Purpose: Measurement of the transverse relaxation time (T2) offers valuable information about the tissue status in a wide range of pathologies. Current approaches for measuring T2 are largely based on multi-echo Carr-Purcell-Meiboom-Gill (CPMG) spin echo sequences, which require long scan times. Fast spin echo (FSE) techniques substantially decreases the scan time, but they introduce unwanted stimulated echo effects. Further, the time-efficient dual-echo FSE sequence generally acquires data with centric ordering of k-space sampling for the first image to enable short echo time (proton-density weighting) and linear ordering for the second long TE (T2 weighting). Differences in k-space profile order can bias T2 values computed from dual-echo FSE.
Methods: An iterative correction of T2 was developed based on Bloch equation simulation using the extended phase graph, accounting for the different profile orders. The amplitude of the echoes corresponding to the center of k-space in each iteration was derived from a simulated echo train using the T2 estimated in the previous iteration. We created a numerical phantom with a range of T2 values similar to normal brain tissues. The acquisition parameters of a clinical dual-echo FSE protocol were used to simulate k-space sampling and signal evolution. T2 maps were created by monoexponential curve fitting, and subsequently with the proposed iterative solution.
Results: The results showed that monoexpnential fitting of dual-echo FSE data may overestimate T2 by 35-45% for T2 values of 60-100 milliseconds. Correction using the proposed iterative solution effectively eliminated the error in T2 map, reducing the error to less than 2%.
Conclusion: The differing k-space sampling order and the stimulated echo contributions cause large errors in T2 values in T2 maps from dual-echo FSE. The proposed correction is important when standardizing or comparing T2 measurements across techniques, and when absolute, rather than relative, values of T2 are desirable.