# Early response prediction during radiotherapy in rectal cancer using sequential diffusion-weighted imaging at a magnetic resonance image-guided linear accelerator

## TL;DR

This study shows that early prediction of rectal cancer treatment response during radiotherapy is possible using diffusion-weighted MRI scans.

## Contribution

The novel use of sequential diffusion-weighted imaging during MR-guided radiotherapy to predict early treatment response in rectal cancer.

## Key findings

- Combining baseline and sequential ADC data improved response differentiation with AUCs up to 0.90 after five fractions.
- Significant ADC differences were observed as early as the first week of treatment.
- Volume-based stratification was only feasible after 15 out of 25 fractions.

## Abstract

•Diffusion images from MR-guided radiotherapy monitor rectal cancer treatment response.•Response to radiotherapy was predicted as early as first treatment week.•Combining baseline and sequential information enhance the response differentiation.

Diffusion images from MR-guided radiotherapy monitor rectal cancer treatment response.

Response to radiotherapy was predicted as early as first treatment week.

Combining baseline and sequential information enhance the response differentiation.

Varying response to chemoradiotherapy (CRT) challenges the treatment of locally advanced rectal cancer (LARC). Our purpose was to timely stratify responders by investigating the predictive potential of sequential, in-treatment diffusion-weighted (DW) magnetic resonance (MR) images.

DW images were acquired with a 1.5 T MR-Linac at baseline and during MR-guided CRT in 15 patients with LARC prescribed long-course CRT, prospectively enrolled between 2018 and 2021. Tumour response was classified as good or poor based on pathologic tumour regression. Changes in mean tumour apparent diffusion coefficient (ADC) were analysed by linear mixed-effects models and compared to modelling using volume. Model coefficient analysis and receiver operating characteristics (ROCs) with area under the curve (AUC) were applied for timepoint investigation of response.

Six patients had good response, with significant difference in median (range) baseline ADC; ADCgood = 1.3 (1.1–1.5) ∙ 10-3 mm2/s, ADCpoor = 1.1 (0.9–1.4) ∙ 10-3 mm2/s, (p = 0.03), and with greater ADC change at all evaluated timepoints. This resulted in AUCs of 0.73–0.85 and the steepest slope (m) after five fractions (mgood = 0.090, mpoor = 0.014). Combining baseline and slope improved the differentiation with AUCs of 0.90, 0.87, 0.85 and 0.83 after 5, 10, 13 or all fractions, respectively. Stratification based on volume changes was feasible after 15/25 fractions.

Early indication of treatment response in LARC was achieved by combining baseline and sequential ADC information. These encouraging results should be validated in a larger cohort.