In vivo detection of hemorrhage in human atherosclerotic plaques with magnetic resonance imaging

Authors

  • Vincent C. Cappendijk MD,

    Corresponding author
    1. Department of Radiology, Cardiovascular Research Institute Maastricht (CARIM), University Hospital of Maastricht, The Netherlands
    • Department of Radiology, University Hospital Maastricht, P. Debijelaan 25, 6229 HX Maastricht (P.O. Box 5800, 6202 AZ), The Netherlands
    Search for more papers by this author
  • Kitty B.J.M. Cleutjens PhD,

    1. Department of Pathology, Cardiovascular Research Institute Maastricht (CARIM), University Hospital of Maastricht, The Netherlands
    Search for more papers by this author
  • Sylvia Heeneman PhD,

    1. Department of Pathology, Cardiovascular Research Institute Maastricht (CARIM), University Hospital of Maastricht, The Netherlands
    Search for more papers by this author
  • Geert Willem H. Schurink MD, PhD,

    1. Department of Vascular Surgery, Cardiovascular Research Institute Maastricht (CARIM), University Hospital of Maastricht, The Netherlands
    Search for more papers by this author
  • Rob J.Th.J. Welten MD, PhD,

    1. Department of Surgery, Atrium Medical Center, Heerlen, The Netherlands
    Search for more papers by this author
  • Alfons G.H. Kessels MD, MSc,

    1. Clinical Epidemiology and Medical Technology Assessment, Cardiovascular Research Institute Maastricht (CARIM), University Hospital of Maastricht, The Netherlands
    Search for more papers by this author
  • Robert J. van Suylen MD, PhD,

    1. Department of Pathology, Cardiovascular Research Institute Maastricht (CARIM), University Hospital of Maastricht, The Netherlands
    Search for more papers by this author
  • Mat J.A.P. Daemen MD, PhD,

    1. Department of Pathology, Cardiovascular Research Institute Maastricht (CARIM), University Hospital of Maastricht, The Netherlands
    Search for more papers by this author
  • Jos M.A. van Engelshoven MD, PhD,

    1. Department of Radiology, Cardiovascular Research Institute Maastricht (CARIM), University Hospital of Maastricht, The Netherlands
    Search for more papers by this author
  • M. Eline Kooi PhD

    1. Department of Radiology, Cardiovascular Research Institute Maastricht (CARIM), University Hospital of Maastricht, The Netherlands
    Search for more papers by this author

  • Oral presentation at the ISMRM meeting 2002, Honolulu, Hawaii

Abstract

Purpose

To investigate the performance of high-resolution T1-weighted (T1w) turbo field echo (TFE) magnetic resonance imaging (MRI) for the identification of the high-risk component intraplaque hemorrhage, which is described in the literature as a troublesome component to detect.

Materials and Methods

An MRI scan was performed preoperatively on 11 patients who underwent carotid endarterectomy because of symptomatic carotid disease with a stenosis larger than 70%. A commonly used double inversion recovery (DIR) T1w turbo spin echo (TSE) served as the T1w control for the T1w TFE pulse sequence. The MR images were matched slice by slice with histology, and the signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) of the MR images were calculated. Additionally, two readers, who were blinded for the histological results, independently assessed the MR slices concerning the presence of intraplaque hemorrhage.

Results

More than 80% of the histological proven intraplaque hemorrhage could be detected using the TFE sequence with a high interobserver agreement (Kappa = 0.73). The TFE sequence proved to be superior to the TSE sequence concerning SNR and CNR, but also in the qualitative detection of intraplaque hemorrhage. The false positive TFE results contained fibrous tissue and were all located outside the main plaque area.

Conclusion

The present study shows that in vivo high-resolution T1w TFE MRI can identify the high-risk component intraplaque hemorrhage with a high detection rate in patients with symptomatic carotid disease. Larger clinical trials are warranted to investigate whether this technique can identify patients at risk for an ischemic attack. J. Magn. Reson. Imaging 2004;20:105–110. © 2004 Wiley-Liss, Inc.

