Identification of Calcification with Magnetic Resonance Imaging Using Susceptibility-Weighted Imaging: A Case Study

Abstract

Susceptibility weighted imaging (SWI) is a new MRI technique that can identify calcification by using phase images. We present a single case with a partially calcified oligodendroglioma, multiple calcified cysticercosis lesions, and multiple physiologic calcifications in the same patient. SWI phase images and computed tomography (CT) images are compared. SWI phase images showed the same calcified lesions as shown on CT and sometimes some new calcifications. Our conclusion is that SWI filtered phase images can identify calcifications as well as CT in this case.

INTRODUCTION

CT has long been considered the gold standard in detecting calcification. When the Hounsfield units (Hu) exceed an established threshold (100 Hu), the source is believed to be calcification (1, 2). In MRI, calcification appears with various signal intensities on conventional spin echo (SE) T1 or T2 weighted images (3, 4, 5), which makes it difficult to identify definitively as calcium. In gradient-echo acquisitions, calcifications usually appear as hypointense and cannot be differentiated from hemorrhage. It has been recognized that using phase helps to discriminate between calcium and iron because calcifications tend to be diamagnetic and iron paramagnetic; thus they appear with the opposite signal intensity in filtered phase images (6). Susceptibility-weighted imaging (SWI) is a fully velocity compensated gradient echo sequence with special magnitude and phase processing (7, 8). The processed phase images provide the possibility that phase information can be used in a standard clinical setting (7, 9). These filtered phase images have had phase variations from the background fields removed, leaving behind the pristine tissue susceptibility changes evident in the final image. In this paper, we will investigate the ability of SWI filtered phase images to demonstrate definitively the existence of calcified tissues.

Intracranial calcification refers to the deposition of crystalline calcium in the parenchyma at various sites in the brain. Calcification can appear in both physiological and pathological conditions. Calcifications of the pineal gland, choroid plexus, basal ganglia and dura mater are commonly seen with aging and are usually not associated with pathological clinical phenomena. However, calcium deposits can be associated with several intracranial pathologies including tumors, cerebrovascular diseases, congenital conditions, trauma and endocrine/metabolic disorders (10). The location and characteristics of the calcification in these lesions are very important indicators in diagnosis and differential diagnosis. Although MRI has been thought to be the best imaging modality in the central nervous system (CNS), there are occasions when a CT scan needs to be obtained to confirm the presence of calcification suspected on MRI when it becomes a critical sign in diagnosis.

Oligodendroglioma is one type of glioma in CNS. Dense, coarse calcifications are common (up to 90%). Neurocysticercosis is the most common parasitic infection of the CNS worldwide (11). Humans are infected by ingestion of ova of the pork tapeworm. The ova develop into larvae (cysticerci) and lodge in soft tissues, especially skin, muscle, and brain. After the larva dies, the capsule thickens and the cyst degenerates. In the final stage, the small cystic lesion completely calcifies and is usually chronic and indolent, not usually associated with edema or with seizures. The number of calcified lesions can be from one to thousands. In this paper, we compare the findings of SWI magnitude and high pass filtered phase images with those from CT data. We illustrate that SWI has the ability to identify calcium components inside lesions and individual deposits of calcium by using phase images.

MATERIALS AND METHODS

The case reported here is that of a 58 year-old Hispanic man with a remote history of neurocysticercosis in early childhood. Twenty-five years ago he was involved in a motor vehicle accident and sustained a minor closed head injury. He was evaluated with a CT scan of the brain which showed multiple small calcifications scattered throughout the supratentorial space consistent with a past history of cysticercosis. No other lesions were seen at that time. The patient had remained stable until five months prior to presentation when he started complaining of ongoing headaches. He was investigated with a CT and MRI of the brain. A new calcified mass lesion was identified in the right fronto-insular region. The mass measured 3.4 × 2.6 × 2.3 cm³ in size and was consistent with a low-grade calcified neoplasm such as oligodendroglioma. He underwent a follow-up MRI 3 months later which showed new areas of focal enhancement suspicious for a malignant transformation into a high-grade glioma. The patient subsequently underwent a frameless stereotactic right fronto-parietal craniotomy for volumetric resection of the enhancing calcified tumor. Histopathology confirmed a WHO grade 3 anaplastic oligodendroglioma with focal calcification.

