LEVERAGING ARTIFICIAL INTELLIGENCE MODELS FOR EARLY DETECTION AND DIAGNOSIS OF NEUROLOGICAL DISORDERS: A COMPARATIVE ANALYSIS OF TECHNIQUES AND APPLICATIONS

Abstract

Detection of Alzheimer's disease (AD) must happen before the onset of neurodegenerative symptoms.  The purpose of this research was to improve AD prediction and early diagnosis through machine learning by analyzing artificial intelligence models (convolutional neural networks, CNN) alongside support vector machines (SVM) and random forests (RF) for neuroimaging data (MRI, PET) assessment.  Researchers conducted a study about the diagnostic power of AI models optimized by combining DNA sequencing results with brain scan and medical history information.  The researchers associated with feature significance analysis to determine which model prediction variables created the most impact by deploying SHAP values alongside other methods.  The study evaluated practical model applications by conducting separate training sessions with AI models on clinical data representing different patient environments.  The validation of each model occurred through statistical measurement consisting of accuracy, precision, recall, F1 score and area under the curve (AUC).  The research showed CNN as the superior model because it achieved highest accuracy at 92% as well as highest precision at 93% and recall at 91% and area under the curve at 0.95.  The study established that CNN serves as an effective technique for detecting Alzheimer's disease at its early stage through the identification of minor imaging deviations.  The accuracy score obtained from CNN was superior to the results found in SVM and RF.  The model displayed important validity through cross-validation tests critical for clinical application.  The application of artificial intelligence models especially CNN has the potential to advance both Alzheimer's disease treatment and diagnosis through prompt identification and improved treatment results.