Machine Learning-based Identification of Prognostic Factors for Surgical Management in Patients With NOS Sarcoma.

Abstract

Background: Non-otherwise specified (NOS) sarcomas, a diverse and diagnostically challenging group of mesenchymal malignancies, pose significant clinical dilemmas due to their variable clinical trajectories and therapeutic responses. This study utilizes advanced machine learning techniques, namely classification and regression trees and Shapley additive explanation (SHAP) values, to identify predictors of survival, metastatic progression, and recurrence within a well-defined patient cohort, aiming to improve risk stratification and individualized care strategies.

Methods: Through the application of classification and regression trees and SHAP values to a cohort of 122 patients with NOS sarcoma, we identified critical factors impacting disease outcomes.

Results: The study findings revealed that age and tumor diameter significantly influenced the development of metastasis, whereas body mass index and tumor grading were key predictors for relapse. Additionally, tumor size, location, and age were identified as influential factors for overall survival in patients with NOS sarcoma. These results have direct clinical relevance and can aid in risk stratification and surgical planning in this challenging patient population.

Conclusions: Considering the comparatively small cohort with which the machine learning algorithm was trained, this study underscores the importance of considering age, tumor size, location, body mass index, and tumor grading in the management of NOS sarcomas, shedding light on factors that may impact clinical outcomes and guide personalized treatment strategies.