IDENTIFICATION OF NON-TRAUMATIC VERTEBRAL COMPRESSION FRACTURES IN CT IMAGES USING A HYBRID DEEP LEARNING MODEL COMBINING DENSENET AND GAN

Vertebral compression fractures are common conditions, particularly in the aging population, often linked to osteoporosis and other degenerative diseases. Non-traumatic vertebral compression fractures (VCFs) can be difficult to identify from medical images, especially those that do not show signs of trauma. This has led to a demand for more effective and automated detection methods. This study proposes a hybrid deep learning approach that uses DenseNet and Generative Adversarial Networks (GANs) to detect non-traumatic VCFs from computed tomography (CT) images. A dataset consisting of patient CT scans was used, including 101 images with confirmed fractures and 99 images without fractures. Our hybrid model demonstrated superior accuracy to conventional methods, showing promising results in distinguishing between fractured and non-fractured vertebrae. This automated method could aid radiologists in early diagnosis and treatment planning by decreasing the time needed for manual image analysis and improving diagnostic accuracy. The combination of DenseNet and GANs demonstrates the effectiveness of using advanced deep-learning techniques for medical image classification, opening the door for future applications in automated medical diagnosis.