Proton MR spectroscopy in patients with pyogenic brain abscess: MR spectroscopic imaging versus single-voxel spectroscopy

European Journal of Radiology 82 (2013) 1299–1307

Contents lists available at SciVerse ScienceDirect

European Journal of Radiology journal homepage: www.elsevier.com/locate/ejrad

Proton MR spectroscopy in patients with pyogenic brain abscess: MR spectroscopic imaging versus single-voxel spectroscopy Shuo-Hsiu Hsu a,1 , Ming-Chung Chou b,2 , Cheng-Wen Ko c,3 , Shu-Shong Hsu d,4 , Huey-Shyan Lin e,5 , Jui-Hsun Fu a,6 , Po-Chin Wang a,f,7 , Huay-Ben Pan a,f,8 , Ping-Hong Lai a,f,∗ a

Department of Radiology, Kaohsiung Veterans General Hospital, Kaohsiung, Taiwan, ROC Department of Medical Imaging and Radiological Sciences, Kaohsiung Medical University, Kaohsiung, Taiwan, ROC c Department of Computer Science and Engineering, National Sun Yat-sen University, Kaohsiung, Taiwan, ROC d Department of Neurosurgery, Kaohsiung Veterans General Hospital, Kaohsiung, Taiwan, ROC e Program of Health-Business Administration, School of Nursing, Fooyin University, Kaohsiung, Taiwan, ROC f School of Medicine, National Yang-Ming University, Taipei, Taiwan, ROC b

a r t i c l e

i n f o

Article history: Received 11 December 2012 Received in revised form 28 January 2013 Accepted 29 January 2013 Keywords: Brain pyogenic abscess Single voxel MR spectroscopy Magnetic resonance spectroscopic imaging

a b s t r a c t Purpose: Single-voxel spectroscopy (SVS) has been the gold standard technique to diagnose the pyogenic abssess. Two-dimensional magnetic resonance spectroscopic imaging (MRSI) is able to provide spatial distribution of metabolic concentration, and is potentially more suitable for differential diagnosis between abscess and necrotic tumors. Therefore, the purpose of this study was to evaluate the equivalence of MRSI and SVS in the detection of the metabolites in pyogenic brain abscesses. Materials and methods: Forty-two patients with pyogenic abscesses were studied by using both SVS and MRSI methods. Two neuroradiologists reviewed the MRS data independently. A  value was calculated to express inter-reader agreement of the abscesses metabolites, and a correlation coefficient was calculated to show the similarity of two spectra. After consensus judgment of two readers, the binary value of metabolites of pyogenic abscesses (presence or absence) was compared between SVS and MRSI. Results: The consistency of spectral interpretation of the two readers was very good ( ranged from 0.95 to 1), and the similarity of two spectra was also very high (cc = 0.9 ± 0.05). After consensus judgment of two readers, the sensitivities of MRSI ranged from 91% (acetate) to 100% (amino acids, succinate, lactate, lipid), and the specificities of MRSI were 100% for detecting all metabolites with SVS as reference. Conclusion: SVS and MRSI provide similar metabolites in the cavity of pyogenic brain abscess. With additional metabolic information of cavity wall and contralateral normal-appearing brain tissue, MRSI would be a more suitable technique to differentiate abscesses from necrotic tumors. © 2013 Elsevier Ireland Ltd. All rights reserved.

1. Introduction

∗ Corresponding author at: Department of Radiology, Kaohsiung Veterans General Hospital, 386 Ta-Chung First Road, Kaohsiung 813, Taiwan, ROC. Tel.: +886 7 3422121x6237; fax: +886 7 3468301. E-mail addresses: [email protected] (S.-H. Hsu), [email protected] (M.-C. Chou), [email protected] (C.-W. Ko), [email protected] (S.-S. Hsu), [email protected] (H.-S. Lin), [email protected] (J.-H. Fu), [email protected] (P.-C. Wang), [email protected] (H.-B. Pan), [email protected] (P.-H. Lai). 1 Tel.: +886 7 3422121x6224; fax: +886 7 3468301. 2 Tel.: +886 7 3121101x235723; fax: +886 7 3113449. 3 Tel.: +886 7 2353535x4330; fax: +886 7 5254301. 4 Tel.: +886 7 3422121x3017; fax: +886 7 3422288. 5 Tel.: +886 7 7811151x7350; fax: +886 7 7824739. 6 Tel.: +886 7 3422121x6251; fax: +886 7 3468301. 7 Tel.: +886 7 3422121x6242; fax: +886 7 3468301. 8 Tel.: +886 7 3422121x8300; fax: +886 7 3468301. 0720-048X/$ – see front matter © 2013 Elsevier Ireland Ltd. All rights reserved. http://dx.doi.org/10.1016/j.ejrad.2013.01.032

