European Journal of Radiology 83 (2014) 2065–2073

Contents lists available at ScienceDirect

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

A comparison of the performance of anatomical MRI and DTI in diagnosing carpal tunnel syndrome Sung Hye Koh a,b,∗ , Bong Cheol Kwon c,∗ , Chanyeong Park a , Su Yeon Hwang d , Joon Woo Lee d , Sam Soo Kim b a Department of Radiology, Hallym University Sacred Heart Hospital, Gwanpyeong-ro 170 beon-gil, Dongan-gu, Anyang-si, Gyeonggi-do, 431-796, Republic of Korea b Department of Radiology, Kangwon National University Hospital, Baengnyeong-ro 156, Chuncheon-Si, Gangwon-Do, 200-722, Republic of Korea c Department of Orthopedic Surgery, Hallym University Sacred Heart Hospital, Gwanpyeong-ro 170 beon-gil, Dongan-gu, Anyang-si, Gyeonggi-do, 431-796, Republic of Korea d Department of Radiology, Seoul National University Bundang Hospital, 82, Gumi-ro, 173 beon-gil, Bundang-gu, Seongnam-si, Gyeonggi-do, 463-707, Republic of Korea

a r t i c l e

i n f o

Article history: Received 14 June 2014 Received in revised form 5 August 2014 Accepted 8 August 2014 Keywords: Carpal tunnel syndrome Diffusion tensor imaging Magnetic resonance imaging Median nerve Wrist

a b s t r a c t Purpose: To compare the performance of anatomical magnetic resonance imaging (MRI) with that of diffusion tensor imaging (DTI) in the diagnosis of carpal tunnel syndrome (CTS). Materials and methods: We performed 3T anatomical MRI and DTI on 42 patients and 42 age-matched controls. The median nerve cross-sectional area (CSA), relative median nerve signal intensity, and palmar bowing of the flexor retinaculum, assessed with anatomical MRI, and fractional anisotropy (FA) and apparent diffusion coefficient of the median nerve, assessed with DTI, were measured at four locations: the hamate level, the pisiform level (P0), the level located 1 cm proximal to the P0 level (P1), and the distal radioulnar joint level (DR). Adding the ratios and differences of the median nerve parameters between the measurements at the DR and other locations to the diagnostic parameters, we evaluated the area under the receiver operating characteristic curves (AUCs) of all the diagnostic parameters of both scans. Results: The AUCs of FA(P1) (0.814) and FA(P0) (0.824) in DTI were larger than the largest AUC for anatomical MRI, CSA(P1) (0.759). However, the receiver operating characteristics of the three parameters were not significantly different (P > 0.1). The sensitivity and specificity of CSA(P1) (76.2% and 73.8%) and FA(P1) (73.8% and 76.2%) increased after inclusive and exclusive combination to 90.5% each. Conclusion: The individual performances of both scans were not significantly different in diagnosing CTS. Measuring both CSA and FA at P1 may be useful and efficient to utilize the merits of both scans and to increase the CTS diagnostic performance. © 2014 Elsevier Ireland Ltd. All rights reserved.

1. Introduction Carpal tunnel syndrome (CTS) is a common peripheral entrapment neuropathy [1]. The diagnosis of CTS is based on both clinical data and nerve conduction studies (NCSs). Magnetic resonance imaging (MRI) provides anatomical information about the carpal tunnels complements NCSs, and is a noninvasive method for imaging of carpal tunnels. There are multiple studies on the use of MRI to diagnose CTS [2–8]. Many of them have used diagnostic parameters such as the cross-sectional area and signal intensity of the median nerve and palmar bowing of the flexor retinaculum. However, these

∗ Corresponding authors. Tel.: +82 31 380 3898; fax: +82 31 380 3878. E-mail address: [email protected] (B.C. Kwon). http://dx.doi.org/10.1016/j.ejrad.2014.08.007 0720-048X/© 2014 Elsevier Ireland Ltd. All rights reserved.

studies have heterogeneous results [9]. There are recent studies on the use of diffusion tensor imaging (DTI) to evaluate the median nerve. In 2013, Lindberg et al. reported that DTI parameters correlate with the NCS results [10]. Some investigators have attempted to discriminate individuals with CTS from controls based on fractional anisotropy (FA) and apparent diffusion coefficient (ADC); however, their methods and results are not consistent, and their sample sizes are relatively small [11–15]. Anatomical MRI and DTI scans have diagnostic parameters for diagnosing CTS, but no study has compared the diagnostic performance of both scans for CTS. In addition, all of the parameters of both scans are location dependent from the distal radioulnar joint level to the distal carpal tunnel, because the median nerve is compressed and flattened within the carpal tunnel; however, it exhibits prestenotic swelling at the proximal portion [2,6,13]. For this reason, several investigations have used ratios or

