Deep learning based mild cognitive impairment diagnosis using structure MR images

• Published in: Neuroscience letters Vol. 730; p. 134971
• Main Authors: Jiang, Jingwan, Kang, Li, Huang, Jianjun, Zhang, Tijiang
• Format: Journal Article
• Language: English
• Published: Ireland Elsevier B.V 21.06.2020
• Subjects: Aged; Aged, 80 and over; Alzheimer Disease - diagnosis; Alzheimer Disease - physiopathology; Brain - pathology; Brain - physiopathology; Cognitive Dysfunction - diagnosis; Cognitive Dysfunction - physiopathology; Convolutional neural network; Deep Learning; Early mild cognitive impairment; Humans; Magnetic Resonance Imaging - methods; Support Vector Machine; Transfer learning; Support vector machine; Transfer learning; Convolutional neural network; Early mild cognitive impairment
• ISSN: 0304-3940; 1872-7972; 1872-7972
• DOI: 10.1016/j.neulet.2020.134971

•The first study to apply CNN to brain sMRI to classify EMCI(early Mild Cognitive Impairment) and NC(Normal Control) Mild cognitive impairment (MCI) is an early sign of Alzheimer's disease (AD) which is the fourth leading disease mostly found in the aged population. Early intervention of MCI will possibly delay the progress towards AD, and this makes it very important to diagnose early MCI (EMCI). However, it is very difficult since the subtle difference between EMCI and cognitively normal control (NC). For improving classification performance, this paper presents a deep learning based diagnosis approach using structure MRI images for exploiting deeply embedded diagnosis features; then a feature selection strategy is performed to eliminate redundant features. A Support Vector Machine (SVM) is further employed to distinguish EMCI from NC. Experiments were performed on the publicly available ADNI dataset with a total of 120 subjects. The classification results demonstrate the superior performance of the proposed method with accuracy of 89.4% for EMCI versus NC.