RPS 1003 - Ischemia and myocardial infarction

Author Block: K. Aisikaier, P. Zhou; Beijing/CN
Purpose: To evaluate the potential mediating role of left ventricular ejection fraction (LVEF) in the relationship between replacement fibrosis (assessed by late gadolinium enhancement (LGE)) and sudden cardiac death (SCD) in patients post-myocardial infarction (MI), and also to assess this mediation effect in subgroups based on LVEF≤35% and>35% according to implantable cardioverter-defibrillator (ICD) selection criterion.
Methods or Background: A retrospective analysis was conducted on 917 post-MI patients (mean age: 56.3±11.0 years, 88.8% male) who underwent cardiac MR from January 2017 to August 2021. The endpoint for SCDs included SCD, aborted SCD, and appropriate ICD discharges. The association of LGE with LVEF was quantified using linear regression models. The associations of LGE and LVEF with SCDs were evaluated using competition risk models. Mediation analysis was then used to decompose the total effect of LGE on SCDs into direct and indirect (mediated through LVEF) effects using accelerated failure time models.
Results or Findings: Over a median follow-up of 63.3 (IQR, 43.6 to 76.6) months, 65 patients (7.1%) experienced SCDs. LGE was significantly associated with lower LVEF (β=-0.35, p<0.001). Both LGE and LVEF independently predicted SCDs (sHR=1.06, p<0.001; sHR=0.95, p=0.03, respectively). Mediation analysis showed that LVEF accounted for 19.7% of the total effect of LGE on SCDs (p<0.001). This mediation effect was 40.4% in patients with LVEF>35% (p=0.02), while no mediation was observed in patients with LVEF≤35% (p=0.08).
Conclusion: LVEF partially mediated the effect of LGE on the SCD, accounting for less than one-fifth of the total effect. LVEF alone inadequately captured the whole SCD risk, irrespective of whether LVEF is greater than 35% or 35% or less.
Limitations: Due to the retrospective nature of the study, the interval between the last MI and CMR varied, which may introduce selection bias.
Funding for this study: This study is supported by the National Key R&D Program of China (Nos. 2021YFF0501400 and 2021YFF0501404), the Key Project of National Natural Science Foundation of China (No. 82430066), and the China International Medical Foundation (No. Z-2014-07-2101).
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study adhered to the principles outlined in the Declaration of Helsinki and received approval from the Institutional Review Board (IRB) and the hospital’s clinical audit department.

Author Block: U. Schmid, A. S. Brendlin; Tübingen/DE
Purpose: Many ACS patients undergo cardiac CT and same-day catheterization, yet only a subset require stenting, leading to possible resource overuse. Photon-counting detector CT (PCCT) provides unique spectral data. We evaluated five classification approaches: clinical data alone (C), radiomics from mixed (M) or spectral high/low energy bins (HL), and early fusion of clinical with mixed (CM) or spectral (CHL) data. We hypothesized PCCT’s inherent informational gain could enhance resource stratification.
Methods or Background: We retrospectively analyzed PCCT cardiac scans from 96 patients undergoing same-day catheterization (15 stented). Clinical data and radiomics from automated AHA myocardial segments (Syngo.via) entered an identical ML pipeline: SMOTETomek, StandardScaler, sequential feature selection (univariate, XGBoost importance, ElasticNet, correlation filtering, RFECV), and an XGBoost classifier with BayesSearchCV tuning. Performance was assessed on a held-out test set via Matthews Correlation Coefficient (MCC), with Bayesian (posterior distributions, BFs, HDIs) and frequentist (bootstrap, adj. p) comparisons.
Results or Findings: Clinical data (C) alone was limited (MCC: 0.427, HDI [0.321–0.530]). Radiomics-only spectral (HL: MCC 0.726, HDI [0.617–0.826]) outperformed mixed (M: MCC 0.513, HDI [0.393–0.626]), confirmed by frequentist (P<0.001) and Bayesian analyses (BF10=2.56). Fused models performed best: CHL (MCC 0.966, HDI [0.929–0.996]) and CM (MCC 0.954, HDI [0.911–0.991]). Bayesian comparisons showed CHL superior to C, M, and HL (BF10>100) but not decisively over CM (Δ=0.012, HDI [−0.049–0.069]; BF10=0.04, P=0.737).
Conclusion: Clinical data alone is insufficient. Spectral radiomics outperform mixed, and fusion (CM, CHL) markedly improves classification. Fusing clinical with spectral radiomics shows highest accuracy, supporting optimized post-CT stratification and cath lab triage.
Limitations: Retrospective single-center design, small cohort, and vendor-specific PCCT/AID implementation. Radiomics extraction and ML pipeline choices may limit generalizability. Clinical endpoints were surrogates without outcome validation. Findings require external validation across centers, scanners, algorithms, and larger populations.
Funding for this study: None
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: University Hospital Tuebingen