2066

S.H. Koh et al. / European Journal of Radiology 83 (2014) 2065–2073

differences between the measurements in the carpal tunnel to the measurement at the distal radioulnar joint for the diagnosis of CTS with anatomical MRI or DTI [2,4,6,15–17]. Evaluating all the parameters of both scans at all locations with differences or ratios to make a diagnosis may be ineffective. Consequently, in this study, we compared the performances of anatomical MRI and DTI in diagnosing CTS to find what parameter of which scan is the best predictor of CTS, with a relatively large sample of subjects compared with the previously performed studies. Additionally, we evaluated the changes in diagnostic performance after the combinatory evaluation of both scans by identifying the appropriate diagnostic parameters among the numerous parameters of both scans for an effective CTS diagnosis. 2. Materials and methods 2.1. Subjects In the study, 42 wrists of 42 patients with idiopathic CTS (31 females: mean age, 52.4 years, range, 32–74 years; 11 males: mean age, 52.6 years, range, 31–64 years) and the wrists of 42 agematched controls (36 females: mean age, 53.4 years, range, 32–72 years; 6 males: mean age, 49.2 years, range, 32–61 years) were included. The exclusion criteria for all study participants were as follows: contraindications for MRI; a previous history of trauma to or surgery on the studied wrist; diseases that could affect the results, such as diabetes mellitus, rheumatoid arthritis, and renal disease; and use of any drugs that could affect the results. The inclusion criteria for the CTS group were clinical presentation and electrodiagnostic results consistent with CTS [1,18]. A normal electrodiagnostic result was mandatory for the controls. This prospective study was approved by our institutional review board, and all subjects provided informed consent. 2.2. Image acquisition MRI scans were acquired with a 3.0-T (Achieva, Philips, Best, Netherlands) scanner, with a dedicated eight-channel wrist coil (Achieva Sense; Philips). Scans were collected from the patients in the prone arm-overhead position. T2 weighted images were acquired as an anatomical MRI using a turbo-spine-echo sequence with the following parameters: TR/TE = 3719/80 ms, turbo-spin-echo factor = 2, field of view (FOV) = 150 mm × 150 mm, matrix = 272 × 264, slice thickness = 3 mm, slice gap = 0.3 mm, slices = 30, acquired voxel size = 0.55 mm × 0.57 mm × 3.00 mm, reconstructed voxel size = 0.2 mm × 0.2 mm × 3.0 mm, acquisition of one signal, and scan time = 2 min 51 s. The DTI scan was acquired using a diffusion-weighted single-shot echo-planar imaging pulse sequence with the following parameters: TR/TE = 7049/91 ms, b value = 1000 mm/sec2 , matrix size = 92 × 88, FOV = 100 mm × 100 mm, slice thickness = 2 mm, slice gap = 0, slices = 45, acquired voxel size = 1.1 mm × 1.1 mm × 2.0 mm, reconstructed voxel size = 1 mm × 1 mm × 2 mm, number of gradient directions = 15, acquisition of five signals, fat suppression = spectral presaturation with inversion recovery, echo-planar imaging factor = 47, sensitivity encoding factor = 2, and scan time = 10 min 28 s. 2.3. Image analysis For each patient, anatomical MRI and DTI scans were evaluated at different times. T2-weighted MR images were analyzed using PACS (Infinitt Co, Ltd., Seoul, Republic of Korea) as an anatomical MRI. The anatomical MRI parameters were cross-sectional area (CSA) of the median nerve, relative median nerve signal intensity

