob- introscopy for visualizing crystal lattices. However,
jects (test tubes containing water, etc.) have been ob- the estimates show that, when C-10 Hz, this would
tained by the method described in Ret. 52. A further require field gradients of the order of 103 T/m (i.e.,
development of this method, which has also been pro- 100 kG/cm), which cannot be produced at present.
posed by Hoult,52 consists in defining the third coordi- Moreover, the narrowing of the NMR lines in solids
nate (y in our example). For this purpose it has been that can be attained by the currently known methods
proposed to employ yet another ac magnetic field requires certain bounds on the total variation of Z?0
parallel to the z axis. The frequency of this field is over the volume of the specimen: thus, in order to
close to ft., while the amplitude B2 has a gradient along obtain 5 * 10 Hz, this variation should not exceed 0.1
they direction. Let us transform to the new, already mT. Finally, as analysis shows, a resolution of the
doubly rotating system of coordinates (the second rota- order of several Angstro'm units requires the mainten-
tion occurs about B, at the frequency Si,). Now we can ance of a fantastic degree of constancy of the gradient
show that the FI signals under these conditions will be of B0 within the specimen. Its relative variations from
marked with three independent variables: B9(x), Bt(z), point to point should not exceed 5*10"6 (!).51 Thus the
and Bi(y). This allows us to construct the complete visualization of a crystal lattice by "NMR diffraction"
three-dimensional image without any switchings of as yet remains outside the bounds of real possibilities.
magnetic fields. Of course, this does not exclude applications of this
technique for studying the macroscopic structure of
h) Prospects of NMR introscopy of solid objects solid objects.
Another possible approach to the problem of NMR
As we have already noted, the strong dipole broad-
introscopy of a solid object is based on the idea of pro-
ening of the NMR lines in solids does not allow one to ducing an artificial narrowing of the NMR line only in
apply to them directly the methods of NMR introscopy a small volume element of the specimen. The latter
described above. Nevertheless, this problem does not will give the decisive contribution to the observed sig-
seem hopeless, since recently a number of radiotech-
nal, whereas the broad and weak lines from the rest of
nical methods has been developed that sharply narrow the object will produce only an insignificant background.
the NMR lines in solid specimens. Here we shall not One of the variants has been proposed by Wind and
touch upon the essence of these methods (one can find
Yannoni.54 It is based on the so-called modulation
a detailed description of them, e.g., in Ref. 53). We
narrowing of NMR in solids.55 In this method a sharp
note only that specially chosen sequences of high-pow-
decrease in 6 is attained only at a definite detuning A
er radiofrequency pulses can decrease the width of between the frequency w of the rf field, which is sub-
NMR lines in solid objects by two to three orders of jected to deep modulation, and the Larmor frequency
magnitude, so that the typical values of 6 under these yJBo of the nuclei. Evidently, in the presence of a gra-
conditions can amount to about 10 Hz. dient G,, the condition for narrowing is satisfied only
At first glance, it can seem that this fully exhausts in a narrow working layer lying perpendicular to the x
the problem by liquidating the difference between NMR axis, so that the NMR spectrum gives a one-dimen-
in solids and in liquids. However, this is actually not sional projection of the spin density, just as in the
completely true, since the inhomogeneous width of the Lauterbur method5 (see Sec. 3 a).
NMR lines is decreased simultaneously with the sup- Here, as in Ref. 5, a set of projections is required
pression of the nuclear dipole interactions. This par- for constructing the complete image. However, this
tially compensates the action of the field gradient B0 can be avoided by using the idea of NMR introscopy in
itself, which is necessary in principle for the con- a rotating system of coordinates, as described in Sec.
struction of the images. 3g. When applied to solid specimens, this method must
This difficulty has been overcome by Mansfield and be modified; one of the possibilities consists in using
Grannell, who have proposed a rather complicated se- the so-called "magic angle"
quence of radiofrequency pulses that suppress the di- =- (7)
pole width of the NMR lines, while conserving the role
of the magnetic-field gradient.51 One cycle of such a between the effective Held Be acting in the RSC and the
sequence consists of eight coherent ir/2 pulses, In z axis- (Fig. 14). As we know, fulfillment of the condi-
which the phase of the radiofrequency field alternates tion (7), which can be written otherwise as
in a definite way, while the direction of the gradient of
B0 is reversed in the intervals between the pulses. a,—£-== (8)

By employing such a sequence of pulses, the authors leads to a considerable decrease in the dipole width of
of Ref. 51 obtained a satisfactory diffraction pattern NMR in solids.56 In the presence of the gradients Gx

852 Sov. Phys. Usp. 24(10), Oct. 1981 Atsarkin et al. 852
 take into account the instrumental aspects of the prob-
lem. From this standpoint, one should compare the
various methods described above in terms of two
parameters: the energy E 0 of the signal input into the
receiving system per unit time, and the overall time
FIG. 14. The effective field in the rotating system of coor- T0 of collecting the information. The first of these
dinates under the conditions of the "magic angle" <Jm between quantities is proportional to the number of volume
the effective field Be and the z axis. elements AxAy&z of the specimen that simultaneously
contribute to the measured NMR signal. Thus, if we
assume for a specimen containing n*n*n elements
and g, we have that the sensitive-point method7 yields E0 = l in arbi-
trary units, then the methods that single out a sensi-
B, == B0 (a), Bi = B! (2).
