Systematic Health Artificial Intelligence (SHAI) - A pathology based NLP model for improved predictive diagnostics in personalised medicine

Abstract

Abstract The Systematic Health Artificial Intelligence (SHAI) model trains on data from medical records and clinical laboratory results to temporally identify disease markers with subsequent pathologies, more efficiently and accurately than is done in the current analog practice. The aim of the SHAI model is to gauge a patient’s medical prognostic status based on a conglomerate of data to predict lurking, occult or comorbid pathologies.Newfound associations and predictions would support clinicians in terms of comprehensively visualising a patient’s health profile, both in real-time and for the future. Proxy findings would also help to establish personalised references ranges for clinical pathological investigations of body fluids. The SHAI model processes EMR progress text-based notes through a NLP ‘Bag of Words’ system, which enables the neural network to train in word representation and ‘weigh’ words of proximity. Using ‘forward propagation’ of the vectors will allow for output activation from hidden and non-hidden layers of the developing neural network architecture, to then use ‘multiclass classification’ as the vector contents grow with new data. This manuscript identifies 8 key questions to be addressed by diagnostic ML models and explains SHAI’s design as it pertains to maximising human benefit and minimising bias. Despite the automaticity of this laboratory medicine solution, physician end-users remain essential to the diagnostic process and final clinical judgements.