Fine-grained precise-bone age assessment by integrating prior knowledge and recursive feature pyramid network

Abstract

Bone age assessment (BAA) evaluates individual skeletal maturity by comparing the characteristics of skeletal development to the standard in a specific population. The X-ray image examination for bone age is tedious and subjective, and it requires high professional skills. Therefore, AI techniques are desired to innovate and improve BAA methods. Most of the BAA method use the whole X-ray image in an end-to-end model directly. Such whole-image-based approaches fail to characterize local changes and provide limited aid for diagnosis and understanding disease progress. To address these issues, we collected and curated a dataset of 2129 cases for the study of BAA with fine-grained skeletal maturity level labels of the 13 ROIs in hand bone based on the expert knowledge from TW method. We designed a four-stage automatic BAA model based on recursive feature pyramid network. Firstly, the palm region was segmented using U-Net, followed by the extraction of multi-target ROIs of hand bone using a recursive feature pyramid network. Given the extracted ROIs, we employed a transfer learning model with attention mechanism to predict the skeletal maturity level of each ROI. Finally, the bone age is assessed based on the percentile curve of bone maturity. The proposed BAA model can automate the BAA. In addition, it provides the detection result of the 13 ROIs and their ROI-level skeletal maturity. The MAE can reach 0.61 years on the dataset with the labeling precision of one year. All the data and annotations used in this paper are released publicly.

Automated Bone Age Assessment: A New Three-Stage Assessment Method from Coarse to Fine

“…Most of the studies used Faster R-CNN in the extraction of ROIs, and these studies all had the phenomenon of missing ROI detection [ 36 , 37 ]. In the second stage, CAP was used to perform 13 bone grade staging, and the average accuracy was 86.93%, which was much higher than the 63.15% accuracy achieved by the recursive feature pyramid network proposed by Jia et al [ 11 ]. In the third stage, the TW3 method was used to estimate bone age, which is more accurate than building a regression model of bone age.…”

“…Alexander et al [ 9 ] won first place in the Pediatric Bone Age Machine Learning Challenge organized by the North American Radiological Society (RSNA) for the Inception V3 [ 10 ] architecture for pixel information in tandem with gender information, with an additional dense layer after concatenation to enable the network to learn the relationship between pixel and gender information, obtaining a mean absolute error (MAE) of 4.27–4.5 months on the RSNA dataset. Yang et al [ 11 ] proposed to perform fine-grained precision assessment by integrating prior knowledge and a pyramid network of recursive features, and the MAE of 1–18 years old was 7.32 months. Liu et al [ 12 ] constructed a two-stage automated assessment method that cascaded coarse-to-fine hands and an integrated convolutional neural network cascade based on stacking, and 5.42/6.58 months of MAE for both males and females was achieved on the RSNA dataset.…”