Deep Learning-Based Bone Age Estimation Using X-Ray Hand Images

This Skeletal bone age estimation measures the development level and maturity of individuals. Investigating genetic, endocrinological, and growth abnormalities in children is a clinical practice. It also helps pediatrics to predict puberty entrance and growth, and identify genetic disorders. The Tanner-Whitehouse (TW) technique or the Greulich and Pyle (G&P) approach are often used for the radiological evaluation of the left hand. Both clinical approaches, however, have several drawbacks. In this research, we present a deep learning technique to automatically determine the age of skeletal bones to enhance and accelerate bone age estimates. This research also provides a comparison between manually crafted and deep learned features. The dataset utilized consisted of 12811 X-ray hand scans of people ranging in age from birth to 19 years old. The results showed a slight deviation between the manual and automatic evaluation of bone age.