End-expiration respiratory gating for a high-resolution stationary cardiac SPECT system

Abstract

Respiratory and cardiac motions can degrade myocardial perfusion SPECT (MPS) image quality and reduce defect detection and quantitative accuracy. In this study, we developed a dual respiratory and cardiac gating system for a high-resolution fully stationary cardiac SPECT scanner in order to improve the image quality and defect detection. Respiratory motion was monitored using a compressive sensor pillow connected to a dual respiratory-cardiac gating box, which sends cardiac triggers only during end-expiration phases to the single cardiac trigger input on the SPECT scanners. The listmode data were rebinned retrospectively into end-expiration frames for respiratory motion reduction or eight cardiac gates only during end-expiration phases to compensate for both respiratory and cardiac motions. The proposed method was first validated on a motion phantom in the presence and absence of multiple perfusion defects, and then applied on 11 patient studies with and without perfusion defects. In the normal phantom studies, the end-expiration gated SPECT (EXG-SPECT) reduced respiratory motion blur and increased myocardium to blood pool contrast by 51.2% as compared to the ungated images. The proposed method also yielded an average of 11.2% increase in myocardium to defect contrast as compared to the ungated images in the phantom studies with perfusion defects. In the patient studies, EXG-SPECT significantly improved the myocardium to blood pool contrast (p < 0.005) by 24% on average as compared to the ungated images, and led to improved perfusion uniformity across segments on polar maps for normal patients. For a patient with defect, EXG-SPECT improved the defect contrast and definition. The dual respiratory-cardiac gating further reduced the blurring effect, increased the myocardium to blood pool contrast significantly by 36% (p < 0.05) compared to EXG-SPECT, and further improved defect characteristics and visualization of fine structures at the expense of increased noise on the patient with defect. The results showed that the proposed methods can effectively reduce motion blur in the images caused by both respiratory and cardiac motions, which may lead to more accurate defect detection and quantifications. This approach can be easily adapted in routine clinical practice on currently available commercial systems.

Figure 1

(a) Illustration of the compressive sensor pillow attached on the abdomen of the patients that detects the respiratory motion, and the experiment setup for the phantom study. (b) The dual gating box with the compressive sensor pillow that records both the cardiac and respiratory signals, but sends only the ECG triggers to the scanner during end-expiration. (c) A representative patient respiratory trace used to drive the QUASAR motion platform in the second phantom experiment.

Figure 2

The schematic illustrates of the end-expiration ECG gating. The highlighted rectangular boxes denote the end-expiration period where the cardiac triggers (the orange bars) are injected in the listmode stream. The histogram of the cardiac triggers reveals multiple frequencies corresponding to the triggers injected at end-expiration (shorter intervals) and inspiration (longer intervals), respectively. A time threshold τ can be determined from the heart rate for the end-expiration rebinning.

Figure 3

The fused SPECT/CT images of the defect phantom in standard views illustrate the locations and sizes of all the defects. The solid line arrows indicate the location of the transmural defects (air filled) in the septal and anterior walls. The dotted line arrow indicates the location of the nontransmural defect (solid insert) on the endocardial surface of the inferior wall. The non-transmural defect is not as well visualized as the transmural defects due to the smaller size. Note, the SPECT images are smaller than CT due to the smaller FOV and reconstruction matrix size of the SPECT scanner.

Figure 4

(a) A sample patient respiratory trace exported from the dual gating box. The circles denote the sharp transition points right after the inspiration peaks determined by the dual gating box prospectively, and the crosses denote the cardiac triggers that were injected into the listmode data. (b) The correlation between the percentages of end-expiration period calculated from the listmode data, and the respiratory traces on 7 patient studies. The discrepancy between these two measurements was mainly due to the synchronization between the emission acquisition and trace recording achieved manually by two operators , and slight synchronization offset between these two acquisitions occurred in some studies.

Figure 5

(a) The reconstructed images of the normal phantom in standard views. The dotted arrow denotes the region that was contaminated by the spill-out activity from the liver. (b) Count profiles from phantom along the vertical (top) and horizontal (bottom) arrows shown on the SA view. The greatest improvement of the myocardium to blood pool contrast is achieved by EXG-SPECT in the vertical direction, which was the primary direction of the phantom motion.

Figure 6

(a) The reconstructed images of the phantom with defects in standard views. The anterior (solid blue), inferior (dashed yellow) and septal (dotted green) wall defects are denoted by the colored arrows. (b) The circumferential count profiles of Static vs NMC-SPECT (left), and Static vs EXGSPECT (right). The yellow solid arrows denote the location of the defects, and white dotted arrows denote the reduced activity in normal myocardium due to respiratory motion.

Figure 7

(a) The myocardial perfusion images of a representative patient in standard views. This patient was injected with 577 MBq ^99mTc-tetrofosmin and imaged for 4 mins. 36% counts were preserved in the EXG-SPECT images. Note 100% counts were used in the NMC images to illustrate the increase of noise due to reduced counts. The solid arrows denote the right ventricle that became better visualized in the EXG-SPECT images. (b) Count profiles for the horizontal and vertical lines drawn on the SA view demonstrate the improvement of the myocardium to blood pool contrast achieved with EXG-SPECT imaging.

Figure 8

Myocardial perfusion images of a representative patient showing a moderate sized defect (denoted by the solid arrows) extending from the mid-ventricle to the base. The images are shown in (a) SA view (top), VLA view (bottom), and (b) polar maps. The end-diastolic phase is shown for the dual gated images. The defect was blurred by respiratory motion in the NMC-M SPECT images. EXGSPECT and dual gated-SPECT yielded a better contrast resolution within the defect. Note the papillary muscle (denoted by the dotted arrow) is only visualized in the dual gated images demonstrating the effectiveness of respiratory and cardiac motion reduction achieved by dual gating. However, dual-gating also yielded a less uniform polar map due to the increase of noise in (b).

Figure 9

Circumferential count profiles of (a) NMC-M (yellow) vs EXG (red) SPECT images, and (b) NMC-M (yellow) vs Dual Gated (red) SPECT images at apical, mid-ventricle, and basal levels for the same patient as shown in Fig. 8. The yellow solid arrows denote the defect. The dashed horizontal lines show the smallest % of maximum achieved by the EXG-SPECT to aid the comparison between EXG and Dual Gated SPECT. The defect size was diminished in the NMC-M SPECT images due to respiratory motion. EXG reduced respiratory motion and yielded a more pronounced defect. Dual-gating revealed an even increased defect size by reducing both respiratory and cardiac motions.

Figure 10

(a) The myocardium to blood pool contrast for all the patients. The definitions of the ROIs in the myocardium and blood pool are overlaid on a sample VLA slice in the top left corner. (b) EXGSPECT and dual gated SPECT improved the mean contrast from 4.38 to 5.43, and 7.37 across 11 patients compared to NMC, respectively. Data are expressed as mean ±SEM. *p<0.05. **p<0.005.

Figure 11

Inter-segment variability measurement across 8 normal studies. On each polar map, the inner, middle and outer rings correspond to apical, mid-ventricle and basal regions, respectively. The value in each segment denotes the percentage deviation of its mean value from the mid-septal wall averaged across 8 normal studies. A uniform polar map should have similar mean value in all the segments, thus smaller deviations. It can be seen that NMC 100 and NMC-M yielded similar variability in all the segments, while EXG yielded less variability than NMC-M in all the regions except in the basal inferior wall.