Journal of Harbin Engineering University ISSN: 1006-7043 Vol 45 No. 5 May 2024 Automatic Auscultation Utilizing Electronic Stethoscope for Early Discovery of Cardiac Diseases N. Babu * Professor, Department of Biomedical engineering, Vinayaka Mission's KirupanandaVariyar engineering College, Vinayaka Mission's Research Foundation (Deemed University) Salem — 636308, India. * Corresponding author email id~N. Babu Abstract ‘A novel embedded system was designed for early detection of cardiac diseases using electronic stethoscope. Phono cardiagram (PCG) signalwas used to detect cardiac diseases as it contains very useful information about the condition of heart: Phono cardio gram signal was recorded using electronic stethoscope which reacts to the sound waves indistingulshably to the customary acoustic stethoscope, which is then given to hardware circuitry for signal processing such as amplification and filtering of noises and to amicrocontoller. A software was developed to classify the heart sounds as normal and abnormal and extracting some Important vital parameters such as total power, Q.factor, time duration of first heart sound (ti), time duration between first ‘and second heart sound (t12) and mean 12 from recorded phono cardiogram signal for finding the cardiac condition. The proposed method is cost effective as itneeds minimal equipmentand it utilizes as it ‘were 5 highlights for heart sounds classification comparesto existing technology which are employing a more noteworthy number of highlights. Keywords ‘Murmurs, Feature extraction, Phonocardiogram, Microcontroller, Heart sounds 1. Introduction Early detection of any disease is very important to diagnose and cute in correct time to avoid further ‘complications in later stage which is very dificult to cure and leads to high health care expenditure land also severe damage to life. In this present workearly detection of heart conditions using Phonocardiogram recording with digital stethoscope and microcontroller with suitable software was developed. Heart and cerebral problems are the major wellbeing issues all over the world which leads to fatality to life, Electro cardiogram is the commonmethod to identify the various cardiac diseases particularly echo cardiogram is used for heart valve related problems in natural and implanted artificial heart valves. Another very common mechanism used to identify the cardiac diseases efficiently and ‘economically is through analysis of heart sounds. ‘Auscultation of heart isthe fundamental tool used In the heart problems diagnasis[3]-The recording. ‘of heart sounds called phonocardiogram (PCG) Contains heart sounds and murmurs in the form of ‘waveform and the machine used to record these sounds called phonocardiograph, This is the dovicewhich records heart condition in audible form. Heart sounds are generated due to opening or closing of the heart valves andblood flow ‘through the heart valves or into the ventricular chambers also produces heart sounds. The four parts of 2 phonocardiogram are S1, $2, $3, and $4, or the first, second, third, and fourth heart sounds. Of these, the two prominent heart sounds, $1 and ‘2, are brought on by valve closure.Ventricular valve vibrations resulting from fist quick filing Provide sound $3, while atria contractions during the second stage of ventricular filing produce sound 54 Due to high cost, limited availability and special skills necessary to record and interpret the results of ECG and echocardiography, heart auscultation rmethodsremain the primary tool for diagnosis and screening in primary health centers and remote areas. However, diagnosis based on heart auscultation heard though conventional acoustic sounds using ordinary stethoscope or electronic stethoscope needs a special skill and it takes very long time to acquire. Experienced doctors, nurses 263Journal of Harbin Engineering University ISSN: 1006-7043 ‘or technicians only can diagnose and interpret the cardiac diseases correctly using the auscultation method. Many primary health physicians and hurses are documented to have a poor ‘auscultation sil 2) Early pathological information's of heart valves are available in phonocardiogram signals and it is most Useful in early detection [3, 4, 5]. The creation of2.3. ‘cardiac soundsisnearly linked to the closing and ‘opening of the atrio-ventricular, pulmonary and aortic valves [6].Many features are available in PCG signal which could be extracted using many. feature extraction methods. These features are very useful in finding the cardiac illness that too automatically. Many feature extraction techniques, 4are available for finding the cardiac diseases using PCG. In order to recognize between normal and aortic stenosis patients, wavelet analysis is the best method for