CT morphological index provides incremental value to machine learning based CT-FFR for predicting hemodynamically significant coronary stenosis

Stenosis Detection and Quantification of Coronary Artery Using Machine Learning and Deep Learning

The aim of this review is to introduce some applications of artificial intelligence (AI) algorithms for the detection and quantification of coronary stenosis using computed tomography angiography (CTA). The realization of automatic/semi-automatic stenosis detection and quantification includes the following steps: vessel central axis extraction, vessel segmentation, stenosis detection, and quantification. Many new AI techniques, such as machine learning and deep learning, have been widely used in medical image segmentation and stenosis detection. This review also summarizes the recent progress regarding coronary stenosis detection and quantification, and discusses the development trends in this field. Through evaluation and comparison, researchers can better understand the research frontier in related fields, compare the advantages and disadvantages of various methods, and better optimize the new technologies. Machine learning and deep learning will promote the process of automatic detection and quantification of coronary artery stenosis. However, the machine learning and the deep learning methods need a large amount of data, so they also face some challenges because of the lack of professional image annotations (manually add labels by experts).

Radiomics features of pericoronary adipose tissue improve CT-FFR performance in predicting hemodynamically significant coronary artery stenosis

OBJECTIVES

To evaluate the value of radiomics-based model of pericoronary adipose tissue (PCAT) combined with CT fractional flow reserve (CT-FFR) in predicting hemodynamically significant coronary stenosis.

METHODS

Patients with suspected or known coronary artery disease, who had coronary computed tomography angiography (CCTA), invasive coronary angiography (ICA), and FFR within 1 month, were retrospectively included. Radiomics features of lesion-based PCAT were extracted. The lesion-specific CT-FFR values, CCTA-derived diameter stenosis, lesion length, and PCAT attenuation were also measured. FFR values were used as the reference standard to assess the diagnostic performance of radiomics model, CT-FFR, and combined model for detection of flow-limiting stenosis.

RESULTS

A total of 146 patients with 180 lesions were included in the study. All lesions were divided into training and validation cohorts at a ratio of 2:1. CT-FFR model exhibited the highest area under the curve (AUC) (0.803 for training, 0.791 for validation) in predicting hemodynamically significant stenosis, followed by radiomics model (0.776 for training, 0.744 for validation). However, no statistically significant difference was found between the AUCs of the above two models (p > 0.05). When CT-FFR was combined with radiomics model, the AUC reached 0.900 for training cohort and 0.875 for validation cohort, which were significantly higher than that of CT-FFR and radiomics model alone (both p < 0.05).

CONCLUSION

The diagnostic performance of PCAT radiomics model was comparable to that of CT-FFR for identification of ischemic coronary stenosis. Adding PCAT radiomics model to CT-FFR showed incremental value in discriminating flow-limiting from non-flow-limiting lesions.

KEY POINTS

• Radiomics analysis of lesion-based PCAT is potentially an alternative method to identify hemodynamic significance of coronary artery stenosis. • Adding radiomics model of PCAT to CT-FFR improved diagnostic performance for the detection of flow-limiting coronary stenosis. • Radiomics features + CT-FFR is a promising noninvasive method for comprehensive evaluation of hemodynamic significance of coronary artery stenosis.

Comparison of coronary CT angiography-based and invasive coronary angiography-based quantitative flow ratio for functional assessment of coronary stenosis: A multicenter retrospective analysis

BACKGROUND

The aim of this study was to evaluate the diagnostic performance of coronary CT angiography (CTA)-based quantitative flow ratio (QFR), namely CT-QFR, and compare it with invasive coronary angiography (ICA)-based Murray law QFR (μQFR), using fractional flow reserve (FFR) as the reference standard.

METHODS

Patients who underwent coronary CTA, ICA and pressure wire-based FFR assessment within two months were retrospectively analyzed. CT-QFR and μQFR were computed in blinded fashion and compared with FFR, all applying the same cut-off value of ≤0.80 to identify hemodynamically significant stenosis.

