Machine learning for dose-volume histogram based clinical decision-making support system in radiation therapy plans for brain tumors

Pawel Siciarz; Salem Alfaifi; Eric Van Uytven; Shrinivas Rathod; Rashmi Koul; Boyd McCurdy

doi:10.1016/j.ctro.2021.09.001

Machine learning for dose-volume histogram based clinical decision-making support system in radiation therapy plans for brain tumors

Clin Transl Radiat Oncol. 2021 Sep 15:31:50-57. doi: 10.1016/j.ctro.2021.09.001. eCollection 2021 Nov.

Authors

Pawel Siciarz^{1

2}, Salem Alfaifi³, Eric Van Uytven³, Shrinivas Rathod^{3

4}, Rashmi Koul^{4

5}, Boyd McCurdy^{4

6

2}

Affiliations

¹ Department of Medical Physics, CancerCare Manitoba, 675 McDermot Avenue, Winnipeg, MB R3E 0V9, Canada.
² Department of Physics and Astronomy, University of Manitoba, Allen Building, Winnipeg, MB R3T 2N2, Canada.
³ Radiation Oncology Resident, Department of Radiation Oncology, CancerCare Manitoba, 675 McDermot Avenue, Winnipeg, MB R3E 0V9, Canada.
⁴ Department of Radiology, University of Manitoba, GA216-820 Sherbrook Street, Winnipeg, MB R3T 2N2, Canada.
⁵ Medical Director and Head, Radiation Oncology Program, Department of Radiation Oncology, CancerCare Manitoba, 675 McDermot Avenue, Winnipeg, MB R3E 0V9, Canada.
⁶ Head of Radiation Oncology Physics Group, Department of Medical Physics, CancerCare Manitoba, 675 McDermot Avenue, Winnipeg, MB R3E 0V9, Canada.

Abstract

Purpose: To create and investigate a novel, clinical decision-support system using machine learning (ML).

Methods and materials: The ML model was developed based on 79 radiotherapy plans of brain tumor patients that were prescribed a total dose of 60 Gy delivered with volumetric-modulated arc therapy (VMAT). Structures considered for analysis included planning target volume (PTV), brainstem, cochleae, and optic chiasm. The model aimed to classify the target variable that included class-0 corresponding to plans for which the PTV treatment planning objective was met and class-1 that was associated with plans for which the PTV objective was not met due to the priority trade-off to meet one or more organs-at-risk constraints. Several models were evaluated using double-nested cross-validation and an area-under-the-curve (AUC) metric, with the highest performing one selected for further investigation. The model predictions were explained with Shapely additive explanation (SHAP) interaction values.

Results: The highest-performing model was Logistic Regression achieving an accuracy of 93.8 ± 4.1% and AUC of 0.98 ± 0.02 on the testing data. The SHAP analysis indicated that the ΔD_99% metric for PTV had the greatest influence on the model predictions. The least important feature was ΔD_MAX for the left and right cochleae.

Conclusions: The trained model achieved satisfactory accuracy and can be used by medical physicists in a data-driven quality assurance program as well as by radiation oncologists to support their decision-making process in terms of treatment plan approval and potential plan modifications. Model explanation analysis showed that the model relies on clinically valid logic when making predictions.