An Enhanced Pediatric Bone Age Estimation Model Based on CNN and Metaheuristic Group Learning Algorithm

Abstract

Pediatric Bone Age estimation is done in the medical field to diagnose the abnormal growth and development in the children. In the recent few years, machine learning algorithms are gained popularity in the field due to better accuracy and overcome human error. In this paper, we have designed an enhanced pediatric bone age estimation model using the convolutional neural network (CNN) and metaheuristic group learning algorithm (GLA) to enhance the assessment accuracy of the model. The group learning algorithm is a metaheuristic algorithm which searches the best region of interest (ROI) in the database images based on the objective function. In this research, k-mean clustering algorithm is taken as the objective function. After that, convolutional neural network is trained and tested with the ROI database images. The proposed model is evaluated on RSNA 2017 database images which contains pediatric bone images. Further, its evaluation is done using three performance metrics such as MAE, RMSE, and RMSPE. The result shows that the proposed model provides minimum value of these performance metrics over the existing models.