RESULTS

Paired comparison between CT-QFR and μQFR was performed in 191 vessels from 167 patients. Average FFR was 0.81 ​± ​0.10 and 42.4% vessels had an FFR ≤0.80. CT-QFR had a slightly lower correlation with FFR compared with μQFR, although statistically non-significant (r ​= ​0.87 versus 0.90, p ​= ​0.110). The vessel-level diagnostic performance of CT-QFR was slightly lower but without statistical significance than μQFR (AUC ​= ​0.94 versus 0.97, difference: -0.03 [95%CI: -0.00-0.06], p ​= ​0.095), and substantially higher than diameter stenosis by CTA (AUC difference: 0.17 [95%CI: -0.10-0.23], p ​< ​0.001). The patient-level diagnostic accuracy, sensitivity, specificity, positive predictive value, negative predictive value, positive likelihood ratio and negative likelihood ratio for CT-QFR to identify FFR value ​≤ ​0.80 was 88%, 90%, 86%, 86%, 91%, 6.59 and 0.12, respectively. The diagnostic accuracy of CT-QFR was 84% in extensively calcified lesions, while in vessels with no or less calcification, CT-QFR showed a comparable diagnostic accuracy with μQFR (91% versus 92%, p ​= ​0.595). Intra- and inter-observer variability in CT-QFR analysis was -0.00 ​± ​0.04 and 0.00 ​± ​0.04, respectively.

CONCLUSIONS

Performance in diagnosis of hemodynamically significant coronary stenosis by CT-QFR was slightly lower but without statistical significance than μQFR, and substantially higher than CTA-derived diameter stenosis. Extensively calcified lesions reduced the diagnostic accuracy of CT-QFR.

Distribution of FFRCT in single obstructive coronary stenosis and predictors for major adverse cardiac events: a propensity score matching study

Background

Fractional flow reserve derived from computed tomography (FFRCT) has been demonstrated to improve identification of lesion-specific ischemia significantly compared with coronary computed tomography angiography (CCTA). It remains unclear whether the distribution of FFRCT values in obstructive stenosis between patients who received percutaneous coronary intervention (PCI) or not in routine clinical practice, as well as its association with clinical outcome. This study aims to reveal the distribution of FFRCT value in patients with single obstructive coronary artery stenosis and explored the independent factors for predicting major adverse cardiac events (MACE).

Methods

This was a retrospective study of adults with non-ST-segment elevation acute coronary syndrome undergoing FFRCT assessment by using CCTA data from January 1, 2016 to December 31, 2020. Propensity score matching (PSM) method was used to account for patient selection bias. The risk factors for predicting MACE were evaluated by a Cox proportional hazards regression analysis.

Results

Overall, 655 patients with single obstructive (≥ 50%) stenosis shown on CCTA were enrolled and divided into PCI group (279 cases) and conservative group (376 cases) according to treatment strategy. The PSM cohort analysis demonstrated that the difference in history of unstable angina, Canadian Cardiovascular Society Class (CCSC) and FFRCT between PCI group (188 cases) and conservative group (315 cases) was statistically significant, with all P values < 0.05, while the median follow-up time between them was not statistically significant (24 months vs. 22.5 months, P = 0.912). The incidence of MACE in PCI group and conservative group were 14.9% (28/188) and 23.5% (74/315) respectively, P = 0.020. Multivariate analysis of Cox proportional hazards regression revealed that history of unstable angina (adjusted odds ratio (adjOR), 3.165; 95% confidence interval (CI), 2.087–4.800; P < 0.001), FFRCT ≤ 0.8 (OR, 1.632;95% CI 1.095–2.431; P = 0.016), and PCI therapy (OR 0.481; 95% CI 0.305–0.758) were the independent factors for MACE.

Conclusions

History of unstable angina and FFRCT value of ≤ 0.8 were the independent risk factors for MACE, while PCI therapy was the independent protective factor for MACE.

