| Name: | Description: | Size: | Format: | |
|---|---|---|---|---|
| 4.07 MB | Adobe PDF |
Advisor(s)
Abstract(s)
A análise de imagens médicas desempenha um papel fundamental no diagnóstico e tratamento de diversas condições de saúde. Entre as várias modalidades de imagens médicas, os raios X são amplamente utilizados devido à sua praticidade, baixo custo e capacidade de fornecer informações valiosas sobre o estado de saúde de um paciente. No entanto, a interpretação precisa e eficiente das radiografias requer conhecimento especializado e experiência clínica, o que pode ser limitado em determinados contextos. Neste contexto, o desenvolvimento de modelos capazes de analisar e interpretar radiografias de forma automatizada e precisa tem se tornado uma área de pesquisa promissora. Estes modelos têm o potencial de auxiliar radiologistas e médicos no diagnóstico precoce e no tratamento de diversas doenças e condições médicas. Existem algumas razões pelas quais o desenvolvimento de um modelo de análise automatizado de raios X é de grande relevância. A necessidade de resposta por serviços de saúde tem aumentado, e o número de radiografias a serem analisadas pode sobrecarregar os profissionais de saúde. Portanto, ter um modelo automatizado pode acelerar o processo de triagem e diagnóstico, permitindo um atendimento mais eficiente e reduzindo o tempo de espera dos pacientes. Foi desenvolvido um modelo de análise automatizado que é capaz de identificar e diagnosticar a presença de problemas e anomalias nas imagens radiográficas. O modelo utiliza técnicas de deep learning, como redes neuronais convolucionais, para extrair características relevantes das radiografias e realizar a classificação adequada. Na construção deste modelo também foi utilizado técnicas de pré processamento de imagem, para preparar todas as imagens que foram usadas no treino deste mesmo modelo, permitindo a melhor análise possível da mesma. No fim, o modelo foi capaz de reconhecer 14 doenças devido ao conjunto de dados usado no treino. Apesar de reconhecer 14 padrões (doenças) diferentes, depois de uma avaliação do modelo, foi percetível que o mesmo sente dificuldades em avaliar alguns destes padrões e retornar uma resposta válida, em contrapartida, tem uma boa performance na análise dos restantes casos principalmente quando a radiografia não apresenta problemas. A aplicação desenvolvida pretende atender aos critérios de desempenho e precisão necessários para torná-la uma ferramenta útil na prática clínica, sempre considerando a importância de uma análise final por parte de um radiologista ou médico especializado. É fundamental realçar que o modelo não substituirá a especialização clínica, mas sim servirá como uma valiosa ferramenta complementar para a tomada de decisão médica, possibilitando assim uma análise mais ágil.
Medical imaging analysis plays a key role in the diagnosis and treatment of various health conditions. Among the various medical imaging modalities, X-rays are widely used due to their practicality, low cost and ability to provide valuable information about a patient's health status. However, accurate and efficient interpretation of x rays requires specialized knowledge and clinical experience, which may be limited in certain contexts. In this context, the development of models capable of analyzing and interpreting x rays in an automated and accurate manner has become a promising area of research. These models have the potential to assist radiologists and medics with the early diagnosis and treatment of various diseases and medical conditions. There are a few reasons why the development of an automated X-ray analysis model isrelevant. Firstly, the need for fast response by healthcare services, the number of x rays to be analyzed can overwhelm healthcare professionals. Therefore, having an automated model can speed up the diagnostic process, increasing the efficient care and reducing the patient waiting time. An automated analysis model was developed that can identify and diagnose the presence of problems and anomalies in x ray images. The model uses deep learning techniques, such as convolutional neural networks, to extract relevant features from the x rays and perform the appropriate classification. In the construction of this model, image pre-processing techniques were also used, to prepare all the images that were used in the training of this same model, allowing the best possible analysis. In the end, the model was able to recognize 14 diseases due to the dataset used in the training. Despite recognizing 14 different patterns (diseases), after an evaluation of the model, it was perceptible that it has difficulties in evaluating some of these patterns and returning a valid answer, on the other hand, it has a good performance in the analysis of the remaining cases, especially when the x ray does not contain any problem. The application developed intends to meet the performance and accuracy criteria necessary to make it a useful tool in clinical practice, always considering the importance of a final analysis by a radiologist or specialized medic. It is essential to highlight that the model will not replace clinical expertise, but will serve as a valuable complementary tool for medical decision making, thus enabling a more agile analysis.
Medical imaging analysis plays a key role in the diagnosis and treatment of various health conditions. Among the various medical imaging modalities, X-rays are widely used due to their practicality, low cost and ability to provide valuable information about a patient's health status. However, accurate and efficient interpretation of x rays requires specialized knowledge and clinical experience, which may be limited in certain contexts. In this context, the development of models capable of analyzing and interpreting x rays in an automated and accurate manner has become a promising area of research. These models have the potential to assist radiologists and medics with the early diagnosis and treatment of various diseases and medical conditions. There are a few reasons why the development of an automated X-ray analysis model isrelevant. Firstly, the need for fast response by healthcare services, the number of x rays to be analyzed can overwhelm healthcare professionals. Therefore, having an automated model can speed up the diagnostic process, increasing the efficient care and reducing the patient waiting time. An automated analysis model was developed that can identify and diagnose the presence of problems and anomalies in x ray images. The model uses deep learning techniques, such as convolutional neural networks, to extract relevant features from the x rays and perform the appropriate classification. In the construction of this model, image pre-processing techniques were also used, to prepare all the images that were used in the training of this same model, allowing the best possible analysis. In the end, the model was able to recognize 14 diseases due to the dataset used in the training. Despite recognizing 14 different patterns (diseases), after an evaluation of the model, it was perceptible that it has difficulties in evaluating some of these patterns and returning a valid answer, on the other hand, it has a good performance in the analysis of the remaining cases, especially when the x ray does not contain any problem. The application developed intends to meet the performance and accuracy criteria necessary to make it a useful tool in clinical practice, always considering the importance of a final analysis by a radiologist or specialized medic. It is essential to highlight that the model will not replace clinical expertise, but will serve as a valuable complementary tool for medical decision making, thus enabling a more agile analysis.
Description
Keywords
Análise de imagens médicas Doenças Redes Neuronais Convolucionais Classificação Medical Imaging Analysis Diseases Convolutional neuronal networks Classification