Interpretable Colorectal Cancer Recurrence Risk Prediction Using a Stacked Cox Survival Framework With Counterfactual Intervention Analysis
IEEE Access · 2026 · 14 · 107603-107626
DOI: 10.1109/access.2026.3706377 ↗Kurz gefasst
Fachartikel in der Zeitschrift IEEE Access, 2026. Verfasst von Rhithu Sri Shankar und 2 weiteren Autor:innen. Die Arbeit nennt 63 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Abstract (Originalsprache)
Colorectal cancer remains one of the leading causes of cancer-related mortality worldwide, with recurrence strongly influencing long-term patient prognosis. Existing machine learning approaches for colorectal cancer prediction predominantly formulate recurrence assessment as a static binary classification task, limiting temporal risk modeling and clinically actionable interpretability. To address these limitations, this study proposes an interpretable colorectal cancer recurrence risk prediction framework based on a stacked ensemble of domain-specific Cox Proportional Hazards survival models integrated with risk stratification and counterfactual intervention analysis. The framework decomposes patient variables into clinical, treatment-related, and lifestyle domains, each modeled using an independent survival estimator and combined through a meta-level Cox model to produce interpretable domain-level risk contributions. Evaluated on a large-scale colorectal cancer dataset containing 89,945 patient records, the proposed framework achieved a concordance index of 0.82, outperforming a standalone Cox model (C-index = 0.76) and conventional classification baselines including Random Forest (AUC = 0.76) and Gradient Boosting (AUC = 0.75). Kaplan–Meier analysis demonstrated statistically significant separation between low-, medium-, and high-risk groups ( $p $ < 0.001). Counterfactual intervention analysis further showed that surgical treatment and combined lifestyle modifications substantially reduced predicted recurrence hazard across high-risk patients. An interactive clinical dashboard was additionally developed to support real-time risk prediction and personalized intervention planning. The proposed framework provides a clinically interpretable and actionable approach for survival-based colorectal cancer recurrence assessment and decision support.
Zitieren
APA
Shankar, R. S., Shankar, R. S., & Panicker, R. O. (2026). Interpretable Colorectal Cancer Recurrence Risk Prediction Using a Stacked Cox Survival Framework With Counterfactual Intervention Analysis. IEEE Access, 14, 107603-107626. https://doi.org/10.1109/access.2026.3706377
Vancouver
Shankar RS, Shankar RS, Panicker RO. Interpretable Colorectal Cancer Recurrence Risk Prediction Using a Stacked Cox Survival Framework With Counterfactual Intervention Analysis. IEEE Access. 2026;14:107603-107626. doi:10.1109/access.2026.3706377
BibTeX
@article{Shankar2026,
author = {Shankar, Rhithu Sri and Shankar, Ruppikha Sree and Panicker, Rani Oomman},
title = {Interpretable Colorectal Cancer Recurrence Risk Prediction Using a Stacked Cox Survival Framework With Counterfactual Intervention Analysis},
journal = {IEEE Access},
year = {2026},
volume = {14},
pages = {107603--107626},
doi = {10.1109/access.2026.3706377}
}
EUVIMED