Title: Dynamic (4D) CT Perfusion Offers Simultaneous Functional and Anatomical Insights

Title: Dynamic (4D) CT perfusion offers simultaneous functional and anatomical insights into pulmonary embolism resolution.

Abstract:

Objective: Resolution and long-term functional effects of pulmonary emboli are unpredictable. This study was carried out to assess persisting vascular bed perfusion abnormalities and resolution of arterial thrombus in patients with recent pulmonary embolism (PE).

Methods and materials: 26 patients were prospectively evaluated by dynamic (4D) contrast enhanced CT perfusion dynamic pulmonary CT perfusion. Intermittent volume imaging was performed every 1.5-1.7 seconds during breath-hold using and perfusion values were calculated by maximum-slope technique. Thrombus load (modified Miller score; MMS) and ventricular diameter were determined. Perfusion maps were visually scored and correlated with residual endoluminal filling defects.

Results: The mean initial thrombus load was 13.1±4.6 MMS (3-16), and 1.2±2.1 MMS (0-8) at follow up. From the 24 CTPs with diagnostic quality perfusion studies, normal perfusion was observed in 7 (29%), and mildly-severely abnormal in 17 (71%). In 15 patients with no residual thrombus on follow up CTPA, normal perfusion was observed in 6, and abnormal perfusion in 9. Perfusion was abnormal in all patients with residual thrombus on follow up CTPA. Pulmonary perfusion changes were classified as reduced (n=4), delayed (systemic circulation pattern; n=5), and absent (no-flow; n=5).

The right ventricle was dilated in 12/25 (48%) at presentation, and normal in all 26 follow up scans. Weak correlation was found between initial ventricular dilatation and perfusion abnormality at follow up (r=0.15).

Conclusions: Most patients had substantial perfusion abnormality at 3-6 months post PE. Abnormal perfusion patterns were frequently observed in patients and in regions with no corresponding evidence of residual thrombus on CTPA. Some defects exhibit delayed, presumed systemic, enhancement (which we have termed ‘stunned’ lung). CT perfusion provides combined anatomical and functional information about PE resolution.

Introduction

Pulmonary embolism (PE) is common in the Western world with an estimated incidence of approximately 3 in 1000 population per year [1] These patients are at increased risk of recurrent thromboembolic episodes and secondary pulmonary hypertension with a reported first year mortality of 5-6% [2,3]. Anticoagulants are the main recognised treatment for PE [1], but are associated with risk of haemorrhage resulting in morbidity and mortality [4].

Despite being a common disease many aspects of the natural course of the condition are poorly understood. It is recognised that pulmonary emboli resolve in an inconsistent manner. The reported rate of PE resolution on computed tomography pulmonary angiography (CTPA) varies from 40-57% in early follow ups (within 2 weeks following the acute episode), to 77-81%, after 3months of the initial PE diagnosis [5,6].

Despite significant improvements, exclusion of pulmonary embolism by CTPA is still restricted by its insensitivity to small, subsegmental emboli [7,8]. The speed of resolution of clots and extent of residual perfusion defects in different patient groups remains to be clarified. How the clot burden and rate of clot resolution relate to persisting micro-vascular perfusion abnormalities is also unclear. Clarification of these issues may help in risk stratification (eg. risk of subsequent pulmonary hypertension), and in guiding treatment strategies. The aim of this study was to use dynamic (4D) computed tomography perfusion studies (CTP) to characterise pulmonary perfusion abnormalities and residual thrombus load in patients with previous pulmonary embolism and to compare the finding with the thrombus load and right heart features as measured on the presentation conventional CTPA studies. The study hypothesis was that CTP would be superior to CTPA in the diagnosis of residual thromboembolic disease in patients recovering from a recent episode of PE.

