nnSAM2: nnUNet-Enhanced One-Prompt SAM2 for Few-shot Multi-Modality Segmentation and Composition Analysis of Lumbar Paraspinal Muscles

2025-10-06

AI Summary

To address the challenge of few-shot segmentation of paraspinal muscles in multi-modal (MRI/CT) lumbar spine imaging, this paper proposes nnSAM2—a framework requiring only one annotated slice per dataset. It synergistically integrates SAM2’s zero-shot prompting capability with nnU-Net’s robust medical image modeling, leveraging single-slice–guided pseudo-label generation, cross-dataset label pooling, and three-stage independent nnU-Net iterative refinement. To our knowledge, nnSAM2 is the first method enabling statistically equivalent body composition analysis—across modalities (MRI/CT), multi-center cohorts, and under extreme low-labeling conditions (1 annotation per dataset). Evaluated on 19,433 unlabeled slices from six diverse datasets, it achieves Dice scores of 0.92–0.96. Automated measurements of muscle volume, fat fraction, and CT attenuation values show no statistically significant difference from expert annotations (two-one-sided t-tests, *P* < 0.05), with intraclass correlation coefficients (ICC) ranging from 0.86 to 1.00.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Multimodal Learning
• Natural Language Processing: Language Grounding & Multi-modal NLP

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Web Mining and Content Analysis: Mining multimedia, multimodal, multilingual, cross-lingual Web data

Abstract

Purpose: To develop and validate No-New SAM2 (nnsam2) for few-shot segmentation of lumbar paraspinal muscles using only a single annotated slice per dataset, and to assess its statistical comparability with expert measurements across multi-sequence MRI and multi-protocol CT. Methods: We retrospectively analyzed 1,219 scans (19,439 slices) from 762 participants across six datasets. Six slices (one per dataset) served as labeled examples, while the remaining 19,433 slices were used for testing. In this minimal-supervision setting, nnsam2 used single-slice SAM2 prompts to generate pseudo-labels, which were pooled across datasets and refined through three sequential, independent nnU-Net models. Segmentation performance was evaluated using the Dice similarity coefficient (DSC), and automated measurements-including muscle volume, fat ratio, and CT attenuation-were assessed with two one-sided tests (TOST) and intraclass correlation coefficients (ICC). Results: nnsam2 outperformed vanilla SAM2, its medical variants, TotalSegmentator, and the leading few-shot method, achieving DSCs of 0.94-0.96 on MR images and 0.92-0.93 on CT. Automated and expert measurements were statistically equivalent for muscle volume (MRI/CT), CT attenuation, and Dixon fat ratio (TOST, P < 0.05), with consistently high ICCs (0.86-1.00). Conclusion: We developed nnsam2, a state-of-the-art few-shot framework for multi-modality LPM segmentation, producing muscle volume (MRI/CT), attenuation (CT), and fat ratio (Dixon MRI) measurements that were statistically comparable to expert references. Validated across multimodal, multicenter, and multinational cohorts, and released with open code and data, nnsam2 demonstrated high annotation efficiency, robust generalizability, and reproducibility.

Problem

• Segmenting lumbar paraspinal muscles with minimal annotated data
• Achieving statistical equivalence with expert manual measurements
• Validating across multi-sequence MRI and multi-protocol CT scans

Innovation

• Uses single-slice SAM2 prompts for pseudo-label generation
• Refines labels with three sequential nnU-Net models
• Combines multi-dataset pooling for enhanced segmentation