Paul B. Kavanagh

An automatic approach to the analysis, quantitation and review of perfusion and function from myocardial perfusion SPECT images.

We have developed a software suite that automatically selects, analyses, quantitates and displays all the key image data in a myocardial perfusion SPECT study. Methods: The files automatically selected (upon specification of the patient name) are rest and stress projections, rest and stress short axis and gated short axis files, and all ‘snapshot’ files. The projection data sets are presented in cine mode for evaluation of patient motion, while the lung/heart ratio at rest and stress is calculated from regions of interest (ROIs) that are automatically derived and overlayed on the LAO 45 images. Left ventricular (LV) cavity volumes at rest and stress are calculated from the short axis data sets, and the related transient ischemic dilation (TID) ratio derived and displayed. Quantitative measurements of global (ejection fraction) and regional function parameters are performed from the gated short axis dataset. All algorithms use the C++, X-Windows and OSF-Motif standards. The overall suite executes in less than 1 minute on a SunSPARC5 with 32 Mb of RAM and no proprietary hardware. Results: The software was validated on 144 patients (118 rest201 Tl/post-stress 99mTc-sestamibi, 18 post-stress 99mTc-sestamibi, 8 rest 201Tl) acquired on a 90° dual detector (ADAC Vertex, 91 patients) and a triple detector camera (Picker Prism 3000, 53 patients). Overall, the individual algorithms for the analysis of projection, short axis and gated short axis images were successful in 622/660 (94.2%) of the images. In 80.5% of the patients (73/91+43/53) all algorithms executed successfully, without significant difference in success rates for201 Tl versus 99mTc-sestamibi images. Conclusion: Our automated approach to myocardial perfusion SPECT analysis and review is highly successful, intrinsically reproducible, and can produce time and cost savings while improving accuracy in a clinical or teleradiology-type environment.

Repeatability of automatic left ventricular cavity volume measurements from myocardial perfusion SPECT

BackgroundThis study sought to assess the repeatability of automatic quantitative measurements of left ventricular (LV) cavity volumes in a large patient population (N=926), to correlate those measurements to similarly obtained LV ejection fraction (LVEF) measurements, and to investigate the relationship between ungated and gated volumes.MethodsAll 926 patients underwent ungated single photon emission computed tomography (SPECT) immediately followed by 8-frame gated SPECT. LV cavity volumes were automatically measured from ungated (V), summed gated (SUMV), end-systolic (ESV) and end-diastolic (EDV) images, and LVEFs derived from the latter 2.ResultsRepeatability (SUMV vs V) was very good overall (6.4%±6.6%), further improving for volumes >25 mL (5.7%±5.5%) and >40 mL (5.2%±5.0%). Exponential regression between ESV and LVEF (r=0.925, SEE=15.0 mL) EDV and LVEF (r=0.802, SEE=24.2 mL), and SUMV and LVEF (r=0.867, SEE=19.7 mL) was also very good. Summed gated volumes were closer to ESV than to EDV (43.3%±8.8% of EDV-ESV range). SUMV <50 mL and SUMV >110 mL were good substitutes for LVEF >50% and LVEF <40% (93.4% and 97.1%, respectively).ConclusionAutomatic quantitative measurements of gated and ungated volumes with our algorithm are repeatable, correlate well with other global myocardial parameters, and may contribute important additional information to that conventionally provided by myocardial perfusion SPECT studies.

Comparison of post-stress ejection fraction and relative left ventricular volumes by automatic analysis of gated myocardial perfusion single-photon emission computed tomography acquired in the supine and prone positions

BackgroundWe have previously described an automatic method for measuring left ventricular ejection fraction (LVEF) for myocardial perfusion single-photon emission computed tomography (SPECT). The repeatability of this method has not been previously described.Methods and ResultsThis study compares LVEF and relative end-systolic and end-diastolic volumes assessed from myocardial perfusion SPECT by our automatic method in 180 consecutive patients undergoing gated myocardial perfusion SPECT with injection of 99mTc-labeled sestamibi in whom the acquisitions were performed sequentially in supine and prone positions. The algorithm operated completely automatically in the prone and supine positions in 178 of the 180 patients. Very high correlations were observed for LVEF (r=0.93), relative left ventricular end-systolic volume (r=0.98), and relative left ventricular end-diastolic volume (r=0.97). The mean paired absolute difference between LVEFs in the prone and supine position was 3.8±3.2, for left ventricular end-systolic volume was 4.9±4.8 ml, and for left ventricular end-diastolic volume was 7.4±6.7 ml. When patients were classified by the extent and severity of stress perfusion defect, there was no significant difference in repeatability for the measurements in any category.ConclusionsOur algorithm for automatic quantification of LVEF and relative end-systolic and end-diastolic volumes from gated 99mTc sestamibi myocardial perfusion SPECT is repeatable. When performed in the prone position, values of ejection fractions and ventricular volumes are essentially identical to those obtained in the supine position.

