Early Detection of Alzheimer's Disease Based on Deep Learning Approach Using Brain MRI

Alzheimer's disease is one kind of neurological condition for which people gradually decrease their mental capacity and memory. The number of Alzheimer's patients nowadays is rising quickly, which is quite concerning for the advancement of humanity. The scenario is much more terrible because this illness has no cure. A professional neurologist needs to examine medical imaging techniques including PET scans, CT scans, and MRIs for diagnosing an Alzheimer's patient. This procedure is time- and resource-intensive. Many people hesitate to treat Alzheimer's disease because of this instability. The potential for early detection of Alzheimer's disease might be achieved by the use of Magnetic Resonance Imaging (MRI). However, the MRI dataset has an issue in class imbalance problem. To address the problems in the dataset, a data balancing technique called SMOTETomek is used and then a lightweight CNN model is developed that uses MRI images as input and determines a patient's stage. In terms of recall, precision, accuracy, ROC curve, and f1 score, the performance of the CNN model has been compared with a few transfer learning algorithms. Therefore, using the proposed model, each Alzheimer's patient may determine what stage of the disease they are now in and take further actions to lower the mortality rate of Alzheimer's patients. The model under consideration demonstrates superior performance when compared to previous models, with an accuracy rate of 97.50%.