Fast Multi-Organ Fine Segmentation in CT Images with Hierarchical Sparse Sampling and Residual Transformer

2025-11-11

AI Summary

To address the high computational cost and the difficulty of balancing speed and accuracy in fine-grained multi-organ segmentation of 3D medical images, this paper proposes a lightweight and efficient segmentation framework based on hierarchical sparse sampling and residual Transformers. Methodologically, we design a multi-resolution hierarchical sparse sampling strategy to preserve critical anatomical context, and integrate residual Transformer modules with low-overhead attention mechanisms to enable efficient modeling and multi-scale fusion of sparse features. The framework supports real-time inference on CPU. Evaluated on an internal dataset of 10,253 CT scans and the TotalSegmentator benchmark, it significantly outperforms existing fast segmenters—achieving a mean inference time of only 2.24 seconds while maintaining high segmentation accuracy. This work establishes a new paradigm for clinically deployable 3D organ segmentation.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Hardware-aware ML
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Security and Privacy: Data transparency and provenance
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Multi-organ segmentation of 3D medical images is fundamental with meaningful applications in various clinical automation pipelines. Although deep learning has achieved superior performance, the time and memory consumption of segmenting the entire 3D volume voxel by voxel using neural networks can be huge. Classifiers have been developed as an alternative in cases with certain points of interest, but the trade-off between speed and accuracy remains an issue. Thus, we propose a novel fast multi-organ segmentation framework with the usage of hierarchical sparse sampling and a Residual Transformer. Compared with whole-volume analysis, the hierarchical sparse sampling strategy could successfully reduce computation time while preserving a meaningful hierarchical context utilizing multiple resolution levels. The architecture of the Residual Transformer segmentation network could extract and combine information from different levels of information in the sparse descriptor while maintaining a low computational cost. In an internal data set containing 10,253 CT images and the public dataset TotalSegmentator, the proposed method successfully improved qualitative and quantitative segmentation performance compared to the current fast organ classifier, with fast speed at the level of ~2.24 seconds on CPU hardware. The potential of achieving real-time fine organ segmentation is suggested.

Problem

• Reducing computation time for 3D medical image segmentation
• Balancing speed and accuracy in multi-organ CT segmentation
• Achieving real-time performance while maintaining hierarchical context

Innovation

• Hierarchical sparse sampling reduces computation time
• Residual Transformer extracts multi-level sparse descriptors
• Achieves fast CPU segmentation in 2.24 seconds