This retrospective study included patients who experienced myocardial infarction type I (ST-segment elevation and non-ST-segment elevation myocardial infarction [7]) from September 2014 to November 2019. These patients underwent CMR prior to hospital discharge, followed by a second CMR on average six months later. Patients with poor image quality at CMR were excluded from the analyses. All clinical data and follow-up (FU) information were sourced from our institutional database.

High-sensitivity cardiac troponin T (hs-cTnT) was measured (cobas pro, Roche Germany Holding GmbH, cutoff value < 0.014 ng/ml) at admission, 8, 16, 24, 48 and 72 h after PCI and patients were monitored using a standardized follow-up protocol, including annual phone calls for up to 8 years to determine adverse events such as death, myocardial infarction, stroke, bleeding, and hospitalizations. All patients gave their written informed consent for the anonymized use of clinical, procedural and follow-up data at the time of the intervention. This study was approved by the institutional review board and complied with the Declaration of Helsinki.

All CMR-examinations were performed on a 3.0 Tesla scanner (Siemens Magnetom Skyra, Siemens Healthineers, Forchheim, Germany) with patients placed in supine position and using a cardiac coil. Images were acquired at end-expiratory breath hold. A bolus of contrast agent was applied (0.2 ml/kg bodyweight, Magnevist®, Bayer Pharma, Berlin, Germany). 5 min after the bolus, retrospectively gated contrast-enhanced steady-state free precision (SSFP) cine images in short-axis (SAX) stack covering the left ventricle from the base to apex, 2-, 3- and 4-chamber view were acquired. Image parameters for SSFP cines were: TE 1.4 ms; TR 2.9 ms; flip angle 60°; image resolution 1.5 × 1.5 × 8 mm; slice gap 0 mm. No parallel imaging was performed to maximize the signal-to-noise ratio (SNR). 15 min after contrast injection, late gadolinium enhancement (LGE) images were acquired in the same planes as cine images with a phase-sensitive inversion-recovery sequence (TE 3.3 ms, TR 7.0 ms, TI 250–500 ms to null the myocardium, 8 mm slice, no gap, matrix 256 × 192).

Image analysis was performed with dedicated post-processing workstations (syngo.via, Siemens Healthineers AG, Forchheim, Germany; CVI42, Circle Cardiovascular Imaging Inc, Calgary, AB, Canada) by two experienced readers (P.B. and P.R. with > 3 years of experience in CMR and both certified with the highest degree in CMR of the German Cardiac Society) independently.

Volume measurements, LVEF, stroke volume, cardiac index and myocardial mass were semi-automatically assessed using the SSFP-cine images. For all analyses, the endocardial and epicardial borders of the left ventricle were manually traced in all short-axis slices in end-diastole and end-systole. The papillary muscles were excluded from the myocardium. Improvement of LVEF was defined as the delta between LVEF in CMR 1 and in CMR 2. As a parameter for left ventricular remodeling we defined the delta between left ventricular end-diastolic volume (LVEDV) in CMR 1 and in CMR 2.

In order to quantify the myocardial edema in the contrast-enhanced SSFP-cine images [8] of first CMR, a semiautomatic delineation using signal-intensity (SI) thresholds of the hyperintense, edematous region (SI > 2 SD exceeding the mean SI of remote myocardium) was performed in all short-axis slices (in systole). The infarct area was semi-automatically assessed in the short-axis LGE images in both CMR (SI > 5 SD exceeding the mean SI of remote myocardium) (Fig. 1). A hypointense signal within the area of LGE representing microvascular obstruction, if present, was included in the analysis. All automatically assessed areas were visually controlled and adjusted if necessary.

Image analysis—quantification of the infarct area: A SI > 3 SD and B SI > 5 SD. Yellow—infarct area, green—epicardial border, red—endocardial border, blue—remote myocardium, grey—visual adjustment/exclusion line. SI signal intensity, SD standard deviation

For GZF analysis, the mean SI of remote myocardium was adopted to 3 SD SI and the area of enhanced myocardium was semi-automatically assessed. GZF was calculated as follows: 3 SD SI—LGE.

All statistical analyses were performed using Stata (StataCorp LCC, Texas, USA, version 18).

Categorical variables are expressed as frequencies and percentages, continuous variables as mean and standard deviation (SD) or median with interquartile range (IQR). For MI-free survival, univariable and bivariable Cox regression models were conducted, and Harrell’s C and Royston and Sauerbrei’s D were calculated as discrimination measures and compared across different potential predictors. For improvements in ejection fraction, univariable and bivariable linear regression models were conducted and the coefficients of determination R2 were calculated and compared. A value of p < 0.05 was considered as statistically significant. Observer agreement was assessed using the intraclass correlation coefficient, with values above 0.90 indicating excellent reliability.