DW-CMR remains challenging due to respiratory and heart motion. Recent developments in cardiac diffusion imaging proposed more robust spin echoes encoding scheme [1-4] (Acceleration Motion Compensation (AMC)) to tackle cardiac motion. In addition, free breathing acquisition with prospective motion correction like slice following technique [5] has been shown to reduce efficiently and significantly the scan time. Quality metrics for objective evaluation of sequence and post-processing performances are therefore of crucial interest. We proposed a method to quantify the remaining cardiac or breathing motion corruption in DW-CMR measurement and evaluated with this new quality metric 3 commonly used diffusion encoding scheme: AMC, Stjekal-Tanner (Monopolar) and Twice Refocused Spin Echo (TRSE). Error maps were also compared to physiological motion indicators: cardiac motion using strain measurement and breathing phase using navigator information.

The same acquisition strategy was applied on 7 volunteers. Before the DWI acquisitions, a mid-level 30 phases short-axis (SA) cine SSFP sequence was performed to extract strain and strain rate curves using feature tracking algorithm (CMR42, Circle) (Figure 1). Then, a 2-min ADC protocol was acquired: 5 slices, 6 diffusion directions and b-values 0, 200 s/mm². Five TDs shifted every 10ms were acquired to access cardiac motion by PCATMIP reconstruction [6]. Monopolar, TRSE was acquired in diastole and AMC in diastole and systole with TE=38, 54, 62ms respectively; TR=5s. Prospective breathing management was achieved using a cross-pair navigator and slice following method with a tracking factor of 0.6. For each single image acquisition, navigator-related position, performed immediately before the acquisition window was recorded (Figure 2).

The error maps was calculated from diffusion image weighted signal S(x,y,i) as:

$$ErrorMaps(x,y,i)=\frac{\max (S(x,y,m))-S(x,y,i)}{\max (S(x,y,m))}+\frac{\overline{S(x,y,m)}-S(x,y,i)}{\overline{S(x,y,m)}}+\frac{S0(x,y)-S(x,y,i)}{S0(x,y)}$$

Where i=0,1….m-1 is the screened dimension, with the 6 directions obtained before the Trace map calculation, and S0(x,y) being a non-weighted reference image (Figure 3).

In-vivo comparison revealed a higher score of artifacts for both the Monopolar and TRSE encoding schemes probably accounting for the higher ADC values found using these techniques: 2.71+/-0.93 and 3.13+/-0.93 * 10^-3 mm²/s respectively (Figure 4). Contrarily, AMC appeared robust to cardiac motion with low corresponding values of artefact criterion, and corresponding ADC values that were lower and coherent with the literature for both diastolic and systolic phase: 1.94+/-0.13 and 1.44+/-0.22 * 10^-3 mm²/s, respectively. The coefficient of correlation between Strain measurements and Error measurements were -0.769 and -0.648 for Monopolar and TRSE encoding schemes respectively while no correlation was found for AMC, 0.017 and -0.222 in diastole and systole condition. There was no correlation found in any volunteer between the breathing phase and motion artefact quantification.Cardiac and bulk motions are very critical for ADC measures and impose adequate management and quality metrics able to account for appropriate motion management. Our evaluation pipe-line allows us to assert that the combination of slice following and second order corrected diffusion encoding scheme provide motion independent ADC estimates while minimizing the acquisition time. In the future, quantification of the reliability of the estimate using quality metrics will be mandatory for standardization purposes and to establish the unique added value of DW-CMR in patients.