tive line31'43 are characterized by the signal energy
Thus the relationship (8) is converted into an equation £ 0 = w ; in recording a defined plane 6 ' 46 we have£ 0 =n 2 ;
relating the x and z coordinates of the volume element and finally, in the method of reconstruction from pro-
being imaged. jections by Fourier transformation of the FID signal
we have£ 0 =n 3 , since here the FID signal arises si-
Now, as is done in Ref. 57, let us observe the NMR
multaneously from the entire specimen.
signal that arises in the rotating system of coordinates
at the frequency On the other hand, the overall time r0 of the experi-
ment is proportional to the number of intermediate ex-
periments. As a rule, the latter consist in measuring
Then the simultaneous solution of the two equations (8) the FID signal for different values of the field gradients
and (9) unambiguously defines both of the cited coordi- Here the sensitive-point methods with w 3 independent
nates. With an appropriate scanning, this enables the measurements and the method of reconstruction from
construction of a two-dimensional image. projections, for which T 0 ocw 2 , yield the poorest re-
sults. The record-making speed belongs to multiplane
As in the method of Hoult, 52 Sec. 3g, one can make
NMR introscopy, 48 in which all the information is col-
the transition to a three-dimensional picture by em-
lected while recording a single FID signal.
ploying resonance in a doubly rotating system of coor-
dinates. We stress that here an even stronger narrow- Since the signal-to-noise ratio is proportional to
ing of the NMR lines should occur in a solid. 58 •/E 0 /r 0 , an ideal method should combine simultaneous
excitation of the entire specimen with application of
In practice the application of the method of Wind and
time-dependent field gradients that would enable an un-
Yannoni has not gone beyond the construction of a one-
ambiguous marking of all the volume elements. Ap-
dimensional projection of a model object amounting to
parently such a program has not yet been fully realized,
two crystals of adamantane, 54 while the variant that we
although multiplane introscopy employing spin echo48
have described last has not yet been realized at all.
approaches rather close to this goal.
Nevertheless, the unswerving and rapid progress of
the technique of NMR in solids allows us to hope for a The other approach yields substantially different con-
successful solution of the problem in the very near clusions. Here purely practical considerations stand
future. uppermost—such as the choice of working frequency,
the attainable homogeneity and stability of the magnetic
4. APPLIED AND TECHNICAL ASPECTS fields, losses in the specimen, etc. Thus, Hoult and
Lauterbur 60 have analyzed the sensitivity in construct-
a) The problem of sensitivity
ing images of biological objects (in particular, the head,
The main defect, which is inherent in all methods of torso, and extremities of a person). They concluded
NMR instroscopy without exception, is their low sensi- that the dielectric losses can be made negligibly small
tivity, which leads to a need for accumulation of sig- with an optimal geometry of the transmitting and re-
nals and substantially increases the time taken by the ceiving coils. In their view, the decisive role is played
measurements. This feature is a fundamental one and by the magnetic losses, which arise from the induc-
it involves the relatively low frequency of NMR as com- tion currents in the conductive medium, and which
pared with other spectroscopic methods, i.e., the low govern the signal-to-noise ratio at a working frequency
quantum energy in the radio range. Therefore, when v2 10 MHz. At lower frequencies (v< 1 MHz) these
one works on refining methods of NMR introscopy, one losses become insignificant, although the sensitivity
must adopt the aim, not so much to reach the sensiti- declines here as vlft. The numerical estimates per-
vity characteristic, say, of x-ray radiography which is formed in this same study 60 show, in particular, that
known to be impossible, as to enable a satisfactory sig- one can obtain FID with a signal-to-noise ratio of about
nal-to-noise ratio to be achieved in a practically ac- 50 for a volume element of 1 cm 3 filled with biological
ceptable time. tissue (brain substance) at a frequency of 1 MHz, and of
about 250 at a frequency of 4 MHz. Other authors81'62
There are two approaches to analyzing the sensitivity
have arrived at similar conclusions.
