Transfer Learning Models for Automated Bone Age Prediction: Evaluating Efficiency and Clinical Applicability

Bone age assessment is critical in pediatric healthcare for diagnosing growth disorders and planning treatments. Traditional methods, like the Tanner-Whitehouse and Greulich-Pyle techniques, are manual, subjective, and time-consuming. Recent advancements in deep learning, particularly convolutional neural networks (CNNs), offer a promising alternative by automating bone age prediction. This research investigates the effectiveness of several pretrained deep learning models, including VGG19, ResNet50V2, DenseNet201, Xception, and InceptionResNetV2, in predicting bone age from hand and wrist radiographs. Utilizing the RSNA 2017 Bone Age Prediction dataset, we explore transfer learning, gender integration, and model performance across various demographics. Our findings show that these models can achieve accuracies comparable to human experts, with the Xception and InceptionResNetV2 models demonstrating superior performance. This study highlights the potential of AI-driven bone age prediction in improving diagnostic accuracy, reducing evaluation time, and enabling personalized treatment in clinical settings.