Knowledge Learning Without Forgetting for the Detection of Alzheimer’s Disease

Abstract

Alzheimer’s disease (AD) is an extremely damaging, slow-progressing neurological disease that causes tremendous inconvenience to patients’ lives. Numerous medical professionals have researched that timely diagnosis and early therapy of AD when it is in its early stages could slow down the progression of AD and even be cured. Therefore, early diagnosis of AD is in urgent need of significant advancement. Nevertheless, there are problems such as brain images with many similar features that are difficult to extract and classify, and insufficient data for training. In this paper, a transfer learning-based knowledge learning without forgetting method we proposed for AD detection, which can preserve the learned knowledge during the transfer process so that it will not be excessively forgotten and this method achieve promising outcomes. The classification accuracy of our method based on resnet50 and resnet18 on the ADNI dataset reached 96.15% and 96.39%, compared to training directly on resnet50 and resnet18, our method increased the classification accuracy by 2.16% and 3.61%, which achieved remarkable results and contributes greatly to the development of AD detection.

This work was supported in part by the Key Research and Development Program of Shaanxi under Grant 2022GY-062, in part by the National Natural Science Foundation of China under Grant 61772401, and in part by the Science and Technology on Communication Information Security Control Laboratory. (R. Liu and Y. Yin — Contributed equally to this work.).