THE COMPOSITION RATHER THAN the size of an atherosclerotic plaque has emerged as a predictor of plaques at risk for thromboembolic events (1). Advanced plaques with fibrous cap disruption and intraplaque hemorrhage are largely responsible for morbidity and mortality (2, 3), and therefore in vivo noninvasive atherosclerotic intraplaque hemorrhage visualization will be of invaluable importance to trace the vulnerable patient. However, there are no established diagnostic tools to visualize intraplaque hemorrhage in vivo with sufficient accuracy. The spatial resolution of nuclear scintigraphy is too limited, and a particular radiotracer is only able to identify one single component of complex atherosclerotic lesions (4). In vivo ultrasound plaque characterization lacks consistency and accuracy (5, 6) and is therefore not very useful for this purpose. Magnetic resonance imaging (MRI) is the most promising noninvasive method for in vivo plaque characterization, as stated in recent reviews (7, 8), and has already been found to detect the major plaque components in vivo, although accurate intraplaque hemorrhage detection is still difficult (7–9). Cai et al (9) have reported a high sensitivity and specificity to identify type VI plaques (complex plaque with possible surface defect, hemorrhage, or thrombus); however, this was based on the high accuracy of a time-of-flight MR pulse sequence to detect the thin or ruptured fibrous cap (10–12). Additionally, Yuan et al (10) described that acute hemorrhage had a high signal intensity in time-of-flight (gradient echo) images, but the accuracy of acute hemorrhage detection could not be estimated due to a limited number of cases. Recently, Moody et al (13) published about an accurate method to detect intraplaque hemorrhage in the carotid artery of patients suffering anterior cerebral circulation ischemia using a T1-weighted (T1w) turbo field echo (TFE) sequence. This sequence is also referred to as a T1w magnetization-prepared three-dimensional gradient echo sequence. High signal intensity identified histologically confirmed complicated plaque. However, the in-plane resolution was limited and no slice-by-slice comparison between MRI and histology was performed. A high in-plane resolution will be needed to detect small areas of intraplaque hemorrhage. The in-plane resolution can be improved considerably by using a small-diameter radio frequency surface coil (14). The purpose of the present study was to test the performance of a T1w TFE MR pulse sequence to detect intraplaque hemorrhage in vivo with high-resolution MRI. The MR slices will be compared on a slice-by-slice basis with histology. It will be investigated whether other tissue components than intraplaque hemorrhage can also result in high signal intensity in the MR images. Because a T1w TSE sequence is a more commonly used T1w sequence for plaque characterization, we also obtained MR scans using this pulse sequence to compare the results.

MATERIALS AND METHODS

Subjects

Eleven patients (mean age = 68 ± 4 years, range = 62–72 years, seven males and four females) with a carotid stenosis of more than 70% as diagnosed by duplex ultrasound and scheduled to undergo carotid endarterectomy were recruited. All patients had suffered one or more transient ischemic attacks or minor strokes within the three months before surgery. Before the operation an MRI scan was performed. The time interval between MRI scan and surgery was 5 ± 4 days (mean ± SD), with a range of 1–13 days. The institutional review committee approved the study and all subjects gave written informed consent.

