Purpose: To develop a machine-learning algorithm to improve hepatic resection selection for metastatic colorectal cancer patients by predicting post-PVE outcomes.
Material & methods: This multicenter retrospective study (2000-2020) included 200 consecutive patients with CRC liver-metastases planned for PVE before surgery. Radiomic features and lab values were collected. Patient-specific eigenvalues for each liver shape were calculated using a statistical shape model approach. After semi-automatic segmentation and review by a board-certified radiologist, the data was split 70/30% for training and testing. Three machine learning algorithms predicting the total liver volume (TLV) after PVE, sufficient FLR%, and the kinetic growth rate % (KGR%) were trained with performance assessed using accuracy, sensitivty, specifity, AUC or RMSE. Significance between the internal and external test sets was assessed by the student's t-test. One institution was kept separate as an external testing set.
Results: A total of 114 (76m; 56y ±12) and 37 (19m; 50y ±11) patients met the inclusion criteria for the internal and external validation. Prediction accuracy (SD) and AUC (SD) for sufficient FLR% or liver growth potential (KGR%> 0%) were high in the internal testing set 0.91 (±0.01), 85.81% (±1.01%) or 0.66 (±0.03), 87.44% (±0.10%)). Similar results occurred on the external testing set 0.88 (±0.00), 79.66% (±0.60%) or 0.69 (±0.01), 72.06% (±0.30%)). TLV prediction showed a discrepancy of 12.56% (95% CI: ±4.20%, p=0.86) internally and 13.57% (95% CI: ±3.76%, p=0.91) externally.
Conclusion: These machine learning-based models can help predict the FLR%, KGR%, and TLV as metrics for successful PVE.
Copyright © 2024. Published by Elsevier Inc.