2

Fundamentals of Microscopy
Kenneth R. Castleman and Ian T. Young

2.1

Origins of the Microscope

During the 1st century ad, the Romans were experimenting with different shapes of clear glass. They discovered that by holding over an object a piece of clear glass that was thicker in the middle than at the edges, they could make that object appear larger. They also used lenses to focus the rays of the sun and start a re. By the end of the 13th century, spectacle makers were producing lenses to be worn as eyeglasses to correct for deciencies in vision. The word lens derives from the Latin word lentil, because these magnifying chunks of glass were similar in shape to a lentil bean. In 1590, two Dutch spectacle makers, Zaccharias Janssen and his father, Hans, started experimenting with lenses. They mounted several lenses in a tube, producing considerably more magnication than was possible with a single lens. This work led to the invention of both the compound microscope and the telescope [1]. In 1665, Robert Hooke, the English physicist who is sometimes called the father of English microscopy, was the rst person to see cells. He made his discovery while examining a sliver of cork. In 1674 Anton van Leeuwenhoek, while working in a dry goods store in Holland, became so interested in magnifying lenses that he learned how to make his own. By carefully grinding and polishing, he was able to make small lenses with high curvature, producing magnications of up to 270 times. He used his simple microscope to examine blood, semen, yeast, insects, and the tiny animals swimming in a drop of water. Leeuwenhoek became quite involved in science and was the rst person to describe cells and bacteria.
Microscope Image Processing Copyright 2008, Elsevier Inc. All rights reserved. ISBN: 978-0-12-372578-3

Fundamentals of Microscopy

Because he neglected his dry goods business in favor of science and because many of his pronouncements ran counter to the beliefs of the day, he was ridiculed by the local townspeople. From the great many discoveries documented in his research papers, Anton van Leeuwenhoek (16321723) has come to be known as the father of microscopy. He constructed a total of 400 microscopes during his lifetime. In 1759 John Dolland built an improved microscope using lenses made of int glass, greatly improving resolution. Since the time of these pioneers, the basic technology of the microscope has developed in many ways. The modern microscope is used in many different imaging modalities and has become an invaluable tool in elds as diverse as materials science, forensic science, clinical medicine, and biomedical and biological research.

2.2
2.2.1

Optical Imaging
Image Formation by a Lens

In this section we introduce the basic concept of an image-forming lens system [17]. Figure 2.1 shows an optical system consisting of a single lens. In the simplest case the lens is a thin, double-convex piece of glass with spherical surfaces. Light rays inside the glass have a lower velocity of propagation than light rays in air or vacuum. Because the distance the rays must travel varies from the thickest to the thinnest parts of the lens, the light rays are bent toward the optical axis of the lens by the process known as refraction.

F I G U R E 2 . 1 An optical system consisting of a single lens. A point source at the origin of the focal plane emits a diverging spherical wave that is intercepted by the aperture. The lens converts this into a spherical wave that converges to a spot (i.e., the point spread function, psf) in the image plane. If df and di satisfy Eq. 2.1, the system is in focus, and the psf takes on its smallest possible dimension.
12

2.2 Optical Imaging

2.2.1.1

Imaging a Point Source

A diverging spherical wave of light radiating from a point source at the origin of the focal plane is refracted by a convex lens to produce a converging spherical exit wave. The light converges to produce a small spot at the origin of the image plane. The shape of that spot is called the point spread function (psf ).
T h e F o c u s E q u a t i o n The point spread function will take on its smallest possible size if the system is in focus, that is, if

1 1 1 df di f

(2:1)

where f is the focal length of the lens. Equation 2.1 is called the lens equation.
2.2.1.2 Focal Length

Focal length is an intrinsic property of any particular lens. It is the distance from the lens to the image plane when a point source located at innity is imaged in focus. That is, df 1 ) di f and by symmetry di 1 ) df f The power of a lens, P, is given by P 1/f ; if f is given in meters, then P is in diopters. By denition, the focal plane is that plane in object space where a point source will form an in-focus image on the image plane, given a particular di . Though sometimes called the object plane or the specimen plane, it is more appropriately called the focal plane because it is the locus of all points that the optical system can image in focus.
Magnication

