Liver and liver tumor segmentation and analysis algorithms

the picture displayed a results of liver tumor analysis and segmentation by algorithm

We are able to develop custom automated and semi-automated algorithms to liver and liver tumor analysis, according to your need. In the example below we present our models for liver and liver tumors segmentation (including primary tumors, secondary tumors and metastases) based on portal venous CT scans.

Our liver segmentation model was trained on a dataset of more than 500 patients. Its results were used to extract the region of interest for tumors segmentation which was trained on more than 400 patients. Our tumor segmentation model is in the current top 10 best solutions in terms of average Dice (0.77) on LiTS2017 Challenge Open Leaderboard. The liver segmentation model provides very accurate results with an average Dice of 0.954 on LiTS2017 test data.

In the example below we vizualize how our models work for a patient from LiTS2017 test set (not used for training). Our algorithms automatically perform volume and RECIST measurements of the tumors with an option to extend it to mRECIST if proper training data are available. It can be a useful tool for monitoring patients in clinical studies.

Visualization description

  • The liver segmentation is marked in blue
  • The tumor segmentation is marked in red.
  • The RECIST measurements are marked in white.
  • The models identified 4 measurable tumors with longest diameters of 26 mm, 18 mm, 16 mm, and 10 mm, and with corresponding volumes of 9 ml, 2.3 ml, 0.7 ml, 0.6 ml. For clarity, only two biggest lesions are marked in the visualizations.
image to content of liver tumor analysis: liver CT image
Liver CT image
Liver CT image with segmented and measured lesions, RECIST
Liver CT scan from project for the pharmaceutical industry with automatically segmented liver and HCC tumors