Automated Quality Control for Segmentation of Myocardial Perfusion SPECT

Left ventricular (LV) segmentation, including accurate assignment of LV contours, is essential for the quantitative assessment of myocardial perfusion SPECT (MPS). Two major types of segmentation failures are observed in clinical practices: incorrect LV shape determination and incorrect valve-plane (VP) positioning. We have developed a technique to automatically detect these failures for both nongated and gated studies. Methods: A standard Cedars-Sinai perfusion SPECT (quantitative perfusion SPECT [QPS]) algorithm was applied to derive LV contours in 318 consecutive 99mTc-sestamibi rest/stress MPS studies consisting of stress/rest scans with or without attenuation correction and gated stress/rest images (1,903 scans total). Two numeric parameters, shape quality control (SQC) and valve-plane quality control, were derived to categorize the respective contour segmentation failures. The results were compared with the visual classification of automatic contour adequacy by 3 experienced observers. Results: The overall success of automatic LV segmentation in the 1,903 scans ranged from 66% on nongated images (incorrect shape, 8%; incorrect VP, 26%) to 87% on gated images (incorrect shape, 3%; incorrect VP, 10%). The overall interobserver agreement for visual classification of automatic LV segmentation was 61% for nongated scans and 80% for gated images; the agreement between gray-scale and color-scale display for these scans was 86% and 91%, respectively. To improve the reliability of visual evaluation as a reference, the cases with intra- and interobserver discrepancies were excluded, and the remaining 1,277 datasets were considered (101 with incorrect LV shape and 102 with incorrect VP position). For the SQC, the receiver-operating-characteristic area under the curve (ROC-AUC) was 1.0 ± 0.00 for the overall dataset, with an optimal sensitivity of 100% and a specificity of 98%. The ROC-AUC was 1.0 in all specific datasets. The algorithm was also able to detect the VP position errors: VP overshooting with ROC-AUC, 0.91 ± 0.01; sensitivity, 100%; and specificity, 70%; and VP undershooting with ROC-AUC, 0.96 ± 0.01; sensitivity, 100%; and specificity, 70%. Conclusion: A new automated method for quality control of LV MPS contours has been developed and shows high accuracy for the detection of failures in LV segmentation with a variety of acquisition protocols. This technique may lead to an improvement in the objective, automated quantitative analysis of MPS.

Effect of the number of projections collected on quantitative perfusion and left ventricular ejection fraction measurements from gated myocardial perfusion single-photon emission computed tomographic images

BackgroundGated myocardial perfusion single-photon emission computed tomographic (SPECT) imaging is currently performed by step-and-shoot detector rotation, resulting in acquisition dead time and lengthened study duration compared with nongated SPECT imaging with continuous or pseudocontinuous rotation. Dead time is particularly undesirable in new fast-gated SPECT imaging protocols with inotropic pharmacologic stress.Methods and ResultsThis article evaluated the influence of projections’ angular spacing on quantitative measurements of left ventricular ejection fraction (LVEF) and perfusion from postexercise 99mTc-labeled sestamibi images. Gated 60-projection data sets from 30 patients were compacted into 30- and 15-projection sets. The three sets (corresponding to 3-, 6-, and 12-degree spacing over 180 degrees) were reconstructed into gated and ungated short-axis image sets. LVEFs were measured from the gated images according to a previously described automatic algorithm, whereas perfusion was assessed from the ungated images by a 20-segment division of their maximal pixel polar maps. LVEF values were essentially unchanged between 60- and 30-projection images (y=0.37+0.996x; r=0.999; standard error of the estimate=0.56) and 60- and 15-projection images (y=1.35+0.987x; r=0.999; standard error of the estimate=0.77) in the 30 patients. Overall, 30- and 15-projection polar maps differed by 1.87%±1.24% and 4.38%±2.25% from the 60-projection polar maps, respectively. Segmental perfusion score agreement between 60- and 30-projection images and between 60- and 15-projection images was 93% (κ=0.92; p<0.001) and 83% (κ=0.81; p<0.001), respectively. Sixty- and 30-projection images were visually undistinguishable, whereas loss of image resolution was noticed in many 15-projection gated and ungated images.ConclusionsThirty-projection gated SPECT imaging is a practical, accurate, and time-saving approach in standard gated protocols and, potentially, fast-gated protocols. Fifteen-projection gated SPECT imaging is not generally recommended and should be considered only for LVEF assessment in conjunction with fast-gated protocols.

