Development of Deep Learning Algorithms to Detect Abnormalities in Mammography Images
Computer-aided detection/diagnosis (CAD) systems have been available in health care settings for many years. These systems have been developed for finding and discriminating between normal and abnormal tissues in many clinical areas, e.g. lung nodules, fracture detection, aneurysm detection, and breast lesions. Many CAD systems use human-driven processes—feature recognition and a-priori knowledge—to accomplish these tasks. Artificial intelligence, and more specifically deep learning methods, shift the rule-based, problem- specific solutions to more generic methods using predictive algorithms and data-based mathematical optimization processes. The Medical Imaging Informatics Lab at Michigan Tech is involved in developing these advanced processing methods. Our research team is working in collaboration with Henry Ford Hospital in Detroit, Michigan for differentiation of breast abnormalities in mammography images using deep learning algorithms.
Location: Medical Imaging Informatics Lab
Unique Medical Biometric Recognition Enforcement of Legitimate and Large-scale Authentication (UMBRELLA)
A new technique of user identification in health care, UMBRELLA consists of algorithms developed to establish a unique health identifier (UHID) based on a user's fingerprints. Facial recognition is then possible by comparing the user to the matching medical record's patient image. If a match is successful, a longitudinal record of the patient is accessible, based on the security provisions set on the record. The algorithms permit inexpensive common phone and web-camera technology to capture fingerprint and facial recognition biometrics. Interoperability and security mechanisms are established to provide an end-to-end accurate national identification and health data exchange.
The UMBRELLA solution aims to reduce user misidentification in health care and enhance the interoperability and security mechanisms affiliated with health data exchange, resulting in lower health care costs, enhanced population health monitoring and improved patient safety.
Location: H2c2 Lab