in constructing NMR images. The first of them, which
has been most fully presented by Brunner and Ernst,59 Bottomley's63 comparison of all the known practical
is of a general, purely theoretical nature. It does not results (up to 1979) on construction of NMR images

853 Sov. Phys. Usp. 24(10), Oct. 1981 Atsarkin et al. 853
TABLE I. Time expenditures In constructing NMR tomograms The construction of the gradient coils involves no fun-
'J\ min damental difficulties; analogous devices have long
Method Ref-
erence Object
v»
MHz
(',
cm Exper- Calc. Calc. been applied in studying diffusion with NMR (see, e.g.,
iment (a) (b) Ref. 9). Usually, in order to create the field gradient
Reconstruction from 12
along the z axis (i.e., in the direction of B 0 ), one em-
Branch of a 3.13 2,8 32 '32 20 67 38
projections (contin- coniferous tree ploys two coaxial coils connected to oppose one another
uous method of NMR
recording) (see Fig. 7), while the gradients G, andG, are created
Reconstruction from 13 5 tubes 30 8 32x32 0.25 50 6
by a system of linear, current-bearing conductors ly-
projections (pulsed
NMR recording)
containing water ing along the field B0.29
Fourier intros-
copy
IB 2 tubes
containing water
_ - 04x64 S - - Thus, since we possess a firm basis in the develop-
Sensitive point 38 Stem of a 60 0.8 28 <28 12H HflJ 130 ment and manufacture of standard NMR Fourier spec-
green onion
Sensitive line 31, ** Human wrist 30 |8 128x128 8 3 0.4 trometers, it is relatively simple to proceed to making
Linear scanning 34 Tumor on the paw $ 5 1.7 64x04 13 13 12
NMR introscopes. This has recently been done by the
of a living rat
well-known "firucker" company (West Germany), which
Selective excitation 48 Toroid filled with 15 1.4 32X16 (1.08 ._ in practically one year since the first successful dem-
of a plane using the oil
spin echo onstration of the biomedical potentialities of NMR in-
j
troscopy, has begun to offer several types of intro-
Magnetic focusing 24 Human chest 2.1836 - 30-36 - —
scopes that differ in the size of the working region and
Sensitive (ine Human head 4.26 55 - i —
39 128x128 2.r. which allow one to study objects from several cm to
i the dimensions of the human body.
In addition, the instrumentation for NMR introscopy
has a number of characteristic features dictated mainly
by its biological and medical applications.
has proved highly instructive. Assuming the final sig-
nal-to-noise ratio in all the experiments to be about the First of all, the dimensions of the specimen to be
same, this author compares the real time expenditures studied, which in routine NMR experiments usually do
T in obtaining an image, taking into account, in con- not exceed several mm, now can be as great as tens
trast to Fef. 59, all the fundamental technical aspects, of centimeters (small experimental animals) or even
including the time for rotation and translation of the a half-meter (cross-section of the human body). To
specimen, computer calculations, etc. Moreover, he enable the rf field to penetrate such objects without
attempts to reduce all the data to a single scale by re- losing its homogeneity, one must choose a low enough
calculating them for two sets of standard conditions: working frequency: from 10-20 MHz in the first of
a) i/0 = 10 MHz; diameter of specimen d = 10 cm; the cited examples to about 1 MHz in the second one.
number of image elements w2 = 64x64, thickness of This sharply contrasts with the tendency toward the
layer fc = 0.5 cm; b) 5 MHz; 40 cm; 64X64; 2 cm, maximal possible increase in v in ordinary NMR spec-
respectively. troscopy, which is necessary to enhance the sensitivity
and resolving power in terms of chemical shifts (see,
Some results of this analysis, supplemented by newer e.g., Ref. 10). For comparison, we recall that the
data, are given in Table I. working frequency of modern commercial NMR spec-
We see from the table that the methods of the sensi- trometers is as much as hundreds of MHz.
tive line and of selective excitation of a plane yield Naturally, the large volume of the specimen requires
the greatest speed. Here, as we have noted, the form- a special design of the magnet that produces the main
er is considerably simpler to realize for large objects. field B0. With account taken of the values of v given
These conclusions flatly contradict the theoretical above, for protons this field must amount to from 0.5
estimates of the sensitivity according to Brunner and T for an aperture of the order of 10 cm to 0.03-0.1 T
Ernst 59 that were discussed above. The discrepancy with a diameter of the object about 50 cm. While stan-
between the theoretical conclusions and their realiza- dard iron-core magnets are still applicable in the for-
tion in experiments evidently indicates the imperfec- mer case, in the latter case one must employ air-core
tion of the contemporary apparatus. In future as tech- solenoids (Fig. 15). Here the coils can be either ordi-
nology progresses, we should expect a radical shorten- nary38"40 or superconductive. 22 ' 26
ing of the time expenditures as compared with the data
of the table. The time is approaching49 when the image
of a cross-section of the human body of quality not in-
ferior to a television image, should take only about a
second.

b) Instrumentation FIG. 15. Solenoid for ob-

taining NMR images of
The basis of an NMR introscope is the nuclear mag- cross-sections of the hu-
netic resonance spectrometer. In most of the methods man body.38
this is a coherent pulsed Fourier spectrometer sup-
plemented with gradient coils and automatic apparatus
for scanning and recording the image on a display.