It has been suggested that 1 H magnetic resonances from succinate (Suc) at 2.4 ppm, acetate (Ac) at 1.9 ppm, and from the three cytosolic amino acids (AAs) consisting of valine, leucine, and isoleucine at 0.90 ppm are potential abscess markers [1–6]. The presence of Ac with or without Suc favors an anaerobic bacterial origin of the abscess; however, this may also be seen in some of the abscesses secondary to facultative anaerobes [5–7]. AA resonances are sensitive markers of pyogenic abscess, which permit the differentiation between cystic brain tumors and intracranial bacterial abscesses [1–6]. However, AAs could not be observed in 20% of abscesses cavity from in vivo 1 H magnetic resonance spectroscopy (MRS), but its absence does not rule out a pyogenic etiology [8]. Radiologically, a brain abscess in the capsule stage appears in MRI as a rim-enhancing mass surrounded by edema which is

1300

S.-H. Hsu et al. / European Journal of Radiology 82 (2013) 1299–1307

similar in appearance to necrotic glioblastomas multiforme (GBM). Spectral patterns recorded for the cystic or necrotic components of GBMs and abscesses caused by aerobic bacteria (aerobic abscesses) were similar in single voxel MRS [9]. Thus, multiple single-voxel spectroscopy (SVS) should be performed in different regions to make differential diagnosis. In contrast, MRSI is known to allow evaluation of the abscess wall, perifocal edema, as well as contralateral normal-appearing brain tissue, in addition to abscess cavity within one measurement, and it may be more suitable than SVS for differential diagnosis between abscess and tumor. A previous MRSI study showed that metabolite ratios, such as maximum Cho/Chon (Cho in the corresponding contralateral normal-appearing (-n) brain), Cho/Cr and Cho/NAA, of contrast-enhancing rim obtained by MRSI were significantly different between abscess and necrotic GBMs, thus MRSI is potentially useful to differentiate abscesses from GBMs [9]. Recent developments in MRS have made it possible to obtain MRSI with high spatial resolution and multiple spectra simultaneously from contiguous voxels [10,11]. MRSI techniques have the advantage over SVS techniques in providing spatial distribution of metabolic information along with multiple spectra within one measurement [12]. However, the disadvantages of MRSI include lower signal to noise ratio (SNR), voxel bleeding effect and worse field homogeneity over larger volume of interest (VOI). Although SVS is considered being able to provide more accurate quantification of metabolites, MRSI could be a preferred technique when the spatial distribution of metabolites is important in clinical diagnosis. Recently, only few abscess studies have been performed using both of SVS and MRSI techniques [9,13], among which MRSI was only acquired to enhance the reliability of SVS. So far, the comparisons of metabolic profile between SVS and MRSI in abscess patients have not been performed. Therefore, the purpose of this study was to investigate the diagnostic efficacy of MRSI patients with pyogenic abscesses in comparison with SVS technique. Our hypothesis was that MRSI could provide similar results of metabolites to SVS technique in the cavity of pyogenic brain abscess, and its spatial distribution of metabolites could help differentiate brain abscess from necrotic tumors. 2. Materials and methods 2.1. Subjects A total of 47 patients with pyogenic brain abscesses from January 2002 to December 2011 were enrolled with in vivo MRS in this prospective study. The diagnosis was based on microscopic examination of culture and histopathologic examination of surgically aspirated or excised lesions. Inclusion criteria were the following: (a) patients had received both SVS and MRSI successfully in the examination; and (b) the size of abscesses was larger than 2 cm in diameter so that VOI can be localized in the center of abscess without partial volume effect caused by surrounding brain tissue. The spectra of 5 patients were obtained in poor quality due to motion artifacts. Therefore, 42 patients (29 men and 13 women; age range, 2 months–82 years) with pyogenic brain abscesses were included in the study. Approval for this study was obtained from our institutional review board, and informed consents were obtained from all of the patients. 2.2. Diffusion-weighted Imaging (DWI) The diffusion imaging sequence for DWI in the axial plane was single-shot spin-echo echo-planar imaging (TR/TE = 10,000 ms/93 ms) with diffusion sensitivities of b = 0 s/mm2 and b = 1000 s/mm2 . The diffusion gradients were

