Deep Learning Applied to Bone Age Analysis

Bone age studies help estimate the maturity of a child's skeletal system, usually done by taking a single X-ray of the left wrist, hand, and fingers. The bones on the X-ray image are compared with X-rays images in a standard atlas of bone development, which is based on data from large numbers of other children of the same gender and age. A difference between a child's bone age and his or her chronological age might indicate a growth problem. RAD Sherpa's AI algorithm analyzes imaging scans from various modalities and predictively diagnoses them for a number of different clinical findings. This can help radiologists provide an accurate and timely diagnosis of conditions that delay or accelerate physical growth and development such as growth hormone deficiency, hypothyroidism, precocious puberty, and adrenal gland disorders genetic growth disorders, such as Turner syndrome (TS) and orthopedic or orthodontic problems in which the timing and type of treatment (surgery, bracing, etc.) must be guided by the child's predicted growth.

The workflow of our model includes preprocessing, inference, and model evaluation. The preprocessing procedure of the model includes extraction, normalization, and rescaling. The inference model uses convolutional neural network to learn complex relationships between extracted features from the X-ray images and bone-age. The model evaluation process shows our model has high accuracy on testing dataset.