CT Examination

The CT acquisition parameters were as follows: x-ray tube current: 350 mA, kvp = 120 kV, convolution kernel H20s (soft tissue window), convolution kernel = H60s (bone window), resolution = 0.45 × 0.45 × 5 mm³ and field-of-view = 230 mm.

MR Imaging

A variety of conventional sequences were collected at 3T (Siemens TRIO, Erlangen, Germany), including: SWI, FLAIR, T2, T1-post contrast, high resolution perfusion weighted imaging, and diffusion weighted imaging. The imaging parameters for the SWI sequence were: TR/TE = 29/20 ms, FA = 15^o, BW = 120 Hz/pixel, spatial resolution = 0.5 × 0.5 × 2.0 mm³ and field-of-view = 256mm. Calcifications appeared as hypointense signal on the magnitude images. On the SWI filtered phase images, veins and blood products showed as hypointense (negative phase) and calcifications showed as hyperintense (positive phase) for a right handed system. This dichotomy makes distinguishing the two very simple. SWI images were also projected over 2 slices to obtain an equivalent 4 mm thick slice to match better with the somewhat thicker CT images (5 mm). All 29 CT slices were compared region for region with the processed SWI data.

Statistical Analysis

Lesion area calculation

We used home-made software, SPIN (Signal Processing in NMR, Detroit, Michigan, USA), to semi-automatically calculate the area of every calcified cysticercosis lesion. Because the lesions we measured were so small, we first identified one lesion and zoomed the image by a factor of 16 centered on the lesion. Then we manually outlined the boundary of the lesion. The software uses a dynamic programming approach to calculate the boundaries (12). The software was set to iterate twenty times where generally it converged to within 3% of the quoted area. We recorded the area of the lesion and the mean and standard deviation in terms of Hounsfield units for the CT data. We used the CT image as a reference and identified the same lesion on the SWI phase image. Then we recorded the area of the lesion and the mean and standard deviation of the phase value using the same method as in the CT image.

Contrast comparison

First, we recorded the mean and standard deviations of the signal intensities or phase values of each lesion when we were calculating the lesion area. Second, we obtained the mean and standard deviation of the background signal. We drew an area in the white matter (where the structure and signal are uniform) ten times on different regions-of-interest (ROI) and took the average value. The contrast was defined as

$${\text{Contrast}}_{\text{CT}}=(\text{CT}{\text{Number}}_{\text{lesion}}-\text{CT}{\text{Number}}_{\text{background}})$$

$${\text{Contrast}}_{\text{MRI}}=(2048-\text{phase}{\text{value}}_{\text{lesion}})$$

The offset of 2048 comes from the fact that Siemens scales the phase from (−π, π) to (0, 4096) with 2048 being zero phase. The white matter happens to have a phase of 0 radians (hence 2048 in Siemens phase units). We compared the CT and MRI results using a paired t-test and plot the results using a scattergram. A p-value less than 0.05 was considered statistically significant. The Pearson correlation coefficient was used to measure the strength of the linear relationship between the CT and MRI results.

RESULTS

A calcified mass lesion was seen in the right fronto-insular region (Figure 1) which was confirmed by histopathological exam as a WHO grade 3 anaplastic oligodendroglioma with focal calcification. The CT showed patchy calcification inside the tumor. The SWI magnitude image showed hypoinstense signal inside the tumor. The area of hypointensity seen in the SWI magnitude image was larger than the calcified part seen on CT. On the SWI filtered phase image, the hyperintensity matched the calcified part of the tumor on CT on a one-to-one basis spatially. The veins (paramagnetic) running in the sulci and along the lateral ventricle appeared dark (negative phase) and are a marker of a right handed system. They serve as a control indicating that the bright signal (positive phase) inside the tumor is a diamagnetic substance, i.e., calcification. In addition, the dark ring structure around the edges of several oval-like parts of the tumor seen in the phase images was enhanced on the post contrast T1 weighted imaging.

