Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Hierarchical Feature Learning for Medical Point Clouds via State Space Model

Guoqing Zhang, Jingyun Yang, Yang Li

TL;DR

This paper tackles the challenge of understanding medical point clouds with scalable multi-scale representations. It introduces a state-space model–based hierarchical framework that performs FPS downsampling, multi-scale KNN neighborhood aggregation, and local-to-global feature learning, aided by coordinate-order and inside-out point serialization and vanilla/group PSSM blocks. The authors also provide MedPointS, a large-scale medical point cloud dataset for anatomy classification, completion, and segmentation, and demonstrate state-of-the-art performance across all tasks. The approach offers efficient, robust modeling of complex anatomical structures and provides a public dataset and codebase to spur further research in medical point cloud analysis.

Abstract

Deep learning-based point cloud modeling has been widely investigated as an indispensable component of general shape analysis. Recently, transformer and state space model (SSM) have shown promising capacities in point cloud learning. However, limited research has been conducted on medical point clouds, which have great potential in disease diagnosis and treatment. This paper presents an SSM-based hierarchical feature learning framework for medical point cloud understanding. Specifically, we down-sample input into multiple levels through the farthest point sampling. At each level, we perform a series of k-nearest neighbor (KNN) queries to aggregate multi-scale structural information. To assist SSM in processing point clouds, we introduce coordinate-order and inside-out scanning strategies for efficient serialization of irregular points. Point features are calculated progressively from short neighbor sequences and long point sequences through vanilla and group Point SSM blocks, to capture both local patterns and long-range dependencies. To evaluate the proposed method, we build a large-scale medical point cloud dataset named MedPointS for anatomy classification, completion, and segmentation. Extensive experiments conducted on MedPointS demonstrate that our method achieves superior performance across all tasks. The dataset is available at https://flemme-docs.readthedocs.io/en/latest/medpoints.html. Code is merged to a public medical imaging platform: https://github.com/wlsdzyzl/flemme.

Hierarchical Feature Learning for Medical Point Clouds via State Space Model

TL;DR

This paper tackles the challenge of understanding medical point clouds with scalable multi-scale representations. It introduces a state-space model–based hierarchical framework that performs FPS downsampling, multi-scale KNN neighborhood aggregation, and local-to-global feature learning, aided by coordinate-order and inside-out point serialization and vanilla/group PSSM blocks. The authors also provide MedPointS, a large-scale medical point cloud dataset for anatomy classification, completion, and segmentation, and demonstrate state-of-the-art performance across all tasks. The approach offers efficient, robust modeling of complex anatomical structures and provides a public dataset and codebase to spur further research in medical point cloud analysis.

Abstract

Deep learning-based point cloud modeling has been widely investigated as an indispensable component of general shape analysis. Recently, transformer and state space model (SSM) have shown promising capacities in point cloud learning. However, limited research has been conducted on medical point clouds, which have great potential in disease diagnosis and treatment. This paper presents an SSM-based hierarchical feature learning framework for medical point cloud understanding. Specifically, we down-sample input into multiple levels through the farthest point sampling. At each level, we perform a series of k-nearest neighbor (KNN) queries to aggregate multi-scale structural information. To assist SSM in processing point clouds, we introduce coordinate-order and inside-out scanning strategies for efficient serialization of irregular points. Point features are calculated progressively from short neighbor sequences and long point sequences through vanilla and group Point SSM blocks, to capture both local patterns and long-range dependencies. To evaluate the proposed method, we build a large-scale medical point cloud dataset named MedPointS for anatomy classification, completion, and segmentation. Extensive experiments conducted on MedPointS demonstrate that our method achieves superior performance across all tasks. The dataset is available at https://flemme-docs.readthedocs.io/en/latest/medpoints.html. Code is merged to a public medical imaging platform: https://github.com/wlsdzyzl/flemme.

Paper Structure

This paper contains 12 sections, 6 equations, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Methods
  3. Hierarchical Architecture
  4. SSM-based Point Feature Learning
  5. Task-Specific Decoders
  6. Experiments
  7. MedPointS Dataset and Evaluation Metrics
  8. Experimental Setup
  9. Results
  10. Conclusion
  11. Acknowledgments.
  12. Disclosure of Interests.

Figures (3)

  • Figure 1: Pipeline of the proposed method. The right part details how the point set is processed at each building block.
  • Figure 2: Illustration of different scanning strategies and PSSM blocks.
  • Figure 3: Visualization of completion (top) and segmentation (bottom) results.