Original title: Fully Automated Deep Learning System for Bone Age Assessment
Bone age is an estimate of how mature a person’s bones look, based on an X-ray. The system finds and prepares the useful area of the image, then uses a trained image-reading program to estimate bone age. On held-out test images, most estimates were within one year of the reference assessment. It was also deployed in a clinical setting, where it completed an interpretation in under two seconds.
Imagine sorting a pile of laundry by how worn each item looks. Some pieces are easy to judge, but others have only a few clues, so you must look closely at the right spots. In this paper, the laundry is an X-ray, and the system estimates bone maturity by finding and examining the useful parts, much as a person would.
Doctors use bone age to help evaluate endocrine and metabolic disorders, but the usual assessment is tedious. This automated system could give clinicians a faster decision aid, while its attention maps show which parts of an X-ray it uses.
The abstract reports test results for the study’s female and male cohorts, but does not describe the test-set size or how well the system performs in other populations or clinical settings. It presents the system as decision support; it does not establish that it can replace clinicians or improve patient outcomes.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
Fully Automated Deep Learning System for Bone Age Assessment — Hyunkwang Lee, Shahein Tajmir, Jenny Lee et al. — Journal of Imaging Informatics in Medicine — doi:10.1007/s10278-017-9955-8
Abstract licence: CC BY
Explanation generated by Synthesis from the abstract — it is not the original paper.