Pegah Saadatmand1
1Medical Physics Department, Iran University of Medical Sciences, Iran
Abstract
Introduction: Radiation-induced acute skin toxicity (AST), is considered a common side effect of breast radiation therapy. The goal of this study was to design dosiomics based machine learning models for prediction of AST, enable creating optimized treatment plans for high-risk individuals.
Methods: 52 patients with breast cancer who underwent radiation therapy with Tomotherapy technique were prospectively included in this study. The superficial layer of the body with a thickness of 2 mm (SL2) was contoured as the equivalent structure of the skin in the CT images. Dosimics features along with dose volume histograms (DVHs) of the SL2 segment were extracted from the Treatment Planning System (TPS). In addition, patient- and treatment-related clinical characteristics (PTR) were collected. Before extracting the dose distribution feature and creating a reliable model, the accuracy of the TPS algorithm in calculating the surface dose distribution was evaluated by comparing the TPS results with the film dosimetry results. Clinical scoring was done using the Common Terminology Criteria for Adverse Events (CTCAE) V4.0 criteria for skin-specific symptoms. Patients were grouped into AST 2+ (CTCAE ≥ 2) and AST 2- (CTCAE < 2) toxicity grades to facilitate AST modeling. They were randomly divided into training (70%) and testing (30%) cohorts. 7 prediction models were created with the characteristics of dosiomics, DVH, and PTR separately, and the combination of dosomics with DVH, the combination of dosomics with PTR, the combination of DVH with PTR, and finally the combination of all three groups of dosomics, DVH and PTR together. After selecting features related to skin complications, each model was created using seven different classification algorithms. The performance of each model was evaluated on the test group using the area under the receiver operating characteristic curve (AUC). The accuracy, precision, and recall of each model were also studied.
Results: Our findings indicate a small difference (3-5%) between measured and calculated skin doses using the EBT3 film and TPS, employing “high” spatial resolution dose calculation in helical and direct Tomotherapy plans. Results showed that 44% of the patients developed AST 2+ after Tomotherapy. The dosiomics model, developed using dosiomics features, exhibited a noteworthy improvement in AUC (up to 0.78), when spatial information is preserved in the dose distribution, compared to DVH features (up to 0.71). Furthermore, a baseline machine learning model created using only PTR features for comparison with DOS models showed the significance of dosiomics in early AST prediction. By employing the Extra Tree (ET) classifiers, the DOS+DVH+PTR model achieved a statistically significant improved performance in terms of AUC (0.83; 95% CI, 0.71-0.90), accuracy (0.70), precision (0.74) and sensitivity (0.72) compared to other models.
Conclusions: This study confirmed the benefit of dosiomics-based ML in the prediction of AST. However, the combination of dosiomics, DVH, and PTR yields significant improvement in AST prediction. The results of this study provide the opportunity for timely interventions to prevent the occurrence of radiation-induced AST.