Long-term prognostic value of the serial changes of CT-derived fractional flow reserve and perivascular fat attenuation index

BACKGROUND

To investigate the serial changes of computed tomography (CT) fractional flow reserve (CT-FFR) and fat attenuation index (FAI), and explore their relationships with long-term clinical outcomes.

METHODS

Consecutive symptomatic patients with an intermediate pretest probability of coronary artery disease 1-4 were prospectively enrolled if coronary CT angiography (CCTA) revealed at least 1 lesion with 30-70% stenosis on major epicardial arteries. Follow-up CCTA was performed at 1 to 1.5-year intervals. All patients were further followed up after the second CCTA until September 2019. The Coronary Artery Disease - Reporting and Data System (CAD-RADS) grade, high-risk plaque features, lesion-specific CT-FFR, and FAI were measured for prognosis analysis.

RESULTS

A total of 263 patients were included in the analysis, and 38 major adverse cardiac events (MACEs) occurred. In the MACE group, the lesion-specific CT-FFR decreased significantly at the follow-up CCTA [0.80 (0.74-0.90) versus 0.85 (0.76-0.93); P=0.01], whereas the FAI did not notably increase (-70.4±8.9 versus -71.3±7.1 HU; P=0.436). In the non-MACE group, lesion-specific CT-FFR increased markedly [0.91 (0.84-0.95) versus 0.90 (0.82-0.94); P<0.001], while the FAI decreased substantially (-74.0±10.8 versus -72.4±11.5 HU; P=0.004). Decreased CT-FFR (adjusted overall hazard ratio =2.455; P=0.023) and increased FAI (adjusted hazard ratio =2.956; P=0.002) were the strongest independent predictors of MACEs. Serial changes of CT-FFR and FAI provided incremental prognostic value (Concordance statistic =0.716; P=0.003; over conventional clinical and imaging parameters (Concordance statistic =0.762; P=0.004).

CONCLUSIONS

Decreased CT-FFR and increased FAI at follow-up CCTA were the 2 strongest predictors of MACEs. Serial changes of CT-FFR and FAI provided incremental prognostic value over conventional clinical and imaging parameters for risk stratification. In addition, decreased CT-FFR provided incremental predictive value for MACEs from 15 months after second CCTA, while increased FAI added prognostic value from the second CCTA onwards.

Prognostic value of CT-derived myocardial blood flow, CT fractional flow reserve and high-risk plaque features for predicting major adverse cardiac events

Background

Myocardial blood flow (MBF), CT fractional flow reserve (CT-FFR) and high-risk plaque (HRP) features have been revealed to be associated with patients' prognosis. However, direct intra-individual comparison of these CT-derived parameters has not been explored yet. The aim of this study was to investigate the prognostic value of CT-derived MBF, CT-FFR and HRP features for predicting major adverse cardiac events (MACEs).

Methods

Consecutive patients with chest pain and intermediate-to-high pre-test probability of coronary artery disease (CAD) were prospectively enrolled. All patients were referred for dynamic CT myocardial perfusion imaging (CT-MPI) + coronary CT angiography (CCTA) and followed up for at least 1 year. MBF_ischemic (mean MBF of all ischemic segments), MBF_ratio (MBF of ischemic segments/MBF of reference segments), CT-FFR and HRP features were measured and multivariate analysis was used to evaluate the predictive value of all above parameters for MACEs.

Results

One hundred and forty-two patients were included into final analysis. MBF_ischemic and MBF_ratio was significantly lower in patients with MACE compared to patients without MACE (87 vs. 153 mL/100 mL/min and 0.64 vs. 0.95, both P<0.001). Similarly, CT-FFR was also markedly lower in patients with MACE (0.58 vs. 0.88, P<0.001) whereas coronary artery calcium score (CACS) was significantly higher (1,038.9 vs. 34.2, P<0.001). According to ROC curve analysis, MBF_ischemic, MBF_ratio and CACS had largest area under curve (AUC =0.872, 0.855 and 0.813 respectively, all P<0.001) for identifying patients with MACE. After adjusted by multivariate analysis, MBF_ischemic (hazard ratio =23.382, P=0.003) and CACS (hazard ratio =3.759, P=0.029) were revealed to be the independent predictors for MACE where CT-FFR and HRP features failed to have prognostic value.