Tracking a therapeutic response: How reliable are serial measurements of LV perfusion and function?

Why are serial measurements of cardiac perfusion and function important? Simply put, because they allow us to make a precise assessment of the worsening of patient conditions due to disease progression, or the improvements brought about by treatment. Our ultimate goal is that of reducing the random “noise” present in the measurements, so that even small differences found between successive studies can be considered significant, and can guide informed decisions about the patients’ clinical pathways. Before examining and discussing the specifics of the issue, it is appropriate to make an important distinction between reproducibility and repeatability. We consider reproducibility of measurements of perfusion and function from gated SPECT and gated PET images to refer to applying a computer algorithm twice to the same image set, and therefore to be proportional to the degree to which the algorithm is automated. Measurements from automated, “push-button” algorithms that operate in a deterministic manner are, by definition, perfectly reproducible, regardless of when or by whom the button is pushed. On the other hand, algorithms that require some degree of operator interaction will have less than perfect intra-operator and inter-operator reproducibility. (1) Conversely, repeatability of measurements is related to applying a computer algorithm to two separately acquired image sets belonging to the same patient (images ideally acquired without changes in the acquisition protocol or patient condition), and measuring the difference between the quantitative results. Low repeatability can be due to either a) changes in the acquisition setup (the patient moved, the radiopharmaceutical uptake pattern changed, gating abnormalities occurred), b) changes in the reconstruction and reorientation parameters used to generate the tomographic images input to the algorithm, c) changes in the way the quantitative algorithm operated, or was applied to the data, and/or d) true physiologic variation in the patient’s state at imaging. In other words, repeatability is a measure of the combined “stability” of the quantitative algorithm, the acquisition protocol itself, and the patient conditions (fig. 1). Fig. 1 While serial measurements fall obviously under the repeatability category, examining the potential limitations of reproducibility offers us a way to determine, in a more controlled environment, the effect of operator interaction on quantitative measurements. Specifically, we can ascribe imperfect reproducibility to variabilities in processing (by which tomographic images are produced from the acquired “raw” projection images) as well as variabilities in quantification Δrepro=Δproc+Δquant The two contributions are obviously correlated, as Δproc will cause changes in the images on which quantification is performed, which may in turn cause changes in the quantitative software’s output. For example, a paper by Knollman et al. investigated the variation in quantitative LV function measurements resulting from a simple 15 degrees change in the reorientation angles during processing of 59 gated SPECT datasets. The patient population comprised a good range of LVEFs (20%–80%) end-systolic volumes (ESVs) and end-diastolic volumes (EDVs), which were quantified using three different software algorithms - and while correlation coefficients were generally high, the most reproducible algorithm still produced differences of 2.8%, 7.5ml and 9ml for LVEF, ESV and EDV, respectively, at the 95% confidence interval level (2). In layman’s terms, 5% of the time one should expect quantitative differences greater than those listed above as a consequence of just altering the reorientation angle by 15 degrees during reconstruction. Changing the type of reconstruction applied to a dataset can impact reproducibility in an even more severe way, as reported by DePuey et al for filtered backprojection vs. different implementations of iterative reconstruction (3). In the extreme, this may