A Two Tier Neural Inter - Network Based Approach to Medical Diagnosis Using K - Nearest Neighbor Classification for Diagnosis Pruning

Artificial Intelligence has always lent a helping hand to the practitioners of medicine for improving medical diagnosis and treatment. In this paper, we propose the design of a two tier Neural Inter-network based Medical Diagnosis System (NIMD) that uses k-Nearest Neighbor Classification for Diagnosis pruning. The system (fig a.) is essentially two tiered with the first tier handling what we term, diagnosis pruning. Each disease is often characterized by a group of symptoms and Diagnosis pruning which is done in the first tier, is the classification of diseases based on the groups of symptoms observed. The second tier consists of separate modules for each disease that handles the actual detection of the disease based on the intensities of the various symptoms reported by the patient. The disease detection modules comprise different classifiers like Neural Networks, Decision Trees, Bayesian Networks etc. depending on the size of the input vector and the characteristics of the training set. NIMD thus provides diagnosis and probability of presence for a number of possible diseases at the same time by checking for each of them individually in tier 2. An analysis of the performance of our hybrid system reveals superior performance and utility compared to other existing approaches.