Fractures are common injuries with diverse presentations and clinical implications. Accurate classification of bone fractures is crucial for guiding treatment decisions and optimizing patient outcomes. In this study, we propose a novel approach to classify bone fractures into three main classes using convolutional neural networks (CNNs).
The dataset used in this study comprises radiographic images of various bone fractures, including avulsion fractures, comminuted fractures, fracture dislocations, greenstick fractures, capillary fractures, impacted fractures, longitudinal fractures, oblique fractures, pathological fractures, and spiral fractures. Each image is labeled according to its corresponding fracture type, resulting in a total of ten classes.
We employ a CNN architecture, specifically tailored for image classification tasks, to learn discriminative features from the radiographic images. Transfer learning is utilized, leveraging pre-trained models such as AlexNet to extract meaningful features from the fracture images.
The dataset is divided into training, validation, and testing sets to facilitate model training and evaluation. We employ rigorous training procedures, including hyperparameter optimization, to enhance the robustness and generalization ability of the CNN model.
Experimental results demonstrate the effectiveness of the proposed approach in accurately classifying bone fractures into three main classes. The trained CNN achieves a classification accuracy of over 80% on the testing set, indicating its capability to distinguish between different fracture types with high precision and reliability.
Our study contributes to the advancement of computer-aided diagnosis in orthopedics, offering a promising tool for automated fracture classification and clinical decision support. The proposed CNN-based approach has the potential to streamline fracture diagnosis workflows, improve treatment planning, and ultimately enhance patient care in orthopedic practice.
ORCID NO: 0000-0001-9105-508X
Anahtar Kelimeler: CNN, Transfer Learning, Deep Learning, Bone Fractures
|
