Evaluation of myocardial CT perfusion in patients presenting with acute chest pain to the emergency department: comparison with SPECT-myocardial perfusion imaging
- Gudrun Maria Feuchtner1,2,
- Fabian Plank2,
- Constantino Pena1,
- Juan Battle1,
- James Min3,
- Jonathon Leipsic4,
- Troy Labounty3,
- Warren Janowitz1,
- Barry Katzen1,
- Jack Ziffer1,
- Ricardo C Cury1
- 1Baptist Cardiac and Vascular Institute, Baptist Hospital Miami, Miami, USA
- 2Department of Radiology, Innsbruck Medical University, Innsbruck, Austria
- 3Cedars Sinai Medical Center, Los Angeles, California, USA
- 4University of British Columbia, Vancouver, Canada
- Correspondence to Dr Gudrun Maria Feuchtner, Department of Radiology, Innsbruck Medical University, Anichstrasse 35, A-6020 Innsbruck, Austria 6020;
Contributors All authors were involved in conception and design, or analysis and interpretation, drafting of the article or revising it critically for important intellectual content and final approval of the version to be published. RC is the guarantor of this study.
- Accepted 10 July 2012
- Published Online First 15 August 2012
Objective To determine whether evaluation of resting myocardial CT perfusion (CTP) from coronary CT angiography (CTA) datasets in patients presenting with chest pain (CP) to the emergency department (ED), might have added value to coronary CTA.
Design, setting 76 Patients (age 54.9 y±13; 32 (42%) women) presenting with CP to the ED underwent coronary 64-slice CTA. Myocardial perfusion defects were evaluated for CTP (American Heart Association 17-segment model) and compared with rest sestamibi single-photon emission CT myocardial perfusion imaging (SPECT-MPI). CTA was assessed for >50% stenosis per vessel.
Results CTP demonstrated a sensitivity of 92% and 89%, specificity of 95% and 99%, positive predictive value (PPV) of 80% and 82% and negative predictive value (NPV) of 98% and 99% for each patient and for each segment, respectively. CTA showed an accuracy of 92%, sensitivity of 70.4%, specificity of 95.5%, PPV 67.8%, and NPV of 95% compared with SPECT-MPI. When CTP findings were added to CTA the PPV improved from 67% to 90.1%.
Conclusions In patients presenting to the ED with CP, the evaluation of rest myocardial CTP demonstrates high diagnostic performance as compared with SPECT-MPI. Addition of CTP to CTA improves the accuracy of CTA, primarily by reducing rates of false-positive CTA.
The management of patients presenting with acute chest pain to the emergency department (ED) is a major challenge for the healthcare system and has a high socioeconomic burden. More than 6 million people a year in the USA1 are affected, with the majority requiring time-consuming2–4 diagnostic investigation, including laboratory tests such as determination of cardiac enzymes2 and imaging techniques.3 ,4 Single-photon-emission CT myocardial perfusion imaging (SPECT-MPI) is the preferred modality in patients with chest pain and suspected coronary artery disease (CAD).3 ,4
Coronary CT angiography5–15 (CTA) has been advocated as a new imaging modality in patients presenting with chest pain16 and a low-to-intermediate risk of acute coronary syndrome (ACS)2 due to high accuracy in evaluation of coronary stenosis >50% in patients with stable17–19 and acute chest pain.5 ,7 ,12 ,20 The accuracy of CTA for detection of significant stenosis is higher than with stress-exercise ECG.21 Further, CTA complements functional imaging like stress MRI in patients with intermediate pre-test likelihood.22 Additionally, CTA used in the ED is both time- and cost-effective.6 A limitation of CTA is overestimation of coronary stenosis in the presence of severe calcification,17 as well as a lack of information about lesion-specific regional myocardial ischaemia.
Recent single-centre studies indicate that the evaluation of myocardial perfusion23–26 and infarcts27–30 by CT is accurate as compared with SPECT-MPI, invasive angiography and cardiovascular magnetic resonance.
Thus, the comprehensive evaluation of myocardial perfusion defects from CTA25 ,26 may overcome the shortcomings of CTA, by providing information about ongoing myocardial ischaemia in patients with acute chest pain.