Histological Processing and Interpretation

After surgery, the carotid endarterectomy plaques, which were removed in one piece, were marked laterally and ventrally in the longitudinal direction with two lines of ink of different colors. Thereafter, the specimens were formalin fixed, and subsequently, they were divided into 3-mm slices, processed, and embedded in paraffin. With on average 3-mm intervals, a section of 4 μm was subjected to histological hematoxylin-eosin staining. High-quality color images were made of the histological slices. Two vascular biologists (K.C. and S.H.) microscopically assessed the histological slices concerning the major plaque components (fibrous tissue, lipid core, calcification, and intraplaque hemorrhage) in consensus and were unaware of the MRI results. The age of the intraplaque hemorrhage was categorized by two pathologists (M.D. and R.S.) as either younger or older than one week. Intraplaque hemorrhage younger than one week was defined as the major component being intact and hemolyzed erythrocytes mixed with sleeves of fibrin with early signs of organization such as ingrowth of myofibroblasts and capillaries. Intraplaque hemorrhage older than one week was defined as the major component being granulation tissue, with or without hemosiderin pigment in macrophages, with remnants of intact and hemolyzed erythocytes mixed with fibrin. Intraplaque hemorrhage without signs of (early) organization was not taken into account because manipulation of the artery during operation resulting in mechanical damage of the arterial segment is most likely responsible for this recent intraplaque hemorrhage. The size (in mm2) of all areas of intraplaque hemorrhage was measured microscopically with a ruler.

Matching Histology With MRI

One investigator (V.C.C.) matched the histological and MRI slices. The carotid bifurcation was used as a landmark to match histology and MRI in the longitudinal direction of the artery. The cross-sectional shape of the plaque, together with the ink markings, was used to adjust the cross-sectional orientation of the slices.

MRI Protocol

The carotid artery was imaged using a 1.5-Tesla whole-body MRI scanner (Intera, release 8.1.2, Philips Medical Systems, Best, The Netherlands). A dedicated, small-diameter radio frequency surface coil with a diameter of 47 mm (14) was fixed to the skin just above the carotid bifurcation. For an optimal scan result the patient was positioned in a head holder. An MR angiogram–balanced three-dimensional TFE, scan parameters: repetition time (TR)/echo time (TE) = 4.4/2.1 msec, flip angle = 60°, field of view (FOV) = 200 × 200 mm, matrix size = 192 × 192, number of signal averages (NSA) = 4, slice thickness = 0.70 mm, over contiguous transverse slices; body coil–was used for the identification of the carotid bifurcation.

Transverse images of the atherosclerotic plaques were obtained from about 7 mm caudal to 2 cm cranial of the bifurcation, a distance that covered the carotid plaque in all subjects. The following pulse sequences were acquired using the surface coil: 1) T1w three-dimensional TFE: TR/inversion time (TI)/TE = 10.3/900/4.0 msec, flip angle = 15°, 20 shots, actual scan percentage = 80%, number of phase encoding steps per shot = 163/20, FOV = 100 × 100 mm, matrix size = 256 × 205 (in-plane resolution = 0.39 × 0.49 mm), NSA = 6, 3.0-mm transverse slices; and 2) T1w two-dimensional double inversion recovery (DIR) turbo spin echo (TSE) sequence: TR/TI/TE = 570/255/14 msec, FOV = 100 × 100 mm, matrix size = 256 × 256 (in-plane resolution = 0.39 × 0.39 mm), NSA = 2, 2.5-mm transverse slices, and 0.5-mm slice gap. For both T1w TFE and T1w TSE images a homogeneity correction was used to correct for the dropoff in sensitivity of the surface coil. The scan durations were 3:37 and 7:05 minutes for nine slices for the T1w TFE and T1w TSE sequences, respectively. Usually the T1w TSE MR images were obtained in two separate scans in order to reduce the scan duration of a single scan and thus to reduce motion artifacts.

MR Image Evaluation

All images were assessed on a Sun workstation using Easyscil (version 4.1.2, Philips Medical Systems, Best, The Netherlands). Signal-to-noise ratio (SNR) and contrast-to-noise ratio (CNR) for both the T1w TFE and T1w TSE images were calculated. The signal was measured (V.C.C.) in an intraplaque hemorrhage user-defined region of interest (ROI). Noise was measured by taking the SD of the signal in air outside the patient corrected for magnitude effects (15). Contrast was calculated as the difference between the signal intensity within an ROI of intraplaque hemorrhage and that of fibrous tissue in plaque immediately adjacent to the intraplaque hemorrhage area. One investigator (V.C.C.) indicated the ROIs, all with a clear match with the corresponding histological area with the same orientation in the plaque, on paper copies of the MRI slices. These indications on paper copies served only as an assurance that two readers assessed the same ROIs. The choice for this study design implies that the readers could not search the images randomly to identify areas of possible hemorrhage. These ROIs were assessed independently by the two MR readers (J.v.E. and M.E.K.) on the workstation concerning plaque areas with signal intensities lower than, equal to, or higher than that of surrounding muscle tissue (qualitative analysis). The area of muscle tissue considered was as close as possible to the ROI. High signal intensity was regarded as a positive test result for intraplaque hemorrhage. The MR readers were not involved in patient enrollment in the study and had no access to the clinical notes and histological sections.

