Predicting radiation-acute esophagitis via machine learning algorithms

Acute esophagitis is a common adverse effect of radiochemotherapy in patients with cancer, and its early detection and prediction of its occurrence can significantly impact patient management and treatment planning. This study aimed to conduct a comparative analysis of various Machine Learning algorithms and explore their practical implementations in predicting acute esophagitis induced by radiochemotherapy.

A comprehensive dataset comprising patient characteristics, treatment parameters, and clinical factors associated with acute esophagitis was collected and preprocessed for analysis. The machine learning classification algorithms were trained and tested using various cross-validation techniques to ensure the robustness and generalizability of the predictive models. Furthermore, practical implementations of the machine learning algorithms were explored, considering real-world scenarios and potential integration into clinical practice.

Various experiments were conducted on the mentioned dataset using different validation and evaluation strategies, in order to perform a comparative analysis of the impact of features on the performance of the classification algorithms. The results showed that dose-volume-related characteristics were more important factors in predicting the occurrence of complications than other information. The machine learning algorithms demonstrated the ability to distinguish and predict various grades of the samples, with more than 90 % F1-scores and accuracies. Moreover, they successfully distinguished high-grade samples from low-grade ones with an AUC exceeding 95 %. These methods demonstrate strong performance in accurately distinguishing between classes.

This study underscores the potential of machine learning to improve patient outcomes and the critical role of Dose-Volume characteristics in predicting acute esophagitis.