Detecção e análise automática do deslocamento vertical do osso hioide em sua amplitude máxima em exames de ultrassonografia da deglutição

• Ano de defesa: 2023
• Autor(a) principal: Silva, Arthur José Lopes da
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal da Paraíba; Brasil; Informática; Programa de Pós-Graduação em Informática; UFPB
• Palavras-chave em Português: Inteligência artificial; Ultrassonografia; Detecção de objetos; Osso hioide; Aprendizado profundo; YOLO; Hyoid bone; Ultrasonography; Object detection; Artificial intelligence; Deep learning; CNPQ::CIENCIAS EXATAS E DA TERRA::CIENCIA DA COMPUTACAO
• Link de acesso: https://repositorio.ufpb.br/jspui/handle/123456789/33071

Located below the mandible, the hyoid bone is present in the upper part of the neck and participates in essential human functions such as speech, swallowing and breathing. When swallowing, the hyoid bone simultaneously lifts and moves forward.The study of movement is of great importance, because if it is compromised and not performed correctly, it can lead to limitations for the individual and help to cause illnesses that compromise their routine.This study is usually based on an analysis of bone displacement, where healthcare professionals carry out the analysis without tools to automate this process. Because it is a small bone (about 42.04 millimeters) and it is moving, the quality of test results or diagnoses for the patient can be compromised. With this in mind, this study aims to evaluate the possibility of detecting and calculating the vertical displacement of the hyoid bone with the aid of deep learning object detection methods, using information from ultrasound scans of young adult patients swallowing water, saliva and pasty liquid, and, from this, develop a tool capable of assisting healthcare professionals. Training and then detection were carried out using YOLO version 7-tiny. Once detected, the vertical displacement of the hyoid bone was calculated using data from the ultrasound scan and information resulting from the detection, focusing on the maximum amplitude, i.e. from rest to the maximum peak of the displacement.After that, all the results were shown and evaluated according to precision and recall, obtaining values of 0.95 and 0.93 respectively, and also according to the displacement calculation, where they were compared to other results in the literature, of three comparisons made, two proved to be closer, in the swallowing of pasty liquid (12,22mm±3,63) x Chi-Fishman e Sonies (2002) (8,50mm±3,31) and liquid (10,33mm±5,77) x Yamazaki et al. (2017) (10,88mm±1,47). This study showed good accuracy in detecting the movement of the hyoid bone and an approximation in relation to the calculation of vertical displacement compared to the literature.