Conclusions

MBF_ischemic derived from dynamic CT-MPI was the strongest predictor for MACE, followed by CACS. MBF_ischemic outperformed HRP features and CT-FFR for prediction of unfavorable clinical outcome.

Clinical prediction models of fractional flow reserve: an exploration of the current evidence and appraisal of model performance

Background

Invasive fractional flow reserve (FFR) is a standard indicator of coronary stenoses' hemodynamic severity. Clinical prediction models (CPMs) may help differentiate ischemic from non-ischemic lesions without using a pressure wire but by integrating related variables. This approach differs from that of physics-based models. However, it is not yet known which CPMs are the most reliable at detecting hemodynamic significance.

Methods

A systematic review was performed of relevant publications that developed or validated any FFR CPMs from inception to April 2019 in the PubMed, EMBASE, and Cochrane Library databases by two independent authors. The risk of bias and applicability were assessed using the prediction model risk of the bias assessment tool (PROBAST).

Results

A total of 11 unique CPMs and 5 subsequent external validation studies were identified. The prevalence of hemodynamically significant lesions (FFR ≤0.80) across the studies had a median of 37.1% (range: 20.7-68.0%). Lesion length, percent diameter stenosis, and minimal lumen diameter were the three most frequently used variables in the CPMs. Of the 11 FFR CPMs, 9 (82%) exhibited strong discrimination [area under the curve (AUC) >0.75], and 5 (45%) had been subject to external validation; however, calibration was only available for 3 models (27%). There was a high degree of applicability; however, none of the studies was assessed as having a low risk of bias. A CPM was identified that had undergone rigorous validation and calibration: the DILEMMA score (three validations; median AUC, 0.83).

Conclusions

Almost half of the existing FFR CPMs had been externally validated. Due to their good discrimination abilities, these FFR CPMs are useful tools that could reduce the need for invasive hemodynamic measurements. Future research that adheres to methodological guidelines should be undertaken to develop high-quality models in this setting. (PROSPERO registration number: CRD42019125011).

Machine Learning Quantitation of Cardiovascular and Cerebrovascular Disease: A Systematic Review of Clinical Applications

Research into machine learning (ML) for clinical vascular analysis, such as those useful for stroke and coronary artery disease, varies greatly between imaging modalities and vascular regions. Limited accessibility to large diverse patient imaging datasets, as well as a lack of transparency in specific methods, are obstacles to further development. This paper reviews the current status of quantitative vascular ML, identifying advantages and disadvantages common to all imaging modalities. Literature from the past 8 years was systematically collected from MEDLINE^® and Scopus database searches in January 2021. Papers satisfying all search criteria, including a minimum of 50 patients, were further analysed and extracted of relevant data, for a total of 47 publications. Current ML image segmentation, disease risk prediction, and pathology quantitation methods have shown sensitivities and specificities over 70%, compared to expert manual analysis or invasive quantitation. Despite this, inconsistencies in methodology and the reporting of results have prevented inter-model comparison, impeding the identification of approaches with the greatest potential. The clinical potential of this technology has been well demonstrated in Computed Tomography of coronary artery disease, but remains practically limited in other modalities and body regions, particularly due to a lack of routine invasive reference measurements and patient datasets.

Diagnostic performance of quantitative, semi-quantitative, and visual analysis of dynamic CT myocardial perfusion imaging: a validation study with invasive fractional flow reserve

OBJECTIVES

To investigate the diagnostic performance of absolute myocardial blood flow (MBF), MBF_ratio, and visual analysis of dynamic CT myocardial perfusion imaging (CT-MPI) for the detection of hemodynamically significant coronary stenosis.