Methods and material:

Patients:

Following institutional approval, 26 adult patients with history of an acute PE, confirmed on CTPA (n=25) and perfusion scintigraphy (n=1) were prospectively recruited. Written informed consent was obtained in all cases. Perfusion CT (CTP) studies were performed 3-6 months following the initial PE episode. Exclusion criteria included pregnancy, inability to undergo computed tomography (CT) scanning, renal failure (serum creatinine >250 µmol/L or estimated glomerular filtration rate <25 mL/min), previous recruitment to the study, known contrast allergy, unable to give informed consent, and background parenchymal lung disease as other potential causes of perfusion changes (emphysema, fibrosis). Patients received anti-coagulation according to local guidelines following an episode of acute PE (Warfarin Sodium for 6 months following the episode of PE, and for lifetime in recurrent PE). Age and gender were recorded. Patient’s body size was classified by measuring the maximum lateral thoracic width (LTW) on AP scout films at the cardiac-diaphragm interface (at the level of liver) [9,10]. Based on the average LTW in our cohort, patient size was graded thin, medium, and large, when the LTW was <32cm, 32-38cm, and >38cm, respectively [10].

Image acquisition and reconstruction:

The CTP scans were performed during a single breath-hold at deep inspiration. Shallow abdominal breathing was permitted at the end stage of acquisition in patients who were unable to hold their breath for the entire perfusion CT data acquisition.

All CT scans were performed on a 320-multidetector computed tomography scanner (Aquilion ONE, Toshiba Medical Systems, Nasushiobara, Japan). Low dose scout imaging was performed to localise the thoracic structures. Intermittent volume imaging was performed every 1.5-1.7 seconds (11-12 volumes in total) with 3 seconds delay after the start of intravenous contrast injection. Scans were performed with 16 cm z-axis coverage (320 x 0.5-mm collimation) with the lowest section at the level of the diaphragm, 100 kVp, 0.5s rotation, and fixed tube current 100 mA (tube current-time product 50 mAs). Variable tube current protocol (mA boost) was used in one patient to reduce image noise in the pulmonary arteries. This included boosting tube current (200mAs) in 3 volumes timed with the peak of the PA enhancement, and standard current (100mA) for the rest of 8 volumes. Toshiba’s iterative reconstruction algorithm (Adaptive Iterative Dose Reduction 3D; AIDR-3D, strong) was used to reconstruct 0.5mm sections from the raw data (Toshiba Medical Systems, Nasushiobara, Japan).

A dual-head power injector (Stellant CT Injection Systems, MEDRAD, Warrendale, USA) was used for bolus injection of 70 mL iodinated contrast agent (Iomeron 400, Bracco SPA, Milan, Italy; 400mg/mL) via a 16G antecubital vein catheter (Vasofix, Braun, Melsungen, Germany) at a rate of 9 mL/sec, followed by 20 mL of saline solution at the same rate.

Post processing of dynamic contrast enhanced perfusion CT Images:

Non-rigid registration of the contrast-enhanced studies was performed using a dedicated commercial workstation to reduce the effect of motion (Vitrea FX v6.3, Vital Images, Minnesota, USA). For the qualitative evaluation of the perfusion, parametric maps of perfusion were produced using the same workstation using the Dual Input Lung Perfusion software by maximum slope method (Vitrea FX v6.3, Vital Images, Minnetonka, USA). Regions of interest (ROIs; 5 mm diameter) were placed in the main pulmonary artery and descending aorta to define the pulmonary and systemic arterial input functions. Same size ROI was placed in the left atrium and the peak enhancement time point was used to differentiate pulmonary circulation from systemic/bronchial circulation [9]. Regions of interest were drawn in the normal lung parenchyma away from main vascular branches, chest wall/mediastinal structures, and dependent/atelectatic lung changes. Computation of parametric perfusion maps was performed using the same workstation (Body Perfusion, (Vitrea FX v6.3, Vital Images, Minnetonka, USA). The perfusion analysis range was set from -300 HU to -1000 HU to restrict the perfusion analysis to lung parenchyma and to exclude major vessels and the soft tissues/bones. Pulmonary blood flow (PBF; mL/mg/min) was calculated as the maximum slope of tissue enhancement curve divided by the maximum arterial enhancement [12]. Parametric 512 × 512 matrix colour-coded maps of the pulmonary flow (PF) and systemic arterial flow (AF), and pulmonary index (PI=PF/PF+AF) were generated automatically [13].

