Neural Network Development for Artery Scan Analysis

This technical report outlines the development of a neural network designed to detect anomalies in artery scan videos. The project was initiated by a major medical equipment vendor in the US, focusing on improving the efficiency of analyzing ultrasound scans for cholesterol plaques. Previously, the analysis process was manual and time-consuming, requiring specialists to review entire five-minute videos. The objective was to automate this process by integrating a more cost-effective portable scanner and developing a machine learning module to identify anomalies. The report details the project's phases, including interface development, device integration, and the automation of scan reading. It describes the use of convolutional neural networks for visual recognition and the challenges faced during scanner integration and testing. The final system enables analysts to receive a set of highlight frames, significantly reducing the time spent on manual review, while ensuring compliance with data protection standards.