Publications
Peer-reviewed publications and conference contributions. Click “Abstract” to expand details.
Implementation of Non-Contact Bed Embedded Ballistocardiogram Signal Measurement and Valvular Disease Detection
Abstract: In a traditional system, ECG leads are connected to the patient's chest to detect the electrical performance of the heart which creates long-term discomfort for the patient. As ballistocardiogram (BCG) and valvular diseases are both mechanical phenomena, we conjectured that valvular disease could be diagnosed from non-contact BCG measurement. We proposed a non-contact way to determine valvular diseases that is favorable for long-term observation. We classified data using ANN and SVM. Data were collected from normal persons and persons affected by mitral and pulmonary valve stenosis. We compared results using overall accuracy, misclassification rate, and fitness, achieving the highest test accuracy of 79.12% using SVM for decomposition level 1.
Diagnosis of Malignant Melanoma using Color and Textural Features from Dermoscopic Images
Abstract: We proposed a feed-forward back-propagation neural network-based technique using color and textural features for early diagnosis of melanoma from dermoscopic images. The pipeline includes image acquisition, pre-processing, lesion segmentation, feature extraction, and classification. The method achieved classification accuracy of 97%, sensitivity of 92.5%, and specificity of 98.12%.