From Prediction to Practice: A Machine Learning–Based Clinical Decision Support Tool for Bevacizumab Risk Stratification in Oncology

Abstract

AI in cancer might enhance decision-making, especially for focused therapy risk management. Its goal was to create and verify a machine learning-based clinical decision support system (CDSS) that could anticipate Bevacizumab or biosimilar problems and turn the prediction model into a clinical tool. A prospective observational research examined 395 solid tumor patients treated with Bevacizumab or biosimilars. Medical records included demographics, medical history, tumor features, and laboratory results ahead of treatment. XGBoost, Random Forest, and logistic regression were trained with data splits of 70/30 and 80/20. These prediction models were compared by accuracy, AUC-ROC, sensitivity, specificity, F1-scores, and error rate. Using the best model, a logistic-based risk score was calculated and put into an interactive HTML form. The improved Random Forest model that was trained using the 80/20 split has the highest accuracy (70.63%), sensitivity (66.67%), specificity (73.85%), and AUC-ROC (0.55). Both the calibration and AUC-ROC value of 0.720 for the logistic risk score were excellent. It found that age > 65, anemia, increased urea, leukocytosis, tumor differentiation, and stage are significant predictors of problems. The final product provides physicians with a straightforward offline form for risk assessment and patient risk stratification. This research shows that real-world data may be used to construct AI-supported cancer forecasting tools. Without supplanting clinical judgment, the interactive form and logistic risk score may assist physicians in customizing targeted therapy selections.

Ähnliche Arbeiten