Avey is an interactive medical self-diagnosis system that has been fully researched, designed, and developed in-house. It utilizes an intelligent inference engine with three major components: (1) a diagnostic algorithm, (2) a finding (i.e., a symptom, an etiology, or an attribute, which is a feature of a symptom or an etiology) recommendation algorithm, and (3) a ranking mechanism. The inference engine taps into a highly sophisticated probabilistic graphical model, namely, a Bayesian network. The above figure demonstrates an actual visualization of Avey’s Bayesian model. The engine's diagnosis algorithm operationalizes the Bayesian model and generates after every patient’s answer (during a session with Avey) a probability for each modelled disease, conditional on the findings that have been discovered or inferred thus far.
Questions are asked during a patient’s session with Avey via the recommendation algorithm of the inference engine. Specifically, after every answer provided by the patient, the algorithm predicts the future diagnostic impact of every relevant finding that has not yet been asked and recommends the one that exhibits the highest impact. The engine asks the recommended finding and continues with the inference process until it converges or hits a maximum number of questions. Afterwards, it applies a ranking mechanism that relies on multiple factors to rank all the possible diseases and outputs them as a differential diagnosis to the patient.