Acoustic Sensor based CardioPulmo Monitoring System
Cardiovascular disorders (CVDs) or Heart diseases remain the 1st cause of mortality globally, responsible for 18.56 million people dying annually whereas acc. to WHO, COPD (only one class of respiratory illness) is the third leading cause of death worldwide, causing 3.23 million deaths in 2019. More
2025-06-28 16:25:00 - Adil Khan
Acoustic Sensor based CardioPulmo Monitoring System
Project Area of Specialization Electrical/Electronic EngineeringProject Summary Context:Cardiovascular disorders (CVDs) or Heart diseases remain the 1st cause of mortality globally, responsible for 18.56 million people dying annually whereas acc. to WHO, COPD (only one class of respiratory illness) is the third leading cause of death worldwide, causing 3.23 million deaths in 2019. More than 75% CVDs and almost 80% of COPD deaths occur in middle and low-income countries, mainly due to a lack of timely diagnosis & treatment.
For CVDs heart sounds are still the primary tool for screening and diagnosing many pathological conditions of the human heart. Using the auscultation technique for heart sound analysis is still insufficient due to human ear limitations and its dependency on highly trained professionals. For lung diseases, primary diagnostic methods include auscultation, percussion, breathing patterns, etc. But again all these require trained professionals to diagnose the disease. Although, we have MRI, X-RAY & CT Scan for early and reliable diagnosis of lung diseases. But these are very expensive and not available at every medicare center.
Phonocardiography (PCG) is one of the non-invasive techniques to diagnose the condition of the human heart, generated by muscle contractions and closure of the heart valves which produces vibrations audible as sounds and murmurs, which qualified cardiologists can analyze. A Phonocardiogram (PCG) can provide the most valuable diagnostic information for evaluating many cardiac abnormalities. One similar non-invasive technique for lung analysis is through Lung Sounds (LS). It produces due to turbulent flow in lung airways. Pulmonary abnormalities can be reflected as wheezes, crackles, squawks, etc in Lung Sounds.
Now when using PCG or LS for heart and lung analysis respectively. One difficulty faced is the presence of lung artifacts when taking heart sounds and heart artifacts when taking lung sounds.
Objective:Our objective is to design a standalone, low-cost & portable system that takes two-channel input of heart sounds and lung sounds simultaneously using acoustic sensors, makes them independent of each other using a novel technique, and then detects & classifies the pathological heart murmur and adventitious (abnormal) lung sounds and generates a comprehensive report on the condition of the heart and lungs having the diagnostic/classification results along with other key metrics and information that may help doctors get diagnostic insights and better understand the current heart and lungs condition.
Our objective is to design a standalone, low-cost & portable system that takes two-channel input of heart sounds and lung sounds simultaneously using acoustic sensors, makes them independent of each other using a novel technique, and then detects & classifies the heart pathological murmur and adventitious (abnormal) lung sounds and generates a comprehensive report on the condition of the heart and lungs having the diagnostic/classification results along with other key metrics and information that may help doctors get diagnostic insights and better understand the current heart and lungs condition.
The Project consists of 3 main modules:
1) Data Acquisition
2) Algorithm Development
3) Hardware Implementation
The Data Acquisition can be performed using the following approaches all of which imply acoustic-based sensors:
- High-quality stethoscopes (By Littmann) with embedded microphones
- Designing our own Stethoscope
- NI Biomedical Instrumentation Kit
The Algorithm Development part for this problem usually consists of the following subparts:
- Preprocessing: In this stage, we retain what we want from our input data, remove what we don't wish to, and enhance our input data to meet our needs. We might have to use decimators/interpolators, Filters/Filter banks/Adaptive Filters, etc. in this part.
- Segmentation: In this part, we segment the input PCG & LS signals. There are various methods that can be used for the segmentation of the PCG & LS signals like Envelope-based methods, Feature-based techniques, Machine Learning Algorithms, HMM-based Approaches, etc.
- Feature Extraction/Reduction & Classification: We may then extract features from our signals for further classification. The typical methods for classification can be artificial neural network-based classification; support vector machine-based classification; HMM-based classification, clustering-based classification, etc.
For Hardware Implementation, we may use FPGA, Arduino, NI myRIO, Raspberry PI, etc, along with a display screen input sensors as mentioned above and other necessary circuitry.
More than 75% CVDs and almost 80% of COPD deaths occur in middle and low-income countries, mainly due to a lack of timely diagnosis & treatment because of a lack of trained professionals or MRI, X-RAY & CT-Scan being highly expensive and their unavailability in every medicare center. Our standalone low-cost and portable CardioPulmo Monitoring System could prove a reliable and possible solution in this situation. It will remove the need for diagnostic methods subjective on human discriminatory ability and highly trained professionals required. It can be used domestically and in hospitals without the need for professionals to operate it as it automates the whole process. It will provide a full-fledged comprehensive report on the condition of the human heart and lungs having the diagnostic/classification results along with other key metrics and information that may help doctors get diagnostic insights and better understand the current heart and lungs condition.
Through this, a timely, cheap and reliable diagnosis can be made possible and millions of lives can be saved through this.
A hardware-software integrated system based on either FPGA Spartan 2/3, NEXYS 4 Artix 7 FPGA, NI myRIO, or Raspberry-pi development boards.
Acoustic sensors will be used for data acquisition. It may be performed using the following approaches:
- High-quality stethoscopes (By Littmann) with embedded microphones
- Designing our own Stethoscope
- NI Biomedical Instrumentation Kit
Finally to display the results and the medical condition reports an LCD display screen will be employed.
As far as the software algorithm is concerned. The algorithm will perform the following tasks.
- Preprocessing: In this stage, we retain what we want from our input data, remove what we don't want, and enhance our input data to meet our needs. We might have to use decimators/interpolators, Filters/Filter banks/Adaptive Filters, etc. in this part.
- Segmentation: In this part, we segment the input PCG & LS signals. There are various methods that can be used for the segmentation of the PCG & LS signals like Envelope-based methods, Feature-based techniques, Machine Learning Algorithms, HMM-based Approaches, etc.
- Feature Extraction/Reduction & Classification: We may then extract features from our signals for further classification. The typical methods for classification can be artificial neural network-based classification; support vector machine-based classification; HMM-based classification, clustering-based classification, etc.
| Item Name | Type | No. of Units | Per Unit Cost (in Rs) | Total (in Rs) |
|---|---|---|---|---|
| Total in (Rs) | 77000 | |||
| Raspberry pi 4 Model B | Equipment | 1 | 32500 | 32500 |
| 7 inch Capacitive Touch LCD screen | Equipment | 1 | 12000 | 12000 |
| Littmann Classic 2 SE Stethoscope | Equipment | 2 | 11500 | 23000 |
| Wires, cables, product casings, additional peripherals, etc | Miscellaneous | 1 | 9500 | 9500 |