DESIGN OF AN EMBEDDED SYSTEM TO RECOGNISE HEART DISEASES BASED ON HEART SOUND ANALYSIS
A PROJECT REPORT
ABSTRACT
Stethoscope is one of the critical tools used to assess a patient’s health. There is a necessity to develop a device which can redefine what a stethoscope can do. It should also be able to recognize the heart disease that a patient might have in the initial stages of examination itself. In practice, only a cardiologist or a highly experienced doctor can recognize the heart disease from a heart sound. Hence there is a necessity to develop a device which can do this function. The system that has been developed involves ambient noise reduction techniques, sound amplification and analysis. Analysis of the recorded auscultation sounds is performed by using stochastic algorithms on a database of recorded heart sounds collected from Michigan Heart Library. The analysis includes identification of a probable heart disease. The system which has been designed contains three modules. A stethoscope, a Signal Conditioning unit and the Analysis and Comparison unit. The processor that has been used is DSK6713 which takes the audio input via a microphone which is fixed into a cardiac stethoscope. The sound has then been cleansed and amplified. This amplified sound had then been compared to the database sounds. The system’s output is recognizing the patient’s state of the heart, i.e., diseased or normal - if diseased, what is the possible heart disease and where the abnormality is occurring.
-
To extract features from heartbeats recorded after removing the noise and amplifying the signal.
-
To develop an embedded system that can analyze the heart sound and further classify it as normal or diseased. If diseased, the “probable” heart disease is identified.
For any living organism, nutrition and oxygen, depending on the fact that it is either aerobic or anaerobic, are continuously necessary for its survival. When it comes to vertebrates and more specifically, human beings, the vehicle for carrying nutrients and oxygen to various body parts is blood. This blood has to reach each and every cell of the body. This has to be collected back from the periphery for purification. All of this is done by heart. Heart is a hollow muscular organ which is controlled by a specific nervous system called, autonomic nervous system. This functions through Brain and Spinal cord.
Heart, is a hollow muscular organ which is specially designed in such a fashion that the muscle fibres in it are also arranged in a specific pattern. The muscle cells do not have a cell boundaries. This in turn enables continuous propagation of an impulse for its functioning.
Heart consists of four chambers –
-
Left and Right Atria
-
Left and Right Ventricles
All of them have individual boundaries. With this, the right heart consists of right atrium and right ventricle which in turn communicate through right atria and ventricular foramen guarded by tricuspid valve or right atrio-ventricular valve . The right atrium receives the used out blood which is deoxygenated and contains less nutrients, i.e the used out blood is received into right atrium for re-oxygenation and replenishment of nutrients.
As said before, this impure blood reaches right atrium from where it flows to right ventricle through tricuspid /right AV valve. From right ventricle the blood is pumped into lungs for oxygenation purpose through pulmonary artery which is guarded by pulmonary valve.
Similarly, the left heart, consists of left atrium and left ventricle which are interconnected through the left atrio-ventricular valve or the bicuspid/mitral valve for unidirectional flow of blood. The left atrium receives fully oxygenated blood from the lungs from where it flows and gets distributed throughout the body. So the blood flows from left atrium to left ventricle. From the left ventricle the blood is flown out for getting distributed throughout the body through aorta which is guarded by a valve called aortic valve. For pumping of blood, the heart continuously contracts and relaxes in a synchronized fashion.
There are right and left atria, both of which contract simultaneously to pump the blood from atria into the ventricles through the tricuspid and bicuspid valves. And this mechanism of atrio-ventricular flow of blood causes the opening of tricuspid and bicuspid valves which will generate the first heart sound. The generation of the First Heart Sound(S1) is a normal physiological process and has a characteristic sound.
When the blood reaches right and left ventricles, it is destined to be flown into the lungs for oxygenation and body parts. The body parts then utilize oxygen and nutrients**.** This mechanism of flow of blood from right and left ventricles into pulmonary artery and aorta, the ventricles are contracted forcibly with the flow of blood into pulmonary artery and aorta through pulmonary and aortic valves. This opening of pulmonary and aortic valves will generate the Second Heart Sound(S2) which is again a normal physiological process with a characteristic sound.
 Figure 1: Cross-Section of a Heart
Thus in a normal heart, under physiological conditions, continuous synchronized contraction, ‘systole’ and relaxation, ‘diastole’ occur along with pumping of blood from both atria into both ventricles and from both ventricles away from the heart, generating first and second heart sounds which is commonly described as “lub-dub”.
When under certain pathological conditions or when the valves become narrow or stenosed on any dilatation or disease process, the unidirectional haemodynamics may be altered, generating abnormal or adventitious sounds called murmurs.
Clinicians auscultate the heart in its four different places i.e where the valves are situated namely –
-
Mitral area
-
Tricuspid area
-
Aortic area
-
Pulmonary area
The Third Heart Sound (S3) also known as the “ventricular gallop”, occurs just after S2 when the mitral valve opens allowing passive filling of the left ventricle. The S3 sound is actually produced by the large amount of blood striking a compliant left ventricle (LV).
Figure 2:Ventricular Gallop
The Fourth Heart Sound (S4) known as the “atrial gallop”, occurs just before S1 when the atria contract to force blood into the LV. If the LV is non-compliant and atrial contraction forces blood through the AV valves, an S4 is produced by the blood striking the LV.
Figure 3: Atrial Gallop
When under certain pathological conditions or when the valves become narrow or stenosed on any dilatation or disease process, the unidirectional haemodynamics may be altered, generating abnormal or adventitious sounds called murmurs. Heart murmurs are produced as a result of turbulent flow of blood strong enough to produce audible noise. They are usually heard as a whooshing sound. The term murmur only refers to a sound believed to originate within blood flow through or near the heart; rapid blood velocity is necessary to produce a murmur.
Rheumatic heart disease has become much less common in developed countries. Degenerative valve disease (aortic stenosis, mitral regurgitation due to chordal rupture) is increasingly common.
Heart murmurs are defined by four characteristics: loudness, quality, location and timing.
The loudness of murmur reflects the degree of turbulence. This relates to the volume and velocity of flow and not the severity of cardiac lesion.
