A carregar...
Publicação
O impacto da inteligência artificial na qualidade de imagem em estudos de joelho por ressonância magnética de 3 Tesla
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 2.8 MB | Adobe PDF |
Autores
Resumo(s)
A ressonância magnética constitui uma modalidade imagiológica fundamental no diagnóstico, particularmente na área músculo-esquelética devido à ausência de radiação ionizante e à sua elevada resolução de contraste. Todavia os tempos de aquisição de ressonância magnética são prolongados e representam uma limitação relevante podendo comprometer o conforto do doente e aumentar a suscetibilidade a artefactos de movimento. O aparecimento da inteligência artificial, nomeadamente os algoritmos baseados em deep learning como o Deep Resolve Boost ®, surgem como uma solução para acelerar a aquisição, mantendo ou melhorando a qualidade de imagem, mas
necessitam de validação com estudos em contextos clínicos.
O presente estudo teve como objetivo avaliar o impacto de algoritmos de inteligência artificial na qualidade de imagem em exames de ressonância magnética de joelho em equipamentos de 3 Tesla. Desenvolveu-se um estudo prospetivo, observacional e comparativo dividido em duas fases: fase experimental, onde foi utilizado um fantoma, e a fase clínica observacional que envolveu 40 participantes. Na fase experimental avaliou-se objetivamente a relação sinal-ruído e a
uniformidade da intensidade do sinal, enquanto que na fase clínica observacional, foi realizada uma avaliação objetiva e subjetiva da qualidade de imagem, recorrendo a escalas de Likert por observadores experientes.
Os resultados demonstram que a utilização do algoritmo de inteligência artificial permite reduzir o tempo de aquisição, mantendo ou melhorando a qualidade de imagem em contexto clínico.
Na fase experimental, a análise objetiva verificou uma pequena redução da relação sinal-ruído e da uniformidade da intensidade de sinal comparativamente à não utilização do Deep Resolve Boost ®.
Contudo em contexto real clínico, a avaliação objetiva evidenciou efeito significativo do Deep Resolve Boost ® na uniformidade da intensidade do sinal (p=0,034), mas não em tecido ósseo (p=0,255), enquanto o fator GRAPPA apresentou efeito significativo na uniformidade da intensidade do tecido ósseo (p < 0,001). Na avaliação subjetiva, a utilização do algoritmo demonstrou efeito estatisticamente significativo na definição de estruturas (p < 0,001), no ruído percebido (p < 0,001) e na qualidade global da imagem (p < 0,001). Estes resultados sugerem que o benefício do algoritmo foi mais evidente na perceção clínica da qualidade de imagem do que nas métricas objetivas isoladas.
Conclui-se então que o Deep Resolve Boost® representa uma ferramenta promissora em ressonância músculo-esquelética, nomeadamente no estudo do joelho em equipamentos de 3 Tesla, permitindo aquisições mais rápidas sem comprometer a qualidade de imagem diagnóstica, mas necessita sempre que esta seja implementada cuidadosamente e integrada de forma crítica na prática clínica para garantir que os ganhos tecnológicos se traduzem efetivamente em benefícios clínicos reais.
Magnetic resonance imaging (MRI) is a fundamental imaging modality in diagnosis, particularly in the musculoskeletal field, due to the absence of ionising radiation and its high contrast resolution. However, MRI scan times are lengthy and represent a significant limitation, which may compromise patient comfort and increase susceptibility to motion artefacts. The emergence of artificial intelligence, notably deep learning-based algorithms such as Deep Resolve Boost®, offers a solution to speed up acquisition whilst maintaining or improving image quality; however, these require validation through clinical studies. The aim of this study was to assess the impact of artificial intelligence algorithms on image quality in MRI scans of the knee using 3 Tesla scanners. A prospective, observational and comparative study was conducted, divided into two phases: an experimental phase, in which a phantom was used, and an observational clinical phase involving 40 participants. In the experimental phase, the signal-to-noise ratio and signal intensity uniformity were objectively assessed, whilst in the observational clinical phase, an objective and subjective evaluation of image quality was carried out using Likert scales by experienced observers. The results demonstrate that the use of the artificial intelligence algorithm reduces acquisition time whilst maintaining or improving image quality in a clinical setting. In the experimental phase, the objective analysis revealed a slight reduction in the signal-to-noise ratio and signal intensity uniformity compared with not using Deep Resolve Boost®. However, in a real-world clinical setting, the objective assessment revealed a significant effect of Deep Resolve Boost® on signal intensity uniformity (p=0.034), but not on bone tissue (p=0.255), whilst the GRAPPA factor showed a significant effect on signal intensity uniformity in bone tissue (p < 0.001). In the subjective assessment, the use of the algorithm demonstrated a statistically significant effect on structural definition (p < 0.001), perceived noise (p < 0.001) and overall image quality (p < 0.001). These results suggest that the benefit of the algorithm was more evident in the clinical perception of image quality than in the objective metrics alone. It can therefore be concluded that Deep Resolve Boost® represents a promising tool in musculoskeletal MRI, particularly in knee imaging using 3 Tesla scanners, enabling faster acquisitions without compromising diagnostic image quality, but it must always be carefully implemented and critically integrated into clinical practice to ensure that technological advances are effectively translated into real clinical benefits.
Magnetic resonance imaging (MRI) is a fundamental imaging modality in diagnosis, particularly in the musculoskeletal field, due to the absence of ionising radiation and its high contrast resolution. However, MRI scan times are lengthy and represent a significant limitation, which may compromise patient comfort and increase susceptibility to motion artefacts. The emergence of artificial intelligence, notably deep learning-based algorithms such as Deep Resolve Boost®, offers a solution to speed up acquisition whilst maintaining or improving image quality; however, these require validation through clinical studies. The aim of this study was to assess the impact of artificial intelligence algorithms on image quality in MRI scans of the knee using 3 Tesla scanners. A prospective, observational and comparative study was conducted, divided into two phases: an experimental phase, in which a phantom was used, and an observational clinical phase involving 40 participants. In the experimental phase, the signal-to-noise ratio and signal intensity uniformity were objectively assessed, whilst in the observational clinical phase, an objective and subjective evaluation of image quality was carried out using Likert scales by experienced observers. The results demonstrate that the use of the artificial intelligence algorithm reduces acquisition time whilst maintaining or improving image quality in a clinical setting. In the experimental phase, the objective analysis revealed a slight reduction in the signal-to-noise ratio and signal intensity uniformity compared with not using Deep Resolve Boost®. However, in a real-world clinical setting, the objective assessment revealed a significant effect of Deep Resolve Boost® on signal intensity uniformity (p=0.034), but not on bone tissue (p=0.255), whilst the GRAPPA factor showed a significant effect on signal intensity uniformity in bone tissue (p < 0.001). In the subjective assessment, the use of the algorithm demonstrated a statistically significant effect on structural definition (p < 0.001), perceived noise (p < 0.001) and overall image quality (p < 0.001). These results suggest that the benefit of the algorithm was more evident in the clinical perception of image quality than in the objective metrics alone. It can therefore be concluded that Deep Resolve Boost® represents a promising tool in musculoskeletal MRI, particularly in knee imaging using 3 Tesla scanners, enabling faster acquisitions without compromising diagnostic image quality, but it must always be carefully implemented and critically integrated into clinical practice to ensure that technological advances are effectively translated into real clinical benefits.
Descrição
Palavras-chave
Deep Resolve Boost® Inteligência Artificial Qualidade de Imagem Ressonância Magnética 3T Tempo de aquisição Acquisition Time Artificial Intelligence Image Quality Magnetic Resonance Imaging