Development and evaluation of an artificial intelligence system to reduce exomass-related metal artefacts in cone-beam computed tomography

• Grant number: 22/12994-0
• Support Opportunities: Scholarships abroad - Research
• Start date: August 01, 2023
• End date: July 31, 2024
• Field of knowledge: Health Sciences - Dentistry - Dental Radiology
• Principal Investigator: Matheus Lima de Oliveira
• Grantee: Matheus Lima de Oliveira
• Host Investigator: Michael M. Bornstein
• Host Institution: Faculdade de Odontologia de Piracicaba (FOP). Universidade Estadual de Campinas (UNICAMP). Piracicaba , SP, Brazil
• Institution abroad: University of Basel, Switzerland

Abstract

The presence of metallic materials in the exomass in cone beam computed tomography (CBCT) is a frequent reality in dentistry and can lead to artifacts that negatively affect the overall image quality. Objective: Considering the great development of artificial intelligence (AI) systems with the potential to correct artifacts, the aim of this study is to develop and evaluate an AI system aimed at reducing artifacts arising from metallic materials in the exomass in CBCT in a pre-clinical model. Methodology: Thirty fresh swine mandibles will have six polypropylene tubes filled with a homogeneous hyperdense solution fixed on the buccal and lingual surfaces and will be scanned using three CBCT scanners. Then, metallic objects will be inserted into the jaws to induce the formation of artifacts arising from metallic materials in the exomass and additional CBCT acquisitions will be obtained. Then, a deep learning AI system based on the MedGAN model - Medical Image Translation using Generative Adversarial Networks -will be trained, validated, and tested with the objective of reducing artifacts ARISING from metallic materials in the exomass in CBCT and then applied to all CBCT images with metals in the exomass. Subsequently, the images without metal will be compared to those with metal before and after the application of the AI model by means of objective and subjective analysis. For the objective analysis, if the data distribution is normal, the multivariate analysis of variance (ANOVA) of repeated measures will be used followed by the post-hoc Sidak's test, otherwise, the Friedman test will be used followed by the post-hoc Tukey test at a significance level of 5% (± = 0.05). Regarding the subjective analysis, the Friedman test will be used with Tukey's post hoc test at a significance level of 5% (± = 0.05). Expected results: It is expected with this project to develop an AI system capable of reducing artifacts arising from metallic materials in the exomass in CBCT, in order to maintain the objective and subjective quality of CBCT exams with metals in the exomass similar to those without metals in the exomass, regardless of the number of metallic objects, which will be very beneficial both for the patient and for companies that develop new solutions based on CBCT images. Furthermore, it is expected that the performance of the developed AI system will not be dependent on the type of CBCT device. (AU)