Explainable Deep Learning Architecture for Alzheimer’s Disease Detection using MRI

Algorithms for machine learning and deep learning have shown to be more successful in detecting diseases, leading to early diagnosis, more accurate findings, and effective therapies. Even though ML and DL algorithms are effective at forecasting outcomes, it is unclear what process the model uses and where the decision-making process originates from. There are situations where it is difficult to understand the predicted outcome. In these situations, XAI is necessary, as it makes the decision-making process transparent and simple to comprehend. The XAI helps the developers ensure that the model satisfies all necessary requirements and functions as intended. Because ML and DL methods are challenging to understand, explainable AI performs tasks enabled by AI. This work compares the behavior of XAI in both scenarios by applying the XAI algorithm LIME (Local Interpretable Model-agnostic Explanations) to DL algorithms. For experimentation, a dataset comprising MRI pictures of the brains of individuals classified as having no dementia, mild dementia, very mild dementia, or moderate dementia is taken into consideration. An accuracy of upto ${9 0. 7 \%}$ is obtained when the LIME method was used in conjunction with the ResNet 50 model.