Author Block: B. Peters¹, R. Symons², A. Nchimi Longang³, O. Ghekiere¹; ¹Schilde/BE, ²Olv-Waver/BE, ³Strassen/LU
Purpose: Intracoronary pressure measurement is the standard of care for assessing vessel specific ischemia of coronary stenosis. New techniques using artificial intelligence emerge to compute Fractional Flow Reserve(FFR) based on coronary computed tomography angiography(CCTA) images(FFRAI).
We assess a new FFRAI Deep Learning Model for the prediction of vessel specific ischemia compared to two validated invasive physiological measurements.
Methods or Background: We retrospectively selected 322 vessels from 275 patients at two different centers who underwent CCTA and intracoronary catheter measurements with FFR or instantaneous wave-free ratio(iFR) within three months. Two radiologist at each center generated curvilinear reformats of each coronary artery for FFRAI≤0.80 prediction. Sensitivity, specificity, positive predictive value(PPV), negative predictive value(NPV) and diagnostic accuracy of FFRAI were calculated using FFR≤0.80(n=224) and iFR≤0.89(n=238) as standards of reference. Radiologists’ concordance was assessed with Cohen’s κ. A multinomial logistic regression model including all FFRAI false-positive and false-negative was built to evaluate the influence of patient- and CCTA-related parameters on the diagnostic value of FFRAI.
Results or Findings: Sensitivity, specificity, PPV, NPV, and diagnostic accuracy of FFRAI in predicting FFR≤0.80 and iFR≤0.89 were 85%(58/68) and 91%(78/86), 82%(128/156) and 78%(119/152), 67%(58/86) and 70%(78/111), 93%(128/138) and 94%(119/127), and 83%(186/224) and 83%(197/238), respectively. Concordance between senior and junior radiologists’ supervision was substantial(κ=0.725,P<0.001). Calcium score of the stenotic vessel has a significant detrimental effect (p<0.001;OR,1.002;95%CI1.001–1.003) on the diagnostic value of FFRAI.
Conclusion: FFRAI has a high diagnostic performance and reliability for the diagnosis of vessel specific coronary ischemia compared to two intracoronary catheter assessments. Missed FFRAI diagnoses were associated with coronary calcifications.
Limitations: Retrospective design.
CT-FFRAI relies on manually adjusted cMPR images, introducing susceptibility to user,.
CT-FFRAI uses a dichotomous outcome with a threshold of 0.80, which overlooks clinical information pertaining to lesions in the grey-zone.
Funding for this study: No funding
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: Ethical approval in 2 centers: Center 1: 2022/130, 2022; Center 2: 13/06/2023

Author Block: T. Kido¹, H. Taniguchi², M. Takeichi², H. Fujii³, T. Fujioka², N. Fujikawa², T. Yamada², S. Okami², S. Kohsaka³; ¹Toon, Ehime/JP, ²Osaka/JP, ³Tokyo/JP
Purpose: Purpose: For patients with suspected chronic coronary syndrome (CCS), selecting appropriate imaging examinations is crucial for accurate diagnosis and improved outcomes. Advancements in CCTA technology, such as higher resolution and lower radiation doses, have led to new guideline recommendations and changes in reimbursement. However, comprehensive analyses of patient trajectories and practice shifts in the contemporary era have been scarcely studied. The JOURNEY study aims to elucidate these patterns of imaging, medication, and revascularization in real-world settings.
Methods or Background: Methods: The JOURNEY study is a retrospective observational study utilising Japanese hospital administrative database provided by Medical Data Vision Co., Ltd. Patients aged ≥40 years who underwent their first imaging examination (i.e., CCTA, SPECT or CAG) for suspected CCS between 01 March 2015 and 31 January 2023 were included. Patients were followed from the date of their first imaging examination to Day 90. Multivariate logistic regression analysis was performed to examine factors associated with the use of CCTA.
Results or Findings: Results: A total of 123,050 patients with suspected CCS were included. The mean age was 70.0 (SD 10.7) years, and 56.2% were male. CCTA was the first imaging examination for 61.2% of patients, followed by SPECT (19.7%) and CAG (19.1%). Among the patients, 15.7% required the second imaging examination. By Day 90, 20.7% initiated antiplatelet therapy, 17.4% were confirmed to have CCS, and 9.0% received revascularization therapy. Recently enrolled patients underwent CCTA more often than those enrolled earlier (index year 2021-2023 versus 2015-2017, adjusted OR: 1.48 [95% CI: 1.43-1.53]).
Conclusion: Conclusion: CCTA was the most commonly performed initial imaging examination for suspected CCS, with increased usage in recent years.
Limitations: Limitations: The MDV data, collected for reimbursement rather than research, may introduce biases in data accuracy and representativeness.
Funding for this study: Funding was provided by Bayer AG (Leverkusen, Germany).
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study was approved by the independent ethics committee of the specified nonprofit organization, MINS (reference number: 240221).

