An Enhanced Approach for Bone-Age Assessment Using EfficientNet-B3 Architecture

The developments in artificial intelligence (AI), especially deep learning methods such as convolutional neural networks (CNNs), EfficientNet-B3 for automated bone age assessment are covered in this research. These methods examine hand radiographs and estimate bone age by identifying and categorizing bone maturity stages. Bone age estimation is a significant task in medical imaging for the diagnosis of developmental anomalies and growth disorders. In this study, a deep learning-based solution using EfficientNet-B3 for automatic bone age estimation is suggested. The method involves three major steps: feature extraction, preprocessing, and prediction. During the preprocessing phase, bone image datasets are normalized by frames, resized, and denoised to improve image quality and uniformity. These optimized images are input into the EfficientNet-3 model, which is a state-of-theart convolutional neural network with optimal depth and width scaling, which allows for stable feature extraction. The EfficientNet-B3 architecture has several convolutional and mobile inverted bottleneck convolution (MBConv) layers, logically organized into blocks for hierarchical learning of features. The proposed model achieves more than 90% of accuracy and it outperform all other existing approaches.