Please use this identifier to cite or link to this item:
http://hdl.handle.net/10400.5/96414
Full metadata record
DC Field | Value | Language |
---|---|---|
dc.contributor.advisor | Pereira, Cristiana Maria Palmela | - |
dc.contributor.advisor | Santos, Rui | - |
dc.contributor.author | Correia, Mariana Delicado Dias Feijão | - |
dc.date.accessioned | 2024-12-17T12:04:37Z | - |
dc.date.available | 2024-12-17T12:04:37Z | - |
dc.date.issued | 2024-10-09 | - |
dc.identifier.uri | http://hdl.handle.net/10400.5/96414 | - |
dc.description.abstract | Introdução: Na fase reconstrutiva da identificação humana médico-legal, a estimativa do sexo é crucial na reconstrução do perfil biológico, podendo ser aplicada tanto em contexto de identificação de vítimas de desastres em massa como na sala de autópsias. Pela subjetividade inerente aos métodos tradicionais, a inteligência artificial (IA) pode surgir como uma alternativa competitiva. Particularmente, as redes neurais convolucionais (CNN) oferecem uma alternativa mais precisa e menos morosa, automatizando a análise de ortopantomografias (OPGs). Objetivos: Este estudo avalia a fiabilidade do modelo VGG16 como um algoritmo preciso da predição forense do sexo e o seu desempenho com recurso a OPGs recolhidas de uma população da Unidade Local de Saúde (ULS) de Santa Maria. Materiais e Métodos: Este estudo incluiu 1050 OPGs de pacientes do Serviço de Estomatologia da ULS, selecionadas com base na idade (16-30 anos) e na presença de estruturas radiopacas. Em linguagem Python, as OPGs foram pré-processadas, redimensionadas e foram criadas cópias semelhantes utilizando métodos de data augmentation. Após o ajuste dos hiperparâmetros, calculou-se a precisão, sensibilidade, F1-score e acurácia, tendo sido criados heatmaps para identificar as regiões relevantes para o modelo. Resultados e Discussão: O treino revelou discrepância entre os valores de perda na validação e no treino. No teste geral, o modelo demonstrou um equilíbrio entre sexos, com F1-scores de 0.89. No teste por grupos etários, contrariamente ao esperado, o modelo obteve a maior acurácia na faixa de 16-20 anos (90%). Em geral, os resultados são ligeiramente inferiores a outros estudos. Além do mento, a análise dos heatmaps mostrou que o modelo não se focou em áreas anatomicamente relevantes, possivelmente devido à falta de aplicação de técnicas de extração de imagem. Conclusão: Os resultados demonstraram que as CNNs possuem acurácia na classificação de cadáveres em relação ao fator genérico sexo, na identificação médico-legal. O modelo obteve uma acurácia total de 89%. Novos estudos serão necessários para melhorar o desempenho dos modelos. | pt_PT |
dc.description.abstract | ABSTRACT Introduction: In the reconstructive phase of medico-legal human identification, the sex estimation is crucial in the reconstruction of the biological profile and can be applied both in identifying victims of mass disasters and in the autopsy room. Due to the inherent subjectivity associated with traditional methods, artificial intelligence (AI) may present an competitive option. Specifically, convolutional neural networks (CNN) offer a more accurate and less timeconsuming alternative, automating the analysis of orthopantomograms (OPGs). Objectives: This study evaluates the reliability of VGG16 model as an accurate forensic sex prediction algorithm and its performance using OPGs collected from a population at the Santa Maria Local Health Unit (ULS). Materials and Methods: This study included 1050 OPGs from patients at the ULS Stomatology Department, selected based on age (16-30 years) and on the presence of radiopaque structures. Using Python, the OPGs were pre-processed, resized and similar copies were created using data augmentation methods. After adjusting the hyperparameters, the model was evaluated for precision, sensitivity, F1-score and accuracy, and heatmaps were created to identify the relevant regions for the model. Results and Discussion: The training revealed a discrepancy between the validation and training loss values. In the general test, the model showed a general balance between the genders, with F1-scores of 0.89. In the test by age group, contrary to expectations, the model was most accurate in the 16-20 age group (90%). In general, the results are slightly lower than in other studies. Apart from the mandibular symphysis, analysis of the heatmaps showed that the model did not focus on anatomically relevant areas, possibly due to the lack of application of image extraction techniques. Conclusions: The results indicate that CNNs are accurate in classifying human remains based on the generic factor sex for medico-legal identification, achieving an overall accuracy of 89%. However, further research is necessary to enhance the models' performance. | pt_PT |
dc.language.iso | por | pt_PT |
dc.rights | openAccess | pt_PT |
dc.subject | Teses de mestrado - 2024 | pt_PT |
dc.subject | Saúde Oral | pt_PT |
dc.title | Avaliação da precisão do desempenho de redes neuronais convolucionais para a diagnose sexual | pt_PT |
dc.type | masterThesis | pt_PT |
thesis.degree.name | Mestrado em Integrado em Medicina Dentária | pt_PT |
dc.identifier.tid | 203720601 | pt_PT |
Appears in Collections: | FMD - Dissertações de Mestrado |
Files in This Item:
File | Description | Size | Format | |
---|---|---|---|---|
Dissertação_MIMD_Mariana_Correia_2023-2024.pdf | 303,1 MB | Adobe PDF | View/Open |
Items in Repository are protected by copyright, with all rights reserved, unless otherwise indicated.