Abstract:
Cardiovascular diseases (CVDs) or heart diseases are the leading cause of deaths around the
world. In Pakistan, 30 to 40 percent of all deaths are due to CVDs [1]. Main reasons for such a
high percentage include failure of early detection of CVDs, expensive treatments, far away
clinics/hospitals from rural areas and less experts available.
In past stethoscopes were used by doctors for auscultation and diagnosis of heart
diseases. However the method was not always reliable as it heavily depends upon expertise of
doctor. Later on devices like ECG/EKG were invented for accurate diagnosis. ECG amplifies the
electrical signals of heart and display the graph. The ECG method is currently market standard.
However ECG devices are expensive and the results are complex to read which require a trained
technician to read them. Moreover an EC device requires 8-10 electrodes to be placed at human
skin at different body part like chest, wrist and leg.
Heart-Lip is a revolutionary, cheap and portable product that will be used to detect
abnormality in heart beats using PCG (Phonocardiogram). PCG is a plot of heart’s mechanical
activity (sound) transduced at chest recorded using a stethoscope. Automatic
detection/classification of heart beats can be helpful in predicting serious heart issues of a
patient. The system is capable of diagnosing fourteen different types of heart conditions and fatal
diseases. We have used machine learning techniques like ANN, CNN, SVM and AdaBoost for
distinguishing a normal heart from abnormal heart and diagnosis of a specific disease. We have
achieved an accuracy of 95% for specific disease diagnosis and 88% percent for normal/abnormal
classification. The accuracy achieved is fairly excellent compared to current available solution in
market.
