- Hand outline segmentation
- Hand outline verification
- Bone segmentation
- Feature extraction
- Bone segmentation verification
ASMA performs age prediction in two different ways the first is to recreate Tanner-Whitehouse stages for each bone and the second is to regress onto chronological age.
Prediction of Tanner-Whitehouse Stages
In  we investigate the use of the extracted features to classify Tanner-Whitehouse stages. One of the reasons that previous automated bone age assessment systems have not gained widespread acceptance is due to a lack of transparency (which features cause an assessment to be given). We believe that by separating the feature extraction from the classification, the system becomes more transparent and explicable to clinicians. An example C4.5 tree built on 100 instances of the distal phalange is shown in Figure 1. This gives us rules to understand why an assessment has been given. So, for example, Stage B is classified by the rule if no epiphysis and phalanx small then B. This work is extended in the paper "Automated Skeletal Maturity Assessment (ASMA): Modelling Bone Age from Hand Radiographs", which is currently under review.
Please send any enquiries to Luke Davis
Regression onto Chronological Age
The paper Automated Skeletal Maturity Assessment (ASMA): Modelling Bone Age from Hand Radiographs is currently under review.