Development of Clinical Decision Support System Using Genetically Optimized Artificial Neural Network

Clinical Decision Support System (CDSS) is a computerized framework designed to aid healthcare professionals in clinical decision-making. One of the most promising applications of computational intelligence is using Artificial Intelligence (AI) and Machine Learning (ML) to make medical data predictions. When solving a complicated optimization issue with a vast and unattainable number of parameters, a Genetic Algorithm (GA) can be a helpful tool. This research work incorporates the use of GA to construct a novel model GA_BPNN which is a hybrid of GA and Back Propagation Neural Network (BPNN) for predicting various diseases like Diabetes, Breast Cancer, and Parkinson’s Disease (PD), where GA is used for optimizing learning rate, momentum, and hidden layers of Neural Network, and neural networks is employed for the purpose of data classification. The main objective of this research work is to create and implement a robust CDSS that can accurately forecast the occurrence of diabetes, breast cancer, and PD. It was found that the BPNN with GA optimization performed better than the BPNN approach with no GA optimization. Four different models were developed for the diagnosis of diseases through CDSS. The first model is the All-Features Back Propagation Neural Network (AF_BPNN); the Second model is the All Features with Optimized Back Propagation Neural Network (AF_OBPNN); the third one used for this study is the Selected Features Back Propagation Neural Network (SF_BPNN). Moreover, the fourth model used is Selected Features with Optimized Back Propagation Neural Network (SF_OBPNN). The hybrid optimized GA-BPNN with selected features improved classification accuracy over GA-BPNN alone with redundant inputs. In the case of diabetes, SF_OBPNN yielded the highest accuracy (92.3%) and precision (90.9%) when compared with the other three models. For Breast cancer, SF_OBPNN yielded the highest accuracy (98.5%) and precision (97.0%). Similarly, the SF_OBPNN model yielded the highest accuracy (98.1%) and precision (95.8%) for PD. It is evident from the results that when GA is used for optimization, the accuracy achieved by the model is the highest. Finally a CDSS named as HealthScan+ was developed which provides interactive way to upload patient data and diagnose diseases.