• Technical Name
  • Hepatitis Ishak Fiborsis Indexing with AI Computer-aided Classification
  • Operator
  • National Cheng Kung University
  • Booth
  • Online display only
  • Contact
  • 詹寶珠
  • Email
  • pcchung@ee.ncku.edu.tw
Technical Description To determine the degree of liver cirrhosis requires the physician to observe the portal structure under the microscope. The process is time-consuming, and the diagnosis varies between doctors. This technique can be used as an objective indicator.
A common marker in hepatitis diagnosis is Ishak fibrosis index. We use fiber segmentation, portal and central vein segmentation, and a circular fibrosis algorithm. We apply these to whole slide images(WSI) with Masson staining.
We develop a multiscale DeepLab v3 network with an object detection cost function. We segment out the fibers and use the circular fibrosis algorithm to analyze them. Nodules are then identified and used to detect the presence of cirrhosis.
We use the Faster-RCNN network to detect portal and central veins. We combine the fiber segmentation results with region-growing to obtain the segmentation of portal and central veins. We then analyze the fiber proliferation around the veins as a reference for Ishak classification.
Scientific Breakthrough This technique develops a series of algorithms for automated analysis of cirrhosis. DeepLabV3, an object segmentation network, is used to accurately segment fibrous areas under Masson stained pathology slides, and the circular fibrosis algorithm developed by this technique is used to detect the presence of nodules(circular area) in the fibers to determine the presence of cirrhosis. The object detection network Faster R-CNN is used to detect the location of portal veins and the central veins to analyze the surrounding fibers. The degree of fiber proliferation is used as a reference for Ishak classification. The algorithm is modeled after the physician's interpretation of cirrhosis to speed up the physician's diagnosis. This will also serve as an objective reference.
Industrial Applicability The cirrhosis classification technique developed using deep learning is used to determine the degree of liver fibrosis. The results of this technique can be used as an objective medical indicator, allowing doctors to more efficiently assess the condition of their patients in order to determine the degree of fibrosis. The ability to make better diagnoses has not only financial benefits, but also health benefits for humanity.
本網站使用您的Cookie於優化網站。繼續瀏覽網站即表示您同意本公司隱私權政策,您可至隱私權政策了解詳細資訊。