require the need for reconstruction-specific normal limits for parameters of cardiac function, and possibly for cardiac perfusion as well. With respect to repeatability, we can consider it as comprising the above-described Δrepro component plus an acquisition-related component Δacq, further composed of an undesirable component Δacq_tech (variabilities related to technical causes), and a desirable component Δacq_clin (variabilities related to actual clinical changes). Δrepeat=Δrepro+Δacq=Δrepro+Δacq_tech+Δacq_clin It is apparent that, while reducing Δrepro is always beneficial in terms of reliability of the measurements, special care is needed to reduce Δrepeat without compromising the sensitivity to detect meaningful clinical changes. In order to investigate the typical Δrepeat in this context, a study was conducted on a population of 100 patients, who were imaged back-to-back with gated SPECT, twice at rest and twice after exercise or pharmacologic stress. Since a) the patients’ status was not expected to have changed between the consecutive rest or the consecutive stress acquisitions, b) each consecutive study pair shared the same radioisotope injection and c) the same cameras and protocols were used for every study pair, this is close to a best-case scenario for repeatability, and the differences measured ought to be directly related to the effect of manual processing and successive, operator-adjusted quantification (4). Not surprisingly, it was found that Δrepeat was generally higher (worse) than the Δrepro reported by Knollman, reflecting the greater variability associated with separate acquisitions. Perhaps more interestingly, repeatability was better (Δrepeat was lower) for stress studies than for rest studies, as a consequence of the formers’ higher statistical quality. Of note, when a 65 patients subsample was considered in which no operator adjustment to the automated quantification algorithm was necessary, repeatability significantly improved for the rest, but not for the stress studies – suggesting a clear benefit of fully automated analysis in the context of lower count images (Table 1). TABLE 1 Repeatability (Δrepeat, 95% confidence interval) of cardiac function measurements from back-to-back gated SPECT acquisitions. Standard processing and quantification were used (4). Repeatability of perfusion SPECT measurements were analyzed in the same 100-patient population using the “total perfusion deficit” (TPD) parameter previously defined by Slomka et al (5), producing 95% confidence intervals of 3.3%, 1.8% and 3.2% for stress TPD, rest TPD and ischemic TPD, respectively (4). Another study by Mahmarian et al investigated the change in “perfusion defect size” (PDS), a parameter similar to the TPD, in a population of 260 patients who underwent serial adenosine stress SPECT within 4 weeks of each other, without changes in clinical status or medications. This test of repeatability is obviously more challenging compared to the single-injection, back to back scenario in Xu’s study, and consequently the mean serial difference in PDS was reported to be (−0.13+/−4.2%), corresponding approximately to an 8.4% variation at the 95% confidence interval level - a result comparable to what was reported by Berman et al in serial studies repeated within 9–22 months (6), and significantly better than that previously reported for visual agreement (7). Summarizing what we know about optimizing repeatability, it is essential that serial studies be required to differ as little as possible in the way of acquisition (same camera, radioisotope, acquisition protocol, comparable count statistics, etc), processing (same reconstruction type, filter type and cutoff, reorientation software, etc) and quantification (same algorithm), while also minimizing the extent of operator intervention throughout the process. Even with these precautions, published data indicates that the best achievable repeatability using conventional approaches can be expected to be about 2–3% for TPD, 5–6% for LVEF, and 10 ml for EDVs and ESVs, all at the 95% confidence interval level.