METHODS

Consecutive patients with chest pain and intermediate-to-high pre-test probability of obstructive coronary artery disease were prospectively enrolled. All patients were referred for dynamic CT-MPI and fractional flow reserve (FFR) measurements within 4 weeks. Absolute MBF, MBF_ratio (mean MBF of stenosis-subtended territories versus that of reference territories), and visually identified perfusion defect were tested for the diagnostic performance with reference to FFR.

RESULTS

Sixty-two patients with 95 target vessels were included for final analysis. The mean radiation dose for dynamic CT-MPI was 3.0 (2.2-4.0) mSv. The mean lesion-based absolute MBF value was significantly lower in ischemic segments than that in non-ischemic segments (78.0 (65.0-86.0) mL/min/100 mL vs. 133.0 (117.5-163.8) mL/min/100 mL, p < 0.001). Similarly, the lesion-based MBF_ratio was also markedly lower in territories with positive FFR results (0.52 (0.44-0.64) vs. 0.93 (0.91-0.97), p < 0.001). According to per-lesion ROC curve analysis, MBF and MBF_ratio had a similar area under the curve (AUC) for detecting hemodynamically significant lesions (AUC = 0.942 vs. 0.956, p = 0.413), which were larger than that of visual analysis (AUC = 0.802, both p < 0.01). The vessel-based sensitivity, specificity, and diagnostic accuracy were 84.3%, 97.7%, and 90.5% for MBF and 96.1%, 93.2%, and 94.7% for MBF_ratio.

CONCLUSIONS

Absolute MBF and MBF_ratio had similarly excellent diagnostic performance with reference to FFR. In addition, these two parameters outperformed visual analysis for the detection of myocardial ischemia.

KEY POINTS

• The mean MBF and MBF_ratio were significantly lower in ischemic segments than those in non-ischemic segments. • Absolute MBF and MBF_ratio had similar AUCs for the detection of hemodynamically significant lesions (AUC = 0.942 vs. 0.956, p = 0.413), which were larger than that of visual analysis (AUC = 0.802, both p < 0.01). • The vessel-based sensitivity, specificity, and diagnostic accuracy were 84.3%, 97.7%, and 90.5% for absolute MBF and 96.1%, 93.2%, and 94.7% for MBF_ratio.

From CT to artificial intelligence for complex assessment of plaque-associated risk

The recent technological developments in the field of cardiac imaging have established coronary computed tomography angiography (CCTA) as a first-line diagnostic tool in patients with suspected coronary artery disease (CAD). CCTA offers robust information on the overall coronary circulation and luminal stenosis, also providing the ability to assess the composition, morphology, and vulnerability of atherosclerotic plaques. In addition, the perivascular adipose tissue (PVAT) has recently emerged as a marker of increased cardiovascular risk. The addition of PVAT quantification to standard CCTA imaging may provide the ability to extract information on local inflammation, for an individualized approach in coronary risk stratification. The development of image post-processing tools over the past several years allowed CCTA to provide a significant amount of data that can be incorporated into machine learning (ML) applications. ML algorithms that use radiomic features extracted from CCTA are still at an early stage. However, the recent development of artificial intelligence will probably bring major changes in the way we integrate clinical, biological, and imaging information, for a complex risk stratification and individualized therapeutic decision making in patients with CAD. This review aims to present the current evidence on the complex role of CCTA in the detection and quantification of vulnerable plaques and the associated coronary inflammation, also describing the most recent developments in the radiomics-based machine learning approach for complex assessment of plaque-associated risk.

Hemodynamic Change of Coronary Atherosclerotic Plaque After Statin Treatment: A Serial Follow-Up Study by Computed Tomography-Derived Fractional Flow Reserve

Background

Whether statin treatment can improve hemodynamic status of coronary atherosclerotic plaque remains unknown. It is of clinical interest to explore the hemodynamic change of coronary lesions after statin treatment.

