A deep learning-based computer-aided diagnosis method of X-ray images for bone age assessment

Abstract

Bone-age assessment from hand-wrist X-ray images is important for diagnosing pediatric growth disorders and supporting individualized treatment decisions, but conventional clinical assessment depends strongly on clinician experience. This study proposes a deep learning-based computer-aided diagnosis method for bone-age assessment. It first localizes informative image regions through an unsupervised learning strategy and image-processing pipeline, then uses a pretrained image backbone with a prediction head to estimate bone age. Gender information is embedded into the learned feature vector in line with clinical practice. Experiments on the public RSNA dataset and an additional dataset show that the approach can reduce assessment error, with the best MobileNetV3-based model achieving mean absolute errors of 6.2 months and 5.1 months on the two datasets.

Recommended citation: Shaowei Li, Bowen Liu, Shulian Li, Xinyu Zhu, Yang Yan, Dongxu Zhang. (2022). "A deep learning-based computer-aided diagnosis method of X-ray images for bone age assessment." Complex & Intelligent Systems, 8(3), 1929-1939. DOI: 10.1007/s40747-021-00376-z.
Publisher page