An automated metaheuristic tunicate swarm algorithm based deep convolutional neural network for bone age assessment model

Abstract

The evaluation of X-ray images of hands serves as the basis for Bone Age Assessment (BAA), a critical component in the prediction and analysis of medical disorders. The key areas of interest in this examination are the epiphyseal ossification centres and the carpal bones. Human BAA models, although necessary, are time-consuming and prone to mistakes, emphasising the need for a more efficient computerised BAA model. This study introduces ODL-BAAM, a novel Deep Learning-based Bone Age Assessment Model, aimed at enhancing efficiency and accuracy in medical image analysis. Given the critical role of Bone Age Assessment (BAA) in predicting medical disorders, particularly based on hand X-ray images, there’s a pressing need for more streamlined and reliable computerized BAA models. Leveraging Deep Learning methodologies over classical Machine Learning approaches, ODL-BAAM offers a comprehensive solution. The model begins with preprocessing steps to standardize and normalize X-ray data, crucial for managing the inherent complexities of such images. By integrating Faster RCNN with MobileNet, feature extraction becomes more effective, while the Tunicate Swarm Algorithm optimizes model hyperparameters. Age determination is facilitated through SoftMax layers applied to feature vectors. Through extensive simulation studies, ODL-BAAM demonstrates promising results, showcasing heightened sensitivity, specificity, and overall accuracy compared to existing BAA models. With a remarkable 96.5% accuracy rate, ODL-BAAM represents a significant advancement in the realm of computerized BAA, effectively addressing prior limitations and setting a new standard for medical image analysis.