1354484 - Comparison between conventional T2 mapping and Cardiac Magnetic Fingerprinting derived T2 in a heterogeneous patient cohort

Myocardial T₂ mapping offers direct quantitative evaluation of myocardium for more accurate detection of myocardial infarction (MI) and diffuse myocardial edema/inflammation.¹ cMR fingerprinting (cMRF) allows simultaneous generation of T₁ and T₂ maps in a single scan, enabling increased efficiency and inherent spatial co-registration of tissues between the maps.² However, direct comparison of cMRF-derived T₂ with conventional, clinically used techniques are limited. This study aims to compare between myocardial T₂ cMRF and conventional T₂ mapping in a clinical setting.



Methods:

Twenty patients were scanned under an institutional review board approved protocol on a 1.5T Philips Achieva scanner. Conventional T₂ mapping was obtained using a vendor-supplied respiratory navigator-gated, ECG-gated, gradient and spin echo (GraSE) sequence with inline reconstruction. T₂ cMRF maps were obtained using a breath-held, 18-heartbeat, ECG-gated research sequence. Reconstruction of cMRF maps was performed offline and included scan-specific dictionaries and local low rank reconstruction.³ Both GraSE and cMRF T₂ maps were generated at the mid-ventricular, cardiac short axis position. First, manually contoured myocardium average slice T₂ values were compared between GraSE and cMRF. Second, average T₂ values of regions of interest (ROI) were compared between methods. ROIs were selected based on late gadolinium enhanced (LGE) lesions with known corroborative disease and non-enhanced septal myocardium. Pearson correlation was used in both analyses. Average motion during cMRF was estimated by the maximum displacement of the inferior right ventricular insertion point among the 18 heartbeats.



Results:

The cohort included patients with disease breakdown shown in Table 1. An example of GraSE T₂, T₂ cMRF, and LGE are shown in Figure 1. Moderate correlation was observed between GraSE T₂ and T₂ cMRF slice average (R=0.50, p = 0.036), and GraSE T₂ and cMRF T₂ ROI average (R=0.64, p < 0.001). Bland Altman plots show biases of 12.4 and 15.6ms in the respective analyses. Average intra-scan motion was 5.7mm, with substantial cardiac or respiratory motion observed in 6 patients (Figure 2).



Conclusion:

Myocardial T₂ values from cMRF showed moderate agreement with GraSE in a clinical setting. Agreement was stronger in the ROI analysis than the slice average analysis, which may be explained by ROI analysis having a larger range of evaluated T₂ values relative to measurement variability. A systematic bias between methods was observed, reinforcing a trend observed in previous studies using other cMRF implementations.⁵ We suspect that motion artifacts in cMRF contributed to this lower agreement. We also note the relatively small sample size, which emphasizes the need for continued patient evaluation in a real-world setting. In future, motion compensated reconstruction, a respiratory navigator, reduced breath-hold time, or inline reconstruction may mitigate motion degradation of cMRF.