Research into hand outline verification Research into hand outline verification

Hand Outline Verification

An incorrect outline can be caused by many different potential problems e.g. a foreign object overlapping the hand or a problem with the outlining algorithm. Since our primary goal is to create a fully automated method for bone age assessment, we need a way to verify if the outline is correct. In [1] and [2]  we investigated the use of a large set of different classifiers on different representations of hand outlines using:

  • Principal Component Transforms;
  • Fast Fourier Transform; and
  • Feature Extraction (No of Tips, No of Webs etc.).

The Random Forest classifier trained on principal components performed best, with an accuracy of ~95%.

Hand 1 Hand 2