feature extraction [7], Short Time Fourier Transform (STFTis the best method for systolie murmur characterization [8]. Using deep learning it is possible to extract features automatically from the original signal and find ‘connectionsbetween prediction value and data [a].Spectrogram and traditional Phono eardiogram2.2. ‘are combined to distinguish pathological murmurs from innocent murmurs [10]-To categorize heart valve ilness, a diagnosis system utilizing SVM (Support Vector Machine) was presented [11]. Mehmet Ali Kobat et al. used a new method for feature generation and extracted 64 discriminative features to automatically find cardiac valve problems(i2]. Many studies uses extraction and using these features doctors/ ‘experienced nurses find the cardiac condition. It is ‘a cumbersome method also needs high skil to diagnose from PCG features. Also, many methods need ECG signal to compare with PCG for accurate detection of cardiac problems. So in addition to PCG ECG machine also needed and it makes the system costly. Many PCG studies not directly gives the cardiac conditions but only the features and it reads skilled technicians to interpret. This is the main idea forthis study.This present method is used only for early detection and uses only five vital parameters to identify the cardiac condition ‘The software used in this study compares the vital, feature of recorded PCG and directly display the heart condition as normal or abnormal so that the feature 2 Vol 45 No. 5 May 2024 Patient itself can understand his heart condition without going to hospital or consulting any ‘medical professionals, As this method is only early detection and not full-ledged PCG device to know the various cardiac diseases, it needs minimal equipment and cost effective, Materials and methods ‘Acquisition of PCG signal and processing Rather than utilizing 2 traditional acoustic stethoscope that reacts to sound waves, an electronic stethoscope detects changes in the PCG signal, which substitutes variations in ait pressure with variations in the electric field.This digital stethoscope wiich picks up heart sounds and act as a transducer which converts sound energy in to electrical signal is fed to analog amplifier for amplifying the signal. Using analog to digital converter, the filtered and amplified signal ic converted to digital signal. After amplification the signal is given to a B-bit CMOS Flash-based ‘Microcontroller (ATMEGA 328) which can process ‘the signal to identify normal or abnormal based on. the five important features such as total power, Q factor, tl, t12 and mean 12, Feature Extraction Generally, many features are used to Identity the PCG signal as normal or abnormal. More than 20 features are used in many studies to diagnose the various heart conditions thoroughly. Frequency domain, statistical features time domain and other classifications are applied to these features. Peak amplitude [33], total power, zero crossing rate [US], Time 1, Time 2, Time 12, Time 22 [16], and other properties are found in the time domain, The frequency domain features are total hormonic distortion, factor, bandwidth, peak frequency[tletcand statistical feature Mean 12. ‘The following are the detailed features used. The explanations for these qualities are provided below, along with potential information that might bbe used to distinguish between normal and pathological behavior. 2.2.1 Time domain features. Peak amplitude: It means, the peak value of the signal. Murmurs have higher peak value compare to normal ‘Total Power: The total power ofthe signal which is related to amplitude, As murmur has. higher 264Journal of Harbin Engineering University ISSN: 1006-7043 amplitude compared to normal, murmur has higher total power compare to normal Zero crossing rate: This is the speed at which a signal transitions from negative to positive or vice versa. Murmur signal is expected 2 larger value of zero crossing rate, Time 1: This parameter isthe first heart sound ($3) time duration. Generally, $1 has shorter duration sounds than murmurs having considerably longer SL Time 2: This parameter is the time duration of second heart sound S2, larger than SI. For murmur signal a longer S2 could be expected considerably than normal Time 42. This is the duration of time between ‘Sand 52. As the region between first and second heart sound is shorter, murmurs is expected to have smaller Time12. Time 21: This Is the duration of time between Ssand Siof nextCycle, As the murmur can be appeared just after S2, the area between the end ‘of S2 to start of next cycle's $1 would be smaller. 