applied sequentially in three orthogonal directions to generate three sets of axial diffusion-weighted images. Sections (5 mm thick) with 2.5-mm intersection gaps, a 24-cm field of view, and a 128 × 256 matrix were used for all images. Analysis of diffusion changes was performed by FuncTool software (General Electric Medical System) for generation and analysis of apparent diffusion coefficient (ADC) maps. Apparent diffusion coefficient of the lesion cavity was calculated. The imaging time was 40 s. 2.3. Proton MR spectroscopy methods 2.3.1. Single voxel spectroscopy (SVS) All proton MRS examinations were performed on a 1.5-T MR scanner (Signa; GE Medical System, Milwaukee, Wis). SVS was performed with conventional point-resolved spectroscopy (PRESS) sequence (TR = 1600 ms, TE = 136 ms, average = 192, bandwidth = 2500 Hz, data point = 2048). In this study, only the long-TE technique was used because it has been observed that the metabolites in brain abscesses have intermediate-to-long T2 values. In addition, at TE = 136 ms, the resonances of AAs (i.e., valine, isoleucine, and leucine) and lactate (Lac) showed phase reversal while the lipid (Lip) resonances remained unchanged [1–6]. Voxel was centered within the lesion, which was determined visually with MR images from three orthogonal directions (sagittal, coronal and axial), in order to avoid partial volume effect as much as possible. The voxel size was ranging from 1.5 cm × 1.5 cm × 1.5 cm to 2 cm × 2 cm × 2 cm according to the lesion size. After automated transmitter and receiver adjustment, the signal over the VOI was shimmed to fulfill a typical full width at a half maximum (FWHM) of 4–8 Hz in all examinations. Optimal water resonance suppression was achieved by pre-irradiation of the water resonance with three chemical-shift-selective RF pulses followed by spoiling gradients. The acquisition time for each sequence was 5 min 45 s. 2.3.2. Magnetic resonance spectroscopic imaging (MRSI) MRSI was performed by using PRESS-localized MRSI sequence along with CHESS water suppression and outer-volume suppression (TR/TE = 1500/136 ms, average = 2, phase encoding matrix = 16 × 16, FOV = 160 mm × 160 mm). Slice thickness was either 10 mm or 15 mm depending on the size of lesion. The VOI was determined by both contrast-enhanced axial T1-weighted and FLAIR images to ensure that voxels were placed to cover the abscess cavity and contrast-enhancing area as well as contralateral normal-appearing brain tissue. To avoid fat contamination from scalp, the VOI was completely enclosed within the brain and positioned at the center of field of view. Automatic prescanning was performed twice before each spectroscopic scan to ensure adequate water suppression. The FWHM was kept under 10 Hz. The acquisition time was 12 min 55 s. 2.3.3. MR spectral analysis The post-processing and spectral analysis of SVS and MRSI data were carried out with FuncTool (GE Medical Systems) by two experienced neuroradiologists. For SVS data, the DC correction, zero filling to 4096 data points, 1.0-Hz exponential apodization, Fourier transformation and zero-order phase correction were performed sequentially for peak assignment based on the literatures [1–6]. The resonance peaks in the cavity including Lac at 1.3 ppm, AAs (leucine, isoleucine and valine) at 0.9 ppm, Ac at 1.92 ppm, Suc at 2.4 ppm and Lip at 0.8–1.3 ppm were analyzed in all patients by both SVS and MRSI. Two experienced neuroradiologists independently reviewed the spectroscopic data of all patients to determine the metabolites of abscesses obtained by SVS and MRSI. The judgments with binary value (presence or absence) of metabolites were made by individual reader and by the consensus of two readers.

S.-H. Hsu et al. / European Journal of Radiology 82 (2013) 1299–1307

The MR spectroscopic criteria for diagnosing pyogenic abscesses are the presence of AAs, such as leucine, isoleucine, valine at 0.9 ppm with or without additional resonances of Lac (1.3 ppm), Suc (2.4 ppm), Ac (1.9 ppm), and Lip (0.9–1.3 ppm) [5–8]. When AAs at 0.9 ppm were detected in MR spectra, it was recognized as a pyogenic abscess. Similarly, the abscess was considered to be of anaerobic origin when AAs along with Ac and/or Suc were identified on in vivo 1 H MRS [5–8]. Due to the J modulation, Lip and Lac as Lac/Lip (LL) mixture were estimated by using the upright Lip peak around 0.9–1.3 ppm and the inverted Lac peak at 1.3 ppm. After post-processing of all spectral data, the Pearson correlation coefficients of SVS and corresponding MRSI spectral data, obtained from the abscess cavity, were calculated to objectively assess the similarity between two spectral data in the range between 0.5 and 2.8 ppm which covered all peaks of interest for diagnosing pyogenic brain abscesses. The mean and standard deviation of correlation coefficients of all subjects were used to show the consistency between two spectral data. 2.3.4. Statistical analysis A Cohen  value was calculated to express inter-reader agreement of the metabolites of pyogenic abscesses. The  values less than 0.20 were interpreted as poor agreement, 0.21–0.40 as fair agreement, 0.41–0.60 as moderate, 0.61–0.80 as good and 0.81–1.00 as very good agreement. In addition, the correlation coefficients greater than 0.8 were interpreted as strong similarity of two spectral data. After consensus judgment of two readers, the binary value of metabolites of pyogenic abscesses (presence or absence) on MRSI was used to calculate sensitivity, specificity, positive and negative predictive values by using the binary values of SVS as the reference. These diagnostic parameters were expressed with a 95% confidence interval with the adjusted Wald method [14]. A  value was calculated to express inter-technique (MRSI and SVS) agreement of the metabolites of pyogenic abscesses. McNemar test was used to compare the discordant pairs between SVS and MRSI in the detection of metabolites for in vivo diagnosis of bacterial abscesses. Statistical analysis was performed by using statistical software (SPSS, version 17.0 for Windows, Chicago, Ill). A p value of less than 0.05 was considered statistically significant. 3. Results Increased signal on DWI was seen in 37 of 42 patients with pyogenic abscesses. The ADC value of the lesion cavity ranged from 0.41 to 0.86 × 10–3 mm2 /s (mean ± SD, 0.66 ± 0.17 × 10–3 mm2 /s). The remaining abscesses in 5 patients appeared isointense to slightly