(MNSI), and palmar bowing of the flexor retinaculum (retinacular bowing). The CSA of the median nerve, including the perineurium, was measured at four levels: (i) the hook of the hamate level (H), (ii) the pisiform level (P0), (iii) the level located 1 cm proximal to the P0 level (P1), and (iv) the distal radioulnar joint level located 2 cm proximal to the P0 level (DR) [2,19,20] (Fig. 1). The MNSI was also obtained by measuring the signal intensity of the median nerve at the four locations and was expressed as a ratio that was calculated with respect to the signal intensity of the hypothenar muscle at the H level (signal intensity of the median nerve/signal intensity of the hypothenar muscle at the H level) [5]. The palmar bowing of the flexor retinaculum was defined as the perpendicular distance to the flexor retinaculum from the line between the hook of the hamate and the trapezium at the H level [19]. Additionally the bowing ratio was calculated based on the retinacular palmar bowing and the distance between the hook of the hamate and the trapezium (palmar bowing/distance of hook of hamate-trapezium) [19]. The DTI scans were evaluated after realignment using the diffusion registration package software provided by the manufacturer (Extended MR Work Space, version 2.5.3.0, Philips Medical Systems, Best, Netherlands). FA and ADC of the median nerve were also obtained at the four levels. FA and ADC were measured for the all of the voxels within the median nerve on the b0 image at each level, and we calculated the median of the FA and ADC measurements of each level (Fig. 2b). The voxels overriding the nerve margin were included due to the drawing ROI method applied and they may have had a significant impact on the measurements of the median nerve [21]. Therefore, we used the voxel selection method and verified that the selected voxels did not extend outside the nerve. The ratios and differences between the measurements related to the median nerve at the DR level and other levels of the anatomical MRI (CSA and MNSI) and DTI (FA and ADC) parameters were included as parameters of each scan (Tables 2 and 3). Anatomical MRI and DTI were analyzed by two radiologists who were blinded to the clinical data. The first observer carried out two series of blind measurements with more than 2 months between measurements to assess the intra-observer agreement. The second radiologist performed the same measurements to assess the interobserver reproducibility. 2.4. Statistical analysis To evaluate the differences between the CTS and control groups in the anatomical MRI and DTI parameters, t tests were performed. Receiver operating characteristic (ROC) curves were obtained for all the anatomical MRI and DTI parameters used to diagnose CTS, and the areas under the ROC curves (AUC) were calculated to find useful parameters for diagnosing CTS, for comparing both scans, and for the combinatory evaluation of both scans. There were 2 types of combinations: exclusive combinations, in which results were considered test-positive if both tests were positive, and inclusive combinations, in which they were considered test-positive if either of the tests was positive. To evaluate the benefit of the combinatory evaluation of both scans, sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), the likelihood ratio for a positive test (LR(+)), the likelihood ratio for a negative test (LR(−)), and the diagnostic odds ratios were calculated after the useful parameters for combinatory evaluation were identified. Intraclass correlation coefficients (ICCs) were used to calculate inter- and intra-observer agreement rates. ICC values of 0.01–0.20 indicated slight agreement; 0.21–0.40, fair agreement; 0.41–0.60, moderate agreement; 0.61–0.80, substantial agreement; and 0.81–1.0, near-perfect agreement [22]. These analyses were performed by using STATA software (version 10.0; Stata Corp, College Station, TX). For all statistical analyses,

S.H. Koh et al. / European Journal of Radiology 83 (2014) 2065–2073

2067

Fig. 1. The four levels used to measure the diagnostic parameters of anatomical MRI and DTI displayed from left to right (A set at each level is shown as superior – coronal, middle – sagittal, lower – axial scan. White arrow = median nerve at each level). (a) The hook of the hamate level (H); (b) the pisiform level (P0); (c) the level located 1 cm proximal to the P0 level (P1); (d) the distal radioulnar joint level located 2 cm proximal to the P0 level (DR). H = hamate, P = pisiform, R = radius, U = ulna, * = hook of the hamate.

a P value of less than 0.05 was considered statistically significant except for the multiple testing of diagnostic parameters of anatomical MRI and DTI for differentiating CTS and controls. Because many tests of significance were conducted within a single study, the statistical significance could be added by chance and cause an inflation of the alpha level [23]. Therefore, the Bonferroni correction was performed, and a Bonferroni-Adjusted P (P-value less than 0.0011) was considered significant [24]. 2.5. SNR of DTI We measured the SNR at the median nerve in the b = 0 image using the following calculation: ‘SNR = mean of the object/standard deviation of the noise in the object’. As the standard deviation of the noise in the object is biased by proper signal variation, the standard deviation of the noise in the object, ‘s’, was estimated from the mean of the noise image at the same location [25,26]. For a Rician distribution in the absence of a signal, ‘s’ was calculated as the mean of the noise image/1.253 [27]. the SNR = Mobject/s = 1.253 × Mobject/Mnoise, Therefore, where Mobject is the mean of the object in the object image and Mnoise is the mean of the noise image. We obtained the mean value of the object at the center of the median nerve and the value of the noise image at the same position according to the method described by Li et al. [28]. 3. Results The age and gender distributions of the CTS and control groups are listed in Table 1. Detailed data on the CSAs, MNSIs, and

retinacular bowings derived from anatomical MRI and on the FAs and ADCs derived from DTI are shown in Tables 2 and 3, including the ratios and differences of the CSA and MNSI of the anatomical MRI and the FA and ADC of the DTI between the DR level and the other levels.

3.1. Anatomical MRI parameters CSA(P1) showed the largest AUC (0.759, 95% CI: 0.655–0.863) among the anatomical MRI parameters and was significantly larger in the CTS group than in the control group (P = 0.0001) (Table 2). CSA(P1-DR) showed significant differences between the control and CTS groups (P < 0.0011), but their AUCs were smaller than those of CSA(P1). The parameters related to MNSI and retinacular bowing revealed no statistically significant differences between the control and CTS groups (P > 0.0011).

