Improving Bone Age Assessment with Inception-V3 and Faster R-CNN

The assessment of hand bone age serves as an accurate indicator of an individual’s developmental stage and level of maturity. The results of bone age assessment in teenagers can serve as a theoretical foundation for understanding their growth and development and predicting their height. Nevertheless, using an artificial intelligence (AI) model for bone age prediction has several constraints and difficulties, including requiring large-scale and varied training datasets and careful validation and testing procedures. The extracted specific characteristics are then integrated with the overall characteristics of the entire picture. Additional studies are required to tackle these obstacles and constraints to guarantee the confidence and precision of AI models for bone age prediction in practical scenarios. This study proposes a computer-aided diagnosis approach based on deep learning—the resnet50 model used for the Faster Region-Convolutional Neural Network mask (Faster R-CNN mask) to select regions of interest and inception v3 for regression bone age. The method utilises a dataset of hand bone radiography images provided by the Radiological Society of North America (RSNA) as the main data for research. This model can automatically identify and extract essential areas based on Tanner Whitehouse (TW3) and specific characteristics in hand bone images for bone age regression. The inceptionv3 model had a mean absolute error (MAE) of 7.5 months for males and 8.3 months for females. These models can enhance bone age prediction’s precision, uniformity and effectiveness.