Machine learning to develop and validate a model for predicting the risk of lymph node metastasis in colorectal cancer patients.

BACKGROUND

The presence of lymph node metastasis (LNM) serves as a critical determinant of prognosis in colorectal cancer (CRC), often correlating with markedly unfavorable clinical outcomes. Consequently, this investigation sought to construct and rigorously validate a machine learning (ML) algorithm capable of estimating LNM probability within the CRC population.

METHODS

Through the application of logistic regression analysis, six independent predictors were identified: extramural vascular invasion (EMVI), T stage, fibrinogen (FIB), systolic blood pressure (SBP), thrombin time (TT), and alpha-fucosidase (AFU). Subsequently, diverse ML architectures were generated and assessed utilizing receiver operating characteristic (ROC) curves, calibration plots, and decision curve analysis (DCA).

RESULTS

Among the evaluated algorithms, the stochastic gradient boosting (gbm) model exhibited superior discriminatory power, yielding area under the curve (AUC) metrics of 0.815 in the training cohort and 0.733 in the external validation set. Calibration assessments revealed a robust concordance between projected probabilities and actual observed events. Notably, the xgbTree algorithm also demonstrated commendable predictive efficacy.

CONCLUSION

We successfully established and verified an ML-based framework designed to forecast LNM risk in individuals with CRC. These computational tools offer clinicians a refined mechanism for the precise identification of nodal involvement, thereby laying the groundwork for formulating personalized therapeutic regimens.