1301

Table 1 Summary of the microorganisms isolated from the culture and metabolites visible on in vivo proton MR spectroscopy (MRS). MRS metabolites

Sterile (n = 10)

Aerobic (n = 6)

Anaerobic (n = 13)

Facultative anaerobic (n = 13)

Lac/Lip AAs + Lac/Lip Ac + AAs + Lac/Lip Suc + Ac + AAs + Lac/Lip

4 4 1 1

4 2 0 0

1 1 5 6

1 4 5 3

Note: AAs, cytosolic amino acids; Ac, acetate; Suc, succinate; Lac, lactate; Lip, lipid.

Table 2 The results of consistency of spectral interpretation of the two readers. MRS metabolites

AAs Ac Suc Lac Lip

Single-voxel

Magnetic resonance spectroscopic imaging





1.00 0.95 1.00 1.00 1.00

95% CI 1.00–1.00 0.86–1.00 1.00–1.00 1.00–1.00 1.00–1.00

1.00 1.00 1.00 1.00 1.00

95% CI 1.00–1.00 1.00–1.00 1.00–1.00 1.00–1.00 1.00–1.00

Note: AAs, cytosolic amino acids; Ac, acetate; Suc, succinate; Lac, lactate; Lip, lipid; CI, confidence interval.

hypointense on DWI and high ADC value of the lesion cavity (2.53 ± 0.07 × 10–3 mm2 /s). In vivo proton MRS characteristics and bacteriologic data obtained from the pus cultures of all patients are summarized in Table 1. Based on the results of bacterial culture, abscesses were grouped as sterile (no bacterial growth), anaerobic, aerobic, and facultative anaerobic. In 32 out of 42 abscess patients (76%), the resonances of AAs with or without other metabolites were observed by both SVS and MRSI, while in the rest of 10 patients (24%) AAs peaks were absent in their MR spectra. In 42 patients, four metabolic patterns of pyogenic brain abscesses obtained by SVS and MRSI were categorized as follows: (A) presence of AAs (leucine, isoleucine and valine) at 0.9 ppm, LL at 0.8–1.3 ppm, Ac at 1.92 ppm, and Suc at 2.4 ppm (Fig. 1); (B) presence of AAs at 0.9 ppm, LL at 0.8–1.3 ppm, and Ac at 1.92 ppm (Fig. 2); (C) presence of AAs at 0.9 ppm and LL at 0.8–1.3 ppm (Fig. 3); (D) presence of LL at 0.8–1.3 ppm, which were consistent with a recent report [8]. The inter-reader agreements for the detection of the metabolites of pyogenic abscesses are listed in Table 2. The spectral interpretation of MRSI data achieved full consistency ( = 1) between two readers in detecting AAs, Ac, Suc, Lac, and Lip. One disagreement between two readers happened on the recognition of Ac ( = 0.95)

Table 3 The diagnostic results on magnetic resonance spectroscopic imaging (MRSI) of the two readers in consensus with the single-voxel spectroscopy (SVS) as reference. MRS metabolites

Sensitivity

Specificity

Positive predictive value

Negative predictive value



AAs

Value 95% CI

100% (32/32) 0.91–1.00

100% (10/10) 0.75–1.00

100% (32/32) 0.91–1.00

100% (10/10) 0.75–1.00

1.00 1.00–1.00

Aca

Value 95% CI

91% (20/22) 0.71–0.99

100% (20/20) 0.86–1.00

100% (20/20) 0.86–1.00

91% (20/22) 0.71–0.99

0.90 0.78–1.00

Suc

Value 95% CI

100% (11/11) 0.77–1.00

100% (31/31) 0.90–1.00

100% (9/9) 0.73–1.00

100% (33/33) 0.91–1.00

1.00 1.00–1.00

Lac

Value 95% CI

100% (41/41) 0.93–1.00

100% (1/1) 0.22–1.00

100% (41/41) 0.93–1.00

100% (1/1) 0.22–1.00

1.00 1.00–1.00

Lip

Value 95% CI

100% (32/32) 0.91–1.00

100% (10/10) 0.75–1.00

100% (32/32) 0.91–1.00

100% (10/10) 0.75–1.00

1.00 1.00–1.00

Note: AAs, cytosolic amino acids; Ac, acetate; Suc, succinate; Lac, lactate; Lip, lipid; CI, confidence interval. a Discordant pairs between SV and MRSI for detecting Ac were not statistically significant (p = 0.5, McNemar test).