3.2. DTI parameters FA(P0) and FA(P1) showed large AUCs over 0.8 and their FA values were significantly lower in CTS than in the controls (P < 0.0001) (Table 3). FA(P0) showed the largest AUC (0.824, 95% CI: 0.735–0.913). FA(DR) and ADC(P0), also demonstrated significant differences between the control and CTS groups (P < 0.0011), but their AUCs were smaller than those for FA(P0). The average number of selected voxels per level to calculate median FA and ADC of the median nerve was five.

2068

S.H. Koh et al. / European Journal of Radiology 83 (2014) 2065–2073

Table 1 Age and gender distribution of the CTS and control groups. Age (years)

CTS

Control

n

Male

Female

Mean

n

Male

Female

Mean

0.1) or between CSA(P1) and FA(P1) (P > 0.1) (Table 4 and Fig. 3).

MNSI(H):FA(H) MNSI (P0):FA(P0) MNSI (P1):FA(P1) MNSI (DR):FA(DR)

−0.1507 (0.6762) −0.1488 (0.6884) −0.2454 (0.1380) −0.0784 (0.9798)

Correlations represent Pearson’s r.

FA(P1) and FA(P0) were not significantly different (P > 0.1) (Table 4). There was no difference between FA(P1) and FA(P0) according to McNemar’s test after the optimal cutoff values of 0.536 and 0.44, respectively (P = 0.68) were obtained. There was a significant negative correlation between CSA and FA at the P1 level, with a Pearson correlation of r = −0.4795 (P < 0.0001) (Table 5). At the other levels, CSA and FA were not significantly correlated (P > 0.1). Additionally, MNSI and FA were not significantly correlated at any level (P > 0.1) (Table 6). 3.5. Combinatory evaluation of anatomical MRI and DTI

3.4. Appropriate location for the combinatory evaluation of anatomical MRI and DTI At the P1 level, both anatomical MRI and DTI yielded large AUCs for CSA and FA (Tables 2 and 3). Their AUCs were the second largest, following that for FA(P0), and they showed significant differences between the control and CTS groups (P < 0.0001). The ROCs of Table 4 ROC comparison between FA(P1) and FA(DR) derived from DTI, CSA(P1) derived from anatomical MR, and FA(P0) and FA(P1) derived from DTI. Parameters

P-value

FA(P0) vs. FA(P1) CSA(P1) vs. FA(P0) CSA(P1) vs. FA(P1)

0.841 0.288 0.374

The sensitivity and specificity of FA at P1 with the cutoff value of 0.536 were 73.8% (95% CI: 58–86.1) and 76.2% (95% CI: 60.5–87.9), respectively (Table 7). For CSA at the P1 level, the most favorable cutoff value was 14.34 mm2 , with a sensitivity of 76.2% (95% CI: 60.5–87.9) and specificity of 73.8% (95% CI: 58–86.1). With the inclusive combination of the parameters CSA(P1) and FA(P1), the sensitivity increased to 90.5% (95% CI: 77.4–97.3), the negative predictive value increased to 86.2% (95% CI: 68.3–96.1), and the likelihood ratio for a negative test decreased to 0.16 (95% CI: 0.061–0.42) for each parameter separately (Table 7). With the exclusive combination of the parameters CSA(P1) and FA(P1), the specificity increased to 90.5% (95% CI: 77.4–97.3), the positive predictive value increased to 86.2% (95% CI: 68.3–96.1), and the likelihood ratio for a positive test increased to 6.25 (95% CI: 2.38–16.4) for each parameter separately. The diagnostic odds

CSA is in mm2 , and FA is unitless. The cutoff values: CSA(P1) = 14.34 mm2 , FA(P1) = 0.536. Numbers in parentheses are the 95% confidence intervals. PPV, positive predictive value; NPV, negative predictive value; LR(+), likelihood ratio for a positive test; LR(−), likelihood ratio for a negative test.

LR (−)

0.323 (0.182–0.571) 0.344 (0.201–0.587) 0.16 (0.061–0.42) 0.447 (0.306–0.654) 2.91 (1.7–4.97) 3.1 (1.75–5.48) 2.24 (1.53–3.27) 6.25 (2.38–16.4)

LR (+) NPV (%)

74.4 (58.8–86.5) 75.6 (59.7–87.6) 69.1 (55.2–80.9) 86.2 (68.3 –96.1)

75.6 (59.7–87.6) 74.4 (58.8–86.5) 86.2 (68.3–96.1) 69.1 (55.2–80.9)

PPV (%)

73.8 (58–86.1) 76.2 (60.5–87.9) 59.5 (43.3–74.4) 90.5 (77.4–97.3) 76.2 (60.5–87.9) 73.8 (58–86.1) 90.5 (77.4–97.3) 59.5 (43.3–74.4) CSA(P1) FA(P1) Inclusive combination: CSA(P1) or FA(P1) Exclusive combination: CSA(P1) and FA(P1)

Specificity (%) Sensitivity (%) Parameter

Table 7 Diagnostic performance of anatomical MRI and DTI in diagnosing CTS by combinatory evaluation at the P1 level.