The quality of murmur relates to its frequency and is best described as low, medium or high pitched. The quality of murmur on the chest walls depends on its site of origin and has led to the description of four valve areas. Clinicians auscultate the heart in its four different areas i.e where the valves are situated namely –
-
Mitral area
-
Tricuspid area
-
Aortic area
-
Pulmonary area
Some murmurs radiate depending on the velocity and direction of blood flow. The sound of the high-velocity systolic flow in aortic stenosis and mitral regurgitation, for example, is directed towards the left sternal edge.
Murmurs are timed according to the phase of systole or diastole during which they are audible. It is inadequate to describe the timing of a murmur and the phase of a systole or diastole during which it is heard; systolic murmurs are either mid-systolic, pan systolic or late systolic; diastolic murmurs are either early diastolic, mid-diastolic or presystolic in timing.^[1]^
Continuous murmurs are audible in both phases of cardiac cycle. A mid-systolic (‘ejection’) murmur is caused by turbulence in the left or right ventricular outflow tract during ejection. It starts following opening of the aortic or pulmonary valve, reaches a crescendo in mid-systole and disappears before the second heart sound.
The murmur is loudest in the aortic area (with radiation to the neck) when it rises from the left ventricular outflow tract, and in the pulmonary area when it arises from the right ventricular outflow tract . It is best heard with the diaphragm of the stethoscope while the patient sits forward. Important causes of aortic ejection murmurs are aortic stenosis.
Aortic Regurgitation also produces an ejection murmur due to the increased stroke volume and velocity of ejection. Pulmonary ejection murmurs may be caused by pulmonary stenosis or infundibula stenosis. Pan systolic murmurs are audible throughout systole from the first to the second heart sounds. They are caused by regurgitation through incompetent atrio-ventricular valves and by ventricular septal defects. The pan systolic murmurs of Mitral Regurgitation are loudest at the cardiac apex and radiates into the left axilla. It is best heard using the diaphragm of the stethoscope with the patient lying on the left side.
The murmurs of Tricuspid Regurgitation and Ventricular Septal Defect are loudest at the lower left sternal edge. Mitral valve prolapse may also produce a pan systolic murmur but, more commonly, prolapse occurs in mid-systole, producing a click followed by a late systolic murmur.
Early diastolic murmurs are high-pitched and start immediately after the second heart sound, fading away in mid-diastole. They are caused by regurgitation through incompetent aortic and pulmonary valves and are beat heard using the diaphragm of the stethoscope while the patient leans forward.
Pulmonary Regurgitation is loudest at the pulmonary area. Mid-diastolic murmurs are caused by turbulent flow through the atrio-ventricular valves. They start after valve opening, relatively late after the second sound, and continue for a variable period during the mid-diastole.
Figure 5: This figure represents the description of various Heart Sounds
Mitral stenosis is the principal cause of a mid-diastolic murmur which is best heard at the cardiac apex using the bell of the stethoscope while the patient lies on the left side. In Mitral or Tricuspid Stenosis, atrial systole produces a presystolic murmur immediately before the first heart sound. The murmur is perceived as an accentuation of the mid-diastolic murmur associated with these conditions. Continuous murmurs are heard during systole and diastole, and are uninterrupted by valve closure.
The most common cardiac cause is Patent Ductus Arteriosus, in which flow from the high-pressure aorta to the low-pressure pulmonary cycle, producing a murmur over the base of the heart which, though continuously audible, is loudest at end systole and diminishes during diastole.Stenosis of the aortic valve is another common heart murmur, a systolic ejection murmur. This is more common in older adults or in those individuals having a two, not a three leaflet aortic valve. Other audible murmurs are associated with abnormal openings between the left ventricle and right heart or from the aortic or pulmonary arteries back into a lower pressure heart chamber.
Physicians have always used patients' chest sounds for diagnosis of diseases. This practice can be tracked back to 1500 B.C. But, it was not until the dusk of the 18th century, when a French doctor named René Théophile Hyacinthe Laënnec (1781–1826) methodically investigated the clinical meanings of both breath and heart sounds. To listen to the sounds of the heart and lungs, he used a simple hollow wooden or an ebony tube. Stethoscope was derived from its Greek roots where ‘stethos’ means chest and ‘skopein’ means to observe or to look at. René understood that the time-honoured method of placing one's ear directly over a patient's chest to be both inadequate and unpleasant. Also, the sounds recorded were muffled when they were recorded from the chests of obese people. Also, many were grimy and lice-ridden, where hygiene was of concern. He hence auscultated (from the Latin auscultare, "to listen") using a hollow tube, which he named “stethoscope”. He also documented that his invention was inspired by the fact that sound is "transported through certain solid bodies, as when we hear the scratch of a pin at one end of a piece of wood, on applying our ear to the other."
For the next few years, Laënnec experimented with a series of hollow tubes that he fashioned out of cedar or ebony, arriving at a model approximately one foot in length and 1.5 inches in diameter, with a 1/4-inch central channel. With this improved tool, he was able to accurately diagnose a number of lung and heart diseases. He had later compiled his findings in his book, “De l'Auscultation Médiate(1819)”. With each passing decade, advances in acoustics and audiology have changed the shape and look of the stethoscope. But it still remains one of the iconic symbols of medicine.
Trained and eminent clinicians can easily identify the timing nature of these adventitious sounds and can diagnose very easily what the underline pathological process is. Modern medicine’s arsenal of non-invasive imaging techniques affords even more stunning glimpses into the human body. A doctor's auscultation skill can be gained only after he or she experiences lots of cases of heart diseases and heartbeat sounds.
Hence, the project aims at assisting a young and less trained internee who is not very well acquainted to identify these abnormal adventitious sounds and murmurs. There is a need of computer-aided equipment to analyze heartbeat sound especially for young doctors to gain a quick learning process. Various recordings of available pathological and abnormal sounds with specific nomenclature are incorporated into one system for analysis. For this purpose, we need accurate and dependable equipment for heartbeat analysis. The hardware and software component used is convenient and accessible that one can monitor his/her heart condition without having an ECG machine. This also makes self-diagnosis possible.
