Statistical analysis and machine learning in multimodal brain imaging of neuropsychiatric disorders

• Main Author: Hu, Mengjiao
• Other Authors: Guan Cuntai
• Format: Thesis-Doctor of Philosophy
• Language: English
• Published: Nanyang Technological University 2021
• Subjects: Engineering::Computer science and engineering
• Online Access: https://hdl.handle.net/10356/153263
• Institution: Nanyang Technological University

Neuropsychiatric disorders are the leading causes of mortality and disability. Specifically, the high prevalence of stroke and high rate of disability due to stroke results in huge socioeconomic burdens. Schizophrenia causes profound effects to patients, their families and the society with its early onset of the disease and its incurable nature with persisting symptoms. Multimodal neuroimaging techniques plays an essential role in investigating brain functional and structural changes in neuropsychiatric disorders. In this thesis, the first study utilized functional magnetic resonance imaging (MRI) to examine the brain alteration post stroke and rehabilitation effect of Brain-computer interface-assisted motor imagery (MI-BCI) or transcranial direct current stimulation (tDCS); the second study utilized structural MRI and diffusion MRI for classification of schizophrenia patients and healthy controls with 3D convolutional neural networks trained from scratch; the third study utilized structural MRI for classification of schizophrenia patients and healthy controls using pre-trained 2D convolutional neural networks. MI-BCI and tDCS have been proven effective in post-stroke motor function enhancement, yet whether the combination of MI-BCI and tDCS may further benefit the rehabilitation of motor functions remains unknown. The first study investigated brain functional activity and connectivity changes using amplitude of low-frequency fluctuation (ALFF), regional homogeneity (ReHo), and functional connectivity (FC) from resting-state functional magnetic resonance imaging (fMRI) data pre- and post- a two-week MI-BCI and tDCS combined intervention in 19 chronic subcortical stroke patients. At baseline, stroke patients had lower ALFF in the ipsilesional somatomotor network (SMN), lower ReHo in the contralesional insula, and higher ALFF/Reho in the bilateral posterior default mode network (DMN) compared to age-matched healthy controls. After the intervention, the MI-BCI only group showed increased ALFF in contralesional SMN and decreased ALFF/Reho in the posterior DMN. In contrast, no post-intervention changes were detected in the MI-BCI+tDCS group. Furthermore, higher increases in ALFF/ReHo/FC measures were related to better motor function recovery (measured by the Fugl-Meyer Assessment scores) in the MI-BCI group while the opposite association was detected in the MI-BCI+tDCS group. Taken together, the findings suggest that brain functional re-normalization and network-specific compensation were found in the MI-BCI only group but not in the MI-BCI+tDCS group although both groups gained significant motor function improvement post-intervention with no group difference. MI-BCI and tDCS may exert differential or even opposing impact on brain functional reorganization during post-stroke motor rehabilitation; therefore, the integration of the two strategies requires further refinement to improve efficacy and effectiveness. The ability of automatic feature learning makes CNN potentially suitable to uncover the complex and widespread brain changes in schizophrenia. Despite that, limited studies have been done on schizophrenia identification using interpretable deep learning approaches on multimodal neuroimaging data. In the second study, I developed a deep feature approach based on pre-trained 2D CNN and naive 3D CNN models trained from scratch for schizophrenia classification by integrating 3D structural and diffusion magnetic resonance imaging (MRI) data. I found that the naive 3D CNN models outperformed the pretrained 2D CNN models and the handcrafted feature-based machine learning approach using support vector machine during both cross-validation and testing on an independent dataset. Multimodal neuroimaging-based models accomplished performance superior to models based on a single modality. Furthermore, I identified brain grey matter and white matter regions critical for illness classification at the individual- and group-level which supported the salience network and striatal dysfunction hypotheses in schizophrenia. The findings underscore the potential of CNN not only to automatically uncover and integrate multimodal 3D brain imaging features for schizophrenia identification, but also to provide relevant neurobiological interpretations which are crucial for developing objective and interpretable imaging-based probes for prognosis and diagnosis in psychiatric disorders. Deep learning has been successfully applied to the classification of both natural images and medical images. However, there remains a gap for applying pre-trained 2D networks to the classification of 3D neuroimaging data, especially for diseases without visible alteration in specific slices such as schizophrenia and depression. In the third study, I propose to process the 3D data by a 2+1D framework so that I can exploit the powerful deep 2D networks pre-trained on huge ImageNet dataset for 3D neuroimaging classification. Specifically, 3D volumes of MRI metrics (grey matter, white matter, and cerebrospinal fluid) are decomposed to 2D slices according to neighouring voxel positions and input to 2D CNN models pre-trained on ImageNet to extract feature maps from three views (axial, coronal, and sagittal). Global pooling is applied to remove redundant information based on sparsely distributed activation patterns from the feature maps. Channelwise and slicewise convolution aggregate the contextual information in the third view dimension unprocessed in the 2D CNN network. Multi-view and multi-metric information are fused for final prediction. The proposed approach outperforms handcrafted feature-based machine learning, deep feature approach with a support vector machine SVM classifier and 3D CNN models trained from scratch with better cross-validation results on publicly available Northwestern University Schizophrenia Dataset and independent testing results on a private dataset. In overall, this thesis illustrated multiple advanced analytic techniques for investigating brain alterations in patients with neuropsychiatric disorders. Statistical analysis examined group-level brain functional activity and connectivity changes and associations with behavioural changes. Machine learning approaches not only constructed conventional classifiers and naïve 3D CNN models for classification of schizophrenia patients and healthy controls but also proposed an efficient framework for applying pre-trained 2D CNN on 3D neuroimaging data. The proposed models and framework lay the foundation for objective imaging-based assays for individual-level classification in neuropsychiatric disorders.