In vivo comparison of DENSE and CSPAMM for cardiac motion analysis
Journal of Cardiovascular Magnetic Resonance volume 11, Article number: P73 (2009)
Tagging has shown great promise for analyzing cardiac motion patterns . Two different methods have been proposed to utilize the phase associated with harmonic modulation of magnetization to enable tissue tracking. Using Displacement ENcoding with Stimulated Echoes (DENSE)  tagged magnetization is demodulated by decoding gradients permitting motion tracking from the signal phase. In contrast, HARmonic Phase (HARP)  analysis of Complementary SPAtial Modulation of Magnetization (CSPAMM) data decomposes tagged data into its harmonic components during post-processing. It has been argued in principle that the information content of DENSE and HARP should be identical . However, a formal comparison of DENSE and HARP has not been undertaken to date. In this study, DENSE and HARP data were obtained consecutively in the same subjects and circumferential shortening, rotation and time to peak motion were evaluated.
Eight healthy volunteers were imaged using 2D CSPAMM  and 2D DENSE  with identical scan duration (~14 sec). Data were read out using an EPI sequence with the following parameters: TR/TE/α = 30 ms/5.3 ms/20°, acquisition matrix of 96ξ42 (CSPAMM) or 48ξ40 (DENSE) reconstructed to 192ξ192, FOV of 320 mmξ253 mm, slice thickness of 8 mm. Tagging was applied using 2× lines (line distance: 8 mm) employing two orthogonal imaging stacks.
Data were analyzed using TagTrack v.1.8 (GyroTools Ltd, Zurich, Switzerland). For DENSE the echo signal was shifted by 20.8% in k-space (corresponding to 8 mm tag line distance), to create a phase image for HARP processing. CSPAMM data were processed directly with the HARP method using peak combination, demodulated peak combination and conventional single peak method .
The mid-contour inside the left ventricular myocardium was tracked starting from an end-diastolic frame. Initial contours were identical for DENSE and CSPAMM data to reduce observer variability. The left ventricle was segmented into six equidistant sectors and resulting curves for circumferential length and rotation were fitted by a fourth order polynomial. Comparison of both methods was performed using a Bland Altman test.
Figure 1 shows Bland Altman plots for the comparison of DENSE and single peak HARP analysis. The difference between two measurements is presented for the time to peak circumferential shortening (A), the relative circumferential shortening (B), the time to peak rotation (C) and the amount of rotation (D), relative to the average of both measurements. Relative motion measurements show a bigger variation than timing measurements. Table 1 gives a detailed overview of the levels of agreement, bias and the first standard deviation of a comparison between DENSE and peak combined HARP, demodulated peak combined HARP and single peak HARP.
This work has presented a direct comparison of CSPAMM/HARP and DENSE. Results indicate that the both methods agree well when considering bias. However, considerable variation in individual values has been found which may partly be attributed to differences in breath hold position for the two different scans and phase unwrapping errors in demodulation of peak combination HARP. Future work is necessary to identify the cause of this variability.
Axel L: Radiology. 1989
Aletras AH: JMR. 1999
Osman NF: MRM. 1999
Kuijer JPA: JMRI. 2006
Fischer SF: MRM. 1993
Gilson WD: MRM. 2004
Ryf S: JMRI. 2004
About this article
Cite this article
Stoeck, C.T., Kozerke, S., Maredia, N. et al. In vivo comparison of DENSE and CSPAMM for cardiac motion analysis. J Cardiovasc Magn Reson 11 (Suppl 1), P73 (2009). https://doi.org/10.1186/1532-429X-11-S1-P73