M y o c a r d i a l P e r f u s i o n E v a l u a t i o n U s i n g O n l y S i x P r o j e c t i o n s 2. Background Efficient computation of high quality 3D densities from projections continues to be a challenge. Early formulations of “statistical reconstruction” [1] are rarely pursued due their overwhelming numerical demands. Recent alternative approaches to reduce CT radiation levels while striving for good image quality include model- based reconstruction and decomposition methods to better account for the nature of the estimation problem. 5. Clinical Relevance, Application Performing tomographic reconstruction under fluoroscopic conditions from a small number of projections implies high statistical efficiency of HECT. Recent numerical improvement optimizing recursions (HECTOR) with an adaptive matrix [4] has further accelerated reconstruction, especially in low-noise settings with many projections. The value of the present technique lies in versatility and efficiency: (i) it may be added to angiographic procedures; and (ii) it may reduce X-ray exposure or MRI measurement times [5] many-fold for traditional 3D tomographic needs. 1. Purpose There is a great need to limit/reduce levels of radiation and measurement time for cardiac evaluation. The purpose here is to add 3-D reconstruction of myocardial blush using the small number of projections available in angiographic procedures. 3. New Solution Building on the Bayesian framework [1] a new numerically efficient Bayesian-derived statistical method is developed by-passing the need for the manipulation of exceedingly large matrices. The new method incorporates constraints such as positive density estimates. The approach allows the use of a small number of projections by combining (i) constraints via non-linear data transformation; (ii) computing near orthogonal data transforms for efficient image density updates; (iii) proceeding with progressive resolution during reconstruction [4]. The major computational burden lies with the few highest resolution iterations – typically 1 to 3. The method incorporates parameters for automated alignment of projections and correction of possible 3D motion artifacts. 36 / SCCT 2012 7. References [1] Wood, S.L., Morf, M., IEEE BME-28, No. 2, pp. 56-68, 1981. [2] Spears, J.R., et al., Cath. Cardiovasc. Diagnosis, Vol. 9, pp. 219-229, 1983. [3] Patent Pending: US-2009-0324047-A1. [4] Patent Pending: US-2010-0278413-A1, and others. [5] Jarisch, W.R., E. Ozarslan, P.J. Basser, ISMRM, Honolulu, Hawaii, 18-24 April 2009. TM Wolfram R. Jarisch Cyber Technology, Potomac, MD 20854, www.4Dtomography.com 6. Acknowledgment This work was performed with the support of J. Richard Spears, MD, at Wayne State University, Detroit, MI, and simulations supported by NIH SBIR grant 1R 43 HL 50187, June-November 1993. 4.1: Angiographic setup at Wayne State University, Detroit, MI. 4.3: Myocardial blush in two consecutive beats. 4.2: Theoretical (a) vs. measured (b) X-ray absorption due to blush. 4.4:Projection of blush after removal of coronary contributions. 4.5: Cut-away view of myocardial blush w/o ligation. 3D tomographic reconstruction was from six projections. 4.6: Left column: control with dye injection into both coronaries; cut-away views. Right column: selective dye injection into the posterior coronary artery; cut-away views. 4.2 Image Pre-Processing For any given projection direction the first four phase points within an R-R interval, free of any blush image defects (e.g. due to late manual injection), were selected for DSA (Fig.4.3). A statistical tracking algorithm applied to DSA images removed the shadow cast by the high- contrast coronaries (Fig. 4.4). 4.3 3D Tomography The new High Efficiency CT (HECT) was applied to sets of the six projection directions for each experimental condition [3]. Progressive resolution made automated mutual alignment of projections very efficient. Figure 4.5 shows a cut-away view of blush (upper portion of the myocardium) during injection of both coronaries. The blush associated with the left ventricle is dominant; blush associated with the right ventricle is too faint to be visible. 4.1 Experimental Procedure Short runs of angiographic data (Fig. 4.1), typically 6-8 beats (Fig. 4.2 a/b) from a mongrel dog preparation [2], were recorded. The data acquisition system (RT-OS Venix) continuously tracked the ECG and gated image buffers with the R-wave. When injection of contrast dye was ready auditory signals suggested manual injection of iodine contrast. During this period the relevant image buffers were saved. With instant playback each fluoroscopic acquisition run was validated (e.g. repeat if chest moved). For any given variation of the animal preparation this acquisition procedure was repeated six times at 30 o increments. Indices show the relative cardiac phase: e.g. 0.75 indicates the state just prior to contraction and 1.00 systole. Estimated dye concentration is indicated by purple shading. Figure 4.6 shows a different set of cuts for blush during injection of both coronaries (left column) as well as posterior coronary only. Both sets are computed for systole. View publication stats View publication stats