9.02 (3.39–24) 9.02 (3.39–24) 14 (4.34–44.3) 14 (4.34–44.3)

S.H. Koh et al. / European Journal of Radiology 83 (2014) 2065–2073

Odds ratio

2070

Table 8 Intra- and inter-observer variance of anatomical MRI and DTI measurements.

Anatomical MRI CSA

MNSI

ICC-intra-observer

ICC-inter-observer

H P0 P1 DR

0.763 0.397 0.540 0.567

0.499 0.431 0.417 0.495

H P0 P1 DR

0.733 0.745 0.669 0.756

0.624 0.627 0.771 0.775

0.799 0.760

0.616 0.671

H P0 P1 DR

0.825 0.908 0.890 0.876

0.539 0.774 0.725 0.718

H P0 P1 DR

0.961 0.847 0.841 0.902

0.529 0.780 0.591 0.782

Retinacular palmar displacement Retinacular bowing ratio DTI FA

ADC

ratios for both the inclusive and exclusive combinations of the two parameters increased from 9.02 to 14 for each parameter separately. 3.6. Intra- and inter-observer variances The intra- and inter-observer variances of the measurements of anatomical MRI and DTI are listed in Table 8. Regarding the FA and ADC measurements of the DTI, the intra-observer agreements were nearly perfect (0.825–0.961), and the inter-observer agreements were fair or moderate (0.529–0.782). However, the intraand inter-observer agreements for the CSA measurements were fair or moderate (0.397–0.763) and lower than DTI parameters. The intra- and inter-observer agreements for MNSI and retinacular bowing were substantial (0.624–0.799). 3.7. SNR of DTI The measured SNR in the median nerve was 24.71 (mean of the ROI in the nerve: 927.42, mean of the ROI in the noise: 47.02). 4. Discussion Our study revealed useful diagnostic parameters among the many anatomical MRI and DTI parameters used to diagnose CTS by comparing the diagnostic performances of both scans, and it revealed useful parameters and the appropriate location for the efficient and diagnostic combinatory evaluation of both scans (Fig. 4). We found that the evaluation of both scans at the P1 level measuring CSA and FA is useful and efficient for CTS diagnosis. To our knowledge, this is the first report of MRI and DTI carpal tunnel evaluation at the P1 level. 4.1. Anatomical MRI We performed T2-weighted imaging in the form of anatomical MRI because the commonly used diagnostic parameters for CTS are CSA, relative signal intensity of the median nerve, and palmar bowing of the flexor retinaculum, and these parameters had been measured on T2-weighted images in previous studies [3,7,16]. Our data showed that the difference in CSA between the two groups was significant only at the P1 level among the four levels.

S.H. Koh et al. / European Journal of Radiology 83 (2014) 2065–2073

Fig. 4. The flow chart of this study.

Although the CSAs at the P0 level, representing the carpal tunnel inlet, were largest in both the CTS and control groups, but they were not significantly different between the two groups (Table 2). Wiesler et al. reported that nerve enlargement just proximal to the carpal tunnel inlet was the most specific finding of CTS using ultrasonography, and they measured the CSA at the distal wrist crease [20]. The distal wrist crease is typically observed near the proximal portion of the pisiform. This level matched the P1 level in our study (Fig. 1). Previous studies generally evaluated the carpal tunnel in at least two or three locations, namely, the P0, H, and/or DR levels but not in the P1, and the study methods and results have been inconsistent [2–7,16,17]. Britz et al. evaluated the median nerve size at P0 and DR, and they considered that the positive finding for CTS was median nerve flattening at P0 and/or proximal swelling [17]. Additionally Radack et al. evaluated the median nerve size at P0 and H, and they considered the finding positive when the median nerve at P0 was larger than at the H level [3]. The former study reported that 65% of CTS was positive, and the later study reported that the finding was not significantly different between the CTS and controls. Mesgarzadeh et al. evaluated median nerve swelling ratios with CSA measurements at the H, P0, and DR levels to obtain CSA(H/DR) and CSA(P0/DR), and the ratios were significantly different between the two groups [2]. However, in the study by Bak et al., the two parameters were not significantly different between groups [4]. We found that the CSA(H/DR) and CSA(P0/DR) were not significantly different between groups (P = 0.901 and 0.839) in our study (Table 1). There are also inconsistent CSA results derived from CTS studies with MR staging. Kleindienst et al. evaluated the CSAs at the H, P0, and DR levels, and they classified CTS into three stages, ranging from early to advanced [5]. CSA(DR) was significantly different only between the controls and the intermediate CTS group and was not significant between the controls and either the mild

