BIOE 489

Credit: 3 OR 4 hours.

The application of machine learning (ML) to medical image data is an area of intense, well-funded research. Due to practical logistics, however, the ideas expounded in published research articles do not necessarily translate perfectly into clinical implementation. The purposeful design and assessment of machine learning experiments will be introduced and revisited throughout the course. The financial cost of training, data acquisition and expert labelling will be considered in the context of product delivery. Some relevant university, corporate and governmental regulatory policies will be presented by expert guest lecturers. Specific issues of clinical implementation and adoption of new technology will be covered. The ethics of using images influenced or analyzed by ML in patient care and/or medical research will be explored in depth. Topics include: diagnosis accuracy, mandated system upgrades, informed consent, patient privacy, researcher/vendor liability and the role and reliability of federal regulations in ethical application of ML to biomedical data. The policies and procedures of NIH study sections and internal review boards will be highlighted and some key issues related to intellectual property will be surveyed.

3 undergraduate hours. 4 graduate hours. Prerequisite: Restricted to students with senior undergraduate or graduate standing and familiarity with machine learning principles.