Assessment of skeletal maturity is an important instrumentality in managing human’s growth problems, especially for assisting the physician decides the best treatment on various skeletal disorders. 

But suffering from limited data and large anatomical variations among different subjects, this task is very challenging when using machine learning method.  

Recently, researchers from the Shenzhen Institutes of Advanced Technology (SIAT) of the Chinese Academy of Sciences and University of Hong Kong introduced an ensemble-based deep learning pipeline to automatically assess the distal radius and ulna (DRU) maturity from left-hand radiographs. The study was published in IEEE Transactions on Systems, Man, and Cybernetics: Systems.  

The dense connection mechanism was combined with the ensemble model to improve the stability and accuracy of a skeletal maturity assessment system. The concept of densely connected mechanism was introduced in the proposed network architecture to reuse features and prevented gradient disappearance.  

Therefore, the model acquired two convincing advantages: First, the model preserved the maximum information flow and has a much faster convergence rate; second, the model avoided over-fitting even if training with limited data. 

The team has done significant quantity of experiments to find the influence of different hyper-parameters on the final result. By applying dense connectivity, the model could achieve better feature transfer efficiency with fewer parameters.

Besides, even training with limited data, the proposed system has less chance to encounter the over-fitting problem, and it was much easier to converge and has greater optimization efficiency.  

"Our results demonstrated the superiority of the proposed model compare with using other models for skeletal maturity stage classification. It can be deployed in the clinical environment to help orthopedist to identify the skeletal maturity via radius and ulna automatically and objectively," said Dr. WANG Shuqiang, first author of the study.  

Figure. Framework of our skeletal maturity assessment system. (Image by WANG Shuqiang) 

Media Contact: 
ZHANG Xiaomin 
Email: xm.zhang@siat.ac.cn