A carregar...
Publicação
Artificial intelligence applied to invasive coronary angiography and physiology
| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 17.45 MB | Adobe PDF |
Resumo(s)
A doença das artérias coronárias (DAC) representa ainda a principal causa de patologia cardíaca, morbidade e mortalidade a nível global. O manejo invasivo da DAC, através da cirurgia de revascularização coronária ou intervenção coronária percutânea, foi viabilizado pelo desenvolvimento da angiografia coronária invasiva (CRG) nas décadas de 1950 e 60. A avaliação da gravidade das lesões em CRG é fundamental, pois aí assenta a decisão de revascularização, mas pouco mudou em mais de 50 anos. A estimativa visual da percentagem de estenose considerando o diâmetro normal (DE) permanece, ainda hoje, a metodologia padrão. Porém, múltiplos estudos, ao longo de 5 décadas, mostraram que esta abordagem é propensa a variabilidade inter-operadores e a sobre (ou mesmo sub)-estimativa. O desenvolvimento da Análise Coronária Quantitativa (ACQ) possibilitou a avaliação objetiva e reprodutível da gravidade da lesão, mas a sua base de evidência é reduzida, sendo a técnica subutilizada. Para ultrapassar estas limitações desenvolveram-se novos métodos baseados no conhecimento da fisiologia coronária, utilizando-se predominantemente dois índices: Fractional Flow Reserve (FFR) e instantaneous wave-free ratio (iFR). O primeiro foi directamente estudado em ensaios de outcomes/desfechos, enquanto o segundo foi principalmente estudado em comparação com o FFR. Contudo, nenhum é perfeito e o iFR poderá ser, por vezes, mais adequado, como na avaliação de lesões em tandem. Múltiplos ensaios clínicos confirmaram a utilidade/superioridade da fisiologia versus angiografia na tomada de decisões de revascularização, com impacto em outcomes/desfechos. Porém, em cenários mais complexos, como doença multivaso e síndrome coronária aguda, a vantagem destas técnicas não está inteiramente estabelecida, provavelmente devido ao acrescido risco de iatrogenia, juntamente com as limitações de fisiologia impostas pelo contexto agudo. Adicionalmente, mesmo no síndrome coronário crónico, a fisiologia coronária é subutilizada. A derivação não invasiva da fisiologia coronária a partir de imagens de CRG poderá superar parte destas limitações. Vários sistemas foram desenvolvidos, mas são semi-automáticos e a fiabilidade pode ser prejudicada pela variabilidade inter-operador e/ou inexperiência. A evidência disponível é ainda insuficiente e os estudos concentram-se quase exclusivamente na derivação do FFR - não do iFR. Adicionalmente, a contribuição da Inteligência Artificial (IA)para melhorar (ou superar) estes sistemas raramente foi explorada, já que a dinâmica de fluidos computacional e a AQC tridimensional têm sido o principal método utilizado nestes sistemas.
Coronary artery disease (CAD) is the dominant cause of cardiac pathology and remains the most relevant cause of morbidity and mortality worldwide. The invasive management of CAD, with revascularization by Coronary Artery Bypass Graft (CABG) surgery or Percutaneous Coronary Intervention (PCI), was made possible by the development of invasive coronary angiography (CAG) in the 1950’s and 60’s. A fundamental step in interpreting CAG is the estimating of lesion severity, as the decision to proceed with revascularization hinges upon it. Notwithstanding, this critical step has hardly changed for over 50 years. Indeed, visual estimation of percentage diameter stenosis (DS) remains the cornerstone of severity assessment, but is prone to operator dependency, resulting in potential over (or even under) estimation of severity, as multiple studies spanning five decades have shown. The development of Quantitative Coronary Analysis (QCA) enabled the objective and reproducible assessment of lesion severity but has been understudied and underused. To address these limitations, coronary physiology based methods were developed. Two main indexes are used: Fractional Flow Reserve (FFR) and instantaneous wave-free ratio (iFR). The former has been studied directly in outcomes studies, whereas the latter was mostly studied as a surrogate of FFR. Nonetheless, no single index is perfect, and iFR may be more suitable in certain scenarios, such tandem lesions assessment. Multiple clinical trials have confirmed the usefulness and superiority of physiology vs angiography-alone in improving revascularization decisions and hence clinical outcomes. However, their usefulness in more complex clinical scenarios, such as multivessel disease and acute coronary syndrome, has been challenged. The increased risk of iatrogeny in such contexts, together with the physiological limitations imposed by the acute setting, likely contributed to these findings. Even in the chronic coronary syndromes setting, coronary physiology remains underused. The non-invasive derivation of coronary physiology from CAG images alone may partially overcome these limitations. Such software systems have been developed, but are semi-automatic and their reliability may be hampered by operator heterogeneity and inexperience. Furthermore, available evidence to support their use is still insufficient and almost all studies have focused solely on the derivation of FFR, not iFR. Lastly, the contribution of AI technology in improving or surpassing these systems has seldom been explored, as most methods have used computational fluid dynamics (CFD) and three-dimensional (3D) QCA to derive physiology.
Coronary artery disease (CAD) is the dominant cause of cardiac pathology and remains the most relevant cause of morbidity and mortality worldwide. The invasive management of CAD, with revascularization by Coronary Artery Bypass Graft (CABG) surgery or Percutaneous Coronary Intervention (PCI), was made possible by the development of invasive coronary angiography (CAG) in the 1950’s and 60’s. A fundamental step in interpreting CAG is the estimating of lesion severity, as the decision to proceed with revascularization hinges upon it. Notwithstanding, this critical step has hardly changed for over 50 years. Indeed, visual estimation of percentage diameter stenosis (DS) remains the cornerstone of severity assessment, but is prone to operator dependency, resulting in potential over (or even under) estimation of severity, as multiple studies spanning five decades have shown. The development of Quantitative Coronary Analysis (QCA) enabled the objective and reproducible assessment of lesion severity but has been understudied and underused. To address these limitations, coronary physiology based methods were developed. Two main indexes are used: Fractional Flow Reserve (FFR) and instantaneous wave-free ratio (iFR). The former has been studied directly in outcomes studies, whereas the latter was mostly studied as a surrogate of FFR. Nonetheless, no single index is perfect, and iFR may be more suitable in certain scenarios, such tandem lesions assessment. Multiple clinical trials have confirmed the usefulness and superiority of physiology vs angiography-alone in improving revascularization decisions and hence clinical outcomes. However, their usefulness in more complex clinical scenarios, such as multivessel disease and acute coronary syndrome, has been challenged. The increased risk of iatrogeny in such contexts, together with the physiological limitations imposed by the acute setting, likely contributed to these findings. Even in the chronic coronary syndromes setting, coronary physiology remains underused. The non-invasive derivation of coronary physiology from CAG images alone may partially overcome these limitations. Such software systems have been developed, but are semi-automatic and their reliability may be hampered by operator heterogeneity and inexperience. Furthermore, available evidence to support their use is still insufficient and almost all studies have focused solely on the derivation of FFR, not iFR. Lastly, the contribution of AI technology in improving or surpassing these systems has seldom been explored, as most methods have used computational fluid dynamics (CFD) and three-dimensional (3D) QCA to derive physiology.
Descrição
Tese de doutoramento em Medicina (Cardiologia), Universidade de Lisboa, Faculdade de Medicina, 2024
Palavras-chave
Artificial intelligence Coronary angiography Coronary artery disease Percutaneous coronary intervention Coronary physiology Inteligência artificial Coronariografia Doença arterial coronária Intervenção coronária percutânea Fisiologia coronária