Adaptive Feature Fusion-based Bone Fracture Detection using Hybrid DenseNet121 and EfficientNetB0 Architecture
Year: 2026
Type: conference-paper
Abstract: Bone fractures are breaks or cracks in the bone structure caused by injury, trauma or excessive force, and accurate fracture detection from X-ray images is crucial for effective medical diagnosis and therapeutic planning. Manual examination of radiographic images is often time-consuming which may lead to diagnostic variations, especially in intricate fracture regions. Artificial intelligence (AI)-based fracture detection systems help radiologists by providing automated and reliable examination of bone X-ray images. This research presents a hybrid deep learning (DL) framework integrating DenseNet121 and EfficientNetB0 for binary bone fracture classification. DenseNet121 captures detailed hierarchical bone features through dense connectivity learning, while EfficientNetB0 extracts optimized spatial feature representations. The extracted feature representations are combined using an adaptive feature fusion block for effective fracture prediction. The proposed framework classifies bone X-ray images into fractured and non-fractured categories, achieving 98.00% classification accuracy with an Area Under the Curve (AUC) value of 0.9815, demonstrating reliable fracture detection capability with minimal prediction error.
Authors: Linchu A M, B Ben Sujitha
Institution: Noorul Islam University
Language: English
Topic: Medical Imaging and Analysis
Subfield: Biomedical Engineering
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7203473835