Precision Fracture Classification: Leveraging Transfer Learning for Enhanced Diagnostics
<p>One of the major challenges to every human, irrespective of age or race, is bone fractures. Bone fractures are life-threatening and endanger public health, with an extensive mortality rate that occurs in all age groups, as the affected sustains hurtful injuries. Meaningfully, the competence of fr...
Guardado en:
| Autores principales: | , , |
|---|---|
| Formato: | Capítulo de libro |
| Lenguaje: | inglés |
| Publicado: |
IntechOpen
2026
|
| Materias: | |
| Acceso en línea: | https://www.intechopen.com/chapters/1240413 |
| Etiquetas: |
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| Sumario: | <p>One of the major challenges to every human, irrespective of age or race, is bone fractures. Bone fractures are life-threatening and endanger public health, with an extensive mortality rate that occurs in all age groups, as the affected sustains hurtful injuries. Meaningfully, the competence of fracture treatment and succeeding patient prognosis are greatly influenced by the speed and accuracy of the diagnosis. Any form of delay or misclassification can result in suboptimal interventions. To respond to these diagnostic issues, this study explores the innovative potential of a transfer learning approach on deep learning techniques to improve the automated and exact classification of bone fractures from medical imaging modalities. The classification task involves EfficientNetB2, VGG-19 and ResNet50, three CNN deep learning models with distinct publicly accessible structures and layers. A dataset comprising 9,473 bone fracture X-ray images was collected, converted and filtered with a Gaussian kernel to reduce noise. The model’s evaluation is based on standard metrics, and the output revealed that EfficientNetB0 has the highest accuracy with 99%, making it a great tool that is capable of detecting the radiological features from the images. VGG-19 showed moderate performance with 87% accuracy, 86% specificity and sensitivity, 82% precision and 82% F1-score. The Naïve Bayes approach yielded 89% accuracy, 87% specificity, 86% precision, 84% sensitivity and an 83% F1-score. This work remains a promising alternative to the traditional methods of reading the films. This model has shown considerable promise for clinical applications in rapid fracture identification.</p> |
|---|---|
| ISBN: | 978-1-80631-382-2 |
| DOI: | 10.5772/intechopen.1015734 |