Figure 1.

CT (a) shows patchy calcification inside the tumor. SWI magnitude image (b) shows hypoinstense signal inside the tumor that is larger than the calcified part on CT. On the SWI filtered phase image (c), the hyperintensity matches CT results reasonably well given the aliasing present. The veins running inside the sulci and along the lateral ventricle appear dark (short arrows). In addition, the dark ring structure (long arrow) around the tumor on phase images matches the enhanced area on the post contrast T1 weighted images (d).

Multiple small calcifications due to cysticercosis were seen scattered throughout the supratentorial space. A total of 89 lesions were seen with CT and the same lesions were identified with SWI. The location and size of individual calcified lesions on MRI matched with the CT perfectly (Figure 2). Although CT and MRI have slightly different slice locations and MRI has a thinner slice thickness than CT, we could always find the corresponding lesions on MRI by checking neighboring slices. In addition, there were 16 very small lesions sharply shown on MRI which was hard to identify from noise on CT. However, because of aliasing on MRI phase images, the area of some of the lesions on MRI was difficult to measure accurately. Therefore, we only took those lesions without aliasing on MRI and plotted the values of the lesion area from both CT and MRI (Figure 3). For these 53 lesions, the mean area measured from CT was 1.11 ± 0.23 (0.66 to 1.76) mm² and for MRI it was 1.72 ± 0.70 (0.73 to 4.22) mm². The difference is significant (P<0.001). The Pearson correlation coefficient of the measured lesion area between CT and MRI is 0.60. The contrast comparison between CT and MRI is shown in Figure 4. The mean contrast for CT was 148.58 ± 72.20 and for MRI it was 674.53 ± 230.77. The difference was significant (P<0.001). The Pearson correlation coefficient of contrast between CT and MRI was 0.30.

Figure 2.

Multiple small calcifications are identified by CT (a). The location and size of the lesions shown on the SWI magnitude image (b) matches the CT data one-to-one. The SWI filtered phase image (c) shows the lesions have opposite signal intensity to the veins along the sulci confirming that the dark areas on the magnitude images are calcification, not hemorrhage.

Figure 3.

Correlation of CT calcification area with MRI calcification area.

Figure 4.

Correlation of CT lesion contrast with MRI lesion contrast for calcification.

SWI revealed physiologic calcifications in the pineal gland (figure 5), choroid plexus in the trigone of the lateral ventricles bilaterally, and left globus pallidus; these regions were corroborated on CT as well.

Figure 5.

A calcified lesion close to the right lateral fissure and pineal gland are shown on CT (a). SWI magnitude images match CT well. On the SWI filtered phase images, the dark regions in the magnitude image shows as bright again (indicating calcium) but with some aliasing or dipole effects as well.

DISCUSSION

The chemical composition of pathological and physiological calcification are crystalline Ca₃(PO4)₂, hydroxylapatite and a miniscule amount of copper (Cu), manganese (Mn), zinc (Zn), magnesium (Mg) and iron (Fe). The morphologies of pathological neurological calcification and calcification of choroid plexus are different (13). The difference of calcium compounds, various concentrations and proportion of calcium might account for the variations of MR signal on SE sequences for example (14). The pathological test using hematoxylin and eosin (HE) stain confirmed that calcium was indeed present in the tumor. Since no chemical analysis was performed on the calcified lesions, we cannot conclude that the calcified lesions are completely composed of crystalline calcium. It is quite possible that there might be a small amount of iron inside the calcified lesions. However, since all lesions show a diamagnetic effect, either the amount of iron is too small to have an effect or it has become MR invisible as in the case of oxyhemoglobin and the calcium present dominates the effect. Further, CT was used as the gold standard, so we can have good confidence in drawing the conclusion that the positive phase objects seen on the phase images which match the CT are predominantly calcium.

