Renata Costa, one of our two Soumac Award finalists, has submitted her final project overview. Please read on for more information about Renata’s Intelligent Stethoscope.
The design and build of the hardware system was developed to feasibly capture the heart and lung sounds and send it to Matlab for analysis. These sound waves were converted into electrical signals and fed to a signal conditioning device that was formed by a pre-amplifier, LPF, a power supply and an audio power amplifier. The audio amplifier was used to convert the output of the sensor into a more suitable form for further processing.
Overall architecture of the Electronic Stethoscope Diagnosis System
Data transfer was achieved through a jack cable that interfaces the hardware with the software. Matlab was used to record the signals as waves for further processing and analysis. The filtering of the heart sounds was achieved with a lower frequency than expected as the main frequency components were below 100Hz. Therefore, the FIR truncated LPF was implemented and the results were satisfactory as most of the noise was eliminated.
The development of a peak detector to calculate the heart rate was successfully implemented using the comparison of three samples where the second has to be greater than the first and third sample. The peak calculation performance was very good. One drawback was the impulse noise as, if present during the BPM calculation, could introduce an error of 6% or more to the results.
The respiratory signal was found to be stronger around the speech frequency range, so a noise reduction system and BPF were used to reduce these noises. The system has shown to be very efficient and has the advantage of being able to adapt to different environments due to the assumption of the noise being white. This allows the implementation of the noise cancelling system in real time by using DSP hardware. Matlab GUI was used to display the heart and lung signals and results are shown below.
Matlab GUI displaying the lung signals
Zoom in of the heart signal before and after filtering
One of the drawbacks of this project is the use of a dual power supply, but it still performs well. Another is the use of a voltage divider that supplies the voltage to the microphone without controlling the current draw by the sensor. A voltage regulator or diodes can help to control the current through the microphone. However, as the current drawn from the microphone was very low (0.5mA), the use of the voltage divider was unnecessary for this task.
Overall this project was very successful. I was able to capture both heart and lung sounds and separate them. The reduction of noise and enhancement of the wanted signal was achieved using DSP technology and the audio files were recorded and stored for the monitoring of a patient’s health. This is a way to provide medical professionals with accurate sounds of the heart and lungs that can help to improve the diagnosis of a patient’s illness and in turn, the monitoring of their treatment. The final product was produced and the final cost was within budget. The final hardware is shown below.
Side view of final product
Although the project was successful, there are a number of things I would have done differently. Firstly, I would spend less time on research and more time on the design, implementation and testing of the full system. I would have also designed a PCB with a dsPIC, with which I would record the signal and transfer it wirelessly to a PC. This would prevent the signal suffering from the noise attributed to the cables and connections, leading to a better quality of signal being achieved when sampling. The ADC can processed by a microcontroller.
Throughout this project I have come across a number of different problems. Some were simple and others proved to be very challenging. The approach that I took to solve any issues was to try to identify the source of the problem by performing limited scope testing. This helped me to solve some problems quickly and efficiently.
I believe that this project helped me to improve my skills in problem solving, hardware and software design, validation and verification. This ensured that the full system met the requirements and stakeholder specifications.
Although I am happy with the results, there is always room for improvement. Many more features could have be implemented; I could have improved the respiratory rate calculations, impulse noise reduction and the PCB design that would include a DSP microcontroller. Also, recording the signal through the use of an SD card or Wi-Fi would allow the signal to be sent directly to the patient’s file for closer monitoring of their health.
More features could include the design and implementation of the circuit into a PCB board, where a DSP microcontroller can be added with higher order LPF to avoid aliasing. The microcontroller could be programmed to filter, reduce noise and aid to classify the signals. The final results would then be displayed using an LCD display. Also, the use of an op-amp with a single power supply and lower voltage such as the TLV2461 would be ideal; whereas two AAA batteries could be used to power the whole board. This would make this device much more efficient and robust.
Stay tuned for our second finalist’s project overview next week!