A patient’s heart sound plays an important role in the primary diagnosis, prognosis and survival analysis of heart diseases. The heart signal contains an important amount of information that can be used in different comportments. It allows for the analysis of anatomic and physiologic aspects of the whole cardiac muscle. Hence, it is possible for only proficient and renowned clinicians to easily identify the timing nature of adventitious sounds caused by heart called murmurs and can diagnose very easily what the underlined pathological process is. A young and less trained doctor who is not very well acquainted to identify these abnormal adventitious sounds and murmurs requires an automated system for Cardio Diagnosis. Present day, there exists no system which can recognize a heart disease automatically after hearing to the heart beat sounds. Hence, a system that can replace a conventional stethoscope is necessary. A system that can ease the diagnosis for medical practitioners at basic levels, like students and junior doctors, is required.
To improve the heartbeat analysis, initially the entire sound is separated into sequences containing two major beat sequences, which are referred as S1 and S2 and then the signal is correlated with the normal heart beat signal. The heart sounds were recorded from a stethoscope using a microphone and were amplified. However when the heart beats were recorded, many kinds of noises were also recorded. When heart sound analysis is considered manually, the slight variations occurring due to the noises are neglected. But these noises can make clinical diagnosis very difficult in the case of automated machines. Hence, in automated machines the baseline must be very clearly defined. So it is important to get signal parameters without the noises. Since the signals are non-stationary in nature, it is difficult to visually analyze them. This unwanted information has to be removed by a process known as denoising. In this project, signal denoising (signal conditioning) of the database signals and the test signals has been performed. The detection of the state of heart (signal processing) is done by using a Digital Signal Processor (DSP) (DSK6713). Processing of heart signal is manifold and requires improvement of measurement accuracy and reproducibility (when compared with manual measurements). DSPs provide a low-cost system support. They are concerned primarily with real time signal processing where the processing has to keep pace with some external events, whereas non-real-time processing has no such timing constraint. The external event to keep pace with is usually the analog input. Whereas analog based systems with discrete electronic components such as resistors can be more sensitive to temperature changes, causing vital part losses. DSP-based systems are less affected by environmental conditions since it is completely digitized. Also, working of analog components under high frequency is unpredictable. As the DSPs are designed and optimized for implementation of various DSP algorithms, most processors share various common features to support the high performance, repetitive, numeric intensive tasks.
The DSK package is powerful, comparatively economical, given the necessary hardware and software support tools for real-time signal processing . It is a complete DSP system. The DSK board, with an approximate size of 5 × 8 inches. shown in Figure.3:1, includes the C6713 floating-point digital signal processor and a 32-bit stereo codec TLV320AIC23 (AIC23) for input and output. The onboard codec AIC23 uses a sigma–delta technology that provides ADC and DAC. It connects to a 12-MHz system clock. Variable sampling rates from 8 to 96 kHz can be set readily. A daughter card expansion is also provided on the DSK board. Two 80-pin connectors provide for external peripheral and external memory interfaces.
Figure 5:TMS320C6713
The DSK board includes 16 MB (megabytes) of synchronous dynamic random access memory (SDRAM) and 256kB (kilobytes) of flash memory. Four connectors on the board provide input and output: MIC IN for microphone input, LINE IN for line input, LINE OUT for line output, and HEADPHONE for a headphone output (multiplexed with line output). The DSK operates at 225MHz.
The internal program memory is structured so that a total of eight instructions can be fetched every cycle . Features of the C6713 include 264 KB of internal memory shared between program and eight functional or execution units composed of six arithmetic-logic units (ALUs) and two multiplier units, a 32-bit address bus to address 4 GB (gigabytes), and two sets of 32-bit general purpose registers. DSK delivers 1800 million instructions per second (MIPs) and 1350 MFLOPS.
The C6713 is capable of both fixed- and floating point processing with 32 bit integer support. JTAG emulation through on-board JTAG emulator with USB host interface.An included 5V external power supply is used to power the board.
Code Composer communicates with the DSK through an embedded JTAG emulator with a USB host interface. The DSK can also be used with an external emulator through the external JTAG connector.
Multichannel Buffered Serial Port is one of the important peripherals of this processor :
The C62x/C67x McBSP is based on the standard serial port interface found on the TMS320C2000 and C5000 platforms.
The standard serial port interface provides:
-
Full-duplex communication.
-
Double-buffered data registers, which allow a continuous data stream.
-
Independent framing and clocking for reception and transmission.
-
Direct interface to industry-standard codecs, analog interface chips (AICs), and other serially connected A/D and D/A devices.
-
External shift clock generation or an internal programmable frequency shift clock and multichannel transmission and reception of up to 128 channels.
-
An element sizes of 8-, 12-, 16-, 20-, 24-, or 32-bit.
-
8-bit data transfers with LSB or MSB first.
-
Highly programmable internal clock and frame generation.
DSK6713 can handle different tasks, since they can be reprogrammed readily for a different applications. As the DSPs are designed and optimized for implementation of various DSP algorithms, most processors share various common features to support the high performance, repetitive, numeric intensive tasks. Its important features include:
The most commonly known and used feature of a DSP is the ability to perform one or more multiply-accumulate operation (also called as “MACs”) in a single instruction cycle. The MAC operation is useful in DSP algorithms that involve computing a vector dot product, such as digital filters, correlation, and Fourier transforms. The MAC operation becomes useful as the DSP applications typically have very high computational requirements in comparison to other types of computing tasks, since they often must execute DSP algorithms (such as FIR filtering) in real time on lengthy segments of signals sampled at 10-100 KHz or higher. To facilitate this DSPs often include several independent execution units that are capable of operating in parallel.
When a small group of instructions is executed repeatedly, the cache is loaded with these instructions thus making the bus available for data fetches, instead of instruction fetches. Due to Harvard architecture in DSPs, i.e. physically separate storage and signal pathways for instructions and data, and pipelined structure the processor is able to fetch an instruction while simultaneously fetching operands and/or storing the result of previous instruction to memory. DSPs also share a feature of efficient memory access i.e. the ability to complete several accesses to memory in a single instruction cycle.
