ESTRO 2020 Abstract book

S156 ESTRO 2020

the prostate was 79.3 Gy (range: 76–80) at 2 Gy per fraction, with 46 Gy delivered to the seminal vesicles. The cohort was split into a training set from 2 centers (n=337, 27 events) and a validation set (3rd center, n=254, 22 events). The classification task was prediction of RB at 3 years after RT. An ML framework was developed consisting of 3 modules to efficiently process multicentric data: 1. covariate shift and imbalance adaptation module relying on SMOTE(EN) [Synthetic Minority Over-sampling Technique -Edited Nearest Neighbours], density estimation ratio and normalization, 2. classification module (implementing 4 ML algorithms: Random Forest, Xtreme Gradient Boosting, LightGBM, CatBoost and 3 DL classifiers: Deep Neural Network, Deep Autoencoder+RF, Deep Variational Autoencoder+RF, as well as majority voting) and 3. Pseudo-labeling module (figure 1). The prediction capability of the proposed method was compared to the prediction capability using “standard” logistic regression using area under the ROC curve (AUC).

Conclusion ML/DL techniques including covariate shift as well as imbalance adaptation can achieve higher predictive ability of RB after RT in PC in a multicentric context, compared to standard modeling approaches (logistic regression). PH-0284 Dose-effect relation for rib fractures after lung SBRT: Planned vs. delivered dose C. Juan-Cruz 1 , B. Stam 1 , J. Belderbos 1 , J. Sonke 1 1 Netherlands Cancer Institute, Radiotherapy, Amsterdam, The Netherlands Purpose or Objective Radiation induced rib fracture (RIRF) is a known toxicity for inoperable early stage non-small cell lung cancer (NSCLC) patients treated with Stereotactic Body Radiotherapy (SBRT). Normal tissue complication probability (NTCP) models have been developed to describe the relation between the planned dose and the risk of RIRF. Anatomical changes during treatment can cause dosimetric differences between the planned and delivered dose, so that the estimated NTCP models may no longer be valid. This study aims to determine if the use of the delivered dose in NTCP models is a better predictor of 353 NSCLC patients treated to a median of 3x18 Gy were included in this study. Cone beam CTs (CBCTs) were acquired prior to each fraction for patient positioning. RIRFs were diagnosed on follow-up (FU) CT scans acquired at 4 months post SBRT, every 6 months for 2 years and annually until year 5. Ribs outside the CBCT field-of-view were not considered (5338 ribs, 63%). Each fraction’s dose distribution was approximated by shifting the planned dose to the daily tumor location followed by conversion to biologically equivalent dose (EQD2) with α/β=3 Gy. The total delivered dose was estimated by deforming each corrected fraction dose onto the planning anatomy (using CBCT-to-planning CT deformable image registration) and accumulating all fractions. Ribs were automatically segmented using atlas based segmentation. A dose volume histogram was computed per rib, and dosimetric parameters D x (dose to volume x; range 0-50%) extracted. NTCP models for both planned and delivered dose were built by maximizing the likelihood to correctly classify fracture, optimizing TD 50 (dose with 50% toxicity risk) and m (steepness of the curve) for all D x . The best dosimetric RIRF predictor for each dose distribution RIRF than planned dose. Material and Methods

Results In the testing set, the best AUC among all the tested methods of 0.68 (sensitivity 0.77, specificity 0.60) was obtained with random forest relying on the proposed modules, combining DVH and clinical variables and successfully predicting RB at 3 years in 17 patients out of 22 with RB (with however 95 false positives). The AUC of the other classifiers ranged from 0.54 to 0.66 (table 1 and table 2). Majority voting did not improve the results. Without the proposed modules(Table 1 and Table 2), all classifiers reached lower AUC (0.42-0.60). Standard Logistic regression performed poorly (AUC 0.52, sensitivity 0.54, specificity 0.53).

Made with FlippingBook - Online magazine maker