Copyright© 2020, Visiana.


The purpose of the Bone Age Rating Test is to assess your ability to manually rate bone age. We will present you with 10 radiographs (i.e. images), each one with an unknown bone age. It’s your task to rate the bone age with the help of the Greulich Pyle (GP) Atlas. When you have rated all 10 images, the app will compute the mean absolute deviation (MAD) of your ratings from reference ratings. The reference ratings are formed as the average of the ratings of forty-six radiologists.


On the left side of the screen, you will be shown 10 images, one at a time. On the right side, you can browse the GP Altas images corresponding to the gender of the current image. You can navigate through different images in the GP Altas by using < and > on the keyboard, or by clicking the left and right arrows on the screen.

Once you have decided on a rating, you type the bone age rating in the box as a decimal number (e.g. 10.5) and press Enter. Note that 10.5 years means “10 and a half year”, not ten years and five months.

When you have entered your rating for a given image, you can press the button “Show/hide reference rating and BoneXpert Result”, which will show the reference bone age. It will also show BoneXpert’s automatic analysis as annotations on the image, including the bone age of each bone and BoneXpert’s overall rating.

You navigate between the 10 images by pressing the Previous and Next buttons. You can review your ratings of previous images, but you cannot modify them.

Once you have rated all 10 images, you can press “Show how well I rated these images”, and the app will compute the Mean Absolute Deviation (MAD) from the reference (in months), and the Root Mean Square Error (RMSE) relative to the reference (in years).

We recommend to use a monitor with a resolution of at least 1920 x 1200. To prevent loss of data, please do not press ”F5”, ”Refresh page” or the Backspace button before completing.




The images originate from the RSNA Bone Age Challenge conducted in 2017.

The reference ratings are formed as the average over 46 ratings done by Dutch radiologists during the 2018 Hands-on AI Day in Nijmegen