Using Unsupervised Learning Methods to Analyse Magnetic Resonance Imaging (MRI) Scans for the Detection of Alzheimer’s Disease

Background: Alzheimer’s disease (AD) is the most common cause of dementia, characterised by behavioural and cognitive impairment. The manual diagnosis of AD by doctors is time-consuming and can be ineffective, so machine learning methods are increasingly being proposed to diagnose AD in many recent studies. Most research developing machine learning algorithms to diagnose AD use supervised learning to classify magnetic resonance imaging (MRI) scans. However, supervised learning requires a considerable volume of labelled data and MRI scans are difficult to label. The aim of this thesis was therefore to use unsupervised learning methods to differentiate between MRI scans from people who were cognitively normal (CN), people with mild cognitive impairment (MCI), and people with AD.
Objectives: This study applied a statistical method and unsupervised learning methods to discriminate scans from (1) people with CN and with AD; (2) people with stable mild cognitive impairment (sMCI) and with progressive mild cognitive impairment (pMCI); (3) people with CN and with pMCI, using a limited number of labelled structural MRI scans.
Methods: Two-sample t-tests were used to detect the regions of interest (ROIs) between each of the two groups (CN vs. AD; sMCI vs. pMCI; CN vs. pMCI), and then an unsupervised learning neural network was employed to extract features from the regions. Finally, a clustering algorithm was implemented to discriminate between each of the two groups based on the extracted features. The approach was tested on baseline brain structural MRI scans from 715 individuals from the Alzheimer’s Disease Neuroimaging Initiative (ADNI), of which 231 were CN, 198 had AD, 152 had sMCI, and 134 were pMCI. The results were evaluated by calculating the overall accuracy, the sensitivity, specificity, and positive and negative predictive values.
Results: The abnormal regions around the lower parts of the limbic system were indicated as AD-relevant regions based on the two-sample t-test (p<0.001), and the proposed method yielded an overall accuracy of 0.842 for discriminating between CN and AD, an overall accuracy of 0.672 for discriminating between sMCI and pMCI, and an overall accuracy of 0.776 for discriminating between CN and pMCI.
Conclusion: The study combined statistical and unsupervised learning methods to identify scans of people with different stages of AD. This method can detect AD-relevant regions and could be used to accurately diagnose stages of AD; it has the advantage that it does not require large amounts of labelled MRI scans. The performances of the three discriminations were all comparable to those of previous state-of-the-art studies. The research in this thesis could be implemented in the future to help in the automatic diagnosis of AD and provide a basis for diagnosing sMCI and pMCI.