Development of algorithms for automatic landmark detection for cephalometric system

Abstract

The diagnostic value of cephalometric analysis depends on the accurate and reproducible identification of clearly defined landmarks on cephalometric radiographs. Manual landmark detection is a very monotonous and time intensive process. It takes around thirty minutes for an orthodontist to analyze one cephalogram using his training and experience. Computer technology is having an increasing impact on the practice of orthodontics. The demand of computerized cephalometric landmark detection, which is a sub field of computer vision, is rising due to the advances, affordability and increased use of digital images in the fields of medicine. It will considerably increase the efficiency of orthodontists if the landmark detection operation could be performed automatically using a personal computer rather than manually. Such a system will reduce the subjectivity involved due to intra and inter examiner variability and save valuable clinical time. Automatic cephalometric analysis has been a subject of research for the past 25 years and attempts to automate the localization of landmarks have been attempted by many researchers with varying degree of success. Although some significant work has been done, there are still some problems, which need to be solved. Previous attempts at locating landmarks have had a limited success. More sophisticated techniques are still required to improve the accuracy of the landmark detection results. The principal difficulty in identifying the landmarks are the poor diagnostic quality of a cephalometric radiograph, large variations in biological features of the patient s skull structure (hard and soft tissues), abnormalities, overlapped complex structures and areas with subtle changes in grayscale. The main aim of our research is to develop an algorithm or a combination of algorithms, which can efficiently and effectively enhance the cephalograms by improving the contrast and reducing the noise and find all the landmarks automatically, has greater ability to cope with a large variety