2.2.2 Frequency domain features Total Harmonic Distortion (THD): Its definition is the ratio between the fundamental Frequency’s powers to the total of all the harmonic ‘components’ powers. This is the measurement of the signal's harmonic distortion. For murmur signals higher THD is expected Bandwidth: It is a measurement of the variation between 2 continuous set of frequencies’ lower ‘and upper frequencies.As murmursare having high Frequency, higher bandwidth is expected. QFactor: Higher Q-factor values correspond to slower Vol 45 No. 5 May 2024 oscillation decay. So, murmur signal has a higher QFactor. Peak Frequency: Peak frequency isthe frequency where the peak amplitude occurs. Sincemurmurs and normal signals varyin frequency amplitude, itis aso a valuable feature 2.2.3 Statistical feature ‘Mean 12: In the regions between Si and S2 or S2 ‘to next cycle’s $1, Visible difference was seenin most of the murmur signals. So, the mean value was calculated in these regions and maximum of ‘these two values lementioned as Mean12 and iti higher for murmur signals. 2.2.8 Selection of Optimal features All the features are not necessary for early detection of cardiac disease, only some vital features are to be extracted from the recorded PCG. Out of many features explained above, only five optimal features were selected based on Ranker and Info Gain Attribute Evaluation [17]. to segment the signals as murmur and normal signal, these § vital features are enough. So, our method requiresonly these 5 features to classify as normal or abnormal and | pase | cee | = Figure 1 Block diagram of PCG module. 265Journal of Harbin Engineering University ISSN: 1006-7043 2.3 An Embedded system for heart sound analysis ‘The proposed system consists of an embedded system contains @ hardware part for signal ‘acquisition and software part for feature classification to. = detect. «cardiac abnormalities Because the features of the PCG sound will vary for different valvular heart disorders, the sound signal can be utilized as an ‘early indicator for the diagnosis of cardiac diseases such as aortic stenosis, mitral stenosis, aortic regurgitation, and mitral regurgitation. [18).In our method, as it is used only for early detection, the software part identifies the normal or abnormal PCG exactly and displays the type of cardiac disorder/disease such as normal or murmur. Figure 1 shows the block diagram of PCG heart sound diagnosing system containing different phases such as acquisition of as heart sound using data acquisition techniques, signal processing such as noise filtering, artifact removing and ‘segmentation of PCG signal Electronic stethoscope is used ae sensor for PCG data acquisition as it has the advantage of ease of use, excellent sound ‘quality, high accuracy, real-time waveform display, small volume, good sensitivity and better features, ‘compared to conventional stethoscope [19]. PCG Vol 45 No. 5 May 2024 Is acquired by placing the stethoscope head into ‘the chest piece which converts sound signal into the electrical signal with the help of microphone [20]. The output of the microphone is fed to the signal pre-processing stoge which amplifies and filters the signal to improve the signal to noise ratio.in pre-processing phase, vital features will be extracted and in final stage the acquired signal will be classified as normal/abnormal The software incorporated in the microprocessor ull identity the normal or abnormal sounds based fon the normal parameter values compared for each five vital featutes. The status of the heart will be displayed in the LCD display (HD44780) as normal / abnormal with heart beat values also. The TTL converter used here is for serial communication. (The AN-USB-TTL module isa cost-effective method to convert TTL signal to a USB interface, From microcontroller, the signal ‘goes to LED which blinks for every heartbeat. At the same time of displaying in LCD, the PCG signal feom microcontroller also connected to speaker ‘through audio output amplifier in which the patient can hear the heart beat sound. T Ses I Figure 2 shows the various steps involved In PCG module. 266Journal of Harbin Engineering University ISSN: 1006-7043 ‘The above flowchart explains the various steps involved like PCG acquisition, signal conditioning, feature extraction and signal classification, to separate the normal and abnormal sounds such as Vol 45 No. 5 May 2024 murmurs. The obtained PCG signal is compared ‘with previously classified (Known normal data) PCG database and identify as normal / abnormal and displayed in LCD display. Figure 3, Circuit diagram of PcG module Figure 3, shows the exact cicult dlagram of the Phono cardio gram device which Includes power supply, ‘microcontroller, LCD display, audio speaker, and LED display. Figure 4, Proto type madel of the PCG module Figure 4 shows the prototype model of the POG module which is incorporated with power supply, Battery, Amplifier for audio, speaker, microcontroller board and relays for switching to ‘audio mode and feature extraction mode. 267Journal of Harbin Engineering University Vol 45 No. 5 ISSN: 1006-7043 May 2024 3. Results Figure 5 (a) Shows the Normal condition of the heart with Heart beat 74. Figure 5 (b) Shows the abnormal condition condition of the heart with Heart beat 92 displayed as tachycardia. Figure 5 (c) Shows the abnormal condition of the heart with Heart beat 58 displayed as bradycardia, Figure 5 (d) shows the abnormal condition of the heart displayed as Murmur. 