Fully Automatic Classification of Skeletal Malformations

The output of the FALA system (i.e. the fully automatically detected cephalometric landmark positions) was used to automatically determine the presence or absence of skeletal malformations as defined by the clinical parameters in Table 1. For this purpose, several measurements between landmark positions were automatically calculated, and every subject was automatically classified into one of three to five groups (C1, C2, C3, C4, and C5) for each of the eight clinical parameters. All calculations were done using custom code developed in Python.

To evaluate the performance of the fully automatic classifications, we compared them to the classifications obtained from the manual ground truth annotations.