NeuroVision: A Comparative Machine Learning Framework for MRI-Based Diagnosis of Alzheimer's and Brain Tumors

Neurological diseases like Alzheimer Disease (AD) and brain tumors are some of the complex diseases that make early diagnosis challenging. Due to such complexities, treatment at the right time often experiences delays in neuro disorders, thus affecting the patient's condition. NeuroVision focuses on improved initial diagnosis accuracy. To achieve the higher precision, the advanced techniques like data augmentation, noise reduction, and hyper-parameter tuning to reduce false positives. Primarily, it is achieved with MRI imaging to be used for more accurate assessment. In the interest of simplicity, four different models of machine learning SVM, Random Forest (RF), Logistic Regression, K-Nearest Neighbors (KNN), and Neural Networks are compared to determine the best model to use in reviewing MRI scans from patients with Alzheimer Disease and brain tumors. Because features between the two conditions share considerable overlap, the aim here simply is to identify the best performing model for attaining diagnostic accuracy. The metrics used to measure the performance in this study are accuracy, precision, recall, Fl-score, and Matthews Correlation Coefficient (MCC). The accuracy of the model on the test set reports an accuracy of over 92 % for Alzheimer's whereas, for the case of brain tumor, the precision is about 82%. Comparing KNN that resulted in being the best model for Alzheimer's disease because this one is simple and accurate, while a neural network performed better when it comes to brain tumor diagnosis for the robust achievement of accuracy and adaptability in recognizing difficult-to-pattern.