Related Experiment Video
Updated: Sep 25, 2025
05:42
Orthopedic Robot-Assisted Femoral Neck System in the Treatment of Femoral Neck Fracture
Published on: March 3, 2023
2.7K
Vision Transformer for femur fracture classification
Leonardo Tanzi1, Andrea Audisio2, Giansalvo Cirrincione3
1DIGEP, Polytechnic University of Turin, Corso Duca degli Abruzzi 24, Torino 10129, Italy.
Injury
|April 26, 2022
Summary
A new Vision Transformer (ViT) system significantly improved bone fracture diagnosis accuracy. Clinicians achieved 29% better diagnostic accuracy when assisted by ViT, demonstrating the power of AI in medical imaging.
Area of Science:
- Medical Imaging
- Artificial Intelligence
- Orthopedics
Background:
- Computer-Aided Diagnosis (CAD) systems using Convolutional Neural Networks (CNNs) have limitations in accurately classifying bone fracture subtypes.
- There is a need for advanced deep learning techniques to enhance diagnostic precision in fracture detection.
Purpose of the Study:
- To evaluate a novel CAD system employing Vision Transformers (ViT) for bone fracture classification.
- To assess the impact of ViT-assisted diagnosis on clinicians' accuracy in identifying proximal femur fractures.
Main Methods:
- Utilized 4207 manually annotated images classified according to AO/OTA fracture types.
- Compared ViT architecture against traditional CNNs and multistage CNNs.
- Analyzed attention maps, performed unsupervised learning comparisons, and conducted a clinical evaluation with 11 physicians classifying 150 fractures with and without ViT assistance.
Main Results:
- The ViT model achieved 83% prediction accuracy on test images, with precision, recall, and F1-scores of 0.77.
- Clinicians demonstrated a 29% improvement in diagnostic accuracy (reaching 97% accuracy, p=0.003) when aided by ViT predictions.
- ViT-assisted diagnosis outperformed the algorithm alone.
Conclusions:
- Vision Transformers show significant potential for improving bone fracture classification, especially for subtypes.
- This study achieved state-of-the-art results in sub-fracture classification.
- Collaborative diagnosis, combining clinician expertise with ViT-powered AI, yields superior outcomes.
