How training doctors with AI could improve patient outcomes
Prof George Shorten presenting at the TechConnect Live exhibition on, 'how training doctors with artificail intelligence (AI) could improve patient outcomes'. AI refers to the capacity of a computer to perform operations analogous to learning and decision making in humans. Machine learning is one advanced application of AI concerned with developing computer programs that automatically improve with experience.
Amalgamation of data from several sources will make it possible to predict clinical outcome for a given procedure more efficiently. The same combined dataset could also be incorporated into the design of guidance systems to optimise a clinician’s personal learning curve. For example, during nerve blocks, such guidance systems could assist the doctor in patient positioning, ultrasound image interpretation, indication of the optimal injection site, needle positioning and evaluation of the dose and spread of local anaesthetic. The doctor’s assisted behaviours (collectively his/her performance) could first be evaluated in the setting of simulation-based training (model or cadaver) and subsequently in the clinical environment. Currently, the manual interpretation of performance data is valuable, but limited by the constraints of expert or trained manual observation or of multimedia information retrieval. By letting ‘the data speak for themselves’ through a machine-learning algorithm, these patient level observations offer a means to predict and assess high- or low quality performance (efficiency and error avoidance) and
clinical outcome. In this application of machine learning, the behaviours identified as predictive of good outcome provide targets for individual deliberate practice, procedure refinement/renewal and medical device design and programming.
British Journal of Anaesthesia, 121 (3): 521e523 (2018)