Identifying the optimal time point for adaptive re-planning in prostate cancer radiotherapy to minimise rectal toxicity using normal tissue imaging biomarkers

TL;DR

This study finds that radiomic features from early treatment scans can help decide the best time to adjust prostate cancer radiotherapy to reduce rectal bleeding.

Contribution

The study identifies week 3 as the optimal time for re-planning in standard fractionation to minimize rectal toxicity using radiomic biomarkers.

Findings

Radiomic features from week 1 showed the strongest standalone predictive performance for rectal bleeding.

Week 3 was identified as the optimal time point for re-planning in patients receiving 74 Gy in 37 fractions.

Radiomic analysis supports biologically informed adaptive radiotherapy beyond anatomy-based methods.

Abstract

•Radiomic features before and during treatment predict late rectal bleeding.•Radiomic features from week 1 showed strongest standalone predictive performance.•Week 3 was optimal for re-planning in patients treated with 74 Gy in 37 fractions.•Radiomics enable biologically informed adaptation beyond anatomy-based methods.•Analysis includes both standard and moderately hypofractionated treatment regimens.

Radiomic features before and during treatment predict late rectal bleeding.

Radiomic features from week 1 showed strongest standalone predictive performance.

Week 3 was optimal for re-planning in patients treated with 74 Gy in 37 fractions.

Radiomics enable biologically informed adaptation beyond anatomy-based methods.

Analysis includes both standard and moderately hypofractionated treatment regimens.

Adaptive radiotherapy (ART) in prostate cancer (PCa), although not yet standard practice, is typically triggered by inter-fractional anatomical changes that emerge progressively during treatment. This study investigates whether radiomics extracted before and during treatment can identify the optimal time point for re-planning, with the goal of reducing late rectal bleeding.

This study included 187 PCa patients from the single-centre, prospectively collected VoxTox dataset (UK-CRN-ID-13716), treated with image-guided radiotherapy using TomoTherapy and daily MVCT. Patients received either 74 Gy in 37 fractions (N = 110) or 60 Gy in 20 fractions (N = 77). Radiomic features were extracted from pre-treatment planning CTs and daily MVCTs. Grade ≥ 1 rectal bleeding was assessed at 2 years post-treatment using CTCAE v4.03. Two analysis strategies were employed: a separate analysis, where weekly features were evaluated independently; and a cumulative analysis, which progressively incorporated features from previous weeks. Logistic regression models with elastic net penalty were trained and evaluated using AUC.

In both groups, week 1 provided the highest standalone predictive performance (test AUC = 0.766 for 74 Gy; 0.734 for 60 Gy). In the cumulative analysis, week 3 was optimal for the 74 Gy group (test AUC = 0.767), balancing performance and timing. For the 60 Gy group, week 1 remained optimal but suffered from reduced generalisability (test AUC = 0.643).

Radiomic analysis of daily imaging can support early, proactive ART in PCa, offering a personalised strategy to reduce late rectal bleeding beyond conventional anatomy-based approaches.

Figures