Multi-Class Alzheimer's Disease Classification Using a Fine-Tuned InceptionV3 Pretrained Model: A Step Towards Diagnosis and Prediction

Alzheimer's disease is a degenerative neurologic condition that results in brain shrinkage and neuronal death. The most prevalent kind of memory loss also known as dementia, Alzheimer's disease gradually deteriorates thinking and social connections by damaging the brain cells. To facilitate early identification of Alzheimer's disease using brain Magnetic Resonance Imaging, it is important to design an automated method for the identification and classification of this disease. This system ought to have the capability to accurately identify and categorize individuals who exhibit symptoms of dementia. This work is focused on applying the technique of transfer learning by implication of segmented MRI images of the brain for detecting cognitive dysfunction as well as Alzheimer's disease at initial and different stages. This technique involves fine-tuning a convolutional network that has already been trained, specifically InceptionV3 for the classification of the segmented brain images. Alzheimer's dataset, which includes various images of the brain for four different stages of Alzheimer's, has been employed for effective training and testing of the proposed framework. The model displays great accuracy outcomes in categorizing the MRI images of 96.01% with a very low loss of 0.27 along with training accuracy of 96.41% with a loss of 0.24. This research can be further utilized in the medical field for diagnosing Alzheimer in its early stages.