Medical diagnosis is a complex subject and sometimes doctors can make errors. Errors of judgement by physicians can be costly because patient can lose his life. Usually complex diagnostic conditions fall in internal medicine domain and consist of about 485 diseases. Manifestation of these diseases overlap group of diseases who share similar features. Approximately 2000 features are shared by these diseases. These conditions are also known as difficult to diagnose cases and mis or missed diagnosis of them is not unusual. Artificial intelligence is used to minimize risks in diagnosis.
Bayesian probabilistic belief networks is popularized as an algorithm in implementation of AI systems. When specifically applied to medical diagnostic domain, running a single diagnosis takes a super computer about 59 years to iteration. Therefore, an approximation method is developed to run the algorithms on a personal computer. Physician assistant artificial intelligence reference system uses variational probabilistic inference as developed by Jaakkola and Jordan (Journal of Artificial Intelligence Research 10, 1999, 291-322). This algorithm enables a diagnostic data for any given patient. The differential diagnosis and investigations suggested helps physician in finding a possible diagnosis. A Natural Language Processor based on SNOMED makes it easy to use PAIRS in selecting features.
Artificial intelligence consisting of both Natural Language Processor and Diagnostic algorithm helps physicians in making diagnosis of complex cases. It helps in minimizing errors in diagnosis. Thus mis or missed cases are avoided and medical errors are minimized. There is potential for saving enormous costs in terms of loss of time and life by using ai-med for patient benefit.