SWI filtered phase images have been used as a means to study tissue susceptibility since 1995 and it has proven to be robust across manufacturer’s systems and across field strengths (15). In Figure 1, the SWI magnitude image shows hypoinstense signal inside the tumor. However the area of hypointensity looks larger than the calcified part on CT. From the magnitude image, one might conclude that the hypointensity inside the tumor is hemorrhage. For advanced glioma, hemorrhage inside tumor occurs quite often. As we know hemorrhage appears as very low signal intensity just like calcium. This makes differentiation between calcium and hemorrhage on magnitude images very difficult. On the SWI filtered phase image, the hyperintensity matches the calcified part of the tumor on CT very well, thus suggesting that the hypointensity in this area comes from the calcified part of the tumor. In addition, the dark ring structure around the tumor on the phase images matches the peripheral enhanced ring structure on the post contrast T1 weighted images. The enhancement of the lesion is due to breakdown of blood-brain barrier (BBB), i.e., leaky microvessels. It is possible that a very small amount of blood leaked out from the small tumor vessels. The dark ring structure on the phase image is a blood product. This is in agreement with the theory that iron in tissue (dark in phase) has the opposite signal from that of calcium (bright in phase). Similarly on Figure 2, SWI magnitude images showed every lesion and matched CT perfectly, but magnitude images alone cannot differentiate between that these black dots are microhemorrhage or calcification where signal loss occurs. On the SWI filtered phase images, these lesions are bright and we can conclude that those regions are calcified lesions. One convenient method for clinicians is to use veins as a reference. Because veins are paramagnetic and ubiquitous inside the brain, if the phase of the region-of-interest is opposite to that of the veins, we can conclude that it is caused by calcium.

Figure 3 shows a plot of the lesion area as seen on CT and MRI. The close correlation between CT and MRI indicates the reliability of SWI in showing calcifications consistently. These small calcified lesions have an average area of 1.11 mm² for CT and 1.72 mm² for MRI. The reason that MRI showed a larger area than CT is likely due to the imperfect reduction of the dipole field by the high pass filter. This can be an advantage when very small lesions are encountered. Also, MRI has a thinner slice (2mm) than CT (5mm). Since we measure the lesion area on the slice with the largest diameter of the lesion, partial volume effects could make the lesions on CT smaller than the real object. This is a major advance over earlier work where it was believed that MRI could not show calcium that is occult on CT (1).

Figure 4 shows the relationship between CT and MRI contrast for the small calcified lesions. Contrast from MRI is higher than that in CT. Thus SWI phase images could be very useful in detecting small amounts of calcifications. Generally, these small lesions appear to have sharp boundaries (high contrast with the surrounding tissues) in the phase images.

The major disadvantage with SWI filtered phase images is aliasing. When the field is large enough that the phase exceeds π radians, it will alias to −π radians and will now appear to be dark rather than bright. As shown on Figure 5, inside the calcified lesion along the lateral fissure, the center of the lesion appears dark. Similar problems are seen in the pineal gland. This effect can be reduced by using a shorter echo time. If one is running SWI on a regular basis, the use of a double echo SWI may prove useful in avoiding this artifact. Further complicating the matter is the dipole field associated with a sphere. This means that both positive and negative phase can appear outside a larger diamagnetic object. Recent advances in reprocessing phase images to extract local magnetic fields or susceptibilities (16) can remove these effects, but a high pass filter or anti-alasing approach (17) cannot do this. On the other hand, a small evenly dispersed concentration of calcium may act to create more of a chemical shift effect as appears to be the case in many of these lesions. This still leads to the usual positive phase shift and a stronger field effect can still cause aliasing in the center of the lesion.

The use of SWI to differentiate the presence of calcium from iron for example may also be useful in evaluating those breast cancers where calcium deposition is a diagnostic indicator of lesion status (18). This is a more difficult problem because of all the inhomogeneities in the breast but would be an interesting area of study for the future. Similarly, imaging calcium in atherosclerosis is an important direction. Again, the advantage of SWI for this would be the ability to differentiate hemorrhage from calcification inside the plaque for example.

In conclusion, SWI filtered phase images proved to be accurate in detecting very small amounts of calcification in this case. Aliasing makes it difficult to show the exact shape of the calcification of a large area or high concentration, although this problem could be alleviated with the use of a shorter echo time or more appropriate phase conversion to local susceptibility (16). CT still remains the best method in the sense that no aliasing occurs for this modality.