Attention module improves both performance and interpretability of four-dimensional functional magnetic resonance imaging decoding neural network.

Decoding brain cognitive states from neuroimaging signals is an important topic in neuroscience. In recent years, deep neural networks (DNNs) have been recruited for multiple brain state decoding and achieved good performance. However, the open question of how to interpret the DNN black box remains unanswered. Capitalizing on advances in machine learning, we integrated attention modules into brain decoders to facilitate an in-depth interpretation of DNN channels. A four-dimensional (4D) convolution operation was also included to extract temporo-spatial interaction within the fMRI signal. The experiments showed that the proposed model obtains a very high accuracy (97.4%) and outperforms previous researches on the seven different task benchmarks from the Human Connectome Project (HCP) dataset. The visualization analysis further illustrated the hierarchical emergence of task-specific masks with depth. Finally, the model was retrained to regress individual traits within the HCP and to classify viewing images from the BOLD5000 dataset, respectively. Transfer learning also achieves good performance. Further visualization analysis shows that, after transfer learning, low-level attention masks remained similar to the source domain, whereas high-level attention masks changed adaptively. In conclusion, the proposed 4D model with attention module performed well and facilitated interpretation of DNNs, which is helpful for subsequent research.

Longitudinal resting-state functional connectivity and regional brain atrophy-based biomarkers of preclinical cognitive impairment in healthy old adults.

BACKGROUND: Intervention against age-related neurodegenerative diseases may be difficult once extensive structural and functional deteriorations have already occurred in the brain. AIM: Investigating 6-year longitudinal changes and implications of regional brain atrophy and functional connectivity in the triple-network model as biomarkers of preclinical cognitive impairment in healthy aging. METHODS: We acquired longitudinal cognitive scores and magnetic resonance imaging (MRI) data from 74 healthy old adults. Resting-state functional MRI (rs-fMRI) analysis was conducted using FSL6.0.1 to examine functional connectivity changes and regional brain morphometries were quantified using FreeSurfer5.3. Finally, we cross-validated and compared two support vector machine (SVM) regression models to predict future 6-year cognition score from the baseline regional brain atrophy and resting-state functional connectivity (rs-FC) measures. RESULTS: After a 6-year follow-up, our results (P < 0.05-corrected) indicated significant connectivity reduction within all the three brain networks, significant differences in regional brain volumes and cortical thickness. We also observed significant improvement in episodic memory and significant decline in executive functions. Finally, comparing the two models, we observed that regional brain atrophy predictors were more efficient in approximating future 6-year cognitive scores (R = 0.756, P < 0.0001) than rs-FC predictors (R = 0.6, P < 0.0001). CONCLUSION: This study used longitudinal data to keep subject variability low and to increase the validity of the results. We demonstrated significant changes in structural and functional MRI over 6 years. Our findings present a potential neuroimaging-based biomarker to detect cognitive impairment and prevent risks of neurodegenerative diseases in healthy old adults.

Decoding and mapping task states of the human brain via deep learning.

Support vector machine (SVM)-based multivariate pattern analysis (MVPA) has delivered promising performance in decoding specific task states based on functional magnetic resonance imaging (fMRI) of the human brain. Conventionally, the SVM-MVPA requires careful feature selection/extraction according to expert knowledge. In this study, we propose a deep neural network (DNN) for directly decoding multiple brain task states from fMRI signals of the brain without any burden for feature handcrafts. We trained and tested the DNN classifier using task fMRI data from the Human Connectome Project's S1200 dataset (N = 1,034). In tests to verify its performance, the proposed classification method identified seven tasks with an average accuracy of 93.7%. We also showed the general applicability of the DNN for transfer learning to small datasets (N = 43), a situation encountered in typical neuroscience research. The proposed method achieved an average accuracy of 89.0 and 94.7% on a working memory task and a motor classification task, respectively, higher than the accuracy of 69.2 and 68.6% obtained by the SVM-MVPA. A network visualization analysis showed that the DNN automatically detected features from areas of the brain related to each task. Without incurring the burden of handcrafting the features, the proposed deep decoding method can classify brain task states highly accurately, and is a powerful tool for fMRI researchers.

Cerebellar Structural Abnormality in Autism Spectrum Disorder: A Magnetic Resonance Imaging Study.

OBJECTIVE: This study uses structural magnetic resonance imaging to explore changes in the cerebellar lobules in patients with autism spectrum disorder (ASD) and further analyze the correlation between cerebellar structural changes and clinical symptoms of ASD. METHODS: A total of 75 patients with ASD and 97 typically developing (TD) subjects from Autism Brain Imaging Data Exchange dataset were recruited. We adopted an advanced automatic cerebellar lobule segmentation technique called CEREbellum Segmentation to segment each cerebellar hemisphere into 12 lobules. Normalized cortical thickness of each lobule was recorded, and group differences in the cortical measures were evaluated. Correlation analysis was also performed between the normalized cortical thickness and the score of Autism Diagnostic Interview-Revised. RESULTS: Results from analysis of variance showed that the normalized cortical thickness of the ASD group differed significantly from that of the TD group; specifically, the ASD group had lower normalized cortical thickness than the TD group. Post-hoc analysis revealed that the differences were more predominant in the left lobule VI, left lobule Crus I and left lobule X, and in the right lobule VI and right lobule Crus I. Lowered normalized cortical thickness in the left lobule Crus I in the ASD patients correlated positively with the abnormality of development evident at or before 36 months subscore. CONCLUSION: These results suggest abnormal development of cerebellar lobule structures in ASD patients, and such abnormality might significantly influence the pathogenesis of ASD. These findings provide new insights into the neural mechanisms of ASD, which may be clinically relevant to ASD diagnosis.