Fast D _M,M calculation in LDR brachytherapy using deep learning methods

Abstract

\textbf{Objective:} The Monte Carlo (MC) method provides a complete solution to the tissue heterogeneity effects in low-energy low-dose rate (LDR) brachytherapy. However, long computation times limit the clinical implementation of MC-based treatment planning solutions. This work aims to apply deep learning (DL) methods, specifically a model trained with MC simulations, to predict accurate dose to medium in medium (D\textsubscript{M,M}) distributions in LDR prostate brachytherapy.\\

\textbf{Approach:} To train the DL model, 2369 single-seed configurations, corresponding to 44 prostate patient plans, were used. These patients underwent LDR brachytherapy treatments in which $^{125}$I SelectSeed sources were implanted. For each seed configuration, the patient geometry, the MC dose volume and the single-seed plan volume were used to train a 3D Unet convolutional neural network. Previous knowledge was included in the network as an r\textsuperscript{2} kernel related to the first-order dose dependency in brachytherapy. MC and DL dose distributions were compared through the dose maps, isodose lines, and dose-volume histograms. Features enclosed in the model were visualized.\\

\textbf{Main results:} Model features started from the symmetrical kernel and finalized with an anisotropic representation that considered the patient organs and their interfaces, the source position, and the low- and high-dose regions. For a full prostate patient, small differences were seen below the 20\% isodose line. When comparing DL-based and MC-based calculations, the predicted CTV D\textsubscript{90} metric had an average difference of -0.1\%. Average differences for OARs were -1.3\%, 0.07\%, and 4.9\% for the rectum D\textsubscript{2cc}, the bladder D\textsubscript{2cc}, and the urethra D\textsubscript{0.1cc}. The model took 1.8 ms to predict a complete 3D D\textsubscript{M,M} volume (1.18 M voxels).\\

\textbf{Significance:} The proposed DL model stands for a simple and fast engine which includes prior physics knowledge of the problem. Such an engine considers the anisotropy of a brachytherapy source and the patient tissue composition.

Aleatoric and epistemic uncertainty extraction of patient-specific deep learning-based dose predictions in LDR prostate brachytherapy

“…The toolkit provides a brachytherapy package with several source geometries and a track-length estimator (TLE) for dose scoring (Williamson 1987). The simulation setup was done following Berumen et al (2023). Patient simulations were performed with the 125 I selectSeed model geometry.…”

“…As recommended by the TG-372 report, the local and global dose differences (equation ( 5)) are used to evaluate the output between advanced dose calculation algorithms in brachytherapy (Beaulieu et al 2023). For LDR prostate brachytherapy patients, Berumen et al reported that 90.1% of voxels had global dose differences in the [−1%, 1%] interval when comparing a single-seed dose predictor (based on a 3D UNet model) and the MC reference (Berumen et al 2023). These results are comparable to those presented in this work for the streamline UNet model.…”