Classification of Alzheimer’s Disease from 18F-FDG and 11C-PiB PET Imaging Biomarkers Using Support Vector Machine

• Published in: Journal of medical and biological engineering Vol. 40; no. 4; pp. 545 - 554
• Main Authors: Yang, Bang-Hung, Chen, Jyh-Cheng, Chou, Wen-Hsiang, Huang, Wen-Sheng, Fuh, Jong-Ling, Liu, Ren‑Shyan, Wu, Cheng-Han
• Format: Journal Article
• Language: English
• Published: Berlin/Heidelberg Springer Berlin Heidelberg 01.08.2020; Springer Nature B.V
• Subjects: Algorithms; Alzheimer's disease; Biomarkers; Biomedical Engineering and Bioengineering; Cell Biology; Classification; Cognitive ability; Cortex (temporal); Dementia; Dementia disorders; Diagnosis; Engineering; Feasibility; Fluorine isotopes; Imaging; Learning algorithms; Machine learning; Magnetic resonance imaging; Medical imaging; Neurodegenerative diseases; Neuroimaging; Original Article; Positron emission; Positron emission tomography; Radiology; Support vector machines; Tomography; FDG; PiB; Biomarker; Support vector machine; Alzheimer's disease; Machine learning
• ISSN: 1609-0985; 2199-4757
• DOI: 10.1007/s40846-020-00548-1

Purpose Alzheimer’s disease (AD) is the most common type of dementia, and its early diagnosis has become a crucial issue. Machine learning provides a systematic and objective approach in classification. Currently, there are many studies using several kinds of neuroimaging modalities to perform classification in dementia. Support vector machine (SVM) is one of machine learning based classification algorithm which is able to retain favorable classification accuracy even with small sample sizes. Our aim is to investigate the feasibility of using dual PET biomarkers in combination with SVM for AD diagnosis in small sample sizes. Methods This study collected PET ( 18 F-FDG and 11 C-PiB) and T1 MRI image of 79 subjects from the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database, including 20 AD, 27 mild cognitive impairment (MCI) subjects, and 32 normal controls (NCs), and performed classification using the SVM algorithm with the quantification of the two PET biomarkers, and finally compared the classification results of each brain region. Results In the classification between diseased (AD and MCI) and NC group, we found that the accuracy, sensitivity and specificity mean in temporal cortex are the highest. Conclusions Overall, using dual PET biomarkers in combination with SVM shows a certain feasibility and clinical value in the diagnosis of AD, especially in the temporal cortex.