An Automated System for Identification of Skeletal Maturity using Convolutional Neural Networks Based Mechanism

Abstract

This paper puts forward a proposition of automated skeletal recognition system that takes an input of left hand-wrist-fingers radiograph and give us an output of the bone age prediction. This system is more reliable, if is successful and time-saving than those laborious, fallible and time-consuming manual diagnostic methods. Here, a Faster R-CNN takes the input of left-hand radiograph giving the detected DRU region from left-hand radiograph. This output is given as an input to a properly trained CNN model. The experiment section provides us with the details regarding the experiments conducted on 1101 radiographs of left hand and wrist datasets and accuracy of model when different optimization algorithms and training sample amounts were utilized. Finally, this proposed system achieves 92% (radius) and 90% (ulna) classification accuracy after the parameter optimization.

Bone age assessment based on deep learning architecture

“…Traditional machine learning and image processing approaches were used in automated BAA procedures. Approaches based on CNNs have been only recently introduced to BAA [14]. Instead of extracting information from specific locations based on clinical expertise, these approaches frequently encode visual features directly [15].…”