The need of processing the digital signals in real time, where in the output (processed samples) have to be produced at the same time at which the input samples are being acquired, evolves the concept of Circular Buffering. Circular buffering allows processors to access a block of data sequentially and then automatically wrap around to the beginning address exactly the pattern used to access coefficients in FIR filter. Circular buffering also very helpful in implementing first-in, first-out buffers, commonly used for I/O and for FIR delay lines.
The instruction sets of the digital signal processors are designed to make maximum use of the processors‟ resources and at the same time minimize the memory space required to store the instructions. Maximum utilization of the DSPs‟ resources ensures the maximum efficiency and minimizing the storage space ensures the cost effectiveness of the overall system. To ensure the maximum use of the underlying hardware of the DSP, the instructions are designed to perform several parallel operations in a single instruction, typically including fetching of data in parallel with main arithmetic operation.
The ESA MCB 51 board requires +5V power and a serial connection to a
PC running Keil µVision debugger. ESA MCB 51 can be operated with AT89C51ED2 or AT89C51RD2 micro controller.
· AT 89C51ED2 / RD2 operating at 11.0592 MHz
On-Chip Memory:
-
CODE MEMORY: 64K Bytes of flash.
-
DATA MEMORY: 256 Bytes of RAM.
-
1792 Bytes of XRAM.
-
2K Bytes of EEPROM.
-
16-bit Timers/Counters.
-
Watch Dog Timer.
-
Programmable Counter Array (PCA) on Port1 i.e. PWM and Capture & Compare.
-
SPI (Serial Peripheral Interface) on Port1.
-
Full duplex enhanced UART.
Nine sources of interrupt (both external and internal).
Two External interrupts INT0 and INT1 are provided with push button switches;
these can also be used as general-purpose switches.
CPU:
All the CPU signals available on a 40-pin connector site.
I/O (Port) Lines:
P0, P1 and P2 Port lines are available on a 26-pin connector,
compatible to ESA Interface Modules.
Four 10-pin connectors for all the 32 I/O lines.
LCD:
LCD compatible signals are available on a 16 Pin flow strip connector.
SERIAL I/O:
The trainer uses the on-chip flash monitor to communicate with Keil Vision Debugger. The Flash monitor allows the user to program and debug the on-chip code. The monitor uses the on-chip UART to communicate with mVision Debugger.
Figure 6: ESA MCB 51 Kit
– Block Diagram
Code Composer Studio is an integrated development environment (IDE) that supports TI's Microcontroller and Embedded Processors portfolio. Code Composer Studio comprises a suite of tools used to develop and debug embedded applications. It includes an optimizing C/C++ compiler, source code editor, project build environment, debugger, profiler, and many other features. The C compiler compiles a C source program with extension .c to produce an assembly source file with extension .asm. The assembler assembles an .asm source file to produce a machine language object file withextension*.obj*. The linker combines object files and object libraries as input to produce an executable file with extension .out. To connect CC Studio with DSK kit, click Debug, Connect as shown in Fig. 15. Then create an application project as seen in Fig. 16 and can “add” the appropriate files to the project such as libraries, hello.cmd, source file, support files etc. Compiler/linker options can readily be specified.
After writing the code, the next step is to compile the code to machine language. Then go to Project Build‟. The Build command will compile all the files that are included in this project and make an executable file for the DSP. A number of debugging features are available, including setting breakpoints and watching variables; viewing memory, registers, and mixed C and assembly code; graphing results; and monitoring execution time. Real-time analysis can be performed using RTDX allows for data exchange between the host PC and the target DSK, as well as analysis in real time without stopping the target. Key statistics and performance can be monitored in real time.
Figure 7: CC Studio - Steps
Useful Types of Files
-
file.pjt: to create and build a project named file.
-
file.c: C source program.
-
file.asm: assembly source program created by the user, by the C compiler, or by the linear optimizer.
-
file.sa: linear assembly source program. The linear optimizer uses file.sa as input to produce an assembly program file.asm
-
file.h: header support file.
-
file.lib: library file, such as the run-time support library file rts6700.lib
-
file.cmd: linker command file that maps sections to memory.
-
file.obj: object file created by the assembler.
-
file.out: executable file created by the linker to be loaded and run on the C6713 processor.
-
file.cdb: configuration file when using DSP/BIOS.
Support files
-
C6713dskinit.c: contains functions to initialize the DSK, the codec, the serial ports, and for I/O. It is not included with CC Studio.
-
C6713dskinit.h: header file with function prototypes. Features such as those used to select the mic input in lieu of line input (by default), input gain, and so on are obtained from this header file.
-
C6713dsk.cmd: sample linker command file. This generic file can be changed when using external memory instead of internal memory.
-
Vectors_intr.asm: a modified version of a vector file included with CCS to handle interrupts. Twelve interrupts, INT4 through INT15, are available, and INT11 is selected within this vector file. They are used for interrupt-driven programs.
-
Vectors_poll.asm: vector file for programs using polling.
-
rts6700.lib, dsk6713bsl.lib, csl6713.lib: run-time, board, and chip support library files, respectively.
-
These files are included with CCS and are located in C6000\cgtools\lib, C6000\dsk6713\lib, and c6000\bios\lib respectively.
-
Finally, to run the program, load the program into the DSP. Go to File Load Program.
-
Load the executable file (.out) that the compiler generated as seen in Fig 17 (generally in the Debug directory of the project) and then run the loaded file.
Through the JTAG, communication with on-chip emulation support occurs to control and monitor program execution. The C6713 DSK board includes a JTAG interface through the USB port.
2. ### MATLAB
MATLAB is a high performance language for technical computing. The fundamental milestone of this project is investigation and implementation of the removal of noise from heart beat sound signal and extract the features necessary for comparison. The MATLAB software provides a variety of functions that makes it easy and flexible while simulating for interactive designing, advance analyzing, exploration and visualizing signals, filters and windows. It provides tools for finite impulse response (FIR) and infinite impulse response (IIR) digital filters’ - design, implementation and analysis. It also provides the toolbox for application such as speech and audio processing, medical imaging and instruction, wired & wireless communication and financial modeling and analysis. MATLAB has set of construct for plotting scientific graphs from raw or computed data.