If the point source moves away from the origin to a position (xo , yo ), then the spot image moves to a new position, (xi , yi ), given by xi Mxo where M di df (2:3) yi Myo (2:2)

is the magnication of the system. Often the objective lens forms an image directly on the image sensor, and the pixel spacing scales down from sensor to specimen by a factor approximately equal
13

Fundamentals of Microscopy

to the objective magnication. If, for example, M 100 and the pixel spacing of the image sensor is 6.8 mm, then at the specimen or focal plane the spacing is 6.8 mm/100 68 nm. In other cases, additional magnication is introduced by intermediate lenses located between the objective and the image sensor. The microscope eyepieces, which gure into conventional computations of magnication, have no effect on pixel spacing. It is usually advantageous to measure, rather than calculate, pixel spacing in a digital microscope. For our purposes, pixel spacing at the specimen is a more useful parameter than magnication. Equations 2.1 and 2.3 can be manipulated to form a set of formulas that are useful in the analysis of optical systems [8]. In particular, f di and df fdi M1 f M di f (2:6) di df di M df M1 di df M 1 fdf f (M 1) df f (2:4) (2:5)

Although it is composed of multiple lens elements, the objective lens of an optical microscope behaves as in Fig. 2.1, to a good approximation. In contemporary light microscopes, di is xed by the optical tube length of the microscope. The mechanical tube length, the distance from the objective lens mounting ange to the image plane, is commonly 160 mm. The optical tube length, however, varies between 190 and 210 mm, depending upon the manufacturer. In any case, di ) df and, M ) 1, except when a low-power objective lens (less than 10 ) is used.
2.2.1.3 Numerical Aperture

It is customary to specify a microscope objective, not by its focal length and aperture diameter, but by its magnication (Eq. 2.3) and its numerical aperture, NA. Microscope manufacturers commonly engrave the magnication power and numerical aperture on their objective lenses, and the actual focal length and aperture diameter are rarely used. The NA is given by (2:7) NA n sina % n a=2df % na=2f where n is the refractive index of the medium (air,immersion oil, etc.) located between the specimen and the lens and a arctan a=2df is the angle between the optical axis and a marginal ray from the origin of the focal plane to the edge of the aperture, as illustrated in Fig. 2.1. The approximations in Eq. 2.7 assume
14

2.3 Diffraction-Limited Optical Systems

small aperture and high magnication, respectively. These approximations begin to break down at low power and high NA, which normally do not occur together. One can compute and compare f and a, or the angles arctan(a=2df ) and arcsine(NA/n) to quantify the degree of approximation.
2.2.1.4 Lens Shape

For a thin, double-convex lens having a diameter that is small compared to its focal length, the surfaces of the lens must be spherical in order to convert a diverging spherical entrance wave into a converging spherical exit wave by the process of diffraction. Furthermore, the focal length, f, of such a lens is given by the lensmakers equation,   1 1 1 n 1 (2:8) f R1 R2 where n is the refractive index of the glass and R1 and R2 are the radii of the front and rear spherical surfaces of the lens [4]. For larger-diameter lenses, the required shape is aspherical.

2.3

Diffraction-Limited Optical Systems

In Fig. 2.1, the lens is thicker near the axis than near the edges, and axial rays are refracted more than peripheral rays. In the ideal case, the variation in thickness is just right to convert the incoming expanding spherical wave into a spherical exit wave converging toward the image point. Any deviation of the exit wave from spherical form is, by denition, due to aberration and makes the psf larger. For lens diameters that are not small in comparison to f, spherical lens surfaces are not adequate to produce a spherical exit wave. Such lenses do not converge peripheral rays to the same point on the z-axis as they do near-axial rays. This phenomenon is called spherical aberration, since it results from the (inappropriate) spherical shape of the lens surfaces. High-quality optical systems employ aspherical surfaces and multiple lens elements to reduce spherical aberration. Normally the objective lens is the main optical component in a microscope that determines overall image quality. A diffraction-limited optical system is one that does produce a converging spherical exit wave in response to the diverging spherical entrance wave from a point source. It is so called because its resolution is limited only by diffraction, an effect related directly to the wave nature of light. One should understand that a diffraction-limited optical system is an idealized system and that real optical systems can only approach this ideal.
15

Fundamentals of Microscopy

2.3.1

Linear System Analysis

It should be clear that increasing the intensity of the point source in Figure 2.1 causes a proportional increase in the intensity of the spot image. It follows that two point sources would produce an image in which the two spots combine by addition. This means that the lens is a two-dimensional linear system [8]. For reasonably small off-axis distances in well-designed optical systems, the shape of the spot image undergoes essentially no change as it moves away from the origin. Thus the system can be assumed to be shift invariant (ignoring magnication effects), or, in optics terminology, isoplanatic, as well as linear. The psf is then the impulse response of a shift-invariant, linear system. This implies that the imaging properties of the system can be specied by either its psf or its transfer function [4, 8]. The optical transfer function is the Fourier transform of the psf. The eld of linear system analysis is quite well developed, and it provides us with very useful tools to analyze the performance of optical systems. This is developed in more detail in later chapters.

