Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Lightweight G-YOLOv11: Advancing Efficient Fracture Detection in Pediatric Wrist X-rays

Abdesselam Ferdi

TL;DR

The paper addresses the need for fast, accurate fracture detection in pediatric wrist X-rays under limited hardware resources. It proposes G-YOLOv11, a lightweight detector that replaces standard convolutions with ghost convolution and C3k2 with C3Ghost to substantially reduce parameters and FLOPs while maintaining competitive accuracy. Results show significant efficiency gains and real-time inference on an NVIDIA A10, with modest reductions in mAP compared to heavier YOLOv11 models. This approach enables practical, real-time CAD support in clinical settings with constrained computational resources.

Abstract

Computer-aided diagnosis (CAD) systems have greatly improved the interpretation of medical images by radiologists and surgeons. However, current CAD systems for fracture detection in X-ray images primarily rely on large, resource-intensive detectors, which limits their practicality in clinical settings. To address this limitation, we propose a novel lightweight CAD system based on the YOLO detector for fracture detection. This system, named ghost convolution-based YOLOv11 (G-YOLOv11), builds on the latest version of the YOLO detector family and incorporates the ghost convolution operation for feature extraction. The ghost convolution operation generates the same number of feature maps as traditional convolution but requires fewer linear operations, thereby reducing the detector's computational resource requirements. We evaluated the performance of the proposed G-YOLOv11 detector on the GRAZPEDWRI-DX dataset, achieving an mAP@0.5 of 0.535 with an inference time of 2.4 ms on an NVIDIA A10 GPU. Compared to the standard YOLOv11l, G-YOLOv11l achieved reductions of 13.6% in mAP@0.5 and 68.7% in size. These results establish a new state-of-the-art benchmark in terms of efficiency, outperforming existing detectors. Code and models are available at https://github.com/AbdesselamFerdi/G-YOLOv11.

Lightweight G-YOLOv11: Advancing Efficient Fracture Detection in Pediatric Wrist X-rays

TL;DR

The paper addresses the need for fast, accurate fracture detection in pediatric wrist X-rays under limited hardware resources. It proposes G-YOLOv11, a lightweight detector that replaces standard convolutions with ghost convolution and C3k2 with C3Ghost to substantially reduce parameters and FLOPs while maintaining competitive accuracy. Results show significant efficiency gains and real-time inference on an NVIDIA A10, with modest reductions in mAP compared to heavier YOLOv11 models. This approach enables practical, real-time CAD support in clinical settings with constrained computational resources.

Abstract

Computer-aided diagnosis (CAD) systems have greatly improved the interpretation of medical images by radiologists and surgeons. However, current CAD systems for fracture detection in X-ray images primarily rely on large, resource-intensive detectors, which limits their practicality in clinical settings. To address this limitation, we propose a novel lightweight CAD system based on the YOLO detector for fracture detection. This system, named ghost convolution-based YOLOv11 (G-YOLOv11), builds on the latest version of the YOLO detector family and incorporates the ghost convolution operation for feature extraction. The ghost convolution operation generates the same number of feature maps as traditional convolution but requires fewer linear operations, thereby reducing the detector's computational resource requirements. We evaluated the performance of the proposed G-YOLOv11 detector on the GRAZPEDWRI-DX dataset, achieving an mAP@0.5 of 0.535 with an inference time of 2.4 ms on an NVIDIA A10 GPU. Compared to the standard YOLOv11l, G-YOLOv11l achieved reductions of 13.6% in mAP@0.5 and 68.7% in size. These results establish a new state-of-the-art benchmark in terms of efficiency, outperforming existing detectors. Code and models are available at https://github.com/AbdesselamFerdi/G-YOLOv11.
Paper Structure (17 sections, 5 equations, 8 figures, 5 tables)

This paper contains 17 sections, 5 equations, 8 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Materials and Methods
  3. Dataset
  4. Proposed Detector (G-YOLOv11)
  5. Input
  6. Backbone
  7. Neck
  8. Head
  9. Comparison with standard YOLOv11
  10. Training details
  11. Evaluation metrics
  12. Results
  13. Discussion
  14. Qualitative Comparison with Standard YOLOv11
  15. Quantitative Comparison with Standard YOLOv11
  16. ...and 2 more sections

Figures (8)

  • Figure 1: Comparisons of real-time object detectors on the GRAZPEDWRI-DX dataset, evaluated in terms of FLOPs-accuracy (left) and size-accuracy (right) trade-offs.
  • Figure 2: Schematic diagram of the proposed CAD system based on the G-YOLOv11 detector for localizing pediatric wrist trauma in X-ray images.
  • Figure 3: Visualization of training labels in the GRAZPEDWRI-DX training set. First row: distribution of classes and visualization of bounding boxes. Second row: statistical distribution of the location and size of bounding boxes.
  • Figure 4: Bounding box-annotated X-ray images of pediatric wrist trauma from the GRAZPEDWRI-DX dataset nagy2022pediatric.
  • Figure 5: Architecture of the proposed G-YOLOv11 detector.
  • ...and 3 more figures