268Journal of Harbin Engineering University ISSN: 1006-7043 ‘The figures 5{a) to 5(d) shows above are output displays of the heart conditions with different condition as Normal (5 a), Tachycardia s(t), bradycardia S{e} and Abnormal conditions for Murmur. As explained above, the conditions ofthe heart can be known directly by simply placing the ‘electronic stethoscope on the chest. So the patient ‘ean take decision himself weather to meet doctor ‘or not by knowing the heart condition, 4. Discussion Cardige auseuitation is one of the fundamental screening tools for clinical diagnosis of heart valve disorders in primary health care centers [21].arly pathological information about heart valves can be found in the phonocardiogram signal, which has been demonstrated to be highly helpful in the ‘early detection of potential cardia issues [22, 23] ‘The generation of heart sounds is linked to the closing and opening of ‘the aortic, atrio ventricular, and pulmonary valves (24, 25]. Many studies have shown auscultation of cardiae function is very useful in early detection of cardiac diseases. But traditional auscultation method provides an assessment of cardiovascular diseases based on the clinician’s knowledge and expertise (261, also this methodology is wasteful and inclined to error [27].Heat valve problems leads to complication if not identified early particulerlydegenerative aortic stenosis (DAS) is associated with a high mortality iF diagnosed late and if no valve replacement therapy is performed [28]. With advancement of embedded system containing hardware and software, it is now easy to acquire the phono cardiogram data’s and extract the features needed to diagnose the cardiac diceaces, Lot of studies ‘were developed so far for feature extraction and ‘early diagnosis using phonocardiogram. Out of these methods, many studies need sophisticated techniques and devices to diagnose the cardiac diseases and some method utlizes ECG to correlate with PCG and makes the methods very costly and cumbersome. Particularly for early diagnosis with out the help of medical experts like doctors and nurses, the patient itself can identity the condition of the heart using a simple technique/ system, it needs a system with cost- effective and simple to use by the patient Vol 45 No. 5 May 2024 ItselfLike glucometer to know the level of blood ‘lucose Level in house itself by diabetic patient and digital blood pressure meter to know the blood pressure by the hyper tension patient, an embedded system was developed to know the condition of the heart whether it is normal or abnormal and the patient can decide whether to meet a doctor or not. The embedded system developed and presented in this study was using only five important features out of many features available as explained in this manuscript for early diagnosis and this is an automatic auscultation system that directly cisplay the heart condition as rotmal or abnormal. This kind of auto auscultation method/device is not yet presented/published as for asthe author is concerned. So, itis a novel and vvery useful method for cardiac patients, Conclusion A cost efficient novel embedded system was developed for phonocardiogram using electronic stethoscope for early detection of cardiac diseases. This system requires minimal equipment and does not require ECG recording. Out of many features available for phonocardiogram, only five vital features used to detect the cardiac abnormalities. This system will be more useful for patients in rural/remote areas where other high- lend instruments are not available particularly for rural health management. It is especially helpful for newborns, for whom other methods such as ECG recording are challenging to utilize. Cardiac Patients who had heart surgery can monitor the status of the heart after the surgery from home itself Those who are having audible problem can Use LED display and who are having visual problem can use audible sound which is incorporated with this system for understanding the heart condition As this system Is very simple to operate and no technical knowledge is needed to understand the heart condition, the patient itself can decide to meet doctor of not based on the result displayed. Funding This project was carried out using funding from \Vinayaka Mision’s Research Foundation (Deemed University) Salem under seed money project scheme. 