2071

or advanced CTS group. CSA(H) and CSA(P0) were significantly different between the control group and all three CTS stage groups. In the study by Uchiyama et al., the CSAs in extreme-stage CTS did not significantly differ from those for the controls at the H level, and the CSAs at the other three locations in the mild, moderate, severe, and extreme-stage CTS groups were significantly different from those in the control group [7]. The differences between the previous studies, including our own, might have been caused by differences in the numbers of subjects and in the proportions of patients with different clinical stages of CTS. The relative MNSI was not a significant parameter for distinguishing between the control and CTS groups in our study, similar to the findings of Radack et al. and Bonél et al. [3,16]. In the study by Kleindienst et al., the MNSIs at the H, P0, and DR levels increased in early-stage CTS but decreased in the advanced stage [5]. This finding was explained by endoneural fibrosis, demyelination, and axonal degeneration in persistent median nerve compression. Because the CTS group was not classified according to symptom duration in our study, increased MNSI in the early stage could be mixed with decreased MNSI in chronic CTS. The previous studies reported that retinacular palmar bowing and the bowing ratio of the hamate hook level were significantly different between the two groups [2,3,6,7,29]. However, in our study, the two parameters were not significant parameters for discriminating between the two groups, again similar to the findings of Bak et al. and Jarvik et al. [4,8]. In the study by Tsujii et al., the degree of retinacular bowing was greater in the early (≤3 months) and chronic (≥13 months) stages compared with the intermediate stage [29]. The authors explained that histological change in the flexor tendon sheath over the disease course is an important factor for the degree of change in palmar bowing of the flexor retinaculum. Thus, the difference between their results and ours might also have been caused by the differences between their studies and ours regarding the CTS disease course. 4.2. DTI Guggenberger et al. and Barcelo et al. reported significant differences in the FA and ADC measurements of the median nerve of the wrist depending on their locations [13,15]. Barcelo et al. reported that FA(H) and FA(H-DR) were good predictors of CTS using DTI. However, in our study, these two parameters were not significantly different between the CTS and controls, but FA(P0) and FA(P1) were good predictors of CTS. Although Barcelo et al. did not evaluate the median nerve at the P1 level, the different results might be caused by the different measuring method for the local FA and ADC. In the study by Guddenberg et al., the AUCs for the mean FA and ADC were 0.773 and 0.741, respectively, without location specification. In our study, the AUCs for FA at P0 (0.824) and P1 (0.814) were higher than their AUC for the mean FA, but the AUCs of the ADC values were slightly lower than that for their mean ADC. Local evaluation of FA may be more effective than obtaining mean values, especially at P0 or P1. 4.3. Comparing MRI and DTI and evaluating their combined results In this study we found that there was no significant difference in the ROC between the parameters exhibiting the largest AUC for each scan (FA(P0) and CSA(P1), P > 0.1), but there were more useful DTI parameters than anatomical MRI parameters for discriminating the CTS group from the controls, and the parameters that exhibited AUCs greater than 0.8 were associated only with DTI (Tables 2 and 3). DTI provides functional data for median nerve evaluation [10,15]. However, when we performed DTI, we also always performed

