# An efficient automated approach for accumulated dose estimation in prostate cancer radiotherapy

## TL;DR

This paper introduces a fast automated method to track radiation dose accumulation in prostate cancer patients during radiotherapy.

## Contribution

A novel automated workflow for efficient accumulated dose estimation in prostate radiotherapy using CBCT and AI-based auto-contouring.

## Key findings

- The workflow estimates accumulated doses in about one minute per fraction.
- Dose estimates agreed within ±2% of CBCT-based recalculations.
- Bladder and rectum dose metrics showed significant interfractional variation.

## Abstract

•Automated one-minute workflow for accumulated dose estimation per fraction.•Interfractional coefficient of variation: bladder 35 %, rectum 21 %, prostate 10 %•Accumulated dose estimates agreed within ± 2 % of cone-beam based recalculations.•Enabled efficient offline adaptive dose monitoring on standard hardware.

Automated one-minute workflow for accumulated dose estimation per fraction.

Interfractional coefficient of variation: bladder 35 %, rectum 21 %, prostate 10 %

Accumulated dose estimates agreed within ± 2 % of cone-beam based recalculations.

Enabled efficient offline adaptive dose monitoring on standard hardware.

Anatomical variations during prostate radiotherapy can cause discrepancies between planned and delivered doses, which may become more pronounced in (ultra-)hypofractionated regimens. However, a standardized method for accumulated dose assessment is lacking. This study presented and validated an efficient automated workflow for retrospective dose accumulation in moderately hypofractionated prostate treatments, avoiding deformable image registration and daily dose recalculation.

Twenty prostate cancer patients were analyzed retrospectively. The workflow combined daily cone-beam computed tomography (CBCT) imaging, online rigid registration, and artificial intelligence-based auto-contouring with planned dose data to estimate accumulated doses. Dose–volume metrics were analyzed for prostate, rectum, and bladder. Anatomical variations were assessed through volume and position analysis. Method accuracy was validated against CBCT-based dose recalculations in selected cases representing minimal and maximal variations.

The workflow required about one minute per fraction. Interfractional volume variability (coefficient of variation) was 35.2 %, 20.6 %, and 9.5 % for bladder, rectum, and prostate, respectively. Mean prostate displacement was 3.49 ± 2.07 mm, and the dice similarity coefficient averaged 0.83 ± 0.08. Bladder dose metrics (V20Gy, V50Gy, mean) differed by > 2 % in 45–49 % of sessions, rectal metrics in 35–39 %. Accumulated dose estimates agreed within ± 2 % of CBCT-based recalculations.

Our automated workflow provides a robust, time-efficient method for monitoring interfractional dose variations in prostate radiotherapy. While constrained by CBCT image quality, the accuracy and practicality of the proposed method support its integration into clinical workflows, potentially improving offline adaptive decision-making.