Machine Learning for Predicting Clinician Evaluation of Treatment Plans for Left-Sided Whole Breast Radiation Therapy

Fiandra, C.; Cattani, F.; Leonardi, M.C.; Comi, S.; Zara, S.; Rossi, L.; Jereczek-Fossa, B.A.; Fariselli, P.; Ricardi, U.; Heijmen, B.

doi:10.1016/j.adro.2023.101228

Cited by 1 publication

(1 citation statement)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Following this initial analysis, we delved deeper into the complex and sometimes nonlinear relationships within radiation therapy planning. To address the complexity of the data, we employed the Random Forest model [25]. This machine learning model, consisting of an ensemble of decision trees, is effective in reducing variability and preventing over tting.…”

Section: Data Evaluation: Statistics and Machine Learningmentioning

confidence: 99%

Analyzing Collimator Rotation Angle Influence on Half-beam VMAT Outcomes for Prostate Cancer: A Comparative Approach Using Statistical and Machine Learning Methods

Kim,

Choi

et al. 2024

Preprint

View full text Add to dashboard Cite

Purpose This study explores the impact of Collimator Rotation Angle (CRA) settings in Half beam Volume Modulated Arc Therapy (HVMAT) for prostate cancer treatment, focusing on dose distribution and treatment efficacy. Materials and Methods Treatment plans (Total 240) for 20 prostate cancer patients were developed using HVMAT. Different CRA settings (n = 12) were employed, specifically comparing 2-arcs and 4-arcs techniques. Data were analyzed using statistical methods and machine learning models, assessing the Mean Relative Error (MRE) across varying CRA settings. Results The analysis revealed no significant impact of CRA settings on the conformity and homogeneity of radiation distribution to the target volume. All treatment plans met the average V95% target for the prescribed dose in the Planning Target Volume (PTV). Machine learning analysis showed consistent predictive accuracy across different CRA settings, with the MRE variance within 2%. Statistical tests further supported these findings, showing no significant differences in treatment plan outcomes based on CRA variations. Conclusion The study demonstrates that CRA settings in HVMAT can be selected with considerable flexibility without compromising the effectiveness of prostate cancer treatment. The results emphasize the importance of employing multi-faceted analysis, including both traditional statistical methods and advanced machine learning techniques, in optimizing HVMAT treatment plans. Although limited by a small sample size and a specific focus on prostate cancer, the findings provide valuable insights into the clinical application of HVMAT and its potential in treatment plan optimization.

show abstract

Section: Data Evaluation: Statistics and Machine Learningmentioning

confidence: 99%