2072

S.H. Koh et al. / European Journal of Radiology 83 (2014) 2065–2073

anatomical MRI as a reference scan, and gross pathological causes of CTS were more easily detected on anatomical MRI than on DTI images. In addition, the sensitivity of CSA(P1) (76.2%) was higher than that of FA(P1) (73.8%) in our study. Both DTI and anatomical MRI have merit for diagnosing CTS. When we evaluated the changes in diagnostic performances after combinatory evaluation, the sensitivity and NPV of the inclusive combination of CSA and FA at the P1 level and the specificity and PPV of their exclusive combination were also higher than they were for each parameter (Table 7). Therefore, it is more beneficial to use both DTI and anatomical MRI for evaluating the median nerve to diagnose CTS compared with using DTI alone. Although the AUC for FA(P1) was the largest among the parameters of DTI and anatomical MRI, it may be more efficient to evaluate CSA and FA values at the same P1 location rather than conducting both scans at different P1 and P0 locations because there was no significant difference between FA(P1) and FA(P0) in diagnosing CTS (Table 4). Only at the P1 level, were both the FA of DTI and the CSA of anatomical MRI significantly different between the CTS and control groups, and they were negatively correlated. At the H and P0 levels, the median nerve was compressed by the flexor retinaculum; thus, nerve swelling might have been partially masked. However, at the P1 level, the nerve swelling was unmasked, and CSA could fully express the nerve swelling without external compression and was negatively correlated with FA. These findings are supported by the study by Naraghi et al., which suggested that FA changes are related to intrafascicular edema and demyelination/remyelination of the median nerve [30]. In our study, we evaluated more anatomical MRI and DTI parameters than did previous studies, and both our CTS and control groups were larger than the samples in previous studies of diagnosing CTS with anatomical MRI and DTI [2–4,8,11–17]. The two groups were matched for age because the DTI parameters are age-dependent [10,14]. The analyses in our study were not conducted according to age group. However, the purpose of our study was to compare the diagnostic performances of anatomical MRI and DTI, rather than to identify any age-specific threshold values for the diagnostic parameters. There are additional limitations in our study. First, we did not calculate either DTI-NCS or anatomical MRI-NCS correlations because this was not the purpose of our investigation. Second, the intra- and inter-observer CSA ICCs were lower than those of FA at all levels. The blurring of the median nerve margin caused by edema of the nerve in CTS and partial-volume artifacts in the surrounding tissue might be the main causes for the lower ICCs for CSA because we only performed anatomical MRI with T2-weighted images with a 3-mm slice thickness. Third, we did not evaluate any changes in CTS diagnostic performance with the combination of other diagnostic anatomical MRI and DTI parameters. Many possible combinations exist that might improve the diagnostic performance over that associated with the use of a single diagnostic parameter. However, the purpose of this study was to identify diagnostic parameters for the effective and highly diagnostic combination of both anatomical MRI and DTI using the merits of each scanning modality.

5. Conclusion The diagnostic parameters that showed relatively large AUCs were FA(P0) and FA(P1) among the DTI parameters and CSA(P1) among the anatomical MRI parameters. There were no significant differences in the ROC values between the three parameters. The AUCs of the other parameters for both scanning modalities, including differences and ratios, were lower than those of the above-mentioned parameters. At the P1 level, the FA and CSA were negatively correlated and exhibited a large AUC. The sensitivity and specificity of FA(P1) (73.8% and 76.2%; cutoff value = 0.536) and CSA(P1) (76.2% and 73.8%; cutoff value = 14.34 cm) increased after

the inclusive and exclusive combinations of both scans to 90.5% each. To take advantage of the merits of both anatomical MRI and DTI and to increase the CTS diagnostic performance, measuring both CSA and FA at P1 may be useful and efficient. Acknowledgements The authors thank Dr. Kwang-Ik Jung (obtaining NCS data), Dr. Soo Joong Choi (study coordination), Haram Moon (study coordination), Yeongmi Yang (data acquisition), Yun Kyung Chung (Statistic analysis) and Joy J. Goto (language help and manuscript editing) for their assistance. This study was supported by the Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education, Science and Technology (grant number 2011-0013773). References [1] Ibrahim I, Khan W, Goddard N, Smitham P. Carpal tunnel syndrome: a review of the recent literature. Open Orthop J 2012;6:69–76. [2] Mesgarzadeh M, Schneck CD, Bonakdarpour A, Mitra A, Conaway D. Carpal tunnel: MR imaging. Part II. Carpal tunnel syndrome. Radiology 1989;171:749–54. [3] Radack DM, Schweitzer ME, Taras J. Carpal tunnel syndrome: are the MR findings a result of population selection bias? AJR Am J Roentgenol 1997;169:1649–53. [4] Bak L, Bak S, Gaster P, Mathiesen F, Ellemann K, Bertheussen K, et al. MR imaging of the wrist in carpal tunnel syndrome. Acta Radiol Stockh Swed 1987 1997;38:1050–2. [5] Kleindienst A, Hamm B, Lanksch WR. Carpal tunnel syndrome: staging of median nerve compression by MR imaging. J Magn Reson Imaging 1998;8:1119–25. [6] Monagle K, Dai G, Chu A, Burnham RS, Snyder RE. Quantitative MR imaging of carpal tunnel syndrome. Am J Roentgenol 1999;172:1581–6. [7] Uchiyama S, Itsubo T, Yasutomi T, Nakagawa H, Kamimura M, Kato H. Quantitative MRI of the wrist and nerve conduction studies in patients with idiopathic carpal tunnel syndrome. J Neurol Neurosurg Psychiatry 2005;76:1103–8. [8] Jarvik JG. MR nerve imaging in a prospective cohort of patients with suspected carpal tunnel syndrome. Neurology 2002;58:1597–602. [9] Pasternack II, Malmivaara A, Tervahartiala P, Forsberg H, Vehmas T. Magnetic resonance imaging findings in respect to carpal tunnel syndrome. Scand J Work Environ Health 2003;29:189–96. [10] Lindberg PG, Feydy A, Viet DL, Maier MA, Drapé J-L. Diffusion tensor imaging of the median nerve in recurrent carpal tunnel syndrome – initial experience. Eur Radiol 2013;23:3115–23. [11] Khalil C, Hancart C, Le Thuc V, Chantelot C, Chechin D, Cotten A. Diffusion tensor imaging and tractography of the median nerve in carpal tunnel syndrome: preliminary results. Eur Radiol 2008;18:2283–91. [12] Stein D, Neufeld A, Pasternak O, Graif M, Patish H, Schwimmer E, et al. Diffusion tensor imaging of the median nerve in healthy and carpal tunnel syndrome subjects. J Magn Reson Imaging 2009;29:657–62. [13] Guggenberger R, Markovic D, Eppenberger P, Chhabra A, Schiller A, Nanz D, et al. Assessment of median nerve with MR neurography by using diffusion-tensor imaging: normative and pathologic diffusion values. Radiology 2012;265:194–203. [14] Wang C-K, Jou I-M, Huang H-W, Chen P-Y, Tsai H-M, Liu Y-S, et al. Carpal tunnel syndrome assessed with diffusion tensor imaging: comparison with electrophysiological studies of patients and healthy volunteers. Eur J Radiol 2012;81:3378–83. [15] Barcelo C, Faruch M, Lapègue F, Bayol M-A, Sans N. 3-T MRI with diffusion tensor imaging and tractography of the median nerve. Eur Radiol 2013;23:3124–30. [16] Bonél HM, Heuck A, Frei KA, Herrmann K, Scheidler J, Srivastav S, et al. Carpal tunnel syndrome: assessment by turbo spin echo, spin echo, and magnetization transfer imaging applied in a low-field MR system. J Comput Assist Tomogr 2001;25:137–45. [17] Britz GW, Haynor DR, Kuntz C, Goodkin R, Gitter A, Kliot M. Carpal tunnel syndrome: correlation of magnetic resonance imaging, clinical, electrodiagnostic, and intraoperative findings. Neurosurgery 1995;37:1097–103. [18] Stevens JC. AAEE minimonograph #26: the electrodiagnosis of carpal tunnel syndrome. Muscle Nerve 1987;10:99–113. [19] Mesgarzadeh M, Schneck CD, Bonakdarpour A. Carpal tunnel: MR imaging. Part I. Normal anatomy. Radiology 1989;171:743–8. [20] Wiesler ER, Chloros GD, Cartwright MS, Smith BP, Rushing J, Walker FO. The use of diagnostic ultrasound in carpal tunnel syndrome. J Hand Surg 2006;31:726–32. [21] Vos SB, Jones DK, Viergever MA, Leemans A. Partial volume effect as a hidden covariate in DTI analyses. Neuroimage 2011;55:1566–76. [22] Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977;33:159. [23] Streiner DL, Norman GR. Correction for multiple testing: is there a resolution? Chest 2011;140:16–8.

