This technology is an automated method to create accurate 3D models of both normal and diseased livers with heterogeneous tissue, using deep learning and artificial intelligence.
The ability to model and separate liver tissue from the surrounding tissue in a CT scan is important for medical interventions such as tumor resection, transplantation, and arthroscopic surgery. The current gold standard uses active contour models or deep learning convolution networks to automatically segment liver tissues. However, active contour models only work well for homogenous liver tissue and perform poorly in heterogenous liver tissue, often seen in livers with tumors, cirrhosis, or partial resection. Convolution networks outperform the active contour models, but they suffer from overfitting of data and thus may not be general enough to model all liver types.
This technology is a software that provides a 3D model of a liver by segmenting images of liver tissue obtained from CT imaging scans. The method used is a combination of both deep learning algorithms and pre-existing deformable models. The technology utilizes convolution neural networks to recognize the liver tissue and active contour models to create a smooth segmentation border through analysis of low-level contour imaging data.
IR CU19417, CU18026
Licensing Contact: Sara Gusik