The Signal processing toolbox of MATLAB provides functions to generate, measure, transform, filter and visualize signals. The toolbox includes algorithms for resampling, smoothing and synchronizing signals, designing and analysing filters, estimating power spectra, and measuring peaks, bandwidth, and distortion. With the help of this, trends in signals can be analysed and compared in time, frequency and time frequency domains, patterns and trends can be compared, features can be extracted and algorithms can be developed, validated and customized to gain insight into our data.
Here, for our project work, MATLAB (R2009a) software was used for initially conditioning the signal and denoising it. The features for comparison were hence extracted after the conditioning.
MATLAB‟s filter design and analysis tool can be invoked by typing
>> “fdatool’’
The Filter Design and Analysis Tool (FDATool) is a powerful user interface for designing and analyzing filters quickly. FDATool enables quick design of digital FIR or IIR filters by setting filter performance specifications, by importing filters from MATLAB workspace, or by adding, moving or deleting poles and zeros. FDATool also provides tools for analyzing filters, such as magnitude and phase response and pole-zero plots. The values and specifications are provided to design the required filter as given in the below fdatool box.
Figure 8: FDATOOL
Filter coefficients are exported to our project in CCS by generating an ANSI C header file that contains those coefficients. The header file defines global arrays for the filter coefficients. When you link the project to the header file, the linker allocates the global arrays in static memory locations in processor memory. Loading the executable file, the processor allocates enough memory to store the exported filter coefficients in its memory and writes the coefficients to the allocated memory and further exports to CC Studio.
Several responses can be chosen as per the necessary conditions from the list below -
Low-pass, Raised cosine, High-pass, Band-pass, Band-stop, Differentiator, Multiband, Hilbert transformer, Arbitrary magnitude.
Additional response types are available if DSP System Toolbox™ software is installed. For example, to design a bandpass filter, select the radio button next to Bandpass in the Response Type region of the GUI. The default filter design method for the response type that you've selected can be used or a filter design method from the available FIR and IIR methods listed in the GUI can be selected. After this, the filter specifications and filter order have to be set.
Any of the following filter response characteristics can be displayed either in the response region or in a separate window.
-
Magnitude response
-
Phase response
-
Magnitude and Phase responses
-
Group delay response
-
Phase delay response
-
Impulse response
-
Step response
-
Pole-zero plot
-
Zero-phase response — available from the y-axis context menu in a Magnitude or Magnitude and Phase response plot.
Typical layout of the flash monitor and Keil debugger is as follows.
Figure 9: Keil Debugger
On-Chip UART signals are available through MAX 232 on a RJ-11 connector. The ESA MCB 51 is built around AT89C51ED2/RD2; this board provides a platform to user to evaluate the on-chip features of 8051 family microcontrollers. The 32 I/O lines of the microcontroller are available to user on different connectors. A 16x2 LCD, which is interfaced to 11 port lines, is available to user. The on-chip UART is used tointerface the board with PC.
A 16x2 LCD is interfaced to the port lines of MCU.
.
LCD accepts characters in ASCII format. Character display font in LCD module is by matrix of dots i.e. each character in LCD module can be represented by 7x5 matrix (7 rows and 5 columns of dots) or 10x5 matrix (10 rows and 5 columns of dots). This interface is built over 16x2 LCD module in which the display data RAM address of LCD module for the first line starts from 00H to 14H and the second line starts from 29H to 3CH.
Figure 10: LCD Description
LCD module has an automatic reset, which is critically dependent upon power supply voltage. Voltage has to rise from 0.2V to 5V within 10 to 15ms for the LCD to reset. Since this is not accurate we choose to give some delay in the application program while initializing it. The Busy Flag of the LCD module will be set while LCD is resetting and also when data or command has been written. At this time user cannot write on to the LCD module. Data can be written on to the module only when busy flag goes low.
The ESA MCB 51 board requires +5V power and a serial connection to a PC running Keil µVision debugger. After the Power up, the board can be connected to Keil IDE to download and execute the user programs. The Keil Monitor-51 Driver is used to download the user application code into on-chip flash, and also to debug the downloaded user application code.
Figure 11: Options for Target in Keil
Figure 12: Microcontroller/ Microprocessor - Device Selection in Keil
1. MODULE 1: *Signal Conditioning:* Amplification of Heart Sound and Noise Filtering
-----------------------------------------------------------------------------------
Hardware:
Audio Amplifier:
Figure 13: Preamplifier Circuit
Figure 14: Amplifier Circuit
-
The electret microphone is attached inside the tubes of the stethoscope.
-
Stethoscope sounds are given as input to the preamplifier circuit which is driven by a 5 V DC supply.
-
The preamplifier circuit essentially reduces the effects of noise and interference.
-
It is used to boost the signal strength to drive the cable to the main instrument without significantly degrading the signal-to-noise ratio (SNR).
-
The preamplifier provides voltage gain (from 10 millivolts to 1 volt) but no significant current gain.
-
The output of the preamplifier circuit is connected to the pin 1 of TDA2040 which takes the input through a capacitor.
-
TDA2040 is a class AB power amplifier. The TDA2040 provides high output current and has very low harmonic and crossover distortion. It also has a built in circuitry for short circuit protection.
-
Pin 5 and 3 are connected to +12V and -12V DC supply.
-
The output is taken at pin 4 of the IC using a speaker, which in a later stage will be given as an input to MATLAB for signal comparison.
TESTING**:**
-
The voltage level of the recorded heart sounds is of the order 10^-6^. So a very huge gain is necessary in order to record the heart sounds.
-
The microphone was too sensitive to ambient noise and lesser sensitive to the heart sound. Though the audio amplifier circuit could amplify signals of amplitude 0.5mV, it wasn’t sensitive enough to amplify the heart sounds.
-
When a lowpass filter was applied across the input before amplification, the noise was reduced but the amplifier wasn’t able to amplify the heart sounds.
-
A huge attenuation in the amplitude of the input signal was observed when the order of filter was increased to make a steep filter.
