Publicação
Computer assisted diagnosis (CAD) with enhanced cognitive architectures (BorisCAD)
| Resumo: | Computer-assisted medical image diagnosis plays a vital role in modern healthcare by enabling accurate and timely detection of various medical conditions. However, conventional image classification techniques often fail to integrate crucial elements such as perception, reasoning, and episodic experiences, essential for achieving optimal performance. This research endeavours to bridge this gap by designing, implementing, and evaluating a cognitive architecture incorporating these elements. The findings of this research will contribute to the development of a deployable cognitive architecture that can provide accurate and reliable diagnoses for a wide range of medical conditions, benefiting both healthcare professionals and patients alike. Furthermore, the potential impact of this research extends beyond medical image analysis, with implications in autonomous systems, robotics, and intelligent decision support systems. By harnessing the potential of cognitive architectures, computer-assisted systems can be revolutionised, leading to improved diagnostic accuracy and fostering innovation in diverse industries reliant on advanced cognitive capabilities. In order to achieve this objective, sub-symbolic methods for perception and symbolic methods for reasoning will integrate the cognitive architecture. The bottom level, responsible for sub-symbolic processing, will incorporate advanced segmentation and classification using convolutional neural networks to handle perception tasks. In contrast, the top level will utilise a decision forest, an ensemble of decision trees, to perform sophisticated reasoning tasks using symbolic data. Additionally, this study will focus on integrating episodic experiences within the architecture by incorporating working and long-term memory mechanisms, enhancing its predictive capabilities. The evaluation of the cognitive architecture demonstrated its effectiveness within the context of the tested datasets and image sizes. However, it is essential to acknowledge that developing a deployable version for medical image diagnosis requires further testing and validation. Expanding the evaluation to include a broader range of pathologies and imaging modalities is crucial to ensure the architecture’s robustness and adaptability in diverse clinical scenarios. By incorporating a more diverse set of pathologies and imaging modalities into the evaluation process, the cognitive architecture can undergo rigorous testing to assess its performance across various medical conditions. This expanded evaluation will help identify potential limitations and areas for improvement, ensuring that the architecture can deliver accurate and reliable diagnoses across a broader spectrum of medical conditions. |
|---|---|
| Autores principais: | Samorinha, Rafael André Escalhão |
| Assunto: | Convolutional neural networks Decision tree Memory Cognitive architecture Image classification Learning Adaptation Medical diagnosis Artificial intelligence Boris Redes neuronais convolucionais Árvore de decisão Memória Arquitetura cognitiva Classificação de imagens Aprendizagem Adaptação Diagnóstico médico Inteligência artificial Engenharia e Tecnologia::Engenharia Eletrotécnica, Eletrónica e Informática |
| Ano: | 2023 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade do Minho |
| Idioma: | inglês |
| Origem: | RepositóriUM - Universidade do Minho |
| Resumo: | Computer-assisted medical image diagnosis plays a vital role in modern healthcare by enabling accurate and timely detection of various medical conditions. However, conventional image classification techniques often fail to integrate crucial elements such as perception, reasoning, and episodic experiences, essential for achieving optimal performance. This research endeavours to bridge this gap by designing, implementing, and evaluating a cognitive architecture incorporating these elements. The findings of this research will contribute to the development of a deployable cognitive architecture that can provide accurate and reliable diagnoses for a wide range of medical conditions, benefiting both healthcare professionals and patients alike. Furthermore, the potential impact of this research extends beyond medical image analysis, with implications in autonomous systems, robotics, and intelligent decision support systems. By harnessing the potential of cognitive architectures, computer-assisted systems can be revolutionised, leading to improved diagnostic accuracy and fostering innovation in diverse industries reliant on advanced cognitive capabilities. In order to achieve this objective, sub-symbolic methods for perception and symbolic methods for reasoning will integrate the cognitive architecture. The bottom level, responsible for sub-symbolic processing, will incorporate advanced segmentation and classification using convolutional neural networks to handle perception tasks. In contrast, the top level will utilise a decision forest, an ensemble of decision trees, to perform sophisticated reasoning tasks using symbolic data. Additionally, this study will focus on integrating episodic experiences within the architecture by incorporating working and long-term memory mechanisms, enhancing its predictive capabilities. The evaluation of the cognitive architecture demonstrated its effectiveness within the context of the tested datasets and image sizes. However, it is essential to acknowledge that developing a deployable version for medical image diagnosis requires further testing and validation. Expanding the evaluation to include a broader range of pathologies and imaging modalities is crucial to ensure the architecture’s robustness and adaptability in diverse clinical scenarios. By incorporating a more diverse set of pathologies and imaging modalities into the evaluation process, the cognitive architecture can undergo rigorous testing to assess its performance across various medical conditions. This expanded evaluation will help identify potential limitations and areas for improvement, ensuring that the architecture can deliver accurate and reliable diagnoses across a broader spectrum of medical conditions. |
|---|