Author, Institution: Mantas Vaitiekūnas, Kaunas University of Technology
Science area, field of science: Technological Sciences, Electrical and Electronics Engineering, T001
Scientific Supervisor: Assoc. Prof. Dr. Darius Jegelevičius (Kaunas University of Technology, Technological Sciences, Electrical and Electronics Engineering, T001)
Dissertation Defence Board of Electrical and Electronics Engineering Science Field:
Prof. Dr. Arminas Ragauskas (Kaunas University of Technology, Technological Sciences, Electrical and Electronics Engineering, T001) – chairman
Assoc. Prof. Dr. Gundega Jakobsone (Ryga Stradins University, Latvia, Medical and Health Sciences, Odontology, M002)
Prof. Dr. Elena Jasiūnienė (Kaunas University of Technology, Technological Sciences, Electrical and Electronics Engineering, T001)
Prof. Dr. Tomas Linkevičius (Vilnius University, Medical and Health Sciences, Odontology, M002)
prof. dr. Darius Viržonis (Kaunas University of Technology, Technological Sciences, Electrical and Electronics Engineering, T001)
The dissertation defence was held remotely.
The doctoral dissertation is available at the library of Kaunas University of Technology (K. Donelaičio g. 20, Kaunas).
Annotation:
Computer-aided three dimensional (3D) processing of medical images is becoming an important way to reduce the time of pathology diagnosis, to confirm an accurate treatment plan, or to evaluate the postoperative follow-up. 3D segmentation is one of the most important steps in medical image processing. With the increasing amount of medical data through the use of various imaging modalities, the time to segment selected volume of an anatomical region should be reduced. However, segmentation is usually based on manual segmentation, where the final result depends on the experience of the user. For this reason, time and experience are the most important factors affecting the quality of segmentation. The same reason is relevant in dentistry, especially, in oral and maxillofacial surgery, orthognathic surgery and orthodontics, where the accuracy of facial bones segmentation is important to ensure the accurate diagnosis of facial asymmetry, to prepare an accurate virtual surgical plan, or to successfully follow the patient’s condition during or after treatment.
The aim of the research – to develop and investigate the automatic method for facial bones segmentation in cone beam computed tomography datasets (CBCT).
In this doctoral thesis, an automatic method for facial bones segmentation in the CBCT dataset was developed. The segmentation of facial bones is automated by using locally assessed distribution of voxel intensities. Three basic elements are used to implement the automatic segmentation: a 3D sliding window, a histogram filter, and thresholding based on Otsu’s method. The proposed automatic method for the facial bones segmentation was investigated and evaluated against the results of reference segmentations performed by an experienced surgeon. The calculated performance five metrics showed high clinical accuracy of segmentation. The proposed automatic method could be applied in the clinical practice (especially in orthognathic surgery). The implementation of the proposed method is simple and fast; it does not require access to a computer with high computing power.
