ESTRO 2025 - Abstract Book

S391

Brachytherapy - Urology

ESTRO 2025

1 Radiotherapy Department, Hospital Clínica Benidorm, Benidorm, Spain. 2 Radiotherapy Department, Hospital Vithas Virgen del Consuelo, Valencia, Spain. 3 Radiotherapy Department, Hospital Universitario y Politécnico La Fe, Valencia, Spain Purpose/Objective: Predicting clinical outcomes after prostate brachytherapy (BT) is challenging. Parameters such as a nadir PSA (nPSA) value <0.2 ng/ml are used. Finding other indicators to improve predictions is inherently complex. Artificial intelligence (AI) tools can help enhance these predictions by using multiple indicators throughout the entire BT process in a quick manner. In this work we have included an AI analysis, in order to assess its capacity to identify prognostic factors. Material/Methods: One-hundred-thirty-nine patients were treated with prostate BT (2005-2019) with exclusive Low dose rate BT (LDRBT) (46%. 160 Gy) or High dose rate BT (HDRBT) (54%. 2 implants of 13.5 Gy each separated 10 days). 69% low risk (LR) and 31% favourable intermediate risk (FIR). nPSA was grouped into two categories: ≤ 0.2 ng/mL and > 0.2 ng/mL. Biochemical failure (BF) was determined using the Phoenix definition. From 108 data-points per patient collected, a set of 32 potentially predictive variables for success/failure of biochemical control were selected, and an AI (Artificial Neural Network “ARN”) was trained with 32 input neurons, 10 neurons in the hidden layer, and one binary output neuron. The patients were randomly divided into 3 groups: training, validation, and test, to apply the cross-validation method. This training was carried out 10 times with different regroupings to improve network-characterization. Results: Median age 69 years (46-84). Median follow-up was 91 months (5-230). The actuarial biochemical failure-free survival (BFFS), local control (LC), overall survival (OS) and cause-specific survival (CSS) were 81%, 89, 68%, 98% at 10 years, respectively. Median nPSA 0,10 ± 0,39 ng/ml. Median time to achieve nPSA 30 months (3-144). One hundred and nine patients (78%) reached a nPSA ≤ 0,2 ng/ml after 5 years of follow-up. Statistical significance between nadir PSA categories were reached in BFFS ( p<0,001 ), LC ( p<0,001 ) and CSS ( p=0,007 ) (Table1). In the univariate analysis there are not differences between LR vs FIR or LDRBT vs HDRBT. The sensitivity of this AI improves that obtained by using exclusively the nPSA (≤ 0.2 ng/mL vs > 0.2 ng/mL) as a predictor (0.712) with p-value <0.05 (Table2). Conclusion: Prostate brachytherapy is an effective treatment. The nPSA ≤ 0,2 ng/ml is a representative value that provide prognostic information. Our data are promising, with a correlation of results between classical statistical analysis and those obtained with AI. They highlight the need for an expanded sample size given that these methods are increasingly being established in clinical practice.