854 Sov. Phys. Usp. 24(10), Oct. 1981 Atsarkinef al. 854
 At first glance, the requirements on the homogeneity "chemical map" of the patient in any chosen cross-
of the main magnetic field (and on constancy of its gra- section is a unique, and perhaps the most attractive
dients) do not seem too severe: one can easily calculate feature of NMR introscopy for medicine.
that one must maintain the given field configuration with
At the same time, the insufficiently high sensitivity
an accuracy of only about 10"6 to attain a resolution of
of the method (see Sec. 4a) leads to relatively long
0.5 mm for B 0 *0.1 T and gradB 0 | =10"2 T/m. This
exposures. This involves prolonged immobilization
value does not seem too frightful if we compare it with
of the patient, impairment of the sharpness of imaging
the relative homogeneity of magnetic fields of 10~8-10"9
of pulsing organs, etc. In this regard, introscopic
attained in the instruments for ordinary high-resolution
practice as yet employs only the NMR signals from the
NMR. Nevertheless, if we take account of the large
most abundant nuclei having the maximal values of the
volume of the working zone of the magnet, a certain
factor y —primarily protons *H. The high content of
effort is needed to satisfy the stated requirements.
Further increase in the constancy and linearity of the hydrogen in organic materials determines the immedi-
ate prospects of applications of NMR introscopy.
field gradients being produced would open up a real
prospect for chemical analysis in vivo by using high- A powerful push in this direction has been given by
resolution NMR spectra taken in an introscopic re- the above-cited discovery of Damadian.25 He establish-
gime (see below, Sec. 4c). ed that the spin-lattice relaxation time T, of protons in
malignantly altered tissues is about twice as long as in
We should also note the high requirements imposed
normal tissues (0.2-0.4 s, as compared with 0.1-0.2 s,
on the long-term stability of the magnetic field, which
respectively). We have already noted that contrast
arises from the prolonged period for accumulation of
involving the times rl arises when the appropriate
signals during the construction of NMR images. Just
method is chosen in NMR imaging. For greater per-
as in ordinary NMR spectrometers, this problem is
spicuity, it is often emphasized by coloring the NMR
solved by automatic stabilization of the field with a
tomogram in false colors coding the values of T ( . A
reference NMR signal.
series of such images demonstrating the development
The large volume of the specimen also requires a of a malignant tumor in an experimental animal was at
special design of the receiving and transmitting high- one time reprinted on the cover of the journal Priroda^
frequency coils. We refer readers interested in this Another example was given in Ref. 24, which showed a
problem to Refs. 39, 40, 52, 60, and 65. transverse section of the chest of a patient—a 42-year-
old woman—at the level of the third rib. The diagnosis
Finally we shall point out that the entire process of
was adenocarcinoma of the breast with metastasis to
obtaining an image, including the application of the de-
the right lung. The malignant tumor is visible as a
fined pulse sequences, the control of the magnetic field
light-blue strip that divides the cavity of the lung in
gradients, the scanning of the "sensitive" region, and
two parts. The image was obtained by the method of
the accumulation and interpretation of the output sig-
magnetic focusing, and the exposure time was 36 min.
nals, etc., must be controlled by a fast on line compu-
ter (see, e.g., Refs. 40, 44). Such automation, which We must note that the reliability of diagnosis of can-
is generally characteristic of modern experimental cer based on measuring T, is apparently yet far from
technique, acquires special significance in this case, 100% (a detailed discussion of this problem is contain-
since this is essentially the only way to reduce the ed in the review 69 ). Nevertheless, in the opinion of the
expenditure of time to a practically acceptable level. specialists, this method can already be useful, e.g.,
in prophylactic examinations.69 Work is also being
c) Applications and prospects carried out to increase its reliability by simultaneous-
Currently the application of NMR introscopy in bio- ly monitoring the values of T ( and T 2 . 70
logy and medicine is exciting the greatest interest. In addition to malignant tumors, the objects of study
Along this line, the problem naturally arises of com- of NMR introscopy can be edema (we recall that the
paring the new method with the more traditional meth- amplitude of the NMR signal is highly sensitive to the
ods, mainly with x-ray reconstructive tomography, 1 ' 66 water content) as well as other pathological changes.
which also yields images of thin layers (sections) of There is no space here to discuss all the possible
the object being studied. applications of this method for diagnostic purposes,
The most evident advantage of NMR introscopy, nor can one do this at present with enough assurance.
which has been pointed out by most of the investigators, Intensive searches and studies are needed (and are
is the absence of radiations harmful to the organism. already being conducted) along this u ne . 67 ' 68 ' 71 ' 72
Apparently this eliminates the restrictions character- Special interest has been shown recently in NMR
istic of the x-ray methods on the exposure time, num- imaging of the brain, 39 • 40 ' 6 '' 68J8 ' U which has been
ber of projections, etc. Yet we should stress that the made possible by the development of NMR introscopes
biological effect of ac and dc magnetic fields has as yet with a working aperture of about 55 cm. 40 Figure 16
been studied insufficiently. However, the first data shows one of the newest results.68 The depicted layer
seem to indicate their safety to the patient. 39 ' 67 ' 68 about 1 cm thick was singled out by the sensitive-plane
It is no less important that the nature of the contrast method by using a magnetic-field gradient oscillating
in NMR images fundamentally differs from all other in time (see Sec. 3d and Fig. 7). The tomogram of this
introscopic methods. The possibility of obtaining a layer was constructed by the method of reconstruction

855 Sov. Phys. Usp. 24(10), Oct. 1981 Atsarkin eta/. 855
FIG. 16. NMB tomogram of a sagittal section of the head of a
patient.68 The arrow indicates the pathological structure—a
thrombus-filled aneurysm in the intracranial segment of the
carotid artery (see text). The image was obtained in 2 min
FIG. 17. NMB tomogram of a cross-section of the head of a
from 128 projections. The working plane was singled out by
patient.68 The plane of the section passes through the eyeballs,
the method of Hinshaw.7
the bridge of the nose, and the conchae of the ears. The light
spot indicated by the arrow is a tumor in the cerebello-pontine
from projections by Fourier transformation of the FI region.
