Banca de DEFESA: Mayra Cristina Ribeiro de Lima

Uma banca de DEFESA de MESTRADO foi cadastrada pelo programa.
STUDENT : Mayra Cristina Ribeiro de Lima
DATE: 29/05/2024
TIME: 14:00
LOCAL: Plataforma Teams
TITLE:

Use of neural transformer network in diagnostic imaging in dentistry: a scoping review

KEY WORDS:

“scoping review; neural transformer network; diagnostic imaging; dentistry”

PAGES: 100
BIG AREA: Ciências da Saúde
AREA: Odontologia
SUMMARY:

“Introduction: Introduction: Artificial intelligence (AI) tools are increasingly being used in dentistry, with the potential to revolutionize clinical practice. Currently, convolutional neural networks (CNNs) are widely used as computer vision algorithms for detecting and classifying diseases in images, such as computed tomography and panoramic radiography. They have also been applied to segment structures in these images and in predictive models. Recently, however, transformer networks, which are widely used in natural language processing, have also been applied to image analysis tasks, with the potential to outperform CNNs. Objective: The aim of this scoping review is to assess the application of the Transformer neural network for diagnosing dental and maxillofacial abnormalities in the most used dental imaging exams. Methodology: A search was conducted in the LILACS, Embase, PubMed/MEDLINE, SCOPUS, Web of Science, and IEEE Xplore databases. Additional research was also conducted in the grey literature, including Google Scholar and Proquest Dissertation & These. The search terms included free terms, (MeSH) descriptors, as well as controlled terms available in the Health Sciences Descriptors (DeCS). The Mendeley and Rayyan applications were used to select and manage the references. The Mendeley and Rayyan applications were used to select and manage the references. The inclusion criteria were: examinations of patients with hard tissue alterations in the jaws and face, studies that used Neural Network Transformer for Diagnosis, and studies that dealt with diagnostic imaging in Dentistry. Clinical trials, observational studies, and systematic reviews were eligible for the study. Results: A total of 5,182 studies were found, and 11 studies were included. Five studies used Panoramic Radiography (PR), four studies used Computed Tomography, two studies used Periapical Radiography, and one study used RP and Bitewing. All studies used Transformer, using Swin Transformer, U-net, VSP Transformer, SWin Unet, FlowgateUnet, Vision Transformer, CTA U-net and M-TransNet. The reference standard included expert evaluation through manual annotation. The total number of the dataset was 39,221, and it was not possible to quantify the total data used for testing, training and validation due to the lack of information in some studies. The studies confirmed, through the results of the metrics used, that the use of the neural network discussed here was favorable in its different applications. Conclusion: The use of the Transformer neural network is a tool with excellent results for application in diagnosis through imaging exams in Dentistry. Its applications are variable, and it can be used not only in the segmentation of teeth, but also to facilitate the identification of important lesions in the hard tissues of the jaws and face, becoming a facilitator in the diagnostic area of dental radiology.”

COMMITTEE MEMBERS:
Presidente - 3437282 - ANDRE FERREIRA LEITE
Interna - 1889805 - CRISTINE MIRON STEFANI
Externa à Instituição - FABIANA TOLENTINO DE ALMEIDA MARQUES - UniAlberta
Interna - 2315081 - NILCE SANTOS DE MELO