Direct Quantification of Left Ventricular Motion and Thickening Changes Using Rest–Stress Myocardial Perfusion SPECT

Changes in myocardial wall motion and thickening during myocardial perfusion SPECT are typically assessed separately from gated studies for the presence of stress-induced functional abnormalities. We sought to develop and validate a novel approach for automatic quantification of rest–stress myocardial motion and thickening changes (MTCs). Methods: Endocardial surfaces at the end-diastolic and end-systolic frames for rest–stress studies were registered automatically to each other by matching ventricular surfaces. Myocardial MTCs were computed, and normal limits of change were determined as the mean and SD for each polar sample. Normal limits were used to quantify the MTCs for each map, and the accumulated sample values were used for abnormality assessments in segmental regions. A hybrid method was devised by combining the total perfusion deficit (TPD) and MTC for each vessel territory. Normal limits were obtained from 100 subjects with low likelihood of coronary artery disease. For validation, 623 subjects with correlating invasive angiography were studied. All subjects underwent a rest–stress 99mTc-sestamibi exercise or adenosine test and coronary angiography within 3 months of myocardial perfusion SPECT. All MTC and TPD measurements were derived automatically. The diagnostic accuracy for detection of coronary artery disease for MTC plus TPD was compared with TPD alone. Results: Segmental normal values were between −1.3 and −4.1 mm for motion change and between −30.1% and −9.8% for thickening change. MTC combined with TPD achieved 61% sensitivity for 3-vessel-disease (3VD), 63% for 2-vessel-disease (2VD), and 90% for 1-vessel-disease (1VD) detection, compared with 32% for 3VD (P < 0.0001), 53% for 2VD (P < 0.001), and 90% for 1VD (P = 1.0) detection using the TPD-alone method. The specificity for the combined method was 71% for 3VD, 72% for 2VD, and 47% for 1VD detection versus 90% for 3VD (P < 0.0001), 80% for 2VD (P < 0.001), and 50% for 1VD detection (P = 0.0625) for the TPD-alone method. The accuracy of 3VD detection by MTC plus TPD was higher (69%) than the accuracy of TPD plus change in ejection fraction (63%) (P < 0.004). Conclusion: We established normal limits and a novel method for computation of regional functional changes between the rest and poststress studies. Compared with TPD alone, the combination of TPD with MTC improved the sensitivity for the detection of 3VD and 2VD.

Improved Quantification and Normal Limits for Myocardial Perfusion Stress–Rest Change

We aimed to improve the quantification of myocardial perfusion stress–rest changes in myocardial perfusion SPECT (MPS) studies for the optimal automatic detection of ischemia and coronary artery disease (CAD). Methods: Rest–stress 99mTc MPS studies (997 cases; 651 consecutive cases with correlating angiography and 346 cases with less than 5% likelihood (low likelihood [LLK]) of CAD) were analyzed. Normal limits for stress–rest changes were derived from additional LLK patients (40 women, 40 men). We computed the global stress–rest change (C-SR) by integrating direct stress–rest changes for each polar map pixel. Additionally, stress–rest change and total perfusion deficit (TPD) at stress were combined in 1 variable (C-TPD) for the optimal detection of CAD. Results: The area under the receiver-operating-characteristic curve (AUC) for C-SR (0.92) was larger than that for stress TPD–rest TPD (0.88) for the identification of stenosis of 70% or more (P < 0.0001). AUC (0.94) and sensitivity (90%) for C-TPD were higher than those for stress TPD (0.91 and 83%, respectively) (P < 0.0001), whereas specificity remained the same (81%). Conclusion: C-SR and C-TPD provide higher diagnostic performance than difference between stress and rest TPD or stress hypoperfusion analysis.

Quantification of 3D regional myocardial wall thickening from gated magnetic resonance images

To develop 3D quantitative measures of regional myocardial wall motion and thickening using cardiac magnetic resonance imaging (MRI) and to validate them by comparison to standard visual scoring assessment.

Myocardial wall thickening from gated magnetic resonance images using Laplace's equation

The aim of our work is to present a robust 3D automated method for measuring regional myocardial thickening using cardiac magnetic resonance imaging (MRI) based on Laplaces equation. Multiple slices of the myocardium in short-axis orientation at end-diastolic and end-systolic phases were considered for this analysis. Automatically assigned 3D epicardial and endocardial boundaries were fitted to short-axis and long axis slices corrected for breathold related misregistration, and final boundaries were edited by a cardiologist if required. Myocardial thickness was quantified at the two cardiac phases by computing the distances between the myocardial boundaries over the entire volume using Laplaces equation. The distance between the surfaces was found by computing normalized gradients that form a vector field. The vector fields represent tangent vectors along field lines connecting both boundaries. 3D thickening measurements were transformed into polar map representation and 17-segment model (American Heart Association) regional thickening values were derived. The thickening results were then compared with standard 17-segment 6-point visual scoring of wall motion/wall thickening (0=normal; 5=greatest abnormality) performed by a consensus of two experienced imaging cardiologists. Preliminary results on eight subjects indicated a strong negative correlation (r=-0.8, p<0.0001) between the average thickening obtained using Laplace and the summed segmental visual scores. Additionally, quantitative ejection fraction measurements also correlated well with average thickening scores (r=0.72, p<0.0001). For segmental analysis, we obtained an overall correlation of -0.55 (p<0.0001) with higher agreement along the mid and apical regions (r=-0.6). In conclusion 3D Laplace transform can be used to quantify myocardial thickening in 3D.