signal (Sec. 3 a). The static gradient of the field £„
was produced in the plane of the cross-section being
imaged and had a magnitude of 60 fiT/cm. The rota- NMR introscopy involves the sensitivity of the NMR
tions of the gradient were carried out electrically with- signals to the velocity of a liquid flowing through dia-
out any mechanical displacements at all of the coils or magnetic tubes,71'75 Model experiments have been re-
of the patient. They employed 128 orientations of the ported 76 on obtaining transverse sections of the pul-
gradient (projections). The main field amounted to 0.1 sating flow of a liquid inside a glass tube placed in a
T, the NMR frequency was 4.26 MHz, the number of liquid of the same type, but at rest. The authors of
elements was 128X128, and the exposure time was this study deem possible similar experiments on the
about 2 min. The patient was a 59-year old woman in aorta and other large vessels of man, and even on
whom a pathological dilatation (aneurysm) had been small vessels when the methodology is perfected. The
found in the intracranial segment of the right carotid realization of this idea would make it possible to mea-
artery. The lightest regions in Fig. 16 correspond to sure the "local pulse" at any point of the living organ-
the most intense NMR signals, which arise from the ism.
fat and muscle of the scalp. The bones of the skull and
the dura mater, which contain practically no mobile NMR based on 19F nuclei is practically not inferior
protons, yield weak signals and look dark. The same in sensitivity to proton NMR, though the biomedical
pertains to the various cavities: the maxillary sinus applications of I9 F imaging are far less promising be-
and the frontal sinus, etc. The cerebral cortex looks cause of the extremely small content of fluorine in the
like a diffuse light band, whereas the central light organism. However, this circumstance also can be
matter of the brain is rendered in various gradations utilized: harmless compounds of fluorine can be intro-
of gray. duced into the patient or into an experimental animal
as a contrast material, which one proposes to trace by
The blood in the vessels looks dark, since its move- NMR imaging. This idea has been advanced by the
ment in the inhomogeneous magnetic field averages and authors of Ref. 32, who have performed successful
weakens the NMR signals. Therefore the lumen (cross- experiments on model objects. They also have pointed
section) of the carotid artery has the form of a black out the possible applications of this method for visual-
spot. The clearly distinguished lighter structure around izing the flow of fluorine-containing blood substitutes.
it constitutes the aneurysm, which is filled with a
As the sensitivity of the apparatus has been increased,
thrombus.68 NMR introscopy based on 3I P nuclei has acquired a
Another NMR tomogram obtained63 by the same meth- serious significance. As we know, phosphorus-con-
od is shown in Fig. 17. The patient is a 36-year-old taining compounds (ATP and a number of others) play
woman with the diagnosis of a tumor in the cerebello- an extremely important role in the energetics and
pontine region near the brain stem. In the tomogram metabolism of living organisms. The sensitivity of
3I
the tumor can be seen as a region of elevated bright- P NMR spectrometers is only an order of magnitude
ness. smaller than for protons, whereas the chemical shifts
of the lines of 31P are much larger, and are strongly
In Refs. 67 and 68 such NMR images of cross-sec-
distinguished for a number of the biologically impor-
tions of the brain are correlated in detail with the x-
tant compounds of phosphorus. This situation allows
ray tomograms. They conclude that the NMR method
us to rely on obtaining high-resolution 31P spectra in
yields valuable information on pathological changes in an NMR introscopic regime, i.e., detailed chemical
the brain, including tumors, and proves no more ex-
maps of metabolic processes in the organism being
pensive than the x-ray method. studied. It would be hard to overestimate the impor-
Another direction in the biomedical applications of tance of this result—it suffices to point out the high

856 Sov. Phys. Usp. 24(10), Oct. 1981 Atsarkin et a/. 856
sensitivity of high-resolution 31P spectra to the patho- 5. CONCLUSION
logical changes in ischemia, myocardial infarction,
etc.69 Intensive work is being directed along this line. The entire field of studies described in this review
began to be revealed distinctly and to take shape in
The first successful experiments have been describ- 1974-1975, while by 1980 the leading foreign instru-
ed in Refs. 77-19. Here the authors78 have developed ment-making firms had developed and shipped to cus-
a special methodology that allows one to distinguish tomers the first NMR-introscope units for biomedical
even relatively small chemical shifts. applications. In the same year the first results of
clinical trials began to be discussed in print. 67 ' 68 Un-
NMR introscopy based on nuclei of low abundance doubtedly the very fact of such a vigorous progress of
( I3 C, 17O, etc.) has not been described, although ideas the new (and incidentally not at all cheap) method indi-
have been expressed on employing these isotopes as cates its high prestige in the eyes of the specialists—
markers for special diagnostic purposes.70'71 developers and customers.