Heart Sound - Signal Conditioning Circuit
Circuit Characteristics:
-
The signal conditioning circuit consists of two elements, audio amplifier (LM386) and filtering circuit.
-
The amplification of this circuit is set to be maximum of 5Vp with frequency filtering is set to 10Hz using an analog passive filter.
-
The analog passive filter used was a low-pass filter of cut-off frequency 10 Hz.
-
noise was removed but heart sounds were not audible.
-
The tube of a stethoscope with 4-5 mm inner diameter was cut in length of 10-12cm, and a small condenser microphone (SONY) was inserted.
-
The sounds were directly recorded to the computer through a collar MIC.
-
The sounds were recorded in a very quiet environment where all the environmental noises were minimised.
-
A filter of 20-250 Hz was applied to separate the heart sounds from the environmental noises. It was further amplified by multiplying with a scalar multiple 100.
Figure 16: Tuning the stethoscope between bell and diaphragm
### Real Time constraints:
-
The stethoscope has to be tuned in between the bell and the diaphragm to record the heart beats. Initial tuning of the Stethoscope is very crucial to record sounds. When the Stethoscope is tuned either to diaphragm or bell completely, the sounds couldn’t be recorded due to presence of very prominent noise.
-
The sensitivity of the collar mike is very high. The microphone is inserted inside the rubber tube of the stethoscope. Hence, noise becomes very loud even if the tube is touched.
The goal of this module is to develop an algorithm for detecting and classifying murmurs. The analysis has been made on 138 various heart sounds. Hence to differentiate between the heart sounds, Stochastic and energy estimations have been carried out in the algorithm. From the frequency estimates, the energy was also calculated. The procedure suggested would provide the display of the heart sound with respect to the time, the category of the murmurs and the types of common heart disorder faced by the patient. The data collected from the stethoscope is processed and analysed. It allows early detection of the common heart disease, and a person does not need to be in the hospital to get them checked.
Explanation:
We use percentile energy as an initial mode of classification. It tells if a signal is abnormal or normal. According to the assumption, the percentile energy of every frame should be constant and that energy should be compared to that of the test signal’s. As per the algorithm, we calculate percentile energy by dividing the total energy per frame with total energy of the signal. This percentile energy calculated is used for comparison between the signals.
2 Beats per frame:
Test Signal: 04 Apex, Mid Sys Click, Supine, ll_0.00-0.08.wav
totaltime = 8.9088 %in Seconds
totalenergy = 1.8778e+009 %total energy of the signal = [abs(fft)]2
percentile_frame = 0.0049 0.0077 0.0031 0.0079 0.0079 0.0049 0.0079
Percentile Energy calculated per frame, where each frame consists of 2 beats)
Figure 17: Energy percentile calculation: test signal : 2beats per frame
Database signal:
02 Apex, Split S1, Supine, bell_0.00-0.07.wav
fs = 44100 %Hz totalenergy = 4.1233e+008 percentile_frame = 0.0070 0.0066 0.0072 0.0108 0.0111 0.0098 0.0114
Figure 18: Energy percentile calculation: database signal
4 beats per frame:
Test Signal: 05 Apex, S3, LLD, Bell_part 5.1 totaltime = 9.0656 %in seconds totalenergy =1.6814e+009 percentile_frame = 0.0214 0.0229 0.0281 0.0227 0.0229 (Percentile Energy calculated per frame, where each frame consists of 4 beats)
Figure 19: Energy percentile Calculations, Test Signal
Database Signal: 06 Apex, Early Sys Mur, Supine, Bell_part 6.1 totaltime =10.0582 %in seconds totalenergy =5.6634e+008 percentile_frame = 0.0137 0.0142 0.0275 0.0143 0.0271 0.0147 0.0271
Figure 20: Energy WHAT IS HERE?
-
The energies of the signal per frame should be almost equal. But unlike that situation the energies of every frame is different.
-
The ratio of energies of the test signal and the database signal should always be either greater than 1 or less than 1. But even this situation keeps changing for every frame. This means that the energy in one frame cannot be compared to energy in another frame.
-
This procedure can at the most tell if the signal is abnormal or normal. It cannot be used to classify into specific category of any heart.
In signal processing, cross-correlation is a measure of similarity of two series as a function of the lag of one relative to the other. This is also known as a sliding dot product or sliding inner-product. It is commonly used for searching a long signal for a shorter, known feature. It also determines the similarity between two signals. For continuous functions f and g, the cross-correlation is defined as:
where f* denotes the complex conjugate of f and
is the lag. There are several
correlation coefficients, often denoted ρ or r, measuring the degree of
correlation.
The taken signal is an abnormal heart sound.
Test Signal: 10 Apex, Sys Click & Late Sys Mur, LLD, Bell_part 10.6.wav
The signal is represented in its time domain. Every set of heart sounds consists of S1,S2 and S3. Also there is another abnormality called Click, which occurs after the systole. Like the name say, the murmur occurs in the late systolic side.
igure 21: Signal in Time domain
Figure 21 depicts the envelope of the signal before rectification. This waveform is similar to that observed in an ECG. All the peaks and valleys can be distinctly noted here.
Figure 22: Envelope of
the Signal
The Figure 22 depicts the rectified version of Figure 21. Here also, the features of the abnormality can be distinctly noted.
Figure 23: Rectified signal for further peak detection
Figure 23 depicts the normalized version of figure 22. While recording sounds there can be differences in the amplitudes. Hence to equalize all those irregularities in amplitude, the signal is normalized.
Figure 24: Normalized
signal
The signal is the divided into sets which consist of data from one S1 to the next S1. This makes correlation more accurate because only required features are compared, hence reducing the execution time .
Figure 25: The separated signal which will be used for correlation
A similar procedure is carried forward for the database signals as well.