Author Block: M. Di Giovanni, L. Száraz, S. Beke, K. Nagy, N. Nagy, H. Balogh, S. Veres, J. Simon, P. Maurovich-Horvat; Budapest/HU
Purpose: The coronary artery calcium score (CACS) is an independent predictor of mortality. However, CACS does not provide information on the spatial distribution of calcified plaques. The aim of the present study is to determine the predictive value of the coronary calcium entropy score proposed by our research group taking into account the distribution of calcified plaques regarding the need for further testing after coronary CT angiography (CCTA).
Methods or Background: We retrospectively included consecutive patients who underwent CCTA for suspected coronary artery disease (CAD). We used multivariate logistic regression to fit three diagnostic models. Model 1 included clinical data (age, gender, type of angina, heart rate, and rhythm), which was augmented by CACS (Model 2) and CACS and calcium entropy score (Model 3). Chi-squared test was used to compare the models.
Results or Findings: After exclusion, data from 325 patients was analyzed (35.1% female, 63.7±10.6 years); 55 patients (16.8%) were recommended for further testing. For the clinical parameters (Model 1), presence of arrhythmia significantly increased the likelihood of further investigation. For Model 2, a CACS greater than 100 was a significant predictor (101-400: p=0.029, 401-1000: p=0.02, >1000: p<0.001). When supplemented with the calcium entropy score (Model 3), even a moderately high entropy score (>33th pc) was found to be a significant predictor (34th-66th percentile: p=0.034, >67th percentile: p=0.003). Overall, Model 3 significantly outperformed Model 2 (p<0.001).
Conclusion: The coronary calcium entropy score was found to be a significant predictor of the need for further testing following CCTA beyond clinical data and CACS. Based on our results, the formula developed for the distribution of calcified plaques may have a potential role in the investigation of CAD.
Limitations: This was a single-center study with a retrospective study design.
Funding for this study: M.D.G. received a Research Excellence Grant from Semmelweis University (EFOP-3.6.3-VEKOP-16-2017-00009) and was supported by the New National Excellence Program of the Ministry of Innovation and Technology of Hungary (ÚNKP-23-2-III-SE-43).
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: The study was approved by the Scientific and Research Ethics Committee of the Hungarian Medical Research Council (IV/665-3/2022/EKU) and was carried out in accordance with the tenets of the Declaration of Helsinki.

