Synapse
⌘+K
Synapse
PulseExploreClubsResearchersJournals
Instagram
HomeClubsExplore
April 5, 2026Cancer Research

Multi-Modal Machine Learning Model Predicts Survival After Hepatectomy in Oligometastatic...

View Full Paper
Ask AI
Bookmark
Share

Key result

Multi-modal machine learning predicts three-year post-hepatectomy survival in colorectal cancer with 0.75 AUROC.

  • AUROC 0.75
  • n=284

Why the study?

There is an unmet need for pre-operative predictive biomarkers to identify which patients with oligometastatic colorectal cancer achieve long-term survival versus early relapse after curative-intent hepatectomy.

Does a multi-modal machine learning model predict overall survival greater than three years after hepatectomy in patients with oligometastatic colorectal cancer?

Population

284 CRC patients with liver-confined oligometastatic disease undergoing partial hepatectomy at Memorial Sloan Kettering Cancer Center

Comparison

Multi-modal machine learning model using clinical, genomic, histopathologic, and lab data vs established Clinical Risk Score

Design

Retrospective cohort study developing and validating an XGBoost machine learning model

Authors

DKDivya KoyyalaguntaMemorial Sloan Kettering Cancer CenterSGStefanie GerstbergerMemorial Sloan Kettering Cancer CenterMLMarion LiuMemorial Sloan Kettering Cancer Center

Discussion

Loading...

Member takes

Implication

May refine risk stratification after hepatectomy for oligometastatic CRC; extends clinical risk scores via multi-modal ML but remains hypothesis-generating pending validation.

Key Points

  • To develop a machine learning model that predicts post-hepatectomy outcomes in patients with oligometastatic colorectal cancer using various data types.
  • Analyzed data from 284 patients with liver-confined oligometastatic colorectal cancer who underwent hepatectomy
  • Collected clinical features, genomic data via targeted sequencing, histopathology features, and lab values pre-hepatectomy
  • Trained an XGBoost machine learning model with nested cross-validation to predict survival greater than three years
  • Evaluated model performance using AUROC and compared with established Clinical Risk Score (CRS).
  • The multi-modal model achieved an AUROC of 0.75, indicating good predictive performance
  • Performance was comparable using only genomics and lab data, achieving an AUROC of 0.73
  • Combining ML predictions with the Clinical Risk Score modestly improved prognostic discrimination
  • Elevated systemic inflammation index correlated with worse prognosis
  • Significant association found between deep deletion in DUSP4 and improved long-term survival.

Study Design

Type

Cohort (n=284)

Multicenter

No

Structured PICO

Does a multi-modal machine learning model predict overall survival greater than three years after hepatectomy in patients with oligometastatic colorectal cancer?

P
Population
284 colorectal cancer patients with liver-confined oligometastatic disease who underwent metastatic resection via partial hepatectomy at Memorial Sloan Kettering Cancer Center
I
Intervention
Multi-modal machine learning model using pre-operative genomic, histopathologic, laboratory, and clinical data
C
Comparator
Established Clinical Risk Score (CRS)
O
Outcome
Overall survival greater than three years after hepatectomyhard clinical

Main Result

Effect estimate: AUROC 0.75

A multi-modal machine learning model integrating genomic, histopathologic, and lab data can predict long-term survival after hepatectomy in oligometastatic colorectal cancer, potentially improving risk stratification.

Cite This Study

Koyyalagunta et al. (2026) conducted a cohort in Oligometastatic colorectal cancer (n=284). Multi-modal machine learning model vs. Clinical Risk Score (CRS) was evaluated on Overall survival greater than three years after hepatectomy (AUROC 0.75). A multi-modal machine learning model using genomic, histopathologic, and lab data predicted overall survival >3 years after hepatectomy in oligometastatic colorectal cancer (AUROC 0.75).

synapsesocial.com/papers/69d1fdf7a79560c99a0a4627https://doi.org/10.1158/1538-7445.am2026-4217
View Full Paper
Ask AI
Bookmark
Share

Also Consider

Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context:

  1. 1Integrative Genomic and Clinical Profiling of Colorectal Cancer Liver Metastases to Guide Personalized Surgery and Liver Transplantation2026
  2. 2Improving preoperative risk stratification in colorectal liver metastases: a multi-institutional evaluation of multimodal prediction models2026
  3. 3Predicting intrahepatic recurrence of colorectal cancer liver metastases after curative hepatectomy using a machine learning model with data integration of ultrasound radiomics and clinicopathological parameters2026
  4. 4Interpretable machine learning models for predicting the risk of metachronous colorectal liver metastases2026
  5. 5Development of multi-algorithm machine learning models integrating novel serum biomarkers for survival prediction in colorectal cancer: a retrospective cohort study2026