Therefore, the primary objective of our study was to assess the diagnostic accuracy of resting myocardial CT perfusion (CTP) in the setting of acute chest pain using SPECT-MPI as reference. As a secondary objective, automated software for quantification of myocardial perfusion was explored.
Methods and Materials
Institutional review board approval was obtained. The study was HIPPA (Health Insurance Portability and Accountability Act) compliant. Written informed consent was waived. The study design was retrospective.
Patients were recruited from an existing prospective study database consisting of consecutive patients presenting with chest pain (atypical or typical lasting <30 min) and a low-to- intermediate risk of ACS according to the thrombolysis in myocardial infarction (TIMI)-risk stratification (score<4)16 to the ED between January 2009 and January 2010. Patients who underwent SPECT-MPI in the ED or during a hospital stay were enrolled. The time interval between CT and SPECT was a mean of 15.4 h (range 1.7–98 h). CT and SPECT-MPI results were compared.
All patients met the clinical indications for cardiac CTA,17 ,20 by having low-to-intermediate risk of ACS and acute chest pain. Patients were recruited only if they had two initial negative cardiac enzymes within a 2 h time interval and negative or non-diagnostic ECG.3
Exclusion criteria were renal dysfunction (glomerular filtration rate <45 ml/min), atrial fibrillation, high heart rate (HR >85 bpm) and contraindications for β blockers or inability to lower the HR.
Patients were examined with a 64-slice CT scanner (VCT 64, GE Healthcare, Milwaukee, Wisconsin, USA) (64×0.625 mm). Prospective ECG triggering was applied in patients with low and regular HRs of <65 bpm and <65 years old. Retrospective ECG gating was applied in those with HR >65 bpm or >65 years old. Tube voltage settings were 120 kV for patients with body mass index (BMI) >25 kg/m2 and 100 kV in patients with low BMI ≤25 kg/m2. Specific mA settings were also defined for patients with BMI ≤25 kg/m2 (maximum mA=500), BMI between 26 and 30 kg/m2 (maximum mA=600) and BMI >30 kg/m2 (maximum mA=700). An arterial iodine phase contrast-enhanced CTA protocol was applied (flow rate 5 ml/s, contrast volume 80 ml, triggered to arterial phase) using an iodine contrast agent (concentration, 320 mg I/ml).
The scan length ranged from the pulmonary bifurcation to the diaphragm in all 73 patients referred for coronary CTA, and from above the aortic arch to the diaphragm in three patients with previous coronary artery bypass grafting.
β-Blockers were given before the scan if the HR was >65 bpm.
CT axial images were reconstructed at 0.75 mm effective slice width (increment 70%) and a medium smooth convolution filter and transferred to a dedicated workstation (GE Advantage, GE Healthcare).
Myocardial perfusion defects were evaluated for each myocardial segment using the American Heart Association 17-segment classification.31 Two observers with 7 years of cardiac CT experience performed the readouts, and a consensus was used for final data analysis. From axial slices, short-axis thick average multiplanar reformations (MPRs) at 8 mm slice thickness were generated. The centre/window setting were adjusted and optimised individually. First, myocardial perfusion defects were defined as hypoattenuated areas contained in a vascular territory. Hypoattenuating artefacts such as beam hardening were identified and recorded for each segment based on their characteristic features, as described previously,28 and regarded as negative.
A four-point image quality score was given (1=poor/non-diagnostic: severe artefacts (eg, beam hardening, motion, image noise, steps); 2=moderate: artefacts but borderline diagnostic; 3=good: mild artefacts; 4=excellent image quality without artefacts). Artefacts were further classified as 1=beam-hardening or streak artefacts; 2=slab artefacts due to ECG; 3=motion artefacts from cardiac or breathing motion. Confidence scores for myocardial perfusion defects were classified as 4=high confidence; 3=good confidence, probably positive; 2=low confidence, borderline;1=very low confidence, uncertain diagnosis.
Automated quantification of myocardial perfusion defects was performed in a subset of 57 patients using the CardIQ Express Reveal (GE Healthcare). This tool compares the distribution of the contrast CT attenuation (HU, Hounsfield units) in the volume of the myocardium with a Gaussian model. Areas of hypodensity were identified by statistical abnormalities in CT attenuation (HU) from this model. The determination of pathological areas is further refined through a heuristic approach, which uses the shape, size and position of the abnormalities to determine which are clinically significant. The pathological areas are displayed by colour maps. The HU of true defects and of the normal myocardium were measured within a region of interest, and image noise (SD of HU) was recorded. Beam-hardening artefacts were identified, as described previously32: (1) loco typico: posterior basal segments of the myocardium, projecting along the spine/aortic elongation line on the axial transversal plane and sometimes propagating into the mid and apical lateral segments; (2) not following a myocardial territory of a coronary artery; using both colour maps and greyscale image compared side by side.
