A inteligência artificial na qualidade da imagem em tomografia computadorizada: uma revisão de escopo
Data
Autores
Título da Revista
ISSN da Revista
Título de Volume
Resumo
A tomografia computadorizada (TC) é um exame de imagem essencial na prática radiológica, porém envolve elevada exposição à radiação ionizante, o que exige rigoroso controle de qualidade e otimização dos protocolos. Diante desse cenário, o presente trabalho teve como objetivo mapear, por meio de uma revisão de escopo, como a inteligência artificial (IA) tem sido aplicada na qualidade da imagem em tomografia computadorizada. A pesquisa foi desenvolvida nas bases PubMed, Scopus e Web of Science, com delimitação temporal de 2020 a 2026, seguindo as recomendações do PRISMA-ScR e da metodologia Joanna Briggs Institute (JBI). Após a remoção de duplicatas e aplicação dos critérios de elegibilidade, foram incluídos 33 artigos na amostra final. Os resultados evidenciaram que a IA tem sido empregada principalmente na reconstrução de imagens e na automação do controle de qualidade, com destaque para o uso de redes neurais convolucionais e algoritmos de deep learning reconstruction. Observou-se que essas tecnologias contribuem para a redução do ruído, manutenção da resolução espacial e preservação da exatidão das unidades Hounsfield, favorecendo protocolos de baixa dose e o princípio ALARA. Apesar dos avanços, ainda existem desafios relacionados à validação em pacientes reais, à heterogeneidade dos equipamentos e à necessidade de supervisão profissional. Conclui-se que a IA consolidou-se como uma realidade transformadora na radiologia, superando limitações dos métodos tradicionais ao oferecer análises quantitativas e objetivas. Contudo, sua aplicação prática ainda exige a superação do distanciamento entre testes em phantoms e a rotina clínica, reforçando que a tecnologia deve atuar de forma assistencial, como suporte ao julgamento e à decisão humana.
Computed tomography (CT) is an essential imaging modality in radiology practice; however, it involves high exposure to ionizing radiation, which requires rigorous quality control and protocol optimization. Given this context, the objective of this study was to map, through a scoping review, how artificial intelligence (AI) has been applied to image quality in computed tomography. The research was conducted using the PubMed, Scopus, and Web of Science databases, with a time frame from 2020 to 2026, following the recommendations of PRISMA-ScR and the Joanna Briggs Institute (JBI) methodology. After removing duplicates and applying the eligibility criteria, 33 articles were included in the final sample. The results showed that AI has been used primarily for image reconstruction and the automation of quality control, with a particular focus on the use of convolutional neural networks and deep learning reconstruction algorithms. It was observed that these technologies contribute to noise reduction, maintenance of spatial resolution, and preservation of the accuracy of Hounsfield units, thereby supporting low-dose protocols and the ALARA principle. Despite these advances, challenges remain regarding validation in real patients, equipment heterogeneity, and the need for professional supervision. It is concluded that AI has established itself as a transformative force in radiology, overcoming the limitations of traditional methods by offering quantitative and objective analyses. However, its practical application still requires bridging the gap between phantom testing and clinical practice, reinforcing that the technology should serve in an assistive capacity, supporting human judgment and decision-making.
Computed tomography (CT) is an essential imaging modality in radiology practice; however, it involves high exposure to ionizing radiation, which requires rigorous quality control and protocol optimization. Given this context, the objective of this study was to map, through a scoping review, how artificial intelligence (AI) has been applied to image quality in computed tomography. The research was conducted using the PubMed, Scopus, and Web of Science databases, with a time frame from 2020 to 2026, following the recommendations of PRISMA-ScR and the Joanna Briggs Institute (JBI) methodology. After removing duplicates and applying the eligibility criteria, 33 articles were included in the final sample. The results showed that AI has been used primarily for image reconstruction and the automation of quality control, with a particular focus on the use of convolutional neural networks and deep learning reconstruction algorithms. It was observed that these technologies contribute to noise reduction, maintenance of spatial resolution, and preservation of the accuracy of Hounsfield units, thereby supporting low-dose protocols and the ALARA principle. Despite these advances, challenges remain regarding validation in real patients, equipment heterogeneity, and the need for professional supervision. It is concluded that AI has established itself as a transformative force in radiology, overcoming the limitations of traditional methods by offering quantitative and objective analyses. However, its practical application still requires bridging the gap between phantom testing and clinical practice, reinforcing that the technology should serve in an assistive capacity, supporting human judgment and decision-making.
Descrição
Palavras-chave
Citação
FOGAÇA, Andrielli Ribeiro. A inteligência artificial na qualidade da imagem em tomografia computadorizada: uma revisão de escopo. 2026. Trabalho de Conclusão de Curso (Superior de Tecnologia em Radiologia) – Instituto Federal de Santa Catarina, Florianópolis, 2026