Data Analysis

Calculations were made with SPSS for Windows, version 10.0. Detection rates with 95% confidence interval (CI) for intraplaque hemorrhage were calculated for T1w TFE and T1w TSE, respectively, using histology as the standard of reference. Statistical difference in the performance reading of T1w TFE vs. T1w TSE images was tested with the McNemar test. The agreement between the two readers was determined with Cohen's K. The statistical difference of SNR and CNR between the two MR pulse sequences was investigated by paired-samples t-test.

RESULTS

High-quality MR images could be obtained of all 11 patients. A total of 59 MRI slices for both pulse sequences were available for comparison with histology. An average of five MRI slices (range = 3–8) were obtained per patient. The vascular biologists scored 43 areas of intraplaque hemorrhage, 29 areas of fibrous tissue, 17 areas of calcification, and 5 areas of lipid core in the corresponding 59 histological slices in mutual agreement. All 43 areas of intraplaque hemorrhage were mainly younger than one week. A few of those 43 areas also contained very small areas of hemorrhage older than one week.

The SNR of the T1w TFE images was significantly better than that of the TSE images (25 ± 10 (mean ± SD) vs. 16 ± 6, respectively; P < 0.01, paired-samples t-test). Additionally the CNR between intraplaque hemorrhage and fibrous tissue was significantly higher for the T1w TFE images (12 ± 9 (mean ± SD) vs. 4 ± 3, respectively; P < 0.001, paired-samples t-test). The superior image quality of the T1w TFE images can be observed in Fig. 1.

Figure 1.

Left: T1w TFE image. Middle: T1w TSE image of the internal (arrow i) and external (arrow e) carotid artery. Right: Magnification of corresponding histological hematoxylin-eosin-stained slice (magnification, ×125). The large area of high signal intensity (*) in the internal carotid artery corresponds to intraplaque hemorrhage, as shown by the histological slice.

The results of both MRI readers for the T1w TFE sequence are summarized in Table 1. Reader 1 (J.v.E.) detected 81% (35/43, 95% CI = 67–92) and reader 2 (M.E.K.) 93% (40/43, 95% CI = 81–99) of the areas of intraplaque hemorrhage. There were 12 false positive (range = 1–5 mm2), seven small false negative (range = 1–5 mm2), and one large false negative (11 mm2) area for the two readers together. All false positive areas contained fibrous tissue and were located within the thickened vessel wall, but outside the main plaque area. The results of both readers for the T1w TSE sequence are presented in Table 2. Reader 1 detected 72% (31/43, 95% CI = 56–85) and reader 2 detected 91% (39/43, 95% CI = 78–97) of the areas of intraplaque hemorrhage. There were 30 false positive (range = 1–22 mm2) and 15 false negative (1–11 mm2) areas for both readers together. These false positive areas contained fibrous tissue (N = 23), lipid core (N = 4), and calcium (N = 3). The interobserver agreement was high for the TFE sequence (Kappa = 0.73, 95% CI = 53–92), but low for the TSE sequence (Kappa = 0.35, 95% CI = 16–54). The areas of intraplaque hemorrhage that were scored correctly by both readers on the TFE images had a size of 10 ± 9 mm2 (mean ± SD; range = 0.3–27 mm2) as measured in the histological slices. Both readers detected intraplaque hemorrhage better with the TFE sequence, but this was not statistically significant (McNemar test). The TFE sequence is significantly better in excluding intraplaque hemorrhage (McNemar test, P < 0.005).