2.4

Incoherent Illumination

Incoherent illumination may be viewed as a distribution of point sources, each having a random phase that is statistically independent of the other point sources [2, 57]. Under incoherent illumination, an optical system is linear in light intensity. The light intensity is the square of the amplitude of the electromagnetic waves associated with the light [2]. In the following, we assume narrow-band light sources. In general, light is a collection of different wavelengths, and modern microscopy makes extensive use of wavelengths between 350 nm (ultraviolet) and 1100 nm (near infrared). The term narrow-band implies using only a small range of wavelengths, perhaps 30 nm wide around some center wavelength. 2.4.1 The Point Spread Function

The spot in the image plane produced by a point source in the focal plane is the psf. For a lens with a circular aperture of diameter a in narrow-band, incoherent light having center wavelength l, the psf has circular symmetry (Fig. 2.2) and is given by  !32 2 r J p 6 1 ro 7 62 ! 7 (2:9) psf(r) h(r) 4 5 r p ro
16

2.4 Incoherent Illumination

F I G U R E 2 . 2 The incoherent point spread function. A focused diffraction-limited system produces this psf in narrow-band incoherent light.

where J1 (x) is the rst-order Bessel function of the rst kind [4]. The intensity distribution associated with this psf is called the Airy disk pattern, after G.B. Airy [9, 10], and is shown in Fig. 2.2. The constant, ro , a dimensional scale factor, is ro ldi a (2:10)

and r is radial distance measured from the origin of the image plane, that is, q r x2 y2 (2:11) i i 2.4.2 The Optical Transfer Function

Since the imaging system in Fig. 2.1 is a shift-invariant linear system, it can be specied either by its impulse response (i.e., the psf ) or by the Fourier transform of its psf, which is called the optical transfer function (OTF). For a lens with a circular aperture of diameter a in narrow-band incoherent light having center wavelength l, the OTF (Fig. 2.3) is given by [4] & !  !' 8 q q < 2 1 1 cos sin cos q # fc OTF(q) F {h(r)} H(q) p 2 fc fc : 0 q $ fc (2:12) where q is the spatial frequency variable, measured radially in two-dimensional frequency space. It is given by p (2:13) q u2 n2
17

Fundamentals of Microscopy

F I G U R E 2 . 3 The incoherent OTF. This is the frequency response of a focused diffraction-limited system in narrow-band incoherent light.

where u and n are spatial frequencies in the x and y directions, respectively. The parameter fc , called the optical cutoff frequency, is given by fc 1 a ro ldi (2:14)

2.5

Coherent Illumination

Some microscopy applications require the use of coherent light for illumination. Lasers, for example, supply coherent illumination at high power. Coherent illumination can be thought of as a distribution of point sources whose amplitudes maintain xed phase relationships among themselves. Diffraction works somewhat differently under coherent illumination, and the psf and OTF take on different forms. Under coherent illumination, an optical system is linear in complex amplitude as opposed to linear in light intensity, as in the incoherent case. 2.5.1 The Coherent Point Spread Function

For a lens with a circular aperture of diameter a in coherent light of wavelength l, the psf has circular symmetry (Fig. 2.4) and is given by hr 2 J1 pr=ro pr=ro (2:15)

where ro is from Eq. 2.10 and r is from Eq. 2.11.

2.5 Coherent Illumination

F I G U R E 2 . 4 The coherent point spread function. A focused diffraction-limited system produces this psf in coherent light.

