# Comparison of deep learning-based segmentation and registration using pre-treatment contours for online rectal delineation in magnetic resonance-guided radiotherapy

## TL;DR

This study compares deep learning methods for improving rectal cancer radiotherapy by using pre-treatment contours to guide online MRI-based target delineation.

## Contribution

The study introduces a comparison of segmentation and registration models for incorporating pre-treatment contours in online rectal delineation for MR-guided radiotherapy.

## Key findings

- Segmentation models outperformed registration models in contour accuracy for mid and cranial regions.
- Using pre-treatment contours improved daily target contour accuracy by 15%–46%.
- Rectal volume changes can reduce the accuracy of both segmentation and registration models.

## Abstract

Deep learning promises accurate target contouring for online adaptive MR-guided radiotherapy (MRgRT) in rectal cancer. However, delineating the mesorectal clinical target volume (CTV) remains challenging. Integrating planning-based contours, delineated offline before treatment, can provide anatomical shape and boundary information. This study evaluated deep learning-based segmentation and registration models to determine the optimal approach for incorporating planning contours into online rectal contouring.

Deep learning-based segmentation and registration models, both U-Nets, were developed using MRI of 104 rectal cancer patients, split into 68, 14, and 22 training, validation, and testing subjects. The segmentation model used the planning CTV and daily fraction MRI, while the registration model used the planning MRI and CTV and the daily fraction MRI. The models were compared in terms of contour accuracy (maximum Hausdorff distance (HD), Dice, and a qualitative score) and robustness against domain shifts.

When incorporating the planning contour, the segmentation and registration models achieved comparable median HD values of 9.3 mm (interquartile range, IQR: 7.1-12.1) and 10.2 (8.2-12.4) (p=0.18), respectively. However, segmentation achieved lower HD values in the middle and cranial regions of the target (HDmiddle: 5.3 mm (4.3-6.6) vs. 6.0 mm (4.8-8.0), p<0.05; HDcranial: 7.6 mm (6.3-10.7) vs. 9.6 mm (7.5-11.9), p<0.05). In addition, segmentation resulted in more clinically acceptable contours (9/10 versus 3/10) and was more robust to rectum volume variations than registration.

Deep learning-based segmentation was identified as the optimal approach for incorporating the planning CTV into online rectal delineation in MRgRT.

•Using pre-treatment contours improved daily target contour accuracy by 15%–46%.•Segmentation vs registration: contour accuracy and robustness to domain shifts.•Segmentation outperformed registration by 11%–20% in mid and cranial regions.•Rectal volume changes can reduce segmentation and registration contour accuracy.

Using pre-treatment contours improved daily target contour accuracy by 15%–46%.

Segmentation vs registration: contour accuracy and robustness to domain shifts.

Segmentation outperformed registration by 11%–20% in mid and cranial regions.

Rectal volume changes can reduce segmentation and registration contour accuracy.