1. iA-HLD: an improved AlexNet for hairline fracture detection in orthopedic images.
- Author
-
Jain, Bhawna, Malik, Diksha, Jagota, Ganiti, Gyanvi, and Chandra, Ishita
- Subjects
- *
CONVOLUTIONAL neural networks , *IMAGE recognition (Computer vision) , *DEEP learning , *CRITICAL currents , *BONE fractures - Abstract
Bone fractures are a substantial health concern affecting approximately 2.7 million individuals annually across six European countries: France, Germany, Italy, Spain, Sweden, and the UK. If left untreated, this issue carries significant health risks, including fatality. It is crucial to accurately identify the types of fractures, especially subtle hairline fractures to mitigate long-term consequences. These fractures are characterized by small breaks where the bone fragments are aligned, and there is no visible displacement. Unfortunately, detecting hairline fractures is a significant challenge in the medical field. This is mainly attributed to the intricate nature of these fractures adding complexity, posing difficulties for both human and machine detection. Additionally, there is a lack of easily accessible datasets focused on hairline fractures. This paper introduces the iA-HLD model, a novel and enhanced approach for detecting hairline fractures. Through architectural modifications, this model exhibits superior capabilities in identifying hairline fractures across all types of human bones using deep learning and stands as the pioneering solution of its kind. A comprehensive comparative analysis is conducted, assessing the performance of the proposed model against established models, including ResNet-50, AlexNet, and convolutional neural network. Evaluation metrics, including accuracy, precision, recall, and F1-score, are used to compare the models. iA-HLD achieved an accuracy rate of 97.6%, highlighting its superior capabilities. In addition, it scored 98% in precision, recall, and F1-score, which surpasses all other models. These results show its improved capabilities as well as its potential for use in real-world applications across many fields. The research is a significant stride in advancing hairline fracture detection and addresses a critical gap in current medical diagnostic methods. [ABSTRACT FROM AUTHOR]
- Published
- 2024
- Full Text
- View/download PDF