Early Diagnosis of Alzheimer Disease using GA optimized Transfer Learning Techniques

Alzheimer Disease (AD) is one of the serious diseases that affects elderly people. MRI images is the commonly used source for the early diagnosis of AD. Deep learning and transfer learning approaches are gaining importance in the field of computer vision, Image and video processing. Previous studies in AD used transfer learning techniques without optimization and it degraded the performance of the algorithm. Hence, this work proposes a modified transfer learning pipeline to detect the different classes of AD as mild demented, moderate demented, non demented, very mild demented. The proposed pipeline includes the following steps: pre-processing, feature learning using deep neural networks and Gray level Co-occurrence Matrix (GLCM), principal feature analysis using principal component analysis (PCA) and classification using Gaussian mixture model (GMM). The proposed pipeline is evaluated on KAGGLE Alzheimer dataset to prove its effectiveness. The experimental results shows that the proposed transfer pipeline produces astounding results in classifying the given MRI images into one of the four classes.