As far as we know, this is the first study to explore the association between deep learning based CT-FFR, PCAT CT attenuation and plaque volume and MACE in patients with PCI. The current study investigated the prognostic potential of deep learning based CT-FFR, PCAT CT attenuation and plaque volume in patients with PCI. Our results demonstrated that PCAT_RCA CT attenuation and plaque volume were independently associated with increasing risk of MACE.
Previous studies have elaborated the prognostic value of invasive FFR (iFFR). The FAME (Fractional Flow Reserve Versus Angiography for Multivessel Evaluation) 1 and 2 studies revealed that a higher post-PCI iFFR value predicted a better clinical outcome (18). Furthermore, Angarwal et al indicated that post-PCI iFFR showed incremental prognostic value beyond clinical and angiographic factors in predicting MACE (19). In additional, meta-analysis has demonstrated that post-PCI iFFR value revealed an inverse relationship with composite MACE (MI, death, revascularization) (20). However, the prognostic value of deep learning based CT-FFR in patients with PCI has not been clearly explored. The present study showed that deep learning based CT-FFR was not associated with MACE in patients with PCI.
PCAT CT attenuation has been regarded as an imaging biomarker of capturing coronary inflammation in coronary CTA. Previous studies have investigated the prognostic value of PCAT CT attenuation (12, 13, 21, 22).the Cardiovascular RISk Prediction using Computed Tomography (CRISP-CT) study showed that PCAT_RCA CT attenuation (HR, 1.49–1.84), PCAT_LAD CT attenuation (HR, 1.77–1.78) and PCAT_LCX CT attenuation (HR, 1.37–1.47) were independently associated with all-cause death in both derivation and validation cohorts during a median follow-up of 72 months (derivation cohort) and 54 months (validation cohort), respectively (12). However, recent studies indicated that only PCAT_RCA CT attenuation was associated with poor clinical outcome. Diemen et al revealed that PCAT_RCA CT attenuation remained as an independently predictor after adjusting for clinical and imaging factors (HR: 2.45, 95%CI: 1.23–4.93, p = 0.011), whereas PCAT_LAD and PCAT_LCX CT attenuation were not associated with the endpoint (21). Tzolos et al further indicated that PCAT_RCA CT attenuation, not PCAT_LAD or PCAT_LCX CT attenuation was predictive of further MI (22). The present study also indicated that only PCAT_RCA CT attenuation was independently associated with MACE after adjusting clinical and imaging factors in patients with PCI. The possible reason for this phenomenon may be because that there is more fat structure around RCA compared to LAD and LCX. In addition, compared with LAD and LCX, there are fewer side branches of RCA. These together made the measurement and analysis of PCAT CT attenuation easier (13, 22).
Plaque volumes derived from coronary CTA have been demonstrated with high prognostic value for adverse cardiovascular events (23–25). Some studies revealed this prognostic value was higher than clinical risk and lumen stenosis factors (26–28). The present study showed similar results. The current measurements of plaque volume in coronary CTA were mainly dependent with various semi-automated research software. Although, these platforms showed high correlations with intravascular ultrasound, the measurements and analysis of plaque volume is time-consuming, as this required a large amount of manual input from expert readers (25). Therefore, it limited its implementation in clinical practice. Our deep learning based-plaque volume measurement improves this process time-saving, thus increasing its potential of clinical application in the future. The present results indicated that plaque volumes of the coronary tree quantified by automatic measurement have an independent and strong prognostic value for MACE in patients with PCI, which has not been reported previously. Moreover, we determined an optimum cutoff (> 20 mm3), exceeding this value leads to a sharp increase in the risk of events.
There are some limitations in the current study. First, the current study is a single center retrospective study. This study lacks information on lifestyle changes, such as exercise, sleep or dietary habits and medical therapy after coronary CTA examination, this information may influence future outcome in the present study, thus may lead to biased results. Second, the high-risk plaque (HRP) has not been evaluated because the current AI version still has shortcomings in identifying and interpreting HRP. Third, other widely classification based on cardio CT scan (coronary artery calcification scores, CACS), or clinical information (SYNTAX score) have not been included. Further large multi-center investigations are needed to determine our findings.
In conclusion, deep learning based RCA PCAT and plaque volume derived from coronary CTA, not CT-FFR was found to be associated with MACE in patients with PCI.