Landmark Detection using a Domain Independent Technique in Cephalograms

The main purpose of this paper is about the relative merits of automated feature detection protocols on cephalometricrelated digital radiology images. While the domain dependent techniques such as handcrafted masks are shown to be more accurate than pixel-based domain independent methods, the process for determining these shapes was unwieldy with respect to time taken and subjectivity. This paper discusses the use of a Pulse-Coupled Neural Network (PCNN) technique as a segmentation technique for parts of a radiology image. The results of this investigation showed a significant improvement in detection performance when compared to the handcrafted shapes when applied to several cephalometric landmarks. This encouraging outcome augurs well for a near-automated feature detection protocol for such medicalrelated radiology images.