1302

S.-H. Hsu et al. / European Journal of Radiology 82 (2013) 1299–1307

Fig. 1. Representative in vivo single-voxel spectroscopy (SVS) and magnetic resonance spectroscopic imaging (MRSI) spectra from a 43-year-old man with stereotactic aspiration proven pyogenic brain abscess in the left frontal region secondary to Bacteroides fragilis (obligate anaerobe) infection. Two spectra have a correlation coefficient of 0.91 between 0.5 and 2.8 ppm. (a) Axial contrast-enhanced T1-weighted MR image (500/30) shows a ring-shaped cystic lesion and surrounding edema. The 2 cm × 2 cm × 2 cm voxel (box) in the center of the lesion represents the 1 H magnetic resonance spectroscopy (MRS) volume of interest (VOI). (b) In vivo 1 H single voxel spectrum (1600/136) from the abscess cavity shows spectral pattern representing succinate (Suc), acetate (Ac), lactate (Lac), and cytosolic amino acids (AAs). At a TE of 136 ms, the phase reversal resonances are well depicted at 1.3 ppm and 0.9 ppm, which confirms the assignment to Lac, and AAs, respectively. (c) Axial contrast-enhanced T1-weighted MR image (500/30) shows the voxel on MRSI. Voxel corresponding to center is highlighted in the contrast-enhanced T1-weighted MR image. The spectra were acquired as a 8 × 8 array with TE = 136 ms and nominal spatial resolution 1 cc. (d) The spectrum shows Suc, Ac, Lac, and AAs peaks in the center, corresponding to the findings of single voxel spectrum.

for SVS due to its small peaks. Besides, the correlation coefficients between SVS and corresponding MRSI spectral data in all subjects were 0.90 ± 0.05. Table 3 summarizes the sensitivity, specificity, positive and negative predictive value and inter-technique  values of the two readers in consensus and the corresponding 95% confidence intervals. The detection of AAs, Suc, Lac, and Lip had the same results by both SVS and MRSI, while Ac was only identified by SVS, but not MRSI, in two patients (Fig. 4). By using the binary values of SVS as reference, both sensitivities and negative predictive values of MRSI ranged from 91% (Ac) to 100% (AAs, Suc, Lac, Lip). The

specificity and positive predictive values of MRSI were 100% for all of five metabolites. The inter-technique  values ranged from 0.90 (for Ac) to 1. Although there existed a discordance in detecting Ac, no significant difference was found between MRSI and SVS (p = 0.5, McNemar test).

4. Discussion The inter-reader agreements in this study were very good for all of the metabolites and techniques ( ranged from 0.95 to 1).

S.-H. Hsu et al. / European Journal of Radiology 82 (2013) 1299–1307

1303

Fig. 2. A 54-year-old man with surgically proven pyogenic brain abscess in the left parietal area secondary to Eubacterium lentum (obligate anaerobe) infection, and two spectra have a correlation coefficient of 0.86 between 0.5 and 2.8 ppm. (a) Axial contrast-enhanced T1-weighted MR image (500/30) shows a ring-shaped cystic lesion. The 1.5 cm × 1.5 cm × 1.5 cm voxel (box) in the center of the lesion represents the 1 H MRS VOI. (b) In vivo 1 H proton single voxel spectrum (1600/136) from the abscess cavity shows spectral pattern representing Ac, Lac, and AAs. At a TE of 136 ms, the phase reversal resonances are well depicted at 1.3 ppm and 0.9 ppm, which confirms the assignment to Lac, and AAs, respectively. (c) Axial contrast-enhanced T1-weighted MR image (500/30) shows the voxel on MRSI. Voxel corresponding to center is highlighted in the contrast-enhanced T1-weighted MR image. The spectra were acquired as a 8 × 8 array with TE = 136 ms and nominal spatial resolution 1 cc. (d) The spectrum shows Ac, Lac, and AAs peaks in the center, corresponding to the findings of single voxel spectrum.

The detection of AAs, Suc, Lac, and Lip showed the same results ( = 1) between SVS and MRSI. The only discordance between SVS and MRSI was in Ac that was only identified by SVS in two patients probably due to its small peak. However, McNemar test for the discordant pairs in detecting Ac showed no statistically significant difference between MRSI and SVS (p = 0.5), given the limited power of this study. Besides, the correlation coefficient test showed very high similarity (=0.90 ± 0.05) between SVS and corresponding MRSI spectral data in patients with brain abscess, suggesting that MRSI produced strong similarity of MRS data to SVS in the range between 0.5 and 2.8 ppm.

In this study, spectra of the enhancing rim of the pyogenic abscess by MRSI typically show decreased NAA and Cr levels, no change to a mild decreased Cho level, a mild elevated Cho/Cr ratio, and elevated Lac/Lip, which are in general agreement with previous reports [1,9,13]. These visible small resonances of Cho, Cr, and NAA of the enhancing rim in the abscess subjects were interpreted as a combination of the resonances representing the cystic components and surrounding contrast-enhancing tissue due to partial volume effects. Moreover, NAA is reduced in the brain abscess capsule due to loss of intact neuronal cells, and Cr is particularly diminished because of general breakdown of energy metabolism.

