Bone Fracture Detection with CNN: A Deep Learning Approach
Shanvi Chauhan
Abstract
This work intends to overcome the significant medical challenges, bone fractures, which necessitates precise and quick diagnosis to guarantee efficient treatment and recovery. This study classifies a complete collection of 10,580 radiographic X-ray images using a $\mathrm{CNN}-\mathrm{AlexNet}$ model including many anatomical areas like the lower limb, upper limb, lumbar, hips, and knees. Carefully curated into training (9,246 images), validation (828 images), and test (506 images) subsets—each comprising both fractured and non-fractured images, make this dataset. This study has trained the CNN -AlexNet model using this vast dataset to correctly distinguish fractured and non-fractured bone X-rays. The proposed model resulted in $96 \%$ accuracy on the testset. These results highlight the opportunities of CNNs, particularly the AlexNet architecture in enhancing diagnosis accuracy for bone fractures, so advancing the field of medical imaging and diagnostics. This study also fits with sustainable development goals by means of boosting health and well-being, enhancing quality education in medical technology, supporting innovation and infrastructure in healthcare, and lastly assisting to lower access to advanced diagnostics equipment.