Technical Name A Regional Boundary Verification Based Multiscale Fatty Liver Analysis System
Project Operator National Cheng kung University
Project Host 詹寶珠
Assessment of liver fat is required prior to the diagnosis of NAFLDliver transplantation. The severity of fatty live is determined by the size of fatty oil dropletstheir area ratio. we developed a deep learning-based method to detect fatty oil droplets in liver pathological slides to quicklyaccurately evaluate the area ratio of fatty oil dropletshelp pathologists to correctly determine the severity of fatty liver.
Scientific Breakthrough
Compared with direct use of MASK-RCNN to detect oil dropletsthe international method of detecting oil droplets, our method shows significant improvement in the detection of both large oil dropletsharder small oil droplets. Using the F1 score to measure the accuracy, our method reaches 9272 F1 score for largesmall oil droplets, respectively.
Industrial Applicability
The system automatically detects fatty oil droplets on the microscope using AI,we can quickly calculate the area of the oil droplets, which has reached the standard of practical use. This technology can help to establish objective criteria for the relationship between the ratio of oil droplet areathe severity of fatty liver, which can be used as a basis for the diagnosis of NAFLDfor more accurate assessment of liver quality in liver transplantation.
Keyword fatty liver pathological slide deep learning watershed algorithm
  • Contact
  • Pau-Choo Chung
other people also saw