2.5.2

The Coherent Optical Transfer Function

For a lens with a circular aperture of diameter a in coherent light of wavelength l, the OTF (Fig. 2.5) is given by [4]   ldi (2:16) H q P q a where q is from Eq. 2.13 and Pq 8 > >1 < > >0 :

1 2 1 jqj > 2 jqj <

(2:17)

F I G U R E 2 . 5 The coherent OTF. This is the frequency response of a focused diffraction-limited system in coherent light.
19

Fundamentals of Microscopy

a 1

y di

r x

0 r= x2 + y2 ro = ldi a 1

1.22ro

2.23ro

3.24ro

fc q= u2 + 2

0 fc = a ldi

fc

F I G U R E 2 . 6 The incoherent point spread function and transfer function. This shows how the psf and OTF depend on wavelength and aperture diameter for a diffraction-limited optical system with a circular aperture.

Notice that, under coherent illumination, the OTF is at out to the cutoff frequency, while under incoherent illumination it monotonically decreases. Notice also that the cutoff frequency in incoherent light is twice that of the coherent case. Fig. 2.6 illustrates the relationships between the incoherent point spread function and transfer function of diffraction-limited optical systems with circular exit pupils.

2.6

Resolution

One of the most important parameters of a microscope is its resolution, that is, its ability to reproduce in the image small structures that exist in the specimen. The optical denition of resolution is the minimum distance by which two point sources must be separated in order for them to be recognized as separate. There is no unique way to establish this distance. The psfs overlap gradually as the
20

2.6 Resolution

points get closer together, and one must specify how much contrast is required if the two objects are to be recognized as distinct. There are, however, two commonly used criteria for comparing the resolving power of optical systems. 2.6.1 Abbe Distance

To a good approximation, the half-amplitude diameter of the central peak of the image plane psf is given by the Abbe distance (after Ernst Abbe [11]),   df 1 di l l 0:5 (2:18) rAbbe l l % M a 2NA NA a 2.6.2 Rayleigh Distance

For a lens with a circular aperture, the rst zero of the image plane psf occurs at a radius   l (2:19) rAiry 1:22ro 0:61 NA which is called the radius of the Airy disk. According to the Rayleigh criterion of resolution (after Lord Rayleigh [12]), two point sources can be just resolved if they are separated, in the image, by the distance d rAiry . (See Fig. 2.7.) In the terminology of optics, the Rayleigh distance denes circular resolution cells in the image, since two point sources can be resolved if they do not fall within the same resolution cell. 2.6.3 Size Calculations

In microscopy it is convenient to perform size calculations in the focal plane, rather than in the image plane as previously, since that is where the objects of

F I G U R E 2 . 7 The Rayleigh criterion of resolution. Two point sources can be just resolved if they are separated, in the image, by the Airy distance.
21

Fundamentals of Microscopy

interest actually reside. The projection implemented by the lens involves a 1808 rotation and a scaling by the factor M (Eq. 2.3). The pixel spacing and resolution can then be specied in units of micrometers at the specimen. Spatial frequencies can be specied in cycles per micrometer in the focal plane. Since df % f for high-magnication lenses (Eq. 2.6), the resolution parameters are more meaningful if we scale them to the focal (specimen) plane rather than working in the image plane. For a microscope objective, the incoherent optical cutoff frequency in the focal plane coordinate system is fc the Abbe distance is rAbbe Ma a 2NA ldi ldf l (2:20)

  df 1 di l l l l % 0:5 M a 2NA NA a

(2:21)

and the Rayleigh distance (resolution cell diameter) is   l dRayleigh 1:22ro 0:61 NA

(2:22)

For l 0:5 mm (green light) and an NA of 1.4 (high-quality, oil-immersion lens), we have fc 5:6 cycles=mm, rAbbe 0:179 mm, and dRayleigh 0:218 mm. The foregoing approximations begin to break down at low power and high NA, which normally do not occur together. Again, one can compute and compare f and a, or the angles arctan(a=2df ) and arcsine(NA=n), to quantify the degree of approximation.

2.7

Aberration

Real lenses are never actually diffraction limited, but suffer from aberrations that make the psf broader, and the OTF narrower, than they would otherwise be [2, 6, 7]. An example is spherical aberration, mentioned earlier. Aberrations in an optical system can never increase the magnitude of the optical transfer function, but they can drive it negative.

