Developing AI Expert System on Dementia for the Elderly

  • 張 駿揚

Student thesis: Doctoral Thesis

Abstract

With the advent of an aging society the number of people who are being afflicted with Alzheimer’s disease is also on a gradual rise However an effective medical treatment to contain the contraction of the disease still does not exist Neuroimaging is becoming a progressively beneficial method in understanding the pathogenesis of AD progress over the recent years and the importance of using deep learning techniques in the field of medical imaging technologies have made progress in both research and clinical care Many researchers have engaged in the development of a auxiliary diagnostic system for medical imaging On the other hand with the features of being prevalent inexpensive and non-invasive SPECT has the whip hand of Neuroimaging and does present diagnostic features of AD to a great extent However diagnosis of SPECT images requires a rigorous process and must have a high level of expertise Physicians need to spend a lot of time viewing images to make a diagnosis Therefore deep learning can assist neurologists diagnose clinically by automatic detection of representations In addition to relying on their own clinical experience physicians can also use the discrimination results of the auxiliary system as the basis for objective reference and thus to reduce the mistakes caused by the limitations of human judgment The study will develop an AI expert system that can analyze the severity of Alzheimer's disease By detecting blood flow in specific areas of the brain the severity of Alzheimer's disease in the subject can be analyzed in real time A method that utilizes deep learning features to enable SPECT images to have a higher recognition accuracy in detection were proposed in this research We use the AutoKeras which is widely implied to reconstruct a deep neural network based on the residual network architecture and the model was trained by the image datasets provided from the Department of Neurology National Cheng Kung University Hospital The study successfully classified the functional SPECT data of subjects with Alzheimer’s disease where the accuracy of validation data and testing data respectively reached 83 17% and 76 39% Experimental results demonstrate that the proposed method of deep neural network is efficient for the diagnosis of Alzheimer's disease with SPECT images and clearly interprets the effectiveness of the proposed method It is expected to facilitate technological advancements in disease-aided diagnosis and help neurologists make correct judgments in the diagnosis of Alzheimer's disease
Date of Award2020
Original languageEnglish
SupervisorChien-Hsu Chen (Supervisor)

Cite this

'