Submillimeter T1 atlas for subject-specific abnormality detection at 7T.

PURPOSE: Studies at 3T have shown that T1 relaxometry enables characterization of brain tissues at the single-subject level by comparing individual physical properties to a normative atlas. In this work, an atlas of normative T1 values at 7T is introduced with 0.6 mm isotropic resolution and its clinical potential is explored in comparison to 3T. METHODS: T1 maps were acquired in two separate healthy cohorts scanned at 3T and 7T. Using transfer learning, a template-based brain segmentation algorithm was adapted to ultra-high field imaging data. After segmenting brain tissues, volumes were normalized into a common space, and an atlas of normative T1 values was established by modeling the T1 inter-subject variability. A method for single-subject comparisons restricted to white matter and subcortical structures was developed by computing Z-scores. The comparison was applied to eight patients scanned at both field strengths for proof of concept. RESULTS: The proposed method for morphometry delivered segmentation masks without statistically significant differences from those derived with the original pipeline at 3T and achieved accurate segmentation at 7T. The established normative atlas allowed characterizing tissue alterations in single-subject comparisons at 7T, and showed greater anatomical details compared with 3T results. CONCLUSION: A high-resolution quantitative atlas with an adapted pipeline was introduced and validated. Several case studies on different clinical conditions showed the feasibility, potential and limitations of high-resolution single-subject comparisons based on quantitative MRI atlases. This method in conjunction with 7T higher resolution broadens the range of potential applications of quantitative MRI in clinical practice.

Whole Tumor Segmentation from Brain MR images using Multi-view 2D Convolutional Neural Network.

In this paper, a study is reported on the popular BraTS dataset for segmentation of brain tumor. The BraTS 2019 dataset is used that comprises four MR modalities along with the ground-truth for 259 high grade glioma (HGG) and 76 low grade glioma (LGG) patient data. We have employed U-Net architecture based 2D convolutional neural network (CNN) for each of the orthogonal planes (sagittal, coronal and axial) and fused their predictions. The objective function is aimed to minimize Dice loss between the binary prediction and its actual labels. Samples having tumor information are considered for each patient data to avoid training on non-informative data. The models are trained on 222 HGG data and tested on 37 HGG data using performance metrics such as sensitivity, specificity, accuracy and Dice score. Test-time augmentation is also performed to improve the segmentation performance. 7-fold cross validation is conducted to analyze the performance on different sets of training and testing data.