Development of a framework for autonomous 3-dimensional craniofacial anthropometric data acquisition / Salina Mohd Asi

Salina , Mohd Asi (2018) Development of a framework for autonomous 3-dimensional craniofacial anthropometric data acquisition / Salina Mohd Asi. PhD thesis, University of Malaya.

PDF (Thesis PhD)
Download (3397Kb) | Preview


    Craniofacial anthropometry (CFA) is the study that measures the human face and head to supports visual description. It describes the craniofacial complex in exact measurements instead of subjective assessment. Researchers and medical professionals use craniofacial anthropometry as a tool to study facial morphology. Many abnormal faces had been quantified against the normal face measurements. This helps the medical professional in understanding various facial syndromes and enable them to plan surgeries for these patients. Conventionally, the measurement is taken directly on the patient's face and human errors could be introduced, besides the need for a well-trained examiner to perform it. Methods of taking facial measurement have evolved and are influenced by the development of the computer hardware and software, and imaging technologies. Nowadays, 3-D images are obtained using stereo-photogrammetry camera system but still needs human input and is susceptible to the human errors. The main objective of this research is to propose for an automated CFA system with expert knowledge embed in it, so that it can simulate the capability of a human expert. The objectives of the project are 1) to determine inter- and intra-examiner errors, 2) to evaluate the reliability and validity of the VECTRA-3D, for further use, 3) to develop an expert system, the craniofacial anthropometry expert system (CFA-ES), 4) to evaluate the reproducibility of CFA-ES anthropometric landmarks localization and the reliability and validity of the CFA-ES against Vectra-3D. Methods: 19 measurements from 30 respondents were directly measured by a well-trained dentist. Then, 100 sets of 19 CFA measurements were acquired indirectly using Vectra-3D from 100 3D facial images. Of these images 30 would be the testing images and 70 would be the training images for the CFA-ES development. Then, the 3D images were pre-processed into 2.5D images and intelligent shape model for normal CFA was constructed. Correlation regression was used to build a mathematical equation to convert pixel unit into millimetre. CFA-ES produced 18 measurements except the tip of nasal protrusion. Paired t-test was performed on 18 measurements to evaluate the accuracy of CFA-ES and ICC was used to analysed the reliability of the measurement. Results: For Vectra-3D, 3 measurements (ex-ex, ex-enL, ex-enR) were not clinically significant where systemic bias was observed. For reliability, 1 measurement was not reproducible, 3 were moderate while 15 were good to excellent. For CFA-ES, all landmark positions were reproducible. The reliability for the positions were good to excellent. For the accuracies of the measurements, a systemic bias was observed at orbital and orolabial. Seven measurements were significant, however, the mean differences were less than 2 mm.

    Item Type: Thesis (PhD)
    Additional Information: Thesis (PhD) – Faculty of Dentistry, University of Malaya, 2018.
    Uncontrolled Keywords: Craniofacial anthropometry; Reliability; Validity; Active shape model; Expert system
    Subjects: R Medicine > RK Dentistry
    Divisions: Faculty of Dentistry
    Depositing User: Mr Mohd Safri Tahir
    Date Deposited: 20 Jun 2019 07:07
    Last Modified: 23 Jun 2021 01:17

    Actions (For repository staff only : Login required)

    View Item