Embarrassingly Simple Scribble Supervision for 3D Medical Segmentation

2024-03-19

arXiv.org

AI Summary

Existing scribble-supervised methods are largely confined to 3D cardiac segmentation (e.g., ACDC/MSCMR), suffering from poor generalizability and overfitting. To address the high cost of dense annotation in medical imaging, this work proposes a lightweight scribble-supervision paradigm. First, we construct the first comprehensive benchmark comprising seven datasets spanning multiple anatomical structures, pathologies, and imaging modalities. Second, we introduce a plug-and-play local loss mechanism that decouples supervision strategy from segmentation architecture, substantially enhancing model transferability. Third, we establish a cross-modal, multi-center scribble annotation protocol. Evaluated on a unified benchmark, our method consistently achieves 92–96% of the Dice score attained by fully supervised nnU-Net, while remaining compatible with any mainstream 3D segmentation architecture. This systematic evaluation demonstrates the efficacy and robustness of scribble supervision in realistic clinical settings.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Multimodal Learning
• Constraint Satisfaction and Optimization: Satisfiability Modulo Theories

Application Category

• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Economics, Online Markets and Human Computation: Data quality aspects of human-annotated datasets

Abstract

Traditionally, segmentation algorithms require dense annotations for training, demanding significant annotation efforts, particularly within the 3D medical imaging field. Scribble-supervised learning emerges as a possible solution to this challenge, promising a reduction in annotation efforts when creating large-scale datasets. Recently, a plethora of methods for optimized learning from scribbles have been proposed, but have so far failed to position scribble annotation as a beneficial alternative. We relate this shortcoming to two major issues: 1) the complex nature of many methods which deeply ties them to the underlying segmentation model, thus preventing a migration to more powerful state-of-the-art models as the field progresses and 2) the lack of a systematic evaluation to validate consistent performance across the broader medical domain, resulting in a lack of trust when applying these methods to new segmentation problems. To address these issues, we propose a comprehensive scribble supervision benchmark consisting of seven datasets covering a diverse set of anatomies and pathologies imaged with varying modalities. We furthermore propose the systematic use of partial losses, i.e. losses that are only computed on annotated voxels. Contrary to most existing methods, these losses can be seamlessly integrated into state-of-the-art segmentation methods, enabling them to learn from scribble annotations while preserving their original loss formulations. Our evaluation using nnU-Net reveals that while most existing methods suffer from a lack of generalization, the proposed approach consistently delivers state-of-the-art performance. Thanks to its simplicity, our approach presents an embarrassingly simple yet effective solution to the challenges of scribble supervision. Source code as well as our extensive scribble benchmarking suite will be made publicly available upon publication.

Problem

• Over-specialization in cardiac domain limits scribble supervision generalization
• Lack of diverse benchmarks misleads scribble-based segmentation performance claims
• Existing methods fail to generalize across broader medical segmentation tasks

Innovation

• Introduces ScribbleBench for diverse dataset benchmarking
• Identifies nnU-Net with partial loss as superior baseline
• Challenges overfitting in cardiac-focused scribble methods