Quick Fire Session
SCMR 22nd Annual Scientific Sessions
Diffusion tensor imaging (DTI) enables whole-heart measurements of microstructural organization and is frequently combined with tractography. Tractography, however, depends on several subjective parameters to generate “myofiber tracts”. Alternatively, each pixel in a tensor field can be defined by three shape-based components (invariant gradients ) and three orientation-based components (rotation tangents; ). Tensor field characteristics derived from these tensor components, such as curvature, dispersion, and twist, have physiological meaning in the context of the heart and can be calculated directly from a tensor field without introducing intrinsic biases from tractography algorithms. Also, this approach does not require an ad hoc local coordinate system as is needed for helix angles calculations. Here, we demonstrate the application of this technique for characterizing the cardiac microstructure. We also simulate the effect of noise on characterizing fiber curvature, dispersion, and twist with different signal-to-noise ratios (SNR) in order to understand measurement feasibility.
Simulated diffusion-weighted images with SNR∈[1,100] were generated as outlined in Fig. 1. Curvature, dispersion, and twist (derived from rotation tangents ) were calculated from synthetic tensor fields. The error as a function of SNR was measured as |(means with noise - means without noise)/means without noise|*100%.
Fresh normal swine hearts (N=13) arrested in diastole underwent ex vivo DT-MRI at 3T (Siemens, Prisma) with spatial resolution 1x1x1 mm3, b-value 1000s/mm2, 30 diffusion gradient directions, and 5 averages. Rotation tangents and invariant gradients were calculated (Matlab) and reported in the LV according to the AHA 16-segment model.
Error in curvature, dispersion, and twist measurements decreases exponentially with increased SNR and stabilize around SNR = 25 to 10% (Fig.2).
The measured SNR of the ex vivo DT-MRI was approximately 15-20. Curvature and twist are higher in the apical regions of the LV compared to the rest of the LV (p = 0.06 and 0.02); twist is significantly higher in the septum compared to the anterior, posterior, and lateral free walls (p<0.01, =0.02, =0.02) (Fig.3).
Curvature, dispersion, and twist derived from rotation tangents are new approaches to diffusion tensor analysis and provide intuitive, quantitative metrics for understanding the microstructural organization of the heart. The bias in these measurements from the current DTI protocol is approximately 10%. Histological studies in normal and pathological conditions will help establish functional implications in changes in rotation tangents. We acknowledge the support from the Sarnoff Foundation, NIH K25 HL135408 and UCLA Radiological Sciences.