Database Signal 1: 03 Apex, S4, LLD, Bell_part 3.3
Database Signal 2 : 04 Apex, Mid Sys Click, Supine, Bell_part 4.5.wav
Database Signal 3 : 05 Apex, S3, LLD, Bell_part 5.4.wav
Database Signal 4: 06 Apex, Early Sys Mur, Supine, Bell_part 6.3.wav
Database Signal 5: 09 Apex, Holo Sys Mur, Supine, Bell_part 9.1
Database Signal 6: 10 Apex, Sys Click & Late Sys Mur, LLD, Bell_part 10.2.wav
Database Signal 7: 01 Apex, Normal S1 S2, Supine, Bell - Copy_part 1.1.wav
As per the waveforms the maximum correlation should be with that of the Last waveform.
The correlations seen are:
Assuming that code doesn’t give any false positives, the correlation match has been categorized into 4 cases. Since a 99% match is set as threshold and since the test signal can be either normal or diseased, the output is divided into four cases-
Peaks: Value of the S1 peak; t : position at which it occurs
peaks1 = 0.9914 0.9818 0.9918 1.0000 0.9649 0.9997 0.9833 0.9986 0.9942 0.9977 t1 = 24 131 242 352 464 571 680 791 904 1014 peaks2 = 0.9857 0.9881 0.9880 1.0000 0.9784 0.9917 0.9763 0.9928 0.9984 t2 = 69 178 284 393 499 611 721 832 941 peaks3 = 0.9522 0.9914 1.0000 0.9898 0.9933 0.9898 0.9974 0.9835 0.9912 0.9812 t3 = 25 157 291 424 559 694 828 961 1096 1232 peaks4 = 0.9992 0.9636 0.9983 0.9569 0.9840 0.9826 0.9766 0.9732 0.9933 0.9658 0.9934 0.9527 0.9728 0.9632 0.9767 0.9738 0.9848 0.9624 1.0000 0.9695 0.9722 t4 = 25 111 157 244 290 378 426 513 559 646 692 780 825 912 958 1044 1089 117 1222 1309 1355 peaks5 = 0.9891 0.9948 0.9957 0.9910 0.9878 0.9969 0.9892 1.0000 0.9921 0.9894 t5 = 26 148 274 396 524 645 774 895 1022 1144 peaks6 = 0.9791 0.9809 0.9817 0.9800 0.9922 0.9754 0.9054 0.9802 0.9825 0.9785 1.0000 t6 = 26 141 266 389 512 629 705 759 882 1005 1129 peaks7 = 0.9617 0.9707 1.0000 0.9811 0.9763 0.9695 0.9765 0.9791 0.9812 t7 = 25 117 209 301 393 485 577 669 761
*10 Apex, Sys Click & Late Sys Mur, LLD, Bell_part 10.1.wav *
Here the rows& columns represent the output match of different sets of S1 to S1 correlated with different database signal sets.
cc1 =
11 11 11 11 11 11 11 11 11
11 11 11 11 11 11 11 11 11
11 11 11 11 11 11 11 11 11
11 11 11 11 11 11 11 11 11
11 11 11 11 11 11 11 11 11
11 11 11 11 11 11 11 11 11
11 11 11 11 11 11 11 11 11
*04 Apex, Mid Sys Click, Supine, Bell_part 4.5.wav *
cc2 =
11 11 11 11 11 11 11 11
11 11 11 11 11 11 11 11
11 11 11 11 11 11 11 11
11 10 11 10 11 11 11 11
11 11 11 11 11 11 11 11
11 11 11 11 11 11 11 11
11 10 11 10 11 11 11 11
11 10 11 10 11 11 11 11
11 11 11 11 11 11 11 11
*03 Apex, S4, LLD, Bell_part 3.3.wav *
cc3 = 10 10 10 10 10 10 10 10 10
10 10 10 10 10 10 10 10 10
10 10 10 10 10 10 10 10 10
10 10 10 10 10 10 10 10 10
10 10 10 10 10 10 10 10 10
10 10 10 10 10 10 10 10 10
10 10 10 10 10 10 10 10 10
*05 Apex, S3, LLD, Bell_part 5.4.wav *
cc4 = 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10 10
*06 Apex, Early Sys Mur, Supine, Bell_part 6.3.wav *
cc5 =
10 10 10 10 10 10 10 10 10
10 10 10 10 10 10 10 10 10
10 10 10 10 10 10 10 10 10
10 10 10 10 10 10 10 10 10
10 10 10 10 10 10 10 10 10
10 10 10 10 10 10 10 10 10
10 10 10 10 10 10 10 10 10
10 10 10 10 10 10 10 10 10
10 10 10 10 10 10 10 10 10
*09 Apex, Holo Sys Mur, Supine, Bell_part 9.1.wav *
cc6 =
11 10 10 10 11 10 10 10 10 10
11 10 10 10 11 10 10 10 10 10
11 10 10 10 11 10 10 10 10 10
11 10 10 10 11 10 10 11 10 10
11 10 10 10 10 10 10 10 10 10
11 10 10 10 11 10 10 10 10 10
11 10 10 10 11 10 10 10 10 10
11 10 11 11 11 10 10 11 11 10
11 10 10 10 11 10 10 10 10 10
*01 Apex, Normal S1 S2, Supine, Bell - Copy_part 1.1.wav *
ccnormal =
101 101 101 101 101 101 101 101
101 101 101 101 101 101 101 101
101 101 101 101 101 101 101 101
101 101 101 101 101 101 101 101
101 101 101 101 101 101 101 101
101 101 101 101 101 101 101 101
101 101 101 101 101 101 101 101
101 101 101 101 101 101 101 101
101 101 101 101 101 101 101 101
***Conclusion: ***
1. The maximum number of 11s are observed in the case of cc1 which corresponds to waveform 10, which is in turn the correct waveform identified.