1304

S.-H. Hsu et al. / European Journal of Radiology 82 (2013) 1299–1307

Fig. 3. A 76-year-old woman with surgically proven pyogenic brain abscess in the left temporal region secondary to Pseudomonas aeruginosa (aerobe) infection, and two spectra have a correlation coefficient of 0.93 between 0.5 and 2.8 ppm. (a) Axial contrast-enhanced T1-weighted MR image (500/30) shows a ring-enhanced cystic lesion. The 2 × 2 × 2 cm voxel (box) in the center of the lesion represents the 1 H MRS VOI. (b) In vivo 1 H SV spectrum (1600/136) from the abscess cavity shows spectral pattern representing Lac and AAs. At a TE of 136 ms, the phase reversal resonances are well depicted at 1.3, and 0.9 ppm, which confirms the assignment to Lac, and AAs, respectively. (c) Axial contrast-enhanced T1-weighted MR image (500/30) shows the voxel on MRSI. Voxel corresponding to center is highlighted in the contrast-enhanced T1-weighted MR image. The spectra were acquired as a 8 × 8 array with TE = 136 ms and nominal spatial resolution 1 cc. (d) The spectrum shows Lac and AAs peaks in the center, corresponding to the findings of single voxel spectrum.

The presence of the resonance of AAs is considered as a sensitive marker of pyogenic brain abscess; however, in our study, we found that AAs were not always present in pyogenic abscesses. It is known that, at an echo time of 136 ms, phase inversion as a result of Jcoupling between Lac and AAs along with the presence or absence of Ac or Suc, can be used to differentiate brain abscess from tumor [1–6,8,15–19]. In our study, the resonance of AAs with or without other metabolites was observed in 32/42 (76%) abscess patients by both SVS and MRSI, but was not detected in 10/42 (24%) cases, which were in accordance with the previous report [8]. The absence of AAs in four of 10 patients with sterile abscesses was probably

due to the fact that these patients were undergoing treatment with antibiotics [4,18,20,21] before the in vivo 1 H MRS, thus no bacterial growth on the culture. The fact that the AAs were not detected in the MR spectra from other six patients in our study may reflect the low concentration in pus of the bacteria-generated end-products and/or more Lip signals in the abscess cavity. Based on the results of this study, it appears that the presence of AAs is a reliable marker of pyogenic brain abscess, but its absence does not rule out the pyogenic etiology. Although SVS has shorter scan time (several minutes vs. 10 min) and better spectral quality than those of MRSI, the lack of ability

S.-H. Hsu et al. / European Journal of Radiology 82 (2013) 1299–1307

1305

Fig. 4. A 74-year-old man with surgically proven pyogenic brain abscess in the left occipital region secondary to Staphylococcus aureus (facultative anaerobes) infection, and two spectra have a correlation coefficient of 0.83 between 0.5 and 2.8 ppm. (a) Axial contrast-enhanced T1-weighted MR image (500/30) shows a ring-enhanced cystic lesion. The 1.5 × 1.5 × 1.5 cm voxel (box) in the center of the lesion represents the 1 H MRS VOI. (b) In vivo 1 H SV spectrum (1600/136) from the abscess cavity shows spectral pattern representing Lac, AAs and small Ac. At a TE of 136 ms, the phase reversal resonances are well depicted at 1.3, and 0.9 ppm, which confirms the assignment to Lac, and AAs, respectively. (c) Axial contrast-enhanced T1-weighted MR image (500/30) shows the voxel on MRSI. Voxel corresponding to center is highlighted in the contrast-enhanced T1-weighted MR image. The spectra were acquired as a 8 × 8 array with TE = 136 ms and nominal spatial resolution 1 cc. (d) The spectrum shows Lac and AAs peaks in the center, corresponding to the findings of single voxel spectrum. Ac was identified by SVS only, but not MRSI, probably due to small peaks.

in determining spatial heterogeneity of spectral patterns limits its ability to differentiate abscess from necrotic tumors. In abscesses caused by aerobic bacteria, the spectra measured in the central cavity may mimic those of intracranial necrotic GBMs, and cannot be differentiated by single measurement of SVS. Since metabolite ratios, maximum Cho/Cho-n, Cho/Cr and Cho/NAA ratios, of contrast-enhancing rim were significantly different between abscess and necrotic GBMs [9], simultaneous evaluations of cystic and rim-enhancing parts of lesions as well as normal parenchyma are helpful for differential diagnosis. MRSI, which acquires twodimensional spectral data from a large VOI in a single measurement,

not only possess similar spectral results with SVS in the center of abscess cavity, as illustrated in Figs. 1–4, but also provides the metabolic information of the cavity wall and contralateral normalappearing brain region, by which the differential diagnosis can be made through the comparison of the metabolites of the cavity wall between abscesses and GBMs [9]. Fig. 5 shows an example of differential diagnosis for abscess and GBMs by MRSI. Another potential advantage of MRSI is its higher spatial resolution, thus the evaluation of metabolites in brain abscess is less affected by the partial volume effect in MRSI [10,11,22]. Therefore, MRSI is potentially more suitable than SVS for the diagnosis of pyogenic brain abscess