Datasets were reviewed for coronary stenosis >50% for each vessel (left anterior descending coronary artery (LAD), right coronary artery (RCA), left circumflex coronary artery) by one experienced observer using interactive oblique and curved MPRs. Image quality of coronary arteries was classified on the same four-point scale as described above. Segments with coronary stents were included. Coronary bypass grafts were evaluated for patency.
SPECT-MPI was performed using a standard clinical protocol. Standard dual-isotope rest and stress SPECT-MPI was performed with technetium-99m sestamibi (Cardiolite). Images were acquired with a dual-headed SPECT gamma camera (Philipps, Best, The Netherlands). Pharmacologically induced stress was achieved by intravenous infusion of adenosine (at 0.14 mg/min/kg) over a period of up to 6 min or intravenous regadenoson (Lexiscan) during a 10 s injection.
The accuracy of myocardial resting CTP for the detection of perfusion defects was compared against resting myocardial SPECT. The accuracy of coronary CTA for the detection of >50% stenosis was compared with stress and rest myocardial SPECT. The combined CTA/CTP approach was rated as positive if either CTP or CTA was positive. The combined CTP/CTA approach was compared with stress/rest SPECT.
Statistical analysis was performed using SSPS Software (SSPS V.14). Quantitative data are expressed as mean±SD. For comparison of CTP with SPECT-MPI for detection of myocardial perfusion defects, the sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) including 95% CIs were calculated, and Cohen's κ applied to express intermodality agreement.
The profile of our study population (76 patients) is shown in table 1.
The time interval between CT and admission was a mean of 47.3 h (range 5.2–144.5) and between SPECT and admission a mean of 34.4 h (range 2.9–180.5 h).
CTP versus SPECT: diagnostic accuracy
Of 76 patients, 13 were positive for resting perfusion defects by SPECT (n=10 with partially reversible defects indicating both ischaemia and infarct, three with fixed defects indicating chronic old myocardial infarcts only without evidence of ischaemia). CTP correctly detected 12 out of 13. There was one false negative (FN) and three false positive (FP) results. The diagnostic accuracy per patient was: 92% sensitivity, 95% specificity, 80% PPV, 98% NPV (table 2). The intermodality agreement between SPECT and CTP was κ=0.82. Figure 1 shows a patient with acute chest pain and an ischaemic perfusion defect caused by RCA stenosis.
The only FN patient on CTP had inferior wall beam-hardening artefact hypoattenuation, CTA showed <50% stenosis, but SPECT was mildly positive.
One FP patient by CTP had two positive segments (S 14/17), resting SPECT was negative, but during stress apical ischaemia was noted (time delay was 4 h and image quality of CTA was limited by severe calcification). The invasive angiogram showed 50–60% LAD stenosis and left ventricular regional wall motion abnormalities (ejection fraction 40%). In another FP patient after LAD-percutaneous transluminal coronary angioplasty, the anterior segments (S 7, 13, 14, S 17) were positive but SPECT negative. The in-stent lumen was non-evaluable. Invasive angiography showed a patent stent and normal fractional flow reserve=0.96.
A 60-year old male after triple coronary artery bypass grafting surgery was FP on CTP for inferior segments, correlating with occluded vein graft to the RCA. SPECT was positive during stress adenosine but not at rest.
For the segments, CTP detected 33 of 37 defects (89% sensitivity, 99% specificity, 82% PPV, 99% NPV) (table 3).
Automated quantification of myocardial perfusion
In a subset of 57 patients, automated quantification of myocardial perfusion defects could be explored by a new prototype. In the remaining, DICOM offline transfer to the prototype failed for technical reasons. All true positive patients using subjective CTP scoring (12/12, 100%) were correctly identified by the automated software. Mean CT attenuation was 46.3 HU for perfusion defects and 101.1 HU for the normal myocardium (figure 2 and figure 3B,C). Mean image noise was 35 HU. Beam-hardening artefacts were identified on colour maps and greyscale CT images based on their typical appearance and location as described by Rodriguez et al32 in 35/57 (61%) patients.
