Evaluation of texture maps performance in improving simulated bone defects detection in condyles by cone beam computed tomography

• Grant number: 19/25977-3
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Effective date (Start): March 01, 2020
• Effective date (End): December 31, 2020
• Field of knowledge: Health Sciences - Dentistry - Dental Radiology
• Principal Investigator: Sérgio Lúcio Pereira de Castro Lopes
• Grantee: Larissa Ferreira Ribeiro
• Host Institution: Instituto de Ciência e Tecnologia (ICT). Universidade Estadual Paulista (UNESP). Campus de São José dos Campos. São José dos Campos , SP, Brazil

Abstract

The aim of this study is to evaluate the performance of texture maps (TM) images generated by texture analysis (TA) as a tool to improve the identification of simulated condyle (Co) bone defects in cone beam computed tomography (CBCT). Additionally, the comparison of the image resolution variation in defect detection will be performed. 03 bone defects will be performed, at a time, in 08 macerated human Co, with spherical diamond drills (0.9; 1.0; 1.2 mm), randomly arranged in two previously drawn lines, joining the lateral/medial (L and M) and anterior/posterior (A and P) poles, passing through a central point (C). These points - L, M, C, A and P - will be marked by hyperdense material, in order to orientate itself in the accomplishment of the 02 slices to be analyzed, determining 04 lines, in each one will be made only 01 defect, and, one of these will not receive the defect. Images will be acquired on an i-CAT Next Generation tomograph, in two resolutions: voxel of 0.20 and 0.40mm. The images in DICOM format will be exported to the OnDeman3D software, where standard and fixed filters and window values (WW and WL) will be applied, drawing two lines with the "curve" tool, generating 02 cuts for each Co: one LCM (paracoronal) and other PCA (parasagital), with the hyperdense point markings as a guide. The cuts will be exported in bitmap format (.bmp) to MaZda 4.60 software, in which a region of interest (ROI) will be selected, encompassing the Co and through TA calculation, images corresponding to different TM will be generated, according to the previously selected parameters. All images, without TM and TM, will be encoded and assembled randomly and individually in Power Point software (Microsoft). Three previously calibrated evaluators will be instructed to analyze them for the presence or absence of defects between each hyperdense point of the images, using scores: absent lesion; present lesion and inconclusive. The agreement of diagnoses of presence or absence of Co defects obtained with images with and without texture map will be measured with the standard template and will be analyzed by Kappa test. The percentages of correct diagnoses in images with and without TM and with different voxel will be compared by the Q-square test. Finally, the influence of defect sizes on diagnostic accuracy will be compared at each resolution (voxel size) and with and without texture maps by the Q-square test.