Abstract:Abstract Background Ankle fractures are prevalent injuries that necessitate precise diagnostic tools. Traditional diagnostic methods have limitations that can be addressed using machine learning techniques, with the potential to improve accuracy and expedite diagnoses. Methods We trained various deep learning architectures, notably the Adapted ResNet50 with SENet capabilities, to identify ankle fractures using a curated dataset of radiographic images. Model performance was evaluated using common metrics like accuracy, precision, and recall. Additionally, Grad-CAM visualizations were employed to interpret model decisions. Results The Adapted ResNet50 with SENet capabilities consistently outperformed other models, achieving an accuracy of 93%, AUC of 95%, and recall of 92%. Grad-CAM visualizations provided insights into areas of the radiographs that the model deemed significant in its decisions. Conclusions The Adapted ResNet50 model enhanced with SENet capabilities demonstrated superior performance in detecting ankle fractures, offering a promising tool to complement traditional diagnostic methods. However, continuous refinement and expert validation are essential to ensure optimal application in clinical settings.

What problem does this paper attempt to address?

The paper primarily aims to address the challenges in diagnosing ankle fractures, particularly in improving diagnostic accuracy. Specifically, the research team hopes to enhance the recognition of ankle fractures by utilizing improved deep learning techniques. The following are the key issues this study attempts to address: ### Research Background and Problem Statement 1. **Epidemiological Characteristics of Ankle Fractures**: Ankle fractures account for approximately 15% of adult fracture cases, and such injuries are usually caused by twisting or rolling movements of the ankle. With the global aging population and increased activity levels, the incidence of such injuries is rising. 2. **Limitations of Traditional Diagnostic Methods**: Although X-rays are the primary tool for diagnosing ankle fractures, their sensitivity to subtle fractures or occult injuries is limited. This may lead to misdiagnosis or missed diagnosis of mild or moderate fractures, thereby affecting treatment plans and patient recovery processes. 3. **Consequences of Misdiagnosis**: Inadequate or incorrect diagnosis can lead to long-term complications, such as post-traumatic arthritis, severely impacting the patient's functional recovery and quality of life. ### Research Objectives - **Propose an Improved Deep Learning Model**: The researchers developed a deep learning model based on the ResNet50 architecture, incorporating Squeeze-and-Excitation Network (SENet) functionalities, aiming to improve the accuracy of ankle fracture image recognition. - **Compare the Performance of Different Models**: By comparing the proposed improved model with the original ResNet50 model and the EfficientNetB5 model, the study evaluates the performance of the models in recognizing ankle fractures. - **Enhance Model Interpretability with Grad-CAM**: By integrating Grad-CAM technology, the study generates heatmaps to visually display the image areas the model focuses on during recognition, providing more detailed diagnostic information for medical professionals. ### Main Contributions - **Model Development**: An improved version of the ResNet50 model adapted for ankle fracture recognition was developed, with enhanced feature extraction capabilities through the addition of SENet components. - **Experimental Validation**: The model was trained and tested using a carefully prepared dataset, including steps such as CT image preprocessing, data augmentation, and normalization. - **Performance Evaluation**: By comparing metrics such as accuracy, precision, recall, and F1 score of different models, the study demonstrates the superiority of the improved model in recognizing ankle fractures. - **Interpretability Analysis**: The heatmap visualization results provided by Grad-CAM technology not only enhance the model's transparency but also confirm that the model can focus on critical pathological areas. In summary, this study aims to address the challenges in diagnosing ankle fractures by introducing advanced machine learning techniques, thereby improving diagnostic accuracy and ultimately enhancing patient treatment outcomes.

Harnessing ResNet50 and SENet for enhanced ankle fracture identification

Foot fractures diagnosis using a deep convolutional neural network optimized by extreme learning machine and enhanced snow ablation optimizer

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

An Innovative Deep Learning Approach to Spinal Fracture Detection in CT Images

Deep neural network improves fracture detection by clinicians

Multiclass datasets expand neural network utility: an example on ankle radiographs

Enhancing diagnosis: ensemble deep-learning model for fracture detection using X-ray images

Long-Bone Fracture Detection using Artificial Neural Networks based on Contour Features of X-ray Images

Lower-extremity fatigue fracture detection and grading based on deep learning models of radiographs

Arm fracture detection in X-rays based on improved deep convolutional neural network

Transfer learning-based ensemble convolutional neural network for accelerated diagnosis of foot fractures

Enhancing fracture diagnosis in pelvic X-rays by deep convolutional neural network with synthesized images from 3D-CT

Artificial Intelligence Model Trained with Sparse Data to Detect Facial and Cranial Bone Fractures from Head CT

Anatomy-Aware Siamese Network: Exploiting Semantic Asymmetry for Accurate Pelvic Fracture Detection in X-ray Images

Application of deep learning algorithm to detect and visualize vertebral fractures on plain frontal radiographs

Automatic Hip Fracture Identification and Functional Subclassification with Deep Learning

Artificial Intelligence Detection of Cervical Spine Fractures Using Convolutional Neural Network Models

An Integrated Approach using YOLOv8 and ResNet, SeResNet & Vision Transformer (ViT) Algorithms based on ROI Fracture Prediction in X-ray Images of the Elbow

AFFnet - a deep convolutional neural network for the detection of atypical femur fractures from anteriorposterior radiographs

Hybrid SFNet Model for Bone Fracture Detection and Classification Using ML/DL

Detection of Hidden Pediatric Elbow Fractures in X-ray Images Based on Deep Learning