Phase Ordering with Automatic Window Selection (PAWS): A Novel Motion Resistant Technique for 3D Coronary Imaging P. Jhooti, P.D. Gatehouse, J. Keegan, N. Bunce, A.M. Taylor and D.N. Firmin Magnetic Resonance Unit, Royal Brompton Hospital, London, UK. Navigatoracceptance imaging methods are hindered by the loss in scan efficiency which results from the changesin the breathing pattern of a subject over time. The Diminishing Variance Algorithm (DVA), which doesnot use a predefined acceptance window, is less influenced by such changes(1). The use of phase ordering and weighting techniques have been shown to significantly improve image quality and attempts ha<e been made to combine the DVA approach with these methods (2,3). However, the use of an acceptancewindow is inherent in all thesi techniquesas a decision to accept or reject data must still be made. This is prohibitive if the breathing pattern changes significantly during the acquisition. A technique is presentedwhich is resistant to changes in breathing whilst allowing the use of phase ordering to provide effective motion artefact reduction in an optimal time. METHOD The proposed technique, Phase-ordering with Automatic Window Selection (PAWS), uses a multi-level approach where no acceptance window is specified. Instead, the initial position of the diaphragm is taken as the reference and all further diaphragm positions are given au index position which is the displacement from this reference. Each, index position is allotted a starting position in a region of k-space, hereafter called a “bin”, as shown in Figure 1. The k-space region is Figure 1. tilled according to the direction of the arrows. Data acquisition for positions on the edgesof k-space is straightforward, phase encode lines are acquired sequentially towards the centre of k-space. For positions which start in the centre of k-space,however, data is more complicated. Each diaphragmposition can be used to completek-spacein one of three possible combinations. For example, for diaphragmposition 4, the three possible combinations which can complete k-spaceare: 2-3-4, 3-4-5 and 4-5-6. All three combinations are checkedto determine which is closest to completion and this information is used to determinewhether the right or left side of k-spaceis acquired. The side which has most unfilled ky lines remaining is acquired. This is demonstrated in Figure 1 using a real exampleduring a scanwhereby the shaded regions indicate phaseencode lines which have been acquired thus far. For the case where the diaphragm is currently at position 4, the window nearest to completion at this moment in time is 4-5-6. As there are more unfilled phase ericode lines remaining between 4-6 than 4-5, the phase encode line toward diaphragmposition 6 is acquired. Image acquisition is complete once the whole of k-space has been acquired by 3 contiguous bins. The use of phaseordering limits motion between successive phaseencode lines and motion artefacts are therefore minimised. Scti time can further be reduced by allowing more than one index position to map to each individual bin. However, image quality may be compromised as the phase ordering within each individual bin is less well defined. To limit motion in the centre of k-space when the bin size is greater than one, a weighted bin schemecan be employed whereby an increased bin size is used for the two outer bins whilst the central bin sizeremainsone. Simulations were carried out to compare the proposed method with the DVA method using respiratory traces acquired from 15 subjects. The DVA method allows a scan to be terminated after a specified time or when the desired range of motion has been achieved. Therefore, for each different window size, two comparisons were made. The first was a comparisonof scantime eachmethod required to acquire a complete data set within the specified acceptance window. The second was a comparison of image quality each method achieved with the scan time equal to that required by the PAWS method. In viva studies have also been carried out with 10 normal subjects. The PAWS method, using a 5mm weighted scheme, was compared with the DVA technique. The thin 3D slab was oriented to lie in the plane of the right coronary artery (RCA). The images were analysed in a blinded fashion by two independent observers and scoredbetween 1 (very good) and 10 (very poor). RESULTS/DISCUSSION The table below shows the percentage difference between cardiac cycles required to acquire 128 phase encode lines using the DVA and PAWS technique compared to the optimum scan time. The PAWS approach achievesa shorter scan time for each window size and, combined with the low standarddeviation achieved by the PAWS method, these results suggestthat PAWS consistently achievesa scantime as good, or almost as good, as the shortestscan time possiblein eachcase. 3mm (PAWS: Bin l-l-l) 5mm (PAWS: Bin 2-l-2) 6mm (PAWS: Bin 2-Z-2) A typical displacement graph for eachmethod is shown in Figure 2. Figure 2. 0 kvline 64 Results in vlvo have also shown to be very promising. Analysis with Wilcoxon signed rank testing showed that the PAWS scores were significantly higher than those achieved by DVA (P<O.Ol). Figure 3 shows a single reconstructed slice for each method. The PAWS image clearly shows the RCA at its origin on the aortic root and several centimetres of its path in the atrio-ventricular groove. The coronary artery is not visible in the DVA image. Figure 3. PAWS DVA Conclusion A method is introduced which is resistant to changes in breathing and allows images to be acquired in the shortest possible scan time with no requirement for operator interaction. The ability to integrate the use of phase ordering with a flexible window selection strategy further allows improvements in image quality. This method has proved to be effective both in improving scan efficiency and reducing respiratory motion ‘tiefacts in our studies. We believe this technique will aid in overcoming many of the current problems faced with navigator acceptance techniques, providing an ordering technique robust against changes in breathing which has previously not beenpossible. REFERENCES 1. Sachs, T.S. et al. MRM34:412-422, 1995 2. Jhooti, P. et al. JM2Z 8:968, 1998 3; Sinkus, R. et al. MM41:148-155, 1999 Proc. Intl. Sot. Mag. Reson. Med. 8 (2000) 403