1306

S.-H. Hsu et al. / European Journal of Radiology 82 (2013) 1299–1307

Fig. 5. (a) and (b) Representative in vivo MR images and spectra from a 74-year-old man with a pathologically proved right frontal region GBM. (a) Axial contrast-enhanced T1-weighted MR image (500/30) shows a ring-enhanced lesion in right frontal region and the area of the spectroscopy measurement (VOI) on MR spectroscopic imaging. Voxels corresponding with the center and enhancing rim of the lesion and corresponding contralateral normal-appearing brain are highlighted in the contrast-enhanced T1-weighted MR image. The spectra were acquired with TE 136 ms and nominal spatial resolution at 1 mL. (b) The spectra from those voxels are shown in detail. The spectra show Lac peak in the center, increased Cho/Cr ratio (maximum, 2.14), and increased Cho/Cho-n ratio (maximum, 1.65) in the rim-enhancing areas of the mass lesion. The contralateral normal-appearing brain region does not show any spectral alterations. (c) and (d) Representative in vivo MR images and spectra from a 46-year-old man with stereotactic aspiration proved pyogenic brain abscess in the right deep basal ganglion region secondary to S aureus (facultative aerobe) infection. (c) Axial contrast-enhanced T1-weighted MR image (500/30) shows a rim-enhanced lesion in the right temporoparietal region and the area of the spectroscopy measurement (VOI) on MR spectroscopic imaging. Voxels corresponding with the center and enhancing rim of the lesion and corresponding contralateral normal-appearing brain are highlighted in the contrastenhanced T1-weighted MR image. (d) The spectra from those voxels are shown in detail. The spectra show Lac and Lip and AAs peaks in the center, mild increased Cho/Cr ratio (maximum, 1.46), and decreased Cho/Cho-n ratio (maximum, 0.51) in the rim-enhancing areas of the mass lesion. The contralateral normal-appearing brain region does not show any spectral alterations.

when the disease under observation is diffuse or covers a large area of anatomy. DWI complements the role of conventional MR imaging in the differentiation of abscesses from necrotic tumors, given that

generally ADC is low in abscess cavities and high in tumor cysts [4]. However, high ADC similar to that found in necrotic tumors has been reported in 5–21% of untreated abscesses [23,24]. On the other hand, necrotic GBMs may demonstrate low ADC mimicking

S.-H. Hsu et al. / European Journal of Radiology 82 (2013) 1299–1307

abscesses [23,25,26]. While comparing DWI and in vivo MRS to differentiate brain abscess from necrotic brain tumor, we showed that DWI requires less imaging time and is superior to MRS when contamination from adjoining tissue or acquisition time may create limitations [4]. There were, however, some conflicting reports in the literature regarding DWI, so we advocated the combined use of MRS and DWI may improve results compared with the use of a single technique to differentiate brain abscess from necrotic tumors [4]. However, there are some limitations of this study. First, the radiologists presumably knew that they were looking at abscesses, so the potential biases toward increased sensitivity of suggesting brain abscess exist in this study. Second, since the voxel sizes of both SVS and MRSI were 2.0 cm × 2.0 cm × 2.0 cm and 1.5 cm × 1.5 cm × 1.5 cm, respectively, the spectra in lesions with smaller size may not be accurately measured due to partial volume effects. Third, the scan time of MRSI was longer than SVS performed in this study, thus it is more likely to have motion artifacts in MRSI. To possibly decrease scan time and keep roughly consistent SNR level, the MRSI could be achieved by using 120 mm × 120 mm FOV with a 12 × 12 matrix and 2 averages for those lesions located more centrally. 5. Conclusion This study compared the spectral results between SVS and MRSI, and found that MRSI can provide similar metabolites of abscess cavity to those of SVS in subjects with pyogenic brain abscesses. With additional metabolic information of the cavity wall and contralateral normal brain, MRSI is able to facilitate the differential diagnosis of brain abscess from necrotic tumors. Therefore, we concluded that MRSI is a more suitable routine protocol for the diagnosis of pyogenic brain abscesses, while SVS is recommended when further quantitative measurement of metabolites is needed. Conflict of interest There is no financial conflict of interest. Funding This work was supported in part by grants VGHKS101-011 and VGHKS102-032 from the Kaohsiung Veterans General Hospital, and NSC 100-2314-B-075B-001 and NSC 101-2314-B-075B-007-MY2 from the National Science Council, Taiwan. Acknowledgments The authors thank Chia-Chi Hsiao, RT, and Hung-Chieh Huang, RT, for patient MRI scanning, Ya-Wen Lin, RN, for patient preparation, and research assistants Tzi-Huay Wu, Shan-Shan Wang, for data management. References [1] Remy C, Grand S, Lai ES, et al. 1H MRS of human brain abscesses in vivo and in vitro. Magnetic Resonance in Medicine 1995;34(4):508–14.