S.H. Koh et al. / European Journal of Radiology 83 (2014) 2065–2073 [24] Aickin M, Gensler H. Adjusting for multiple testing when reporting research results: the Bonferroni vs Holm methods. Am J Public Health 1996;86:726–8. [25] Gudbjartsson H, Patz S. The Rician distribution of noisy MRI data. Magn Reson Med 1995;34:910–4. [26] Kim HS, Yoon YC, Kwon JW, Choe B-K. Qualitative and quantitative assessment of isotropic ankle magnetic resonance imaging: three-dimensional isotropic intermediate-weighted turbo spin echo versus three-dimensional isotropic fast field echo sequences. Korean J Radiol 2012;13:443–9. [27] Dietrich O, Raya JG, Reeder SB, Reiser MF, Schoenberg SO. Measurement of signal-to-noise ratios in MR images: influence of multichannel coils, parallel imaging, and reconstruction filters. J Magn Reson Imaging 2007;26:375–85.

2073

[28] Li C, Chen W, Beatty P, Brau A, Hargreaves B, Busse R, et al. SNR quantification with phased-array coils and parallel imaging for 3d-fSE. In: 19th annual meeting & exhibition. 2010. p. 552. [29] Tsujii M, Hirata H, Morita A, Uchida A. Palmar bowing of the flexor retinaculum on wrist MRI correlates with subjective reports of pain in carpal tunnel syndrome. J Magn Reson Imaging 2009;29:1102–5. [30] Naraghi A, da Gama Lobo L, Menezes R, Khanna M, Sussman M, Anastakis D, et al. Diffusion tensor imaging of the median nerve before and after carpal tunnel release in patients with carpal tunnel syndrome: feasibility study. Skeletal Radiol 2013;42:1403–12.

A comparison of the performance of anatomical MRI and DTI in diagnosing carpal tunnel syndrome.

To compare the performance of anatomical magnetic resonance imaging (MRI) with that of diffusion tensor imaging (DTI) in the diagnosis of carpal tunne...
1MB Sizes 1 Downloads 5 Views