Publicação
Deep Learning Techniques for Medical Image Classification
| Resumo: | In recent years, artificial intelligence (AI) has been applied in many fields to address complex and critical real-world tasks. Deep learning rises as a subfield of AI, where artificial neural networks (ANN) are used to map complicated functions, which can be challenging even for experienced users. One of the ANN variants is called convolutional neural network (CNN), which has shown great potential in image processing by providing state-of-the-art results for many significant image processing challenges. The medical field can significantly benefit from AI usage, especially in the medical image classification domain. In this doctoral dissertation, we applied different AI techniques to analyze medical images and to give the physicians a second opinion or reduce the time and effort needed for the image classification. Initially, we reviewed several studies that were published to discuss the transfer learning of CNNs. Afterward, we studied different hyperparameters that need to be optimized for CNNs to be trained accurately. Lastly, we proposed a novel CNN architecture to help in the classification of histopathology images. |
|---|---|
| Autores principais: | Kandel, Ibrahem Hamdy Abdelhamid |
| Assunto: | Image classification Convolutional neural networks Transfer learning Deep learning Medical images SDG 3 - Good health and well-being |
| Ano: | 2021 |
| País: | Portugal |
| Tipo de documento: | tese de doutoramento |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade Nova de Lisboa |
| Idioma: | inglês |
| Origem: | Repositório Institucional da UNL |
| Resumo: | In recent years, artificial intelligence (AI) has been applied in many fields to address complex and critical real-world tasks. Deep learning rises as a subfield of AI, where artificial neural networks (ANN) are used to map complicated functions, which can be challenging even for experienced users. One of the ANN variants is called convolutional neural network (CNN), which has shown great potential in image processing by providing state-of-the-art results for many significant image processing challenges. The medical field can significantly benefit from AI usage, especially in the medical image classification domain. In this doctoral dissertation, we applied different AI techniques to analyze medical images and to give the physicians a second opinion or reduce the time and effort needed for the image classification. Initially, we reviewed several studies that were published to discuss the transfer learning of CNNs. Afterward, we studied different hyperparameters that need to be optimized for CNNs to be trained accurately. Lastly, we proposed a novel CNN architecture to help in the classification of histopathology images. |
|---|