# CoRe: An Automated Pipeline for The Prediction of Liver Resection Complexity from Preoperative CT Scans

**Authors:** Omar Ali, Alexandre Bone, Caterina Accardo, Omar Belkouchi, Marc-Michel Rohe, Eric Vibert, Irene Vignon-Clementel  
**arXiv:** [2210.08318](https://arxiv.org/abs/2210.08318)  
**Date:** 2022-10-15  
**Summary page:** https://www.summarizepaper.com/en/arxiv-id/2210.08318v1/

## Key points

- Liver cancer is a serious health concern and surgical resections are the most common curative treatment for primary liver cancer.
- Tumors located in critical positions can make liver resections complex and increase the risk of intra- and postoperative complications.
- CoRe is an automated medical image processing pipeline that predicts postoperative liver resection (LR) complexity from preoperative CT scans using imaging biomarkers.
- The CoRe pipeline segments the liver, lesions, and vessels with two deep learning networks and prunes the liver vasculature to define the hepatic central zone (HCZ).
- From HCZ, a new imaging biomarker called BHCZ is derived along with additional biomarkers used to train and evaluate an LR complexity prediction model.
- An ablation study shows that HCZ-based biomarkers are essential in predicting LR complexity with an accuracy of 77.3%, F1 score of 75.4%, and AUC of 84.1%.
- This approach could help improve standard care by reducing intra- and postoperative complications associated with complex LRs while also providing guidance for surgical planning in specialized medical centers where experienced surgeons may not be available or accessible.

## Summary

Liver cancer is a serious health concern and surgical resections are the most common curative treatment for primary liver cancer. However, tumors located in critical positions can make liver resections complex and increase the risk of intra- and postoperative complications. To address this need, a team of researchers led by Omar Ali has proposed CoRe - an automated medical image processing pipeline that predicts postoperative liver resection (LR) complexity from preoperative CT scans using imaging biomarkers. The CoRe pipeline first segments the liver, lesions, and vessels with two deep learning networks. Then, based on a topological criterion it prunes the liver vasculature to define the hepatic central zone (HCZ), which is a convex volume circumscribing major liver vessels. From HCZ, a new imaging biomarker called BHCZ is derived. Additional biomarkers are extracted and used to train and evaluate an LR complexity prediction model. An ablation study shows that HCZ-based biomarkers are essential in predicting LR complexity with an accuracy of 77.3%, F1 score of 75.4%, and AUC of 84.1%. This approach could help improve standard care by reducing intra- and postoperative complications associated with complex LRs while also providing guidance for surgical planning in specialized medical centers where experienced surgeons may not be available or accessible. Overall, this research presents a promising step forward towards developing automated tools that can assist clinicians in making informed decisions about patient care while also improving outcomes for patients undergoing complex surgical procedures like LRs for primary liver cancer.