S1043
Clinical – Sarcoma, skin cancer, malignant melanoma
ESTRO 2026
(Time_diag), lesion dimension and site, SBRT dose and fractionation, type of concomitant treatment with ipilimumab (IPI). A decision tree (DT) supervised machine learning models was trained and internally validated with 5-fold cross validation. Model performances were assessed using ROC, AUC and calibration curves. An explainable approach based on SHapley Additive exPlanations (SHAP) method was deployed to generate individual explanations of the model decisions. Results: 21.4% of lesions had a complete or partial response and were used as ground truth for the DT models. SHAP analysis strongly associated complete response with three variables, namely in order of importance the type of treatment with ipilimumab, the time from primary diagnosis to brain lesions discovery and the BRAF status. The mean AUCs for complete response in the 5-fold cross validation was 0.89 (0.87-0.93). In particular, the DT classified patients with concomitant ipilimumab treatment to have a probability of complete response of 75.0%. Any other temporal choices for ipilimumab decreased the complete response down to 17.4%. In this last group of patients, a Time_diag greater than 34 months increases the probability of complete response to 40.0%, otherwise it drops to 10.7%. Lastly, a wild-type BRAF status still decreases the CR probability to 4.4%. Conclusion: A machine-learning based DT model enables a reliable prediction of the treatment response of oligometastatic lesions receiving SBRT. This approach may assist radiation oncologists to tailor more individualized treatment plans in this clinical setting. An external validation of the results is mandatory. References: 1. Borzillo V, Di Franco R, Giannarelli D, Cammarota F, Scipilliti E, D'Ippolito E, Petito A, Serra M, Falivene S, Grimaldi AM, Simeone E, Festino L, Vanella V, Trojaniello C, Vitale MG, Madonna G, Ascierto PA, Muto P. Ipilimumab and Stereotactic Radiosurgery with CyberKnife® System in Melanoma Brain Metastases: A Retrospective Monoinstitutional Experience. Cancers (Basel). 2021 Apr 13;13(8):1857. doi: 10.3390/cancers13081857. PMID: 33924595; PMCID: PMC8068853. Keywords: Explainable A.I. model; SBRT; brain metastases
Conclusion: Modern mixed reality imaging is showing early promise and may improve skin radiotherapy planning set up and reproducibility. Keywords: Skin, planning, imaging
Digital Poster 3226
Explainable machine learning predictive model for treatment response after SBRT in melanoma brain metastases Savino cilla 1 , Donato Pezzulla 2 , Rossella Di Franco 3 , Carmela Romano 1 , Valentina Borzillo 3 , Gabriella Macchia 2 , Esmeralda Scipilliti 3 , Gianluca Ametrano 3 , Simona Mercogliano 3 , Rocco Mottareale 3 , Marcello Serra 3 , Francesco Deodato 2,4 , Vincenzo Ravo 3 1 Medical Physic Unit, Responsible Research Hospital, Campobasso, Italy. 2 Radiation Oncology Unit, Responsible Research Hospital, Campobasso, Italy. 3 Department of Radiation Oncology, Istituto Nazionale Tumori-IRCCS-Fondazione G. Pascale, Napoli, Italy. 4 Istituto di Radiologia, Università cattolica del Sacro Cuore, Roma, Italy Purpose/Objective: To develop an explainable artificial intelligence (XAI) model to predict the complete/partial response of oligometastatic brain lesions from melanoma after two months from radiosurgical or fractionated stereotactic treatments. Material/Methods: 60 consecutive patients with a total 117 MBMs treated with CyberKnife® SRS/SRT from December 2012 to December 2018 were selected within the Rabbit study [1]. Local response was evaluated on brain MRI performed during follow-up using response evaluation criteria in solid tumors (RECIST) and revised using response criteria for brain metastases from the response assessment in neuro-oncology criteria (RANO) group. For each treated lesion the following variables were considered: number of treated lesions, the LDH pre-treatment level, previous treatments (surgery or whole brain RT), BRAF status, time from primary diagnosis to brain lesions discovery
Made with FlippingBook - Share PDF online