In-context learning for medical image segmentation

2024-12-17

arXiv.org

AI Summary

To address the high annotation cost and inter-slice boundary inconsistency in medical image (MRI/CT) segmentation, this paper proposes In-Context Segmentation (ICS), a fine-tuning-free, context-cascaded segmentation method. Inspired by in-context learning, ICS requires only a few initially annotated slices and dynamically updates its support set using inference outputs, enabling cross-slice feature propagation and bidirectional consistency modeling. Its core innovations include: (i) the first serialized context enhancement mechanism, which systematically analyzes the critical impact of support slice count and spatial positioning on performance; and (ii) zero-training few-shot segmentation within the UniverSeg framework. Evaluated on the HVSMR benchmark for eight-region cardiac segmentation, ICS achieves an average 4.2% mDice improvement over baselines and significantly enhances inter-slice boundary continuity—demonstrating clinical-grade robustness under minimal annotation cost.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Semi-Supervised Learning
• Search and Optimization: Learning to Search

Application Category

• Search and Retrieval-Augmented AI: Personalized, context-aware and across-device search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization

Abstract

Annotation of medical images, such as MRI and CT scans, is crucial for evaluating treatment efficacy and planning radiotherapy. However, the extensive workload of medical professionals limits their ability to annotate large image datasets, posing a bottleneck for AI applications in medical imaging. To address this, we propose In-context Cascade Segmentation (ICS), a novel method that minimizes annotation requirements while achieving high segmentation accuracy for sequential medical images. ICS builds on the UniverSeg framework, which performs few-shot segmentation using support images without additional training. By iteratively adding the inference results of each slice to the support set, ICS propagates information forward and backward through the sequence, ensuring inter-slice consistency. We evaluate the proposed method on the HVSMR dataset, which includes segmentation tasks for eight cardiac regions. Experimental results demonstrate that ICS significantly improves segmentation performance in complex anatomical regions, particularly in maintaining boundary consistency across slices, compared to baseline methods. The study also highlights the impact of the number and position of initial support slices on segmentation accuracy. ICS offers a promising solution for reducing annotation burdens while delivering robust segmentation results, paving the way for its broader adoption in clinical and research applications.

Problem

• Reduces annotation workload for medical image segmentation
• Improves segmentation accuracy in complex anatomical regions
• Ensures inter-slice consistency in sequential medical images

Innovation

• In-context Cascade Segmentation minimizes annotation workload
• ICS uses iterative support set updates for consistency
• ICS improves boundary consistency in complex anatomical regions