Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

A multi-stage semi-supervised learning for ankle fracture classification on CT images

Hongzhi Liu, Guicheng Li, Jiacheng Nie, Hui Tang, Chunfeng Yang, Qianjin Feng, Hailin Xu, Yang Chen

TL;DR

The paper tackles the challenge of ankle fracture diagnosis by proposing a CT-based, multi-stage pipeline that first segments the tibia and fibula, then registers the fractured bone masks to healthy references to isolate the syndesmosis region, and finally applies a semi-supervised classifier that leverages unlabeled data via a Relational Weight Network and MMD-based loss. A tibia-fibula segmentation dataset and an ankle fracture dataset with labeled and unlabeled samples are constructed to support training. The method achieves state-of-the-art segmentation performance (mean Dice ≈ 0.943) and outperforms existing semi-supervised baselines for fracture type classification, particularly when labeled data are limited. This approach advances data-efficient, explainable AI-assisted diagnosis in orthopedics and lays groundwork for handling more complex fracture patterns in the future, with potential clinical impact in reducing annotation burden and supporting junior clinicians.

Abstract

Because of the complicated mechanism of ankle injury, it is very difficult to diagnose ankle fracture in clinic. In order to simplify the process of fracture diagnosis, an automatic diagnosis model of ankle fracture was proposed. Firstly, a tibia-fibula segmentation network is proposed for the joint tibiofibular region of the ankle joint, and the corresponding segmentation dataset is established on the basis of fracture data. Secondly, the image registration method is used to register the bone segmentation mask with the normal bone mask. Finally, a semi-supervised classifier is constructed to make full use of a large number of unlabeled data to classify ankle fractures. Experiments show that the proposed method can segment fractures with fracture lines accurately and has better performance than the general method. At the same time, this method is superior to classification network in several indexes.

A multi-stage semi-supervised learning for ankle fracture classification on CT images

TL;DR

The paper tackles the challenge of ankle fracture diagnosis by proposing a CT-based, multi-stage pipeline that first segments the tibia and fibula, then registers the fractured bone masks to healthy references to isolate the syndesmosis region, and finally applies a semi-supervised classifier that leverages unlabeled data via a Relational Weight Network and MMD-based loss. A tibia-fibula segmentation dataset and an ankle fracture dataset with labeled and unlabeled samples are constructed to support training. The method achieves state-of-the-art segmentation performance (mean Dice ≈ 0.943) and outperforms existing semi-supervised baselines for fracture type classification, particularly when labeled data are limited. This approach advances data-efficient, explainable AI-assisted diagnosis in orthopedics and lays groundwork for handling more complex fracture patterns in the future, with potential clinical impact in reducing annotation burden and supporting junior clinicians.

Abstract

Because of the complicated mechanism of ankle injury, it is very difficult to diagnose ankle fracture in clinic. In order to simplify the process of fracture diagnosis, an automatic diagnosis model of ankle fracture was proposed. Firstly, a tibia-fibula segmentation network is proposed for the joint tibiofibular region of the ankle joint, and the corresponding segmentation dataset is established on the basis of fracture data. Secondly, the image registration method is used to register the bone segmentation mask with the normal bone mask. Finally, a semi-supervised classifier is constructed to make full use of a large number of unlabeled data to classify ankle fractures. Experiments show that the proposed method can segment fractures with fracture lines accurately and has better performance than the general method. At the same time, this method is superior to classification network in several indexes.
Paper Structure (17 sections, 5 equations, 6 figures, 2 tables)

This paper contains 17 sections, 5 equations, 6 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related works
  3. Ankle fracture classification
  4. Medical image segmentation
  5. Semi-supervised classification
  6. Method
  7. Overview
  8. Tibia and fibula segmentation network
  9. Ankle masks registration
  10. Semi-supervised fracture classification
  11. Experimental results
  12. Datasets and preprocessing
  13. Implementation details
  14. Evaluation metrics
  15. Tibia and fibula segmentation
  16. ...and 2 more sections

Figures (6)

  • Figure 1: An example of a local injury in an ankle fracture patient.
  • Figure 2: The flowchat shows the process from input raw CT images to segmentation masks.
  • Figure 3: Healthy and fracture ankle joint segmentation mask registration.
  • Figure 4: The semi-supervised model for ankle mask classification based on the self-training strategy.
  • Figure 5: Comparison of qualitative results of different segmentation methods. The red is the fibula and the green means tibia.
  • ...and 1 more figures