Of course, the development of NMR introscopy is not
Other applications of NMR introscopy have as yet at all finished, while its future is yet not fully clear.
been weakly developed, and they do not extend beyond The statements on its significance vary from the ec-
preliminary ideas and proposals. Among them we can static proclamation of a "new era in medicine" (Dama-
note the possible observation of biological processes in dian et a/. 23 ) to the restrained, and sometimes even
selected cross-sections of living plants in the course skeptical attitude on the part of the professionally cau-
of vegetating; harmless radiography of seeds with local tious practicing physicians. Apparently the truth lies
measurements of concentrations and characteristics somewhere between these extreme viewpoints. We
of oils, proteins, and other substances, etc. must agree with one of the founders of x-ray recon-
structive tomography, Nobel prize winner G. N. Houns-
As we know, the "ordinary" broad-line NMR spec- field, who thinks that NMR introscopy will not so much
troscopy, as well as the measurement of nuclear re- compete with the other tomographic methods as essent-
laxation times, are being successfully applied for ially supplement them.66
quality control of margarine, butter, cheese, and
other food products (see, e.g., Refs. 80, 81), and In order to reveal fully the actual practical poten-
also for measuring the moisture content of various tialities of NMR introscopy, we need first of all global
materials.82 The combination of these methodologies biomedical and physiological studies. Along with this,
with the technique of introscopy would enable nonde- we must work on the further improvement of the sen-
structive monitoring throughout the volume of a pro- sitivity, speed, and spatial resolution of NMR intro-
duct in its commodity form. The topic can also arise scopes—perhaps by fundamentally new solutions.
of technical defectoscopy of nonmetallic (e.g., poly- Finally, the problem is becoming timely of producing
meric) wares. However, here the competition from commercial instrumentation for NMR introscopy. In
the traditional methods will apparently be more suc- particular, the latter is associated with the increased
cessful than in biology and medicine. production of pulsed NMR spectrometers, which are a
necessary constituent part of the introscopes. Taking
Granted further progress in spatial resolution, NMR into account the high cost of such instruments.(about
introscopy can prove useful in studying microscopic 100 thousand rubles), we can address not only the
objects and structures, such as the cells of plants and scientific and technical problem, but also the problem
animals, impurities and defects in crystals, etc. Here, of the national economy.
apparently, the methods inspiring the greatest hopes
are selective excitation (Sees. 3 e , f ) , and also a trans- In closing, we should like to call the attention of the
fer of the ideas of NMR introscopy into electron para- readers to the extreme fruitfulness of the magnetic-
magnetic resonance (EPR). 83 ' 84 Thus, the spatial dis- resonance methods, which stem from the famous dis-
tribution of the nitrogen impurity in diamond crystals covery of E. K. Zavoiskii.86 Repeatedly in the past de-
has been found by EPR introscopy, 83 and a spatial cades, the skeptics have predicted the exhaustion of
resolution of about 10 nm has been attained 84 in a this field of science. However, each time it has turned
model object. to us a new, of ten unexpected facet and has given life to new
and topical scientific approaches. For example, it
The same methods can be used also in nuclear quad- suffices to recall the founding of solid-state quantum
rupole resonance (NQR). In particular, the develop- electronics, 8 ' the development of the methods of dy-
ment of an NQR introscopic technique based on 14N nu- namic polarization of nuclei, 88 and the rise of high-
clei has been developed.85 One can use it to detect the resolution NMR in solids.53 Now NMR introscopy ari-
presence of nitrogen compounds at distances up to 80 ses in the same sequence, eloquently indicating that the
cm from the high-frequency coil. The authors85 pro- potentialities of magnetic-resonance methods are yet
posed to use this for detecting explosives in closed far from exhaustion.
volumes.
Undoubtedly, the list presented above of possible 1
E. Gordon, G. T. Herman, and S. A. Johnson, Scientific
applications of the methods of NMR introscopy does American 233, No. 4, 56 (1975).
2
not exhaust its possibilities. It will expand as the A. Abrag-am, The Principles of Nuclear Magnetism, Claren-
methodology and instrumentation are perfected. don Press, Oxford, 1961 (Russ. Transl., IL, M., 1963).

857 Sov. Phys. Usp. 24(10), Oct. 1981 Atsarkin et al. 857
3
A. Loesehe, Nuclear Induction [Buss. Transl., IL, M., Phys. C 7, L457 (1974).
1963], 44
P. Mansfield, A. A. Maudsley, andT. Baines, ibid., Ser. E
4
A. G. Lundin and E. I. Fedin, Yadernyi magnitnyi rezonans. 9, 271 (1976).
45
Osnovy i prlmeneniya (Nuclear Magnetic Resonance. Founda- P. Mansfield and A. A. Maudsley, ibid., Ser. C 9, 1409
tions and Applications), Nauka, Novosibirsk, 1980. (1976).
