Bidirectional Mapping Perception-enhanced Cycle-consistent Generative Adversarial Network for Super-resolution of Brain MRI images.

As an effective tool for visualizing neurodegeneration, high-resolution structural magnetism facilitates quantitative image analysis and clinical applications. Super-resolution reconstruction technology allows to improve the resolution of images without upgrading the scanning hardware. However, existing super-resolution techniques relied on paired image data sets and lacked further quantitative analysis of the generated images. In this study, we proposed a semi-supervised generative adversarial network (GAN) model for super-resolution of brain MRI, and the synthetic images were evaluated using various quantitative measures. This model adopted the cycle-consistency structure to allow for a mixture of unpaired data for training. Perceptual loss was further introduced into the model to preserve detailed texture features at high frequencies. 363 subjects with both high-resolution (HR) and low-resolution (LR) scans and 217 subjects with HR scans only were used for model derivation, training, and validation. We extracted multiple voxel-based and surface-based morphological features of the synthetic and real 3D HR images for comparison. We further evaluated the synthetic images in the differential diagnosis of diseases. Our model achieved superior mean absolute error (0.049±0.021), mean squared error (0.0059±0.0043), peak signal-to-noise ratio (29.41±3.71), structural similarity index measure (0.914±0.048). Eight morphological metrics, both voxel-based and surface-based, showed significant agreement (P<0.0001). The gap of accuracy in disease diagnosis between synthetic and real HR images was within 5% and significantly outperformed the LR images. Our proposed model enables the reconstruction of HR MRI and could be used accurately for image quantification.Clinical relevance- Quantitative evaluation of the synthetic high-resolution images was used to determine whether the synthetic images have sufficient realism and diversity.

Early Diagnosis and Biomarkers of Alzheimer's Disease Based on Spatio-temporal Graph Convolution Network.

Functional magnetic resonance imaging (fMRI) could detect the dynamic activity of brain function and communication. Previous studies have found reduced brain functional connectivity in Alzheimer's disease (AD) patients. In this study, we proposed to process fMRI data by spatio-temporal graph convolution network (ST-GCN) to achieve an early differential diagnosis of AD and to extract image markers using gradient-weighted class activation mapping (Grad-CAM). The data used in this study were from the Alzheimer's Disease Neuroimaging Initiative (ADNI) database, Xuanwu Hospital, and Tongji Hospital. The study included 1105 normal controls and 790 patients with mild cognitive impairment (MCI). The grid search method of K-fold cross-validation was used to train the model. In addition, we used Grad-CAM to extract image markers and carried out visualization analysis. This model obtains better AD diagnosis power: accuracy = 0.92, sensitivity = 0.97, specificity = 0.89, and area under the curve=0.96. Salient brain regions extracted by Grad-CAM include the paracentral lobule, inferior occipital gyrus, middle frontal gyrus, superior temporal gyrus, cuneus, posterior cingulate gyrus, and superior parietal gyrus. Our proposed ST-GAN model will help to explore objective markers that can be used for the early diagnosis of AD.Clinical relevance- Our proposed model shows great potential for enhancing the understanding of the pathology of AD by detecting functional connectivity interruptions.