1H MRSI can assess glioma infiltration margins and malignant invasion but technical limitations prevent widespread use. In this study we used 2D 1H MRSI to determine voxels of specific tumour tissue type from which we extracted multimodal MRI (M-MRI) image characteristics. Subsequently, we applied superpixel segmentation and Bayesian statistical analysis to M-MRI alone to derive nosologic tumor images of these same tissue types with whole brain coverage. We obtained 100% classification accuracy for overall glioma grade, and an average 0.77 Dice overlap coefficient with the manual segmentation volume. Such methodology could aid prognostic assessment, surgical treatment and radiotherapy dose planning.
This abstract and the presentation materials are available to members only; a login is required.