Table 1. Concordance With Histology and Interobserver Agreement of Intra-plaque Hemorrhage Detection With High Resolution MRI, Using a T1w TFE Sequence
Reader 1Histology, intra-plaque hemorrhageHistology, no intra-plaque hemorrhage
Reader 2Reader 2
HemorrhageNo hemorrhageTotalHemorrhageNo hemorrhageTotal
Hemorrhage35035404
No hemorrhage53883947
Total40343123951
Table 2. Concordance With Histology and Interobserver Agreement of Intra-plaque Hemorrhage Detection With High Resolution MRI, Using a T1w TSE Sequence
Reader 1Histology, intra-plaque hemorrhageHistology, no intra-plaque hemorrhage
Reader 2Reader 2
HemorrhageNo hemorrhageTotalHemorrhageNo hemorrhageTotal
Hemorrhage2833114418
No hemorrhage11112122133
Total39443262551

DISCUSSION

The present study demonstrates that in agreement with the recent study by Moody et al (13), hemorrhage can be identified with a high detection rate and high interobserver agreement in atherosclerotic plaques of the carotid artery of patients with a recent history of transient ischemic attack or minor stroke with a commercially available state-of-the-art whole-body MR scanner. More than 80% of the histologically proven areas of intraplaque hemorrhage could be detected by both readers on the T1w TFE MR slices with a high interobserver agreement. Since plaques containing hemorrhage are considered to be at risk (1–3), this technique could be important in the selection of patients at risk for stroke. Additionally, in the present high-resolution study, small areas of hemorrhage could be detected. Furthermore, it was investigated whether other tissue components than hemorrhage could result in high signal intensity in the MR images as well. Finally, the performance of the T1w TFE sequence was compared both qualitatively and quantitatively with that of the T1w TSE sequence, which is more commonly used for plaque characterization.

The performance of the TFE sequence was superior to the TSE sequence, because the TFE sequence had a higher detection rate for intraplaque hemorrhage, a reasonable low number of false positive results, and a much better interobserver agreement. Additionally, the T1w TFE images had a significantly higher SNR and CNR, which resulted in better image quality (Fig. 1). Furthermore, another disadvantage of the TSE sequence was that it suffered more often from motion artifacts due to a longer scan duration. Both sequences have nearly the same in-plane resolution.

False positive results might result in unnecessary carotid endarterectomy if intraplaque hemorrhage is used as the criterion to select the patients, and therefore they need to be very low in number. All false positive areas, i.e., areas of high signal intensity, represented fibrous tissue in the T1w TFE images (12/12), although fibrous tissue was in most cases an iso- or hypointense signal (17/29). The false positive areas in the TFE sequence were all located within the thickened vessel wall, but outside the main plaque area, suggesting that it is important to interpret high MR signal intensity together with its location to prevent false positive results. Hemorrhage outside the main plaque area is rare, and we observed it only once, which was scored incorrectly by both MR readers. The number of false positive results was much higher for the TSE images. One of the reasons for this higher number was that fibrous tissue far more often had a higher signal intensity in the TSE than in the TFE images (23/29 vs. 12/29 for the TSE and TFE images respectively). Earlier it was postulated by Xu et al (16) that fibrous tissue of an atherosclerotic plaque shows low signal intensity in gradient echo images (TFE) because the layered collagen-rich structure of the cap reduces T2*. However, the signal intensity is high in a T1w spin echo pulse sequence (TSE) because the latter lacks a T2* effect. An example of a false positive area in a TSE MR image is given in Fig. 2. In this figure an area of fibrous tissue is hyperintense in the TSE MR image, but isointense in the TFE MR image.

Figure 2.