2.8

Calibration

Making physical size measurements from images is very often required in the analysis of microscope specimens. This can be done with accuracy only if the
22

2.8 Calibration

pixel spacing at the focal plane is known. Making brightness measurements in the image is also useful, and it requires knowledge of the relationship between specimen brightness and gray levels in the digital image. 2.8.1 Spatial Calibration

The pixel spacing can be either calculated or measured. Calculation requires knowledge of the pixel spacing at the image sensor and the overall magnication of the optics (recall Eq. 2.3). Often it can be calculated from dx Dx Mo Ma (2:23)

where dx and Dx are the pixel spacing values at the specimen and at the image sensor, respectively, and Mo is the magnication of the objective. Ma is the magnication imposed by other optical elements in the system, such as the camera adapter. Usually this is quoted in the operators manual for the microscope or the accessory attachment. The image sensor pixel spacing is quoted in the camera manual. Too often, however, the numbers are not available for all the components in the system. Pixel spacing must then be measured with the aid of a calibrated target slide, sometimes called a stage micrometer. This requires a computer program that can read out the (x, y) coordinates of a pixel in the digital image. One digitizes an image of the calibration target and locates two pixels that are a known distance apart on the target. Then D dx q x2 x1 2 y2 y1 2 (2:24)

where dx is the pixel spacing, D is the known distance on the calibration target, and (x1 , y1 ) and (x2 , y2 ) are the locations of the two pixels in a recorded image. For precision in the estimate of dx, the two points should be as far apart as possible in the microscope eld of view. 2.8.2 Photometric Calibration

Photometric properties that can be measured from a microscope image include transmittance, optical density, reectance, and uorescence intensity. Optical density calibration, as well as that for reectance, requires a calibration target. The procedure is similar to that for spatial calibration [8]. Fluorescence intensity calibration techniques are covered in Chapter 12.
23

Fundamentals of Microscopy

2.9

Summary of Important Points

1. Lenses and other optical imaging systems can, in most cases, be treated as two-dimensional, shift-invariant, linear systems. 2. The assumptions involved in the use of linear analysis of optical systems begin to break down as one moves far off the optical axis, particularly for wide-aperture, low-magnication systems. 3. Under coherent illumination, an optical system is linear in complex amplitude. 4. Under incoherent illumination, an optical system is linear in intensity (amplitude squared). 5. An optical system having no aberrations is called diffraction-limited because its resolution is limited only by the wave nature of light (diffraction effects). This is an ideal situation that real systems can only approach. 6. A diffraction-limited optical system transforms a diverging spherical entrance wave into a converging spherical exit wave. 7. The point spread function of an optical system has a nonzero extent because of two effects: the wave nature of light (diffraction) and aberrations in the optical system. 8. The optical transfer function is the Fourier transform of the point spread function. 9. The point spread function is the inverse Fourier transform of the optical transfer function. 10. For a circularly symmetric lens, the incoherent point spread function is given by Eq. 2.9. 11. For a circularly symmetric lens, the incoherent optical transfer function is given by Eq. 2.12. 12. For a circularly symmetric lens, the coherent point spread function is given by Eq. 2.15. 13. For a circularly symmetric lens, the coherent transfer function is given by Eq. 2.16.

24

References

References
1. WJ Croft, Under the Microscope: A Brief History of Microscopy, World Scientic Publishing Company, 2006. 2. M Born and E Wolf, Principles of Optics: Electromagnetic Theory of Propagation, Interference and Diffraction of Light, 7th ed., Cambridge University Press, 2000. 3. EL ONeill, Introduction to Statistical Optics, Courier Dover Publications, 2004. 4. JW Goodman, Introduction to Fourier Optics Third Edition, Roberts and Company, 2005. 5. C Scott, Introduction to Optics and Optical Imaging, IEEE Press, 1998. 6. E Hecht, Optics, 4th ed., Addison-Wesley, 2002. 7. DJ Goldstein, Understanding the Light Microscope, Academic Press, 1999. 8. KR Castleman, Digital Image Processing, Prentice-Hall, 1996. 9. GB Airy The Diffraction of an Annular Aperture, Philos. Mag. Ser., 1841. 10. GB Airy, On the Diffraction of an Object-Glass with Circular Aperture, SPIE milestone series, Society of Photo-optical Instrumentation Engineers, 2007. 11. E Abbe, Archiv. Mikroskopische Anat. 9:413468, 1873. 12. L Rayleigh, On the Theory of Optical Images, with Special Reference to the Microscope, Philosophical Magazine, 42(5):167, 1896.

25