I have performed Instance segmentation of Corpus Callosum using a Novel CNN architecture, build on top of U-Net Segmentation Architecture. I have generated the ground Truth on the fly using a novel parcellation algorithm based on witelson scheme. It outputs a Multi-Class Dice Similarity Score of 0.932. This is the first time in my knowledge when a Corpus Callosum present in the brain has been Segmented into various portions by utilising Deep learning. This can be useful in producing the Bio-markers present in Corpus Callosum for diagnosing the Autistic or Alzheimer Suffer.
siddharth-shrivastava7 / corpus-callosum-instance-segmentation Goto Github PK
View Code? Open in Web Editor NEWNovel CNN Architecture || Ground Truth Generated on the Fly using Parcellation algorithm
License: Apache License 2.0