ScribbleDose: Scribble-Guided Dose Prediction in Radiotherapy

2026-05-12

AI Summary

This work addresses the high annotation cost of conventional radiotherapy dose prediction, which relies on meticulously delineated anatomical structure masks. The authors propose the first dose prediction framework that operates directly from sparse scribble annotations. Their approach introduces a Scribble Completion Module (SCM) to generate dense anatomical masks and integrates a Structure-Guided Dose Generation Module (SGDGM) to establish an end-to-end “scribbles → masks → dose” learning pipeline. Innovatively leveraging sparse scribbles as structural priors, the method incorporates supervoxel regularization to preserve anatomical boundary consistency and explicitly enforces a strong association between structural cues and dose distribution. Evaluated on the GDP-HMM dataset, the framework achieves competitive dose prediction performance using only sparse scribble annotations.

Abstract

Anatomical structure masks are widely adopted in radiotherapy dose prediction, as they provide explicit geometric constraints that facilitate structure-dose coupling. However, conventional manual delineation of these masks requires precise annotation of structure boundaries relevant to radiotherapy, which is time-consuming and labor-intensive. To address these limitations, we propose a scribble-guided dose prediction framework that relies solely on anatomical structures annotated with sparse scribbles. Specifically, we design a Scribble Completion Module (SCM) to generate dense anatomical masks by propagating sparse scribble labels to semantically similar voxels. During the propagation process, a supervoxel-based regularization is introduced to preserve geometric boundary consistency to ensure anatomical plausibility. Furthermore, we propose a Structure-Guided Dose Generation Module (SGDGM) to strengthen the correspondence between sparse structural cues and dose distribution. The completed dense masks derived from scribbles serve as structural guidance to condition dose prediction, forming a scribble-mask-dose learning pipeline under sparse annotation. Experiments on the GDP-HMM dataset demonstrate that ScribbleDose achieves competitive dose prediction performance using only sparse structural annotations. The source code and reannotated scribble annotations are publicly available at https://github.com/iCherishxixixi/ScribbleDose.

Problem

• radiotherapy
• dose prediction
• scribble annotation
• anatomical structure
• sparse labeling

Innovation

• scribble annotation
• dose prediction
• anatomical mask completion
• supervoxel regularization
• radiotherapy planning