Human age prediction from bone X-ray Scan images by using advanced Deep CNN

Abstract

The pace of innovation has enabled the development of computerized systems that do have a variety of uses in the medical field, including pediatrics. Fully automated bone age calculation from left-hand X-ray scanning has been one of the applications, which enables radiologists and pediatricians to make decisions about the skeletal bone growth status of youngsters. However, one of the most difficult aspects of building a computerized system is determining the best technique for getting more accurate estimation, especially in dealing with massive volumes of data. The earlier conventional techniques use hand skeletal bone scans and depend upon an optical examination of the bone picture scans. Conventional approaches like these have a lot of variation within observers, are exceedingly time-consuming, and depend upon an experienced domain expert who assesses skeletal bone age by taking an age annotated bone scan as a reference. Various automatic systems have been employed to reduce such variation. Neural network algorithms such as Convolution neural network (CNN) algorithms and their altered versions are employed to address this, resulting in much more accurate outcomes in significantly less time. Efficient Net is one such architecture that we use in this study. This design substantially replaces the usual Convolution architectures with the considerably more efficient Compound scaled architectures which optimize both accuracy and efficiency and result in drastically decreases in time it takes to generate a trained model. In this article, we have used a collection of left-hand scans provided by the RSNA. The testing results demonstrate a 6.03 month disparity in age in the bone estimation process.

Ähnliche Arbeiten