Methods and Results

Consecutive patients with intermediate pre‐test probability of coronary artery disease were prospectively enrolled and underwent baseline coronary computed tomography angiography (CCTA) as well as follow‐up CCTA. The primary end point was to determine the lesion‐specific change of △computed tomography‐derived fractional flow reserve (△CT‐FFR, defined as the change of CT‐FFR value across each lesion) after rosuvastatin treatment. The secondary end point was to compare the change of other plaque characteristics according to serial CCTA findings. 152 patients (mean age: 67.1±9.7 years, 100 men, mean follow‐up duration of 13.9±2.5 months) were finally included. In non‐calcified plaque subgroup, △CT‐FFR was significantly lower at follow‐up compared with baseline (0.051±0.010 versus 0.035±0.012, P=0.013). All other parameters were not found to be significantly different between baseline and follow‐up CCTA measurements. In calcified plaque and mixed plaque subgroups, all parameters showed no significant differences between baseline and follow‐up CCTA groups (P>0.05 for all). According to multivariate regression analysis, non‐calcified plaque was >2 times more likely than calcified plaque to observe the decrease of △CT‐FFR (adjusted hazard ratio: 2.05 [1.03–4.09], P=0.042).

Conclusions

In patients with mild to intermediate coronary stenosis, rosuvastatin treatment resulted in a reduction in lesion‐specific △CT‐FFR at mid‐term follow‐up. This hemodynamic improvement was mainly observed for non‐calcified lesions.

Comparison of Machine Learning Computed Tomography-Based Fractional Flow Reserve and Coronary CT Angiography-Derived Plaque Characteristics with Invasive Resting Full-Cycle Ratio

BACKGROUND

The aim is to compare the machine learning-based coronary-computed tomography fractional flow reserve (CT-FFRML) and coronary-computed tomographic morphological plaque characteristics with the resting full-cycle ratio (RFRTM) as a novel invasive resting pressure-wire index for detecting hemodynamically significant coronary artery stenosis.

METHODS

In our single center study, patients with coronary artery disease (CAD) who had a clinically indicated coronary computed tomography angiography (cCTA) and subsequent invasive coronary angiography (ICA) with pressure wire-measurement were included. On-site prototype CT-FFRML software and on-site CT-plaque software were used to calculate the hemodynamic relevance of coronary stenosis.

RESULTS

We enrolled 33 patients (70% male, mean age 68 ± 12 years). On a per-lesion basis, the area under the receiver operating characteristic curve (AUC) of CT-FFRML (0.90) was higher than the AUCs of the morphological plaque characteristics length/minimal luminal diameter4 (LL/MLD4; 0.80), minimal luminal diameter (MLD; 0.77), remodeling index (RI; 0.76), degree of luminal diameter stenosis (0.75), and minimal luminal area (MLA; 0.75).

CONCLUSION

CT-FFRML and morphological plaque characteristics show a significant correlation to detected hemodynamically significant coronary stenosis. Whole CT-FFRML had the best discriminatory power, using RFRTM as the reference standard.

Machine Learning for Assessment of Coronary Artery Disease in Cardiac CT: A Survey

Cardiac computed tomography (CT) allows rapid visualization of the heart and coronary arteries with high spatial resolution. However, analysis of cardiac CT scans for manifestation of coronary artery disease is time-consuming and challenging. Machine learning (ML) approaches have the potential to address these challenges with high accuracy and consistent performance. In this mini review, we present a survey of the literature on ML-based analysis of coronary artery disease in cardiac CT. We summarize ML methods for detection and characterization of atherosclerotic plaque as well as anatomically and functionally significant coronary artery stenosis.

Diagnostic performance of perivascular fat attenuation index to predict hemodynamic significance of coronary stenosis: a preliminary coronary computed tomography angiography study

OBJECTIVE

This study aimed to investigate the association between perivascular fat attenuation index (FAI) and hemodynamic significance of coronary lesions.

METHODS

Patients with stable angina who underwent coronary computed tomography (CT) angiography and invasive fractional flow reserve (FFR) measurement within 2 weeks were retrospectively included. Lesion-based perivascular FAI, high-risk plaque features, total plaque volume (TPV), machine learning-based FFR_CT, and other parameters were recorded. Lesions with invasive FFR ≤ 0.8 were considered functionally significant.