Analysis of CT pulmonary angiography (CTPA) data:

The CTPA from the acute PE presentation episode and the follow up CTPA from the dynamic contrast enhanced perfusion study were evaluated in consensus by two thoracic radiologists (SM and EJVB) who were blinded to patients’ details. From the 11 CT volumes, the angiogram(s) with visually densest contrast within the pulmonary artery were used to identify residual intravascular thromboembolic material.

Overall image quality of each CTPA was rated on a 4-point scale (1: non-diagnostic; 2, fair image quality; 3: good image quality; and 4: excellent image quality). Causes for the compromised image quality were recorded (eg. motion, image noise, beam hardening).

Two thoracic radiologists (EJVB & JTM) objectively assessed in consensus the pulmonary arterial obstruction using the Modified Miller Scoring system (MMS) on both the initial CTPA and on the angiogram from the perfusion scan [14]. The modified Miller Score (MMS) is a score of thrombus load that was originally proposed by Miller et al. for conventional angiography and adapted for CTPA scan [14]. The right pulmonary artery was assigned nine segmental arteries (three to the upper lobe, two to the middle lobe, and four to the lower lobe), whereas the left pulmonary artery was assigned seven segmental arteries (two to the upper lobe, two to the lingula, and three to the lower lobe). Each occluded segmental pulmonary artery that is given a score of 1. Any more proximal occlusion scores the number of segmental branches distal to the occlusion thereby giving a MMS of 0 (no thrombus) to 16 (thrombus in all segmental arteries or saddle embolism) [14].

Diameter of the right ventricle (RV) and left ventricle (LV) was measured on the axial sections in the basal segment at the level of mitral valve. The RV was defined as dilated when the RV/LV ratio was 1.

Analysis of perfusion data:

Parametric perfusion maps were objectively assessed in consensus by 2 thoracic radiologist with expertise in CTP studies (SM & EJVB). Four patterns of pulmonary perfusion were identified on parametric maps and were confirmed by examination of the time-density curves in the regions of interest. The perfusion patterns were classified as: normal perfusion, reduced perfusion (reduced enhancement slope with delayed peak compared to normal lung; Fig. 1), systemic perfusion (delayed peak enhancement that is synchronised with the aortic enhancement peak), and absent (only image noise observed, no real enhancement).

The two radiologists objectively scored lobar perfusion. The agreed score was to reflect the volume of lung in each lobe. Compared to the modified Miller’s score, the upper lobes were assigned lower scores compared to the lower lobes since the 16cm detector area did not cover the apices of the upper lobes in any patient. Each of the upper lobes, right lower lobe, left lower lobe, middle lobe and the lingual were assigned, 3, 4, 3, 2 and 1 score (total of 16 score). An objective perfusion index (OPI) was calculated as total scored perfusion divided by 16. The lung perfusion was then graded as: severely abnormal (OPI <0.6), moderately abnormal (OPI 0.6-0.8), mildly abnormal (OPI >0.8), and normal (OPI =1).

Statistical analysis:

All results were expressed as mean ± standard deviation (SD) unless indicated otherwise. Correlations between the presence of RV dilatation at presentation and perfusion abnormality in follow up was performed using Pearson’s correlation and expressed as r2 values. SPSS for Windows (v10.0.1) was used for all statistical analysis.

Results:

Twenty-six consecutive patients with recent PE and a mean age of 61 (±16.5 years; range 23-86) were recruited (male/female=16/10). The average lower thoracic diameter was 38cm (±4.7; 30-50). Based on the lower thoracic diameter measurements, 3, 10, and 13 patients were classified as small, medium, and large, respectively. The average time interval between the initial CTPA (when the diagnosis of acute PE was made) and the follow up CTP was 22 ±16 weeks.