5 46
P. C. Lauterbur, Nature 242, 190 (1973). P. Mansfield and A. A. Maudsley, J. Magn. Reson. 27, 101
6
P. Mansfield and P. K. Grannell, J. Phys. C 6, L422 (1973). (1977).
7 41
W. S. Hinshaw, Phys. Lett. A 48, 87 (1974). P. Mansfield and A. A. Maudsley, Brit. J. Radiol. 50, 188
8
R. Damadian, L. Minkoff, M. Goldsmith, M. Stanford, and (1977).
48
J. Koutcher, Science 194, 1430 (1976). P. Mansfield and I. L. Pykett, J. Magn. Reson. 29, 355
9
T. Farrar and E. Bekker, NMR Pulsed and Fourier Spec- (1978).
49
troscopy (Buss. Transl., M., 1973]. Mansfield et al,, ibid., 33, 261 (1979).
10 50
J. W. Emsley, 3. Feeney, and L. H. Sutcliffe, High Reso- P. Mansfield, I. Pykett, P. Morris, and R. Coupland, Brit.
lution Nuclear Magnetic Resonance Spectroscopy, 1st edn., J. Radiol. 51, 921 (1978).
51
Pergamon, Oxford, 1965-1966 (Russ. Transl., Mir, M., P. Mansfield and P. K. Grannell, Phys. Rev. B 12, 3618
1968*. (1975).
U 52
B. Gordon and C. T. Herman, Comm. Assoc. Comput. D. I. Hoult, J. Magn. Reson. 33, 183 (1979); 38, 369 (1980).
53
Mach. M, 759 (1971). U. Haberlen and M. Mering, High-Resolution NMR in Solids
12
P. C. Lauterbur, Pure and Appl. Chem. 40, 149 (1974). (Buss. Transl., Mir, M., 1980].
13 54
H. R. Brooker and W. S. Hinshaw, J. Magn. Reson. 30, H. A. Wind and C. S. Yannoni, J. Magn. Reson. 36, 269
129 (1978). (1979).
14 55
P. Lauterbur et al., J. Am. Chem. Soc. 97, 6866 (1975). C. S. Yannoni and H.-M. Vieth, Phys. Rev. Lett. 37, 1230
15
C.-M, Lai, J. Shook, and P. Lauterbur, Chem. Blomed. (1976).
56
Envir. Instr. 9, 1 (1979). M. Lee and W. I. Goldburh, Phys. Rev. A 140, 1261 (1965).
16 57
P. Bendel, C.-M. Lai, and P. Lauterbur, J. Magn. Reson. A. E. Mefed and V. A. Atsarkin, Zh. Eksp. Teor. Fiz. 74,
38, 343 (1980). 720 (1978) ISov. Phys. JETP 47, 378 (1978)].
17 58
P. Lauterbur, Abstracts of Joint ISMAR-AMPERE Intern. V. A. Atsarkin, A. E. Mefed, and M. I. Rodak, Fiz. Tverd.
Conference on Magnetic Resonance, Delft, The Netherlands, Tela (Leningrad) 21, 2672 (1979) ISov. Phys. Solid State 21,
1980, p. 56. 1537 (1979)].
18 59
A. Kumar, D. Welti, and R. R. Ernst, J. Magn. Reson. 18, P. Brunner and H. Ernst, J. Magn. Reson. 33, 83 (1979).
60
69 (1975); Naturwissenschaften 62, 34 (1975). D. I. Hoult and F. C. Lauterbur, ibid., 34, 425 (1979).
19 81
H. Freeman and G. A. Morris, Bull. Magn. Reson. 1, 5 P. A. Bottomley and E. R. Andrew, Phys. Med. Biol. 23,
(1979). 630 (1978).
20 W
G.-J. Be*ne* et al., C. R. Acad. Scl. Ser. B 284, 141 (1977); J. M. Liboveand J. R. Singer, J. Phys. E 13, 38 (1980).
63
B. Borcard et al., ibid. 288, 41 (1979). P. A. Bottomley, J. Magn. Reson. 36, 121 (1979).
2t 64
G.-J. Be'ne', Phys. Rep. 58, 213 (1980). I. L. Pykett and P. Mansfield, Phys. Med. Biol. 23, 961
22
R. Damadian et al., Naturwissenschaften 65, 250(1978). (1978).
23 65
R. Damadian, M. Goldsmith, and L. Minkoff, Physiol. T. Baines and F. Mansfield, J. Phys. E 9, 809 (1976).
66
Chem. Phys. 9, 97 (1977). G. N. Hounsfield, Science 210, 22 (1980).
24 67
R. Damadian, M. Goldsmith, and L. Minkoff, ibid. 10, W. Moore, G. Holland, and L. Kreel, CT: Comp. Tomo-
285 (1978). graphy 4, 1 (1980).