RESULTS

This study included 167 patients with 219 lesions. Diameter stenosis (DS), lesion length, TPV, and perivascular FAI were significantly larger or longer in the group of hemodynamically significant lesions (FFR ≤ 0.8). In addition, smaller FFR_CT value was associated with functionally significant lesions (0.720 ± 0.11 vs 0.846 ± 0.10, p < 0.001). No significant difference was found between the hemodynamically significant and insignificant subgroups with respect to CT-derived high-risk plaque features. According to multivariate analysis, DS, TPV, and perivascular FAI were significant predictors of lesion-specific ischemia. When integrating DS, TPV, and perivascular FAI, the area under the curve (AUC) of this combined method was 0.821, which was similar to that of FFR_CT (AUC, 0.821 vs 0.850; p = 0.426). The diagnostic accuracy of FFR_CT was higher than that of the combined approach, but the difference was statistically insignificant (79.0% vs 74.0%, p = 0.093).

CONCLUSIONS

Perivascular FAI was significantly higher for flow-limiting lesions than for non-flow-limiting lesions. The combined use of FAI, TPV, and DS could predict ischemic coronary stenosis with high diagnostic accuracy.

KEY POINTS

• Perivascular FAI was significantly higher for flow-limiting lesions than for non-flow-limiting lesions. • Combined use of FAI, plaque volume, and DS provided diagnostic performance comparable to that of machine learning-based FFR _CTfor predicting ischemic coronary stenosis. • No significant difference was found between the hemodynamically significant and insignificant subgroups with respect to CT-derived high-risk plaque features.

The best predictor of ischemic coronary stenosis: subtended myocardial volume, machine learning-based FFRCT, or high-risk plaque features?

OBJECTIVES

The present study aimed to compare the diagnostic performance of a machine learning (ML)-based FFR_CT algorithm, quantified subtended myocardial volume, and high-risk plaque features for predicting if a coronary stenosis is hemodynamically significant, with reference to FFR_ICA.

METHODS

Patients who underwent both CCTA and FFR_ICA measurement within 2 weeks were retrospectively included. ML-based FFR_CT, volume of subtended myocardium (V_sub), percentage of subtended myocardium volume versus total myocardium volume (V_ratio), high-risk plaque features, minimal lumen diameter (MLD), and minimal lumen area (MLA) along with other parameters were recorded. Lesions with FFR_ICA ≤ 0.8 were considered to be functionally significant.

RESULTS

One hundred eighty patients with 208 lesions were included. The lesion length (LL), diameter stenosis, area stenosis, plaque burden, V_sub, V_ratio, V_ratio/MLD, V_ratio/MLA, and LL/MLD⁴ were all significantly longer or larger in the group of FFR_ICA ≤ 0.8 while smaller minimal lumen area, MLD, and FFR_CT value were noted. The AUC of FFR_CT + V_ratio/MLD was significantly better than that of FFR_CT alone (0.935 versus 0.873, p < 0.001). High-risk plaque features failed to show difference between functionally significant and insignificant groups. V_ratio/MLD-complemented ML-based FFR_CT for "gray zone" lesions with FFR_CT value ranged from 0.7 to 0.8 and the combined use of these two parameters yielded the best diagnostic performance (86.5%, 180/208).

CONCLUSIONS

ML-based FFR_CT simulation and V_ratio/MLD both provide incremental value over CCTA-derived diameter stenosis and high-risk plaque features for predicting hemodynamically significant lesions. V_ratio/MLD is more accurate than ML-based FFR_CT for lesions with simulated FFR_CT value from 0.7 to 0.8.

KEY POINTS

• Machine learning-based FFR _CT and subtended myocardium volume both performed well for predicting hemodynamically significant coronary stenosis. • Subtended myocardium volume was more accurate than machine learning-based FFR _CT for "gray zone" lesions with simulated FFR value from 0.7 to 0.8. • CT-derived high-risk plaque features failed to correctly identify hemodynamically significant stenosis.