Stress T1-mapping is a novel technique for ischaemia testing without the need for gadolinium-based contrast agents. We investigate pitfalls of blinded T1-map analysis, using inter-method agreement between MOLLI and ShMOLLI T1-maps as an illustrative example of their impact on stress T1 results.
We prospectively scanned 18 patients (61±15 years; 72% male) referred for clinical stress CMR for inducible ischaemia, with varied pathology. All patients underwent CMR at 1.5 Tesla including cine, rest and Regadenoson stress T1-mapping using both ShMOLLI and MOLLI 5(3)3, perfusion and LGE imaging. Rest and stress T1-maps were analysed in the mid-ventricular short-axis view by a single human operator, blinded to other CMR images. T1 reactivity (T1stress - T1rest)/T1rest was analysed using original and eroded (mid-wall myocardium) contours on the segmental basis (n=108 segments).
Stress T1 responses were clearly detectable (Table 1). ShMOLLI T1 measurements had a higher T1 reactivity (4%±4%) compared with MOLLI (3%±4%; p<0.01), with both methods showing similar intragroup variability. Head-to-head comparison of ShMOLLI versus MOLLI revealed low positive correlation R=0.15. Minimizing the partial volume as one known source of noise in T1, improved inter-method correlation to R=0.27. Inevitably linked reduction in sample volumes increased the SD and slightly reduced the statistical significance of the apparent difference in reactivity between methods (p<0.03). Inspection of cases with major discrepancies, intended to shed light on unexpectedly weak inter-method correlation, revealed pitfalls in blinded T1-map analysis, such as significant variations in the ROI thickness and failures to include subendocardial infarctions at the blood-myocardium interface (Figure 1).
T1-map analysis blinded to LGE imaging by a human operator affects the stress T1 results using both ShMOLLI and MOLLI techniques. Clear pitfalls identified include significant variations in the ROI wall thickness and missing subendocardial infarctions. Multi-stage image analysis approaches with robust quality control are required to deliver more reliable stress T1-mapping for clinical applications.