Coronary CTA versus rest and stress myocardial SPECT
CTA results are presented at tables 4–6. Table 4 shows the severity of CAD by CTA. There were 19 arteries with moderate and 18 with severe stenosis, six occluded coronary arteries, six patents grafts and one occluded RCA graft.
Table 5 shows the diagnostic accuracy of CTA >50% stenosis versus rest/stress SPECT.
Of eight FN results by CTA, seven were negative with both CTA and CTP, one was positive with CTP.
There was no coronary stenosis; however, a perfusion defect was seen at the inferolateral mid-ventricular point (figure 4) on both CTP and resting SPECT, with myocardial thinning representing an old myocardial infarct with reperfused distal OM2. This patient was positive with CTP but negative with CTA (see figure 4).
None of the seven patients with negative CTA results but positive on SPECT developed a major cardiac adverse event-ACS during 30 days, and none of those underwent invasive angiography, suggesting that the doctor's impression of the patients, despite a positive SPECT result, was not sufficiently serious to refer the patient for invasive angiography. One positive patient on SPECT was finally diagnosed as having dilated cardiomyopathy with a patchy non-ischaemic perfusion defect pattern.
When CTP and CTA results were combined, the number of FP results on CTA was reduced from nine to seven, and the PPV improved to 90.1%, the sensitivity improved to 74% (table 6).
A total of 21/76 (27.6%) patients had evidence of ischaemia by SPECT: 11 (14%) had completely reversible defects, 10 (13%) patients had partially reversible (old infarct and superimposed ischaemia) defects. Only three (3.9%) of 76 patients had fixed defects only, indicating chronic myocardial infarcts without ischaemia (in nine of 1292 myocardial segments, 0.7%).
CTP/CTA examination: image quality and diagnostic confidence
Image quality scores and confidence scores for the CTP and coronary CTA evaluations are shown in tables 7 and 8. Image quality of CTP was diagnostic in 100% (excellent in 54 (71.1%), good in 11 (14.5), moderate in 11 (14.5%), and the diagnostic confidence for detection of myocardial perfusion defects was very low in 1 (1.3%), uncertain in 9 (11.8%), good in 25 (32.9%) and high in 41 (53.9%) (table 7). Table 8 shows the image quality of coronary arteries on CTA.
There were no artefacts in 41 (53.9%) examinations, and 21 datasets (27.6%) had beam-hardening (most common location: inferior and inferolateral basal segments S4/S5, and less commonly, S9/S10 or S16) or streak artefacts—for example, from a right ventricular pacemaker (n=1) or other metallic devices such as sternal cerclage (n=1), 13 (17.1%) examinations with step artefacts, and 1 (1.3%) dataset with motion artefacts.
Prospective ECG triggering was used in 17 patients (22%) with a stable HR of <65 bpm, and retrospective ECG gating in 60 (79%). Heart rate was a mean of 64.5 bpm, and all patients were in sinus rhythm. Mean centre/window levels were 150/351 (range 67/167–600/350). Interobserver agreement was 92%.
This study demonstrates a high accuracy of CTP for the detection of rest myocardial perfusion defects in patients presenting with acute chest pain at the ED from coronary CTA datasets, as compared with the reference method, SPECT-MPI.
Moreover, the accuracy of coronary CTA improves when CTP analysis and coronary CTA are combined as compared with SPECT-MPI, mainly through a reduction of the FP rate. FP results were caused by known limitations of coronary CTA such as a high coronary calcium load leading to an overestimation of coronary stenosis,18 ,19 or difficulties in the visualisation of in-stent lumen through metal artefacts.
The main advantage of using cardiac CT in the ED is the complementary evaluation of both coronary arteries and myocardial perfusion from the same CT,25 ,26 without added radiation or contrast agent exposure.