269Journal of Harbin Engineering University Vol 45 No. 5 ISSN: 1006-7043 May 2024 References 15, Rabiner, L, and Shafer, R. 2008, 2nd Edition. 1. Reed TR, Reed NE, Fritzson P. Heart sound Digital Processing of Speech Signals. Prentice Hal ‘analysis for symptom detection and computer- New Jersey, USA, aided diagnosis. Simul. Model. Pract. Theory. 2004; 16. Jiang, Z, and Choi, S. 2006. A cardiac sound 12:129-146, doi: 10.1016} simpat.2003,11.005. characteristic waveform method for in-home 2. Roy DL. The pediatrician and cardiac auscultation. heart disorder monitoring with electric Pediatric Child Health, 2003 Nov;8(2):561-3 stethoscope. Expert Systems with Applications 3. Oh St, et al. Classification of heart sound signals 2006;31:286-208 using 3 novel deep WaveNet model. Comput 17. Mandeep Singh and Amandeep Cheema. Heart Methods Programs Biomed. 2020; 196:105604, Sounds Classification using Feature Extraction of 4. Yuenyong S, et al. & framework for automatic Phonocardiography Signal. International Journal of heart sound analysis without segmentation. Computer Applications. 2013;77() 13-17. Biomed Eng Online, 2011;10(21):1-23 18, Unger, ®, Clavel, MA, Lindman, 8. Ry 5, Deng M, et al. Heart sound classification based on Mathieu, P., &Pbarot, P. Pathophysiology and improved MFC features and convolutional management of multivalvular disease. Nature recurrent neural networks. Neural Netw. 2020; Reviews Cardiology, 2006;13 (7): 429-40. 130:22-32. 19, Pinto, C, Pereira, D., Ferreira-Coimbra, J, 6. Son G-Y, Kwon S. Classification of heart sound Portugués, J, Gama, V., & Coimbra, M.A signal using multiple features. Appl Sei comparative study of electranie stethoscopes for 2018;8(12):2344 cardige auscultation. In 2017 39th Annual 7. Shul, L. 2004. AM13 Analysis of Heart Sound. International Conference of the IEEE Engineering “Thesis. National University of Singapore. In Medicine and Biology Society (EMBC),2017: 8. Segaler, M., Lilja, O., Lukkarinen, S, Orme, L, 2610-2613, Sepponen, R,, and Pesonen, £, 2005, 20, Jiang, Q, Liv, G., Hu, N., Gao, B., Chen, P., Hu, Computer-Based Detection and Analysis of G, & Chen, L Development of a signal Heart Sound and Murmur. Annals of Biomedical collecting system for double-channel auscultation. Engineering, 2005; 33(7}: 937-942. Journal of Biomedical Science and 9. US,etal. A review of computer-aided heart sound Engineering, 2013; 6(07), 712. detection techniques. BioMed Res Int. 24. Sharma, LN. Multiscale analysis of heart 2020, hitps://doi org/10.1155/2020/5846191, sound for segmentation using multiscale Hilbert 10, Noponen, AL, Lukkarinen, 5S, Angerla, A.et envelope, In 2015 13th International Conference ‘a, Phono-spectrographic analysis of heart murmur fon ICT and Knowledge Engineering (ICT & in children, amc Pediatr 2007;7:23 Knowledge Engineering 2015; 33-27 nttosi//doi org/10,1186/1471-2831-7-23, 22, Yuenyong 5, et al. A framework for automatic 11, Maglogiannis, L, Loukis, E, Zafiropoulos, £., and heart sound analysis without segmentation. Statis, A. Support Vector Machine-based Biomed Eng Online. 2013;10(1):1-23, Identification of heart valve diseases using Heart 23, Deng M, etal. Heart sound elaceication based on Sounds. Elsevier, Computer methods and Improved MFC features and convolutional programs in Biomedicine 95. 47-61. recurrent neural networks. Neural Netw. 2020; 12, Kobat MA, Dogan S. Novel three kernelled binary 130:22-32. pattern feature extractor based automated PCG 24. Son G-¥, Kwon S. Classifcation of heart sound sound classification method. ApplAcoust. 2021; signal using multiple features. Appl Sci 179:208040. 2018;8(12)-2344, 13, Amit, G, Gaviiely, N., and Intrator, N. Cluster 25. Randhawa SK, Singh M. ClassiRcation of heart ‘analysis and Classification of Heart Sounds. sound signals using multi-modal features. Elsevier, Biomedical Signal Processing and Control, ProcediaComput Sc. 2015; 58:165-71. 12009; 4:26-36 26. Xu W, et al. Automatic pediatric congenital heart 44, Muruganantham. 2003. Methods for disease classiRcation based on heart sound signal. Classification of Phonacardiogram. TENCON, ‘Antifintell Med. 2022; 126:102257. 270Journal of Harbin Engineering University ISSN: 1006-7043 27. Ganguly A, Sharma M. Detection of pathological heart murmurs by feature extraction of, Phonocardiogram signals. J Appl Adv Res. 2017;2(4):200-5. 28, Kapadia SR, Leon MB, Makkar RR, et al. 5-year ‘outcomes of transcatheter aortic valve replacement compared with standard treatment for patients with Inoperable aortic stenosis (PARTNER 1}: a randomised controlled trial Lancet 2015; 385:2485-91. Vol 45 No. 5 May 2024 am.