Example of a false positive T1w TSE result. Left: T1w TFE image. Middle: T1w TSE image. Right: Magnification of corresponding histological hematoxylin-eosin-stained slice (magnification, ×125). Part of the carotid external artery appears with high signal intensity in T1w TSE, but not in T1w TFE (arrows). This area corresponds to fibrous tissue (f), as shown by the upper histological image. The area of high signal intensity (*) compared to muscle tissue (m) in both T1w TFE and T1w TSE in the internal carotid artery corresponds to intraplaque hemorrhage, as shown by the lower histological image.

The most commonly accepted explanation for the high signal intensity of intraplaque hemorrhage in T1w MR images is the presence of methemoglobin, a breakdown product of hemoglobin, which is formed about 12 hours after the hemorrhage. This produces a shortening of T1 and hence an increase in signal intensity in T1w images. Not much is known about the development over time of the MR signal intensity of hemorrhage in human atherosclerotic plaque. In swine, the peak signal intensity in artificial luminal thrombi was observed after one week using a T1w spin echo sequence (17). The signal relapsed to normal in three weeks' time. It needs to be emphasized that the intraplaque hemorrhage detected in the present study needs to be distinguished from the detection of luminal thrombi (17, 18) because from a pathophysiological point of view luminal thrombi are quite different from intraplaque hemorrhages. Intraplaque hemorrhages can clearly be differentiated from luminal thrombi in histological slices. Bradley (19) reported a very hyperintense signal in T1w MR images of parenchymal brain hematomas with an age between 3 and 14 days. MR signal intensity of carotid intraplaque hemorrhage also depends on age, as described by a number of groups (8, 9, 20, 21). Another well-recognized problem by these authors is partial-volume effects of small or inhomogeneous areas of hemorrhage. In the present study the false negative results were mainly small areas of less than 5 mm2.

The pathophysiology of advanced atherosclerotic disease is quite complex. Clot maturation in an atherosclerotic plaque may be impaired, and therefore the time elapsed between clot formation and full regeneration may vary considerably (22). Supplementary capillaries, which invade the intraplaque hemorrhage or atherosclerotic plaque thrombus within about 14 days (3, 22), are easily damaged and are responsible for new spots of hemorrhage. One might assume that with the technique used in the present study only the existence of intraplaque hemorrhage can possibly be predicted with MRI in vivo, especially the presence of methemoglobin, but not the age of the hemorrhage. This was supported with the histological observation that all areas with intraplaque hemorrhage were less than one week old, while the mean delay between MRI and operation was five days. The mean delay and range was the same for the group of false negatives. The amount of methemoglobin will decrease in older areas of hemorrhage, and therefore, it is likely that in vivo MRI using a T1w TFE sequence will detect mainly young areas of intraplaque hemorrhage.

In the present study, no multisequence MRI was used since the objective was to investigate the performance of the T1w TFE sequence for hemorrhage detection. However, it needs to be emphasized that for an accurate separation of the other plaque components, multisequence MRI is essential (10, 23, 24).

A limitation of the present study is that in order to be able to compare the performance of both MR readers, the ROIs on the MR images could not be chosen randomly by the readers. This implies that the diagnostic performance might be somewhat biased. A second limitation is the limited number of subjects, which were included in the present investigation, and therefore, larger studies are warranted to confirm the current promising results. Additionally, it needs to be investigated whether patients with high signal intensity on the T1w TFE images are at risk for an ischemic attack.

In conclusion, the present study showed that noninvasive high-resolution MRI, using a T1w TFE sequence, is very promising for the detection of intraplaque hemorrhage. If the location of an area with high signal intensity in the vessel wall will be taken into account, the TFE sequence will be even more accurate. Because of both the much lower amount of false positive results and the much higher interobserver agreement in TFE, this sequence will be more valuable for surgical decision making than the TSE sequence. Future in vivo studies need to elucidate the time frame of sustained high signal intensity on T1w TFE images and its correlation with histology and patients' symptoms. The asymptomatic patients with high-grade stenosis are another challenging group that needs to be assessed.

Ancillary