Abstract:
Alzheimer's disease is a progressive neurodegenerative disorder that affects millions of people worldwide. The earliest detection and precise diagnostic methodologies associated with Alzheimer’s are essential for the management of the disease and necessary measures aimed at its prevention. Nowadays, medical imaging especially MRIs have demonstrated impressive prospects for early diagnosis.
This final year project report proposes an innovative method for early diagnosis of Alzheimer’s through MRI scan using a hybrid model in combination with CNN. A new model, which hopes to exploit CNN’s strengths for improved predictions is being proposed here.
This was a special software which accepted MRI image file as the input and produced live results of its processing on the application system’s interface. This process starts with pre-processing of the MRI data and identifying the appropriate features for analysis. Thereafter, a CNN model is trained using an adequately curated set of MRI scans as well as corresponding clinical information.
In particular, the CNN portion of this model learns to find high-level patterns in MRI data and, based on that knowledge, makes predictions about the probability of early-stage Alzheimer’s disease. The model's training process focuses on optimizing the performance of the CNN component to enhance overall predictive accuracy.
The dataset that has been used in project is obtained from Kaggle which is a benchmark dataset that has been used in multiple research work. The final accuracy of experimentation is 94%. The state-of-the-art comparison is also given with 2 more approaches.
