An interpretable schizophrenia diagnosis framework using machine learning and explainable artificial intelligence
Samhita Shivaprasad, Krishnaraj Chadaga, Cifha Crecil Dias, Niranjana Sampathila, Srikanth Prabhu
Abstract
Schizophrenia is a complicated and multidimensional mental condition marked by a wide range of emotional, cognitive, and behavioural symptoms. Although the exact root cause of schizophrenia is unknown, experts believe that a complex interaction of genetic, environmental, neurobiological, neurodevelopmental, and immune system dysfunctional elements are the contributing factors. In healthcare, artificial intelligence (AI) is used for analysing big datasets, enhance patient care, personalize treatment regimens, improve diagnostic accuracy, and expedite administrative duties. Hence, ML has been used to diagnose Schizophrenia in this study. The term ‘explainable artificial intelligence' (XAI) describes the development of AI systems that are able to provide understandable explanations for their choices as well as behaviours. In our research paper, we harnessed the power of five diverse XAI methodologies: LIME (Local Interpretable Model-agnostic Explanations), SHAP (Shapley Additive exPlanations), ELI5 (Explain Like I'm 5), QLattice, and Anchor. According to (XAI), the most significant attributes include age range, sex, the presence of a triradius on the left thumb, the total number of triradii, and the left thenar region's palmar pattern. By enabling early intervention, automatic identification of schizophrenia using XAI can benefit patients, assisting doctors in making precise diagnoses, assisting medical personnel in maximizing resource allocation and care coordination.