Machine Learning Methods in Classification of Prolonged Radiation Therapy in Oropharyngeal Cancer: National Cancer...

Area of Science:

• Oncology
• Medical Informatics
• Data Science

Background:

• Oropharyngeal squamous cell carcinoma (OPSCC) treatment often involves radiation therapy (RT) or chemoradiation.
• Prolonged radiation treatment duration (RTD) can impact patient outcomes and resource allocation.
• Accurate risk stratification for prolonged RTD is crucial for personalized treatment planning.

Purpose of the Study:

• To evaluate the accuracy of various machine learning (ML) algorithms in predicting prolonged RTD (>50 days) in OPSCC patients.
• To compare the performance of ML models against traditional logistic regression for RTD risk stratification.
• To identify the most effective ML algorithm for identifying patients at risk of prolonged RTD.

Main Methods:

• Retrospective cohort study utilizing the National Cancer Database (NCDB) from 2004-2016.
• Analysis included 3152 OPSCC patients treated with primary RT or chemoradiation.
• Eight ML algorithms were trained (70%) and tested (30%) against logistic regression using performance metrics.

Main Results:

• Random Forest (RF) demonstrated superior accuracy in predicting prolonged RTD compared to other ML algorithms and logistic regression.
• The study included 1928 patients with prolonged RTD and 1224 without.
• Performance metrics confirmed RF as the most effective predictive model.

Conclusions:

• Random Forest (RF) significantly outperforms traditional logistic regression in classifying OPSCC patients at risk of prolonged RTD.
• ML algorithms, particularly RF, show potential for identifying high-risk patients.
• Early identification of high-risk patients can facilitate timely interventions, potentially improving survival rates.