Another study observation was that in patients with unknown ‘silent’ (ie, clinically occult and non-diagnosed33) old myocardial infarcts after subsequent coronary recanalisation, CTA can be negative but CTP positive (figure 4). Clinically asymptomatic infarcts are mostly caused by small-vessel occlusion, and with an estimated 175 000 cases annually in the USA34 and a prevalence of 2–16%.33
Adenosine stress CTP23–26 allows for imaging of reversible myocardial ischaemia by direct visualisation of hypoattenuating myocardial perfusion defects. Myocardial blood flow is reduced in ischaemic myocardial regions; subsequently, the iodine attenuation is decreased leading to hypoattenuating myocardial areas. Similarly, perfusion defects occur in patients with old or reperfused myocardial infarcts, as scar tissue is also hypoperfused. Our data are in line with some previous studies on patients with stable chest pain and known CAD with significant stenosis,35–37 in which myocardial perfusion defects have been linked to areas of infarct or ischaemia. A good sensitivity of CT rest perfusion in 98 patients with known ischaemic heart disease and stable symptoms was reported,35 and good correlation between quantitative scores for CTP and SPECT-MPI in 44 patients.36 A similar study37 defined myocardial perfusion defects by CT as areas <50 HU, which is in accordance with our measurements. Furthermore, patients with >50% and >70% stenosis had a higher prevalence of myocardial perfusion defects (60% and 32.4%) than those with <50% and <30% stenosis (4.8% and 0%)37 and disappeared after revascularization; the majority of defects disappearing were perfusion defects.
Beam-hardening artefacts were commonly noted (27.6%), and mostly appeared in the inferior or inferolateral basal myocardium. These artefacts have a lower CT attenuation of 53.5 HU,32 as compared with other normal myocardial regions with a mean of 97 HU, and a typical appearance with the following criteria: first, they are mostly found within the basal inferolateral segment, and sometimes propagate into the mid and apical lateral segments. Second, in contrast to an ischaemic perfusion defect, artefacts do not follow the entire vascular territory of a coronary artery (figure 3), but appear along the descending aorta or thoracic spine axis. In some cases, however, a clear differentiation between artefacts and true defects can be difficult and may lead to FP findings. Dedicated CT data and image postprocessing for beam-hardening artefact correction as recently proposed, and currently under development, is desirable.38
In a minority of patients in our study, stair step and right ventricular pacemaker lead artefacts affected image quality.
Finally, dedicated postprocessing is of importance to ensure the accuracy of CTP, such as using thick average MPR 8 mm slices39 and narrow centre/window settings, adjusted individually.
Automated quantification of myocardial perfusion was explored in our study, which was helpful for differentiation of artefacts and true defects.
We acknowledge a low prevalence in our study population of positive defects meeting clinical indications of coronary CTA16; patients mainly had a low-to-intermediate likelihood of ACS. These patients formed the majority of our study population, while a minority had known CAD. The majority of positive patients on CTP (10/13, 77%) had partially reversible defects on SPECT, indicating both an old myocardial infarction and superimposed ischaemia within one myocardial territory. Only a minority (n=3) of patients had chronic old infarcts without ischaemia. Our study population represents the ‘real-world’ scenario in the ED.
One needs to take into account technical intermodality differences between CTP and SPECT in the higher spatial resolution (namely, CTP) and contrast resolution (namely, SPECT), leading to slight differences in the visualisation of the extent of perfusion defects. Fortunately, these differences appear to be minor not affecting the territory-based and segment-based evaluation of CTP in our study.
We acknowledge that to date, the automated perfusion defect segmentation software does not allow for quantification of perfusion defect size (ie, extent), which would be a valuable predictive parameter.
In conclusion, the evaluation of resting myocardial CTP on coronary CTA datasets is accurate compared with SPECT-MPI in patients presenting with chest pain to the ED. Evaluating CTP in addition to CTA improves diagnostic accuracy of CTA and reduces FP results.
The evaluation of resting myocardial CT perfusion (CTP) on coronary CT angiography (CTA) datasets is accurate compared with single-photon emission CT myocardial perfusion imaging in patients presenting with chest pain to the emergency department.
Evaluating CTP in addition to CTA improves the diagnostic accuracy of CTA and reduces the false-positive rate.
We acknowledge the contribution of the Baptist Cardiac and Vascular Institute research and outcomes team, particularly the special assistance of Margaret Kovacs, Karen Rickard and Pamela Blattner for data collection and institutional review board assistance.
Funding GE Healthcare provided a research grant to support this study.
Competing interests None.
Ethics approval Obtained.
Provenance and peer review Not commissioned; externally peer reviewed.