2. Hence the signal is identified as a Late systolic murmur, which corresponds to a “MITRAL VALVE COLLAPSE”
This module implements “Signal Conditioning” and “Analysis and Comparision” modules in the Digital Signal Processor. The Module 3 has been converted into C code to be possible for processing by the DSK6713 processor. Hence the algorithms for the correlation and signal smoothening remain the same. As shown in Chapter 3, CCS has been used to code in C Language which can be processed in DSK6713. The processor uses DSK6713 AIC23 codec to handle the real time operations. The codec allows the user to take input and give outputs through 3.5mm jacks either via MIC IN/OUT and LINE IN/OUT. A MIC is fixed into the tube of the stethoscope. This MIC output is given as an input to the AIC23 codec via MIC IN. The DSK6713 cleanses the noise by filtering the signal in between the range 2-250 Hz, to remove noise due to lungs sounds and external environment. MATLAB’s FDATool is used to derive the coefficients of the required filter for denoising the signal. Once the denoised signal is thru, it is passed through a function to extract the envelope of the signal. Once the envelope is taken, it is rectified and the heart beat sets are separated from one maximum to the next maximum. These sets of sounds are compared to the already separated sets, extracted and exported from MATLAB to the processor’s memory. The correlation is performed over the test signal and database sounds and the output is taken based on the percentage match with the database sounds. Based on the output, the DSK6713 writes appropriate binary code relating to each case to its GPIOs. These GPIOs are connected to Port 1 of 89C51ED2 in the MCB ESB 51 board, through a voltage convertor to set low of the output to 0V and high to 5V. The Port 1 of 8051 is configured as an input port. The Port 2 of the LCD is the port which carries the data and commands to LCD. P3.6,P3.7 and P3.5 of the board are used for configuring the LCD to command or data display mode. Hence, based on the inputs given by GPIOs of DSP into 8051, the 8051 takes the input from Port 1 and displays the corresponding output via a switch case on the LCD interfaced to 8051 in the board.
Figure 34: Working
Screenshot of the project
The test signal has been obtained through MICIN of DSK6713. It has been manually played in laptop. The acquired signal is then filtered in the preprocessing module then sent for comparision through correlation function.
Test Signal: X11: 10 Apex, Sys Click & Late Sys Mur, LLD, Bell_part 10.1.wav
Figure 35: Test Signal: X11: 10 Aortic, Sys Click & Late Sys Mur, LLD, Bell
The database signals are preprocessed and embedded into the code after separating them into sets of heartbeats.
Database Signal 1: 04 Apex, Mid Sys Click, Supine, Bell_part 4.5.wav
Figure 36: Database Signal 1: 04 Apex, Mid Sys Click, Supine, Bell_part 4.5.wav
Database Signal 2: 10 Aortic, Sys Click & Late Sys Mur, LLD, Bell_part 10.5.wav
Figure 37: **Database
Signal 2:** 10 Apex, Sys Click & Late Sys Mur, LLD, Bell_part 10.5.wav
Database Signal 3: 03 Apex, S4, LLD, Bell_part 3.3.wav
Figure 38: Database
Signal 3: 03 Apex, S4, LLD, Bell_part 3.3.wav
Database Signal 4: 05 Apex, S3, LLD, Bell_part 5.4.wav
Figure 39: Database Signal 4: 05 Apex, S3, LLD, Bell_part 5.4.wav
The results obtained after correlating the test signal with Database signals.
Figure 40: Correlation with Database signal 1:
Max correlation value noted: 30.8712
Figure 41: Correlation
with Database signal 2:
Max Correlation Value Noted: 32.41663
Figure 42: Correlation with Database signal 3:
Max Correlation Value Noted: 23.26266
Figure 43: Correlation with Database signal 4:
Max Correlation Value Noted: 19.89651
Figure 44: Output shown on LCD display through 8051
-
Maximum correlation is obtained with Database signal 2.
-
The input is a diseased heart sound.
-
The Identified disease is “Aortic Systolic Murmur”, it rightly corresponds to the database signal 2 which is “10 Aortic, Sys Click & Late Sys Mur, LLD, Bell_part 10.5.wav”.
The system that has been developed involves ambient noise reduction techniques, sound amplification and analysis. Analysis of the recorded auscultation sounds is performed by using stochastic algorithms (i.e. correlation) on a database of recorded heart sounds collected from Michigan Heart Library. The analysis included identification of a probable heart disease. The system which has been designed contained three modules. A stethoscope, a Signal Conditioning unit and the Analysis and Comparison unit. The processor that has been used is DSK6713 which takes the audio input via a microphone which is fixed into a cardiac stethoscope. The sound has then been cleansed and amplified. This amplified sound had then been compared to the database sounds. The system correctly identified the patient’s state of the heart, i.e., diseased or normal - if diseased, what is the possible heart disease and where the abnormality is occurring.
- Written by Hutchison edited by Michael Glynn and William M Drake ,“ Hutchison's Clinical Methods,” 23rd Edition An Integrated Approach to Clinical Practice With student consult Online Access ,Pub Date: 2012 , Publisher: Saunders Ltd.
-
Rulph Chassaing, “DSP Applications Using C and the TMS320C6x DSK**,**” A Wiley–Interscience Publication,Published:Januray2002.
-
Indu Udai , Lekshmi P R , Sherin K Mathews , Tinu Maria Daie and Manu T S. “ECG Signal Processing Using DSK TMS320C6713,” IOSR Journal of Engineering (IOSRJEN) Volume 2, Issue 10 (October 2012), PP 24-43 .
-
R. M. Potdar and Nishi Shahnaj Haider “Removal of Heart Sound from Lung Sound using LabVIEW 8.6, ” International Journal of Engineering Research and Applications (IJERA) Vol. 2, Issue 3, May-Jun 2012, pp.1313-1319 1313.
-
Anas Mohd Noor, Mohd Faiz Shadi ” The Heart Auscultation. From Sound To Graphical,” Journal of Engineering and Technology ,2013.
-
Zaiton Sharif, Mohd Sharmian Zainal, , Ahmad Zuri Shaameri andSheikh Hussain Shaikh Salleh” Analysis And Classification Of Heart Sounds And Murmurs Based On Tee Instantaneous Energy And Frequency Estimations,”IEEE Conference,2000.
-
Masaru Yamashita, Masataka Himeshima, and Shoichi Matsunaga “Robust Classification Between Normal And Abnormal Lung Sounds Using Adventitious-Sound And Heart-Sound Models”, IEEE International Conference on Acoustic, Speech and Signal Processing (ICASSP),2014.
-
Database heart sound from “University of Michigan Heart Sound and Murmur Library”[Online]Available at