25 68
R. Damadian, Science 171, 1151 (1971). R. Hawkes, G. Holland, W. Moore, and B. Worthington,
26
L. Minkoff et al., Physiol. Chem. Phys. 9, 101 (1977). J. Comp. Assist. Tomography 4, 577 (1980).
27 69
M. Goldsmith et al., ibid. 9, 105 (1977). D. P. Hollis, Bull. Magn. Reson. 1, 27 (1979).
28 70
W. S. Hinshaw, in: Proc. 18th AMPERE Congress, J. Koutchar, M. Goldsmith, and R. Damadian, Cancer 41,
Nottingham, 1974, eds. P. S. Allen, E. R. Andrew, and C. 174 (1978).
71
A. Bates, North-Holland, Amsterdam, 1975, Vol. 2, p. 433. C. M. Lai, W. House, and P. Lauterbur, Nuclear Magnetic
29
W. S. Hinshaw, J. Appl. Phys. 47, 3709 (1976). Resonance Zeugmatography for Medical Imaging: Preprint of
30
E. R. Andrew, Phys. Bull. 28, 323 (1977). State University of New York at Stony Brook, 1980.
31 72
W. S. Hinshaw, P. A. Bottomley, and G. N. Holland, Nature J. Mallard, J. Hutchison, W. Edelstain, et al., Philos.
270, 722 (1977). Trans. R. Soc. London Ser. B 289, 519 (1980).
32 73
G. N. Holland, P. A. Bottomley, and W. S. Hinshaw, J. G. N. Holland, R. C. Hawkes, and W. S. Moore, J. Comp.
Magn. Reson. 28, 133 (1977). Assist. Tomography 4, 429 (1980).
33 74
P. A. Bottomley et al., Phys. Med. Biol. 22, 71 (1977). G. De Chiro, ibid. , p. 210.
34 75
W. S. Hinshaw et al., Brit. J. Radiol. 51, 273 (1978). A. Garroway, J. Phys. D 7, L159 (1974).
35 76
W. S. Hinshaw et al., Neuroradiology 16, 607 (1978). P. Lauterbur and C. M. Lai, in: Proc. of 1977 Devices
36
P. A. Bottomley, Cancer Res. 39, 468(1979). and Techn. Branch Contractors Conference Program, Gov-
37
E. R. Andrew et al., in: Magnetic Resonance and Related ernment Printing Office, Washington, 1978.
Phenomena; Proc. 20th AMPERE Congress, Tallinn, 1978, "P. Bendel, C. M. Lai, and P. Lauterbur, J. Magn. Reson.
eds. E. Kundla, E. Lippmaa, and T. Saluvere, Springer- 38, 343 (1980).
78
Verlag, Berlin, 1979, p. 53. S. J. Cox and P. Styles, ibid., 40, 209 (1980).
38 79
E. R. Andrew, Philos. Trans, R. Soc. London Ser. B 289, P. Bottomley and R. Nunnally, see Ref. 17, p. 21.
80
471 (1980). P. Mansfield and C. Horn, J. Food Techn. 7, 53 (1972).
39 W
W. Holland, W. Moore, and R. Hawkes, J. Comp. Assist. A. F. Babkin et al., Pishchevaya Tekhnologiya 1976, No. 6,
Tomography 4, 1 (1980); Brit. J. Radiol. 53, 253 (1980). 130.
ffi
*°W. Moore and G. Holland, Philos. Trans. R, Soc. London S. P. Gabuda, Yadernyi magnitnyi rezonans v kristallogid-
Ser. B 289, 133 (1980). ratakh i gidratirovannykh belkakh (Nuclear Magnetic Reson-
41
F. T. Meiere and F. C. Thatcher, J. Appl. Phys. 50, 4491 ance in Crystal Hydrates and Hydrated Proteins), Nauka,
(1979). Novosibirsk, 1978.
42 83
E. I. Fedin, Priroda 1980, No. 4, 77. M. Hoch and A. Day, Solid State Commun. 30, 211 (1979).
43 M
A. N. Garroway, P. K. Grannell, and P. Mansfield, J. W. Karthe and E. Wehrsdorfer, J. Magn. Reson. 33, 107

858 Sov. Phys. Usp. 24(10), Oct. 1981 Atsarkin ef al. 858
 (1979). Transl., IL, M., 1961).
K 88
T. Hirschfeld and S. M. Klainer, J. Mol. Struct. 58, 63 c. D. Jeffries, Dynamic Nuclear Orientation, Interscience,
(1980). ^ New York, 1963 (Russ. Transl., Mir, M., 1965).
^E. K. Zavoiskii, Doctoral dissertation, Institute of Physics ''L. G. Dubitskii, Defektoskopiya 1971, No. 6, 53.
of the Academy of Sciences of the USSR, 1944; J. Phys.
USSR 9, 245 (1945).
87
J. R. Singer, Masers, Wiley, New York, 1959 (Russ. Translated by M. V. King

859 Sov. Phys. Usp. 24(10), Oct. 1981 Atsarkin etal. 859