1307

[2] Dev R, Gupta RK, Poptani H, Roy R, Sharma S, Husain M. Role of in vivo proton magnetic resonance spectroscopy in the diagnosis and management of brain abscesses. Neurosurgery 1998;42(1):37–43. [3] Grand S, Lai ES, Esteve F, et al. In vivo 1 H MRS of brain abscesses versus necrotic brain tumors. Neurology 1996;47(3):846–8. [4] Lai PH, Ho JT, Chen WL, et al. Brain abscess and necrotic brain tumor: discrimination with proton MR spectroscopy and diffusion-weighted imaging. American Journal of Neuroradiology 2002;23(8):1369–77. [5] Garg M, Gupta RK, Husain M, et al. Brain abscesses: etiologic categorization with in vivo proton MR spectroscopy. Radiology 2004;230(2): 519–27. [6] Lai PH, Li KT, Hsu SS, et al. Pyogenic brain abscess: findings from in vivo 1.5-T and 11.7-T in vitro proton MR spectroscopy. American Journal of Neuroradiology 2005;26(2):279–88. [7] Himmelreich U, Dzendrowskyj TE, Bourne R, Mountford CE, Sorrell TC. MR spectroscopy and infectious diseases. MAGMA 2000;11:197. [8] Pal D, Bhattacharyya A, Husain M, Prasad KN, Pandey CM, Gupta RK. In vivo proton MR spectroscopy evaluation of pyogenic brain abscesses: a report of 194 cases. American Journal of Neuroradiology 2010;31(2): 360–6. [9] Lai PH, Weng HH, Chen CY, et al. In vivo differentiation of aerobic brain abscesses and necrotic glioblastomas multiforme using proton MR spectroscopic imaging. American Journal of Neuroradiology 2008;29(8): 1511–8. [10] Vigneron DB, Nelson SJ, Murphy-Boesch J, et al. Chemical shift imaging of human brain: axial, sagittal, and coronal P-31 metabolite images. Radiology 1990;177(3):643–9. [11] Wald LL, Moyher SE, Day MR, Nelson SJ, Vigneron DB. Proton spectroscopic imaging of the human brain using phased array detectors. Magnetic Resonance in Medicine 1995;34(3):440–5. [12] Skoch A, Jiru F, Bunke J. Spectroscopic imaging: basic principles. European Journal of Radiology 2008;67(2):230–9. [13] Burtscher IM, Holtas S. In vivo proton MR spectroscopy of untreated and treated brain abscesses. American Journal of Neuroradiology 1999;20(6): 1049–53. [14] Agresti A, Coull BA. Approximate is better than “exact” for interval estimation of binomial proportions. American Statistician 1998:119–26. [15] Kim SH, Chang KH, Song IC, et al. Brain abscess and brain tumor: discrimination with in vivo 1 H MR spectroscopy. Radiology 1997;204(1): 239–45. [16] Poptani H, Gupta RK, Jain VK, Roy R, Pandey R. Cystic intracranial mass lesions: possible role of in vivo MR spectroscopy in its differential diagnosis. Magnetic Resonance Imaging 1995;13(7):1019–29. [17] Martinez-Perez I, Moreno A, Alonso J, et al. Diagnosis of brain abscess by magnetic resonance spectroscopy. Report of two cases. Journal of Neurosurgery 1997;86(4):708–13. [18] Chang KH, Song IC, Kim SH, et al. In vivo single-voxel proton MR spectroscopy in intracranial cystic masses. American Journal of Neuroradiology 1998;19(3):401–5. [19] Grand S, Passaro G, Ziegler A, et al. Necrotic tumor versus brain abscess: importance of amino acids detected at 1 H MR spectroscopy – initial results. Radiology 1999;213(3):785–93. [20] Harada M, Tanouchi M, Miyoshi H, Nishitani H, Kannuki S. Brain abscess observed by localized proton magnetic resonance spectroscopy. Magnetic Resonance Imaging 1994;12(8):1269–74. [21] Akutsu H, Matsumura A, Isobe T, et al. Chronological change of brain abscess in (1)H magnetic resonance spectroscopy. Neuroradiology 2002;44(7): 574–8. [22] Shulman RG, Blamire AM, Rothman DL, McCarthy G. Nuclear magnetic resonance imaging and spectroscopy of human brain function. Proceedings of the National Academy of Sciences of the United States of America 1993;90(8):3127–33. [23] Reddy JS, Mishra AM, Behari S, et al. The role of diffusion-weighted imaging in the differential diagnosis of intracranial cystic mass lesions: a report of 147 lesions. Surgical Neurology 2006;66(3):246–50 [discussion 250–1]. [24] Lee EJ, Ahn KJ, Ha YS, et al. Unusual findings in cerebral abscess: report of two cases. British Journal of Radiology 2006;79(947):e156–61. [25] Hartmann M, Jansen O, Heiland S, et al. Restricted diffusion within ring enhancement is not pathognomonic for brain abscess. American Journal of Neuroradiology 2001;22(9):1738–42. [26] Lai PH, Hsu SS, Ding SW, et al. Proton magnetic resonance spectroscopy and diffusion-weighted imaging in intracranial cystic mass lesions. Surgical Neurology 2007;68(Suppl. 1):S25–36.