Author Block: M. Moretti¹, F. Troise², A. Maiorana¹, L. Blasi¹, D. Verna¹, T. D'Angelo³, M. Olivieri⁴, R. Faletti¹, M. Gatti¹; ¹Torino/IT, ²Bari/IT, ³Messina/IT, ⁴Chieti/IT
Purpose: To evaluate the agreement between cardiac computed tomography (CCT) and cardiac magnetic resonance (CMR) in assessing left ventricular (LV) function in post-STEMI patients, and to investigate the impact of CCT acquisitions on restricted R–R intervals, with a view toward potential prospective protocols.
Methods or Background: Forty-eight patients (mean age 62 ± 11 years, 87.5% male) were prospectively enrolled in the multicenter CT-STEMI study (NCT05941585, November 2023 – October 2024). Patients were randomized to undergo either CCT or CMR at 6.8 ± 1.8 days after STEMI, followed by the complementary exam within 5.3 ± 2.8 days. LVEDV, LVESV, and LVEF were measured using CVI42 software. The inclusion of true systolic and diastolic phases was assessed across 40–80%, 30–90%, and 30–100% R–R ranges, along with their corresponding functional estimates.
Results or Findings: Full-cycle CCT demonstrated excellent concordance with CMR (LVEF: 46 ± 11% vs 45 ± 10.9%, p=0.40). The 40–80% interval included 26/48 systolic and 2/48 diastolic phases, with a mean LVEF underestimation of –8.7% (p<0.0001). The 30–90% interval captured all systolic but only 5/48 diastolic phases, leading to an underestimation of –3.4% (p<0.0001). The 30–100% range consistently included both systolic and diastolic phases in all patients (48/48), providing functional measurements comparable to CMR.
Conclusion: CCT shows high agreement with CMR in the evaluation of LV function. However, a 40–80% acquisition window significantly underestimates LVEF. When precise functional assessment is required, extending the reconstruction range to 30–100% allows comprehensive and reliable analysis.
Limitations: These results should be interpreted in light of study limitations, including the small sample size, preliminary nature of the data, variable imaging timing after STEMI, reliance on a single software platform, and lack of outcome or radiation analysis.
Funding for this study: This research was supported by a grant from the Italian Ministry of Health under the “Ricerca Finalizzata 2021 – Giovani Ricercatori” program, project number GR-2021-12372092. The funding was allocated to the project titled “Cardiac Computed Tomography for Comprehensive Risk Stratification of Arrhythmic, Atherothrombotic, and Heart Failure Events Following Reperfused ST-Segment Elevation Myocardial Infarction”. The funder had no role in the design, data collection, analysis, or interpretation of this study
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: No additional information

Author Block: F. Pisu¹, D. Vignale¹, D. Serra¹, A. Palmisano¹, L. Saba², A. Esposito¹; ¹Milan/IT, ²Cagliari/IT
Purpose: Perivascular imaging seeks to quantify inflammation, fibrosis, and remodeling linked to coronary artery disease (CAD). Fat-related biomarkers (e.g., mean attenuation of perivascular adipose tissue) are promising, yet their sensitivity to scanner, protocol and body habitus variations limits reproducibility. Coronary wall (CW) radiomics may reveal more reproducible and subtler remodeling beyond mean attenuation. We tested whether unsupervised machine learning (ML) of CW radiomics can separate obstructive from non-obstructive CAD (ObCAD/NObCAD) and generalize across cohorts.
Methods or Background: We analyzed 40 CCTA cases (public ASOCA dataset: 20 patients with CAD, median stenosis 30% [IQR 25–70], 20 controls). CW radiomics were extracted from the proximal 5 cm of the LAD, RCA, and LCx and harmonized for site/scanners differences. We clustered patients using two principal components from the seven most informative features. Alignment with ObCAD status (≥50% stenosis) and Agatston coronary calcium (CAC) was assessed. Generalizability was tested by projecting 49 institutional patients (38 CAD patients ≤70% stenosis, CCTA for suspected CAD) into the learned cluster space.
Results or Findings: Two radiomic phenotypes separated CAD status with high rule-out performance (specificity 94%, NPV 91%). In the external cohort, 34/49 (69%) mapped to the NObCAD-like and 15/49 (31%) to the ObCAD-like phenotype (specificity 78%, NPV 85%). Misclassified cases had borderline stenosis (50–65%), older age (median 75y) and had more CV risk factors. CAC did not differ between phenotypes (median CAC 0[0–130] vs 18[0–95], p=0.76), supporting remodeling beyond calcium burden.
Conclusion: Machine learning of CW radiomics yields data-driven coronary phenotypes that separate obstructive from non-obstructive CAD and generalize across cohorts, indicating a feasible pathway to subclinical risk stratification beyond stenosis.
Limitations: Small, retrospective cohorts with limited clinical data and no adjudicated outcomes; obstructive CAD defined by stenosis only; residual site/scanner effects despite harmonization.
Funding for this study: The project PNRR-MAD-2022-12376633 is funded by the European Union– Next Generation EU – PNRR M6C2 - Investimento 2.1 Valorizzazione e potenziamento della ricerca biomedica del SSN
Has your study been approved by an ethics committee? Yes
Ethics committee - additional information: All study methods received approval from the ethics committee of IRCCS San Raffaele Scientific Institute (study
registration number CE 158/INT/2022) and adhered to relevant guidelines.