Source: Deep Learning on Medium
CNN for ECG-based Stethoscope Tracking
Cardiac auscultation is the auditory detection of heart sounds to diagnose abnormalities, a crucial skill that is both efficient and cost-effective in medical practice. However, due to increased prevalence and usage of expensive cardiac technologies, many new physicians and trainees have difficulty performing essential cardiac examinations on their patients. A virtual pathology stethoscope is a simulation-based solution that trains students to perform cardiac examinations by listening to abnormal heart sounds in otherwise healthy Standardized Patients. This study reports the accuracy of an electrocardiogram (ECG)-based virtual pathology stethoscope tracking method for placing virtual symptoms in correct auscultation landmarks of patients. A one dimensional convolutional neural network (1D-CNN) algorithm is used to classify the location of the stethoscope on each of four primary chest sites. A 91% accuracy was obtained showing promising performance gain over our pervious method. This finding would significantly aid in extending the capabilities of SPs by placing virtual symptoms in correct auscultation regions.