In 3 patients, the initial CTPA had to be repeated due to poor contrast timing and hence sub-optimal enhancement of the pulmonary arteries. The mean total dose length product (DLP) of the initial CTPA (including the repeat scans) and the CTP was 724±328 and 615±159 mGy·cm, respectively. The DLP for the mA boost protocol in a borderline large patient (LTW) was 690 mGy·cm. The estimated effective doses for the CTPA and CTP studies were 10.1±4.6 and 8.6±2.2 mSv, respectively (conversion factor=0.014 mSv/mGy.cm). The entire lung was covered in the standard CTPA whilst the 16cm field of view coverage did not cover the whole upper lobe in any patient.

The quality of pulmonary angiography on the initial CTPA was rated “excellent” in 22 cases, and “good” in 3. This does not take into account that the initial CTPA in 3 patients was non-diagnostic and had to be repeated. The quality of CT pulmonary angiogram in the CTP series was rated as “excellent” in 12, “good” in 7, “sub-optimal but diagnostic” in 4, and non-diagnostic in 3 cases. The causes for sub-optimal quality of the angiograms were excessive noise in 3, beam hardening in 3, and motion in 1 case. Excessive noise and beam hardening was almost exclusively seen in larger patients (mean LTW 39.5cm ±4). The mA boost in a patient with LTW of 37cm resulted in excellent improvement of the quality of pulmonary angiogram, when compared to standard mA.

The pulmonary arterial obstruction, as evaluated by the Modified Miller Scoring system (MMS) was 13.1±4.6 (3-16) in the initial CTPA, and 1.2±2.1 (0-8) in the angiogram from the CTP series. From the 25 diagnostic follow up CTPA, MMS was reported as 0 in 16 patients (64%), 1-3 in 7 patients (28%), and 7-8 in 2 patients (8%) (Fig. 2).

In 4 patients there was CT evidence of ischemic lung injury (peripheral wedge-shape consolidation in the territory of thrombosed arteries). All the changes resolved on follow up examination and no scaring seen.

Perfusion analysis was non-diagnostic in 2 patients, sub-optimal but diagnostic in 6 patients, and good in 18. The common causes for quality compromises were the effect of motion on calculated perfusion values (n=7), excessive image noise (n=3), and beam hardening (n=1). Significant image noise and beam hardening in a very large patient with LTW of over 50cm, and significant motion that was not corrected by image registration accounted for the non-diagnostic perfusion studies.

Figure 2 shows the percentage of patients with residual thromboembolic by pulmonary arterial obstructive score and CT perfusion abnormality at follow up. From the 24 patients with diagnostic CTP, normal perfusion was observed in 7 (29%) patients (OPI=1), 10 (42%) had mild perfusion abnormalities (OPI= 0.8-0.99), 5 (20%) had moderate perfusion abnormalities (OPI=0.6-0.79), and 2 (8%) had severe perfusion abnormalities (OPI <0.6) (Fig. 2). Perfusion was normal in 6 patients, but abnormal in 9 patients with normal follow up CTPA. In 2 patients who had significant residual thromboembolic material in the post treatment CT, there was a discrepancy between the abnormalities on CTPA and CTP in both patients. One patient with significant residual thromboembolic material (MMS=8) had severe perfusion abnormalities (OPI=0.56), whereas the other patient with significant residual thromboembolic material (MMS=7) showed only mild perfusion abnormalities. In contrast, in another patient with severe perfusion abnormalities (OPI=0.58), only mild residual thromboembolic material (MMS=2) was observed. Figures 3-4 show examples of patients with various levels of PE resolution on CTPA at follow up, and inconsistent level of perfusion resolution on CTP. There was weak correlation between the MMS at presentation and perfusion abnormality at follow up (r=0.11; Pearson's Product-Moment Correlation).