An Efficient Hybrid Deep Learning Model for Detecting Musculoskeletal Abnormalities

Deep Convolutional Neural Networks (DCNNs) have made tremendous advances in healthcare, but their limited perceptual capacities make it difficult to capture extensive structural details. To solve this restriction, this study presents a new technique which integrates the benefits of EfficientNet to perform local data extraction with DenseNet201 for processing global characteristics in pictures. The MURA dataset, which contains 40,005 radiographic pictures, is used in the project with the goal of enhancing radiographic image categorization, notably for identifying Musculoskeletal disorders. The goal of the project is to accomplish a more thorough analysis of radiological pictures by successfully collecting global as well as local information via the fusion of EfficientNet and DenseNet201. The proposed approach shows positive outcomes, with the combined approach obtaining an overall accuracy of 96.38% on the test dataset and a sensitivity of 0.90.