B7: Computational Bioengineering IV

MULTICLASS CLASSIFICATION OF APG SIGNALS USING ELM FOR CVD RISK IDENTIFICATION: A REAL-TIME APPLICATION


Manjunath Gaonkar1, Niranjana Krupa1, Kunal Bharathi1, Sai Karun1, Suhan Surendranath1, Mohd Alauddin Mohd Ali2


1PES University, India;
2Universiti Kebangsaan Malaysia, Malaysia

 

Cardiovascular diseases are one of the leading causes of deaths in the world today, accounting for 17.5 million cases every year. This paper presents a non-invasive method of classifying a subject’s health as “Healthy” or “At Risk” of cardiovascular disease (CVD). The novelty of the work lies in recognizing the rare case of a young subject with cardiovascular disease as well as old subjects who are healthy, and the real-time implementation of CVD risk analysis. In the proposed work, 30 healthy and 30 pathological signals were considered. Empirical Mode Decomposition (EMD) was used to remove high-frequency noise, following which the analysis of the acceleration plethysmogram (APG) signals was carried out.

An APG signal is a second derivative of the PPG signal. APG allows more accurate recognition of the inflection points and easier interpretation of the characteristics/features. The signal is analyzed and seven features of the wave contour are extracted. With these features and considering the actual age of
the subject, 4 classes were identified using an extreme learning machine (ELM) classifier, and we grouped them as, Healthy Young, Unhealthy Young, Healthy Old, Unhealthy Old. Implementation of the proposed system is done on a Raspberry Pi 2 using the Python programming language. The training of the classifier and prediction of CVD risk group, using the extracted features, takes on average 17.83 milliseconds. The overall accuracy of the system is 86%. Therefore, the developed system is suitable for real-time identification of CVD risk from APG signals. 

Organised by

Endorsed by

 

        Supported by
 

 

                          
     

       

                             
           
 

        Supporting Media