Comparative Analysis of Methods for Obtaining Median Nerve Cross-Sectional Area Using Ultrasound

Abstract

Background:

Ultrasound measurement of median nerve cross-sectional area (CSA) is widely used in carpal tunnel syndrome (CTS) diagnosis. This study compared 3 common measurement methods: A-B, Ellipse, and Trace to evaluate their reliability and accuracy.

Methods:

Two observers (experienced hand surgeon and orthopedic surgery resident) measured median nerve CSA in 10 patients with suspected CTS. Each method was applied in standard and zoom-enhanced views: A-B method (measuring major/minor axes), Ellipse method (fitting ellipse to nerve boundaries), and Trace method (manually tracing nerve boundary). Measurements were compared to ImageJ software analysis as reference standard. Reliability and accuracy were assessed using intraclass correlation coefficients (ICCs), Bland-Altman analysis, and standard detectable difference (SDD).

Results:

All methods showed high reliability (ICC: 0.940-0.994) but consistently underestimated CSA compared to ImageJ (12.15 ± 5.13 mm²). The Trace method showed smallest mean difference from ImageJ (−1.26 mm²), followed by A-B (−1.46 mm²) and Ellipse (−1.96 mm²). The Trace method demonstrated highest accuracy without zoom (85.5%) and lowest SDD (1.57 mm²). Interrater reliability was significantly higher in zoom condition for all methods, with Trace method showing highest interrater reliability in both zoom and nonzoom conditions.

Conclusion:

All 3 ultrasound methods demonstrated high reliability in measuring median nerve CSA, with Trace method showing modest advantages in accuracy and precision. Zoom enhanced views improved interrater reliability but reduced measurement accuracy. Based on our findings, we recommend nonzoom Trace method for consistent longitudinal measurements, though maintaining proficiency in one selected method is more important. Further validation with larger studies is warranted.

Level of Evidence:

Level 2

Introduction

Carpal tunnel syndrome (CTS) is a compressive neuropathy of the median nerve that presents with numbness, tingling, and/or pain in the median nerve distribution of the hand as well as potential muscle atrophy in later stages.^1-3 Of the available confirmatory tests, ultrasound is among the most widely used as a diagnostic tool to evaluate the cross-sectional area (CSA) of the carpal tunnel. Not only is an increased CSA of the median nerve on ultrasound at the carpal tunnel inlet sensitive and specific for the diagnosis of CTS, but using ultrasound also offers significant advantages, including cost-effectiveness, noninvasiveness, less patient discomfort compared to electrodiagnostic testing, widespread availability, time efficiency, and easy learning curves while simultaneously enabling real-time tracking of nerve courses and assessing positional variations.^4-12

Three different methods are most commonly used for the measurement of CSA obtained via ultrasound: A-B method, Ellipse method, and Trace method. Both the A-B method and Ellipse method hinge on the geometric approximation technique that assumes the CSA is elliptical and relies on the measurement of an ellipse to determine the CSA of the carpal tunnel. ¹³ The A-B method uses a ruler or caliper to measure the length and width of the CSA. Which is then used to calculate the area of an ellipse using the equation, A_ellipse = πab, where “a” is the length of the semimajor axis and “b” is the length of the semiminor axis. The Ellipse method places a best-fit ellipse around the given CSA, and the measurement of that ellipse is taken as the CSA. Finally, the Trace method requires a best-fit line to be drawn along the outline of the CSA, and the area within this best-fit line is calculated. Software-based analysis tools like ImageJ (National Institutes of Health, Bethesda, MD, USA) have been used in previous studies to provide standardized measurements and serve as reference standards when evaluating ultrasound-based methods. ¹⁴

While ultrasound is known to be a reliable method for defining the CSA of the median nerve, to the best of our knowledge there is no consensus on which method is best in clinical practice.^8,15 In addition, the use of ultrasound for the diagnosis of CTS relies on accurate measurement of the CSA of the median nerve at the inlet of the carpal tunnel and is both method-dependent and operator-dependent. With these factors known, and no current consensus on the best method for obtaining the CSA when using ultrasound, our study set out to describe which method is the most reliable for measuring CSA of the median nerve when using ultrasound. Our null hypothesis is that there is no difference in CSA measurements among techniques when compared to a uniform reference standard.

Methods

This study was a retrospective analysis of data collected from a cohort of subjects seen at a single center. All subjects were evaluated by a single provider (JF) for suspected CTS. The study protocol was approved by the Institutional Review Board of UPMC, and all participants provided informed consent. Participants were randomly selected from a database of patients who underwent ultrasound examination for suspected carpal tunnel. Demographic and clinical data were collected from electronic health records for all participants. All ultrasound examinations were performed using a GE Venue Go ultrasound system equipped with a 4 to 20 MHz linear array transducer (Figure 1). The median nerve was identified at the level of the carpal tunnel inlet, defined as the proximal margin of the flexor retinaculum.

Figure 1.

Median nerve image on ultrasound using transducer from using a GE Venue Go ultrasound system equipped with a 4 to 20 MHz linear array transducer.

Cross-sectional area measurements were obtained by 2 authors; a fellowship trained hand surgeon (JF), labeled as observer A, and an orthopedic surgery resident (CO), labeled as observer B. Three different methods were used:

1. A-B method: The major (a) and minor (b) axes of the nerve were measured, and the CSA was calculated using the formula π(ab) (Figure 2a).

2. Ellipse method: An ellipse was fit to the nerve boundaries, and the ultrasound software calculated the enclosed area (Figure 2b).

3. Trace method: The nerve boundary was manually traced, and the ultrasound software calculated the enclosed area (Figure 2c).

Figure 2.

Ultrasound methods for measuring median nerve cross-sectional area. (a) Ultrasound image demonstrating the A-B method for measuring median nerve cross-sectional area. The major (a) and minor (b) axes of the nerve are measured using calipers, and the CSA is calculated using the formula π(ab). (b) Ultrasound image demonstrating the Ellipse method for measuring median nerve cross-sectional area. An ellipse is fit to the nerve boundaries, and the ultrasound software calculates the enclosed area. (c) Ultrasound image demonstrating the Trace method for measuring median nerve cross-sectional area. The nerve boundary is manually traced following the inner hyperechoic rim of the epineurium, and the ultrasound software calculates the enclosed area.

For each method, measurements were taken twice by each provider: once using the standard ultrasound view and once using a zoom-enhanced view. The order of standard and zoom-enhanced measurements was also randomized to mitigate potential bias.

The measurement for each sample was made on the clearest image showing optimal nerve visualization, as would be selected in routine clinical ultrasound assessment. To compare the observers’ measurements to a standard reference measurement, the CSA of the median nerve was also determined using ImageJ software (National Institutes of Health, Bethesda, MD, USA). This was done by first calibrating the software to the image dimensions to convert pixels to millimeters. The user then manually traced the inner border of the nerve, following the inner hyperechoic rim of the epineurium, allowing the software to calculate the CSA enclosed within the tracing. Both evaluators were blinded to the results of the ImageJ analysis.

Statistical Analysis

Statistical analysis was conducted using Stata 15 (StataCorp. 2017, Stata Statistical Software: Release 15; StataCorp, College Station, Texas, USA) and Excel Version 16.0 (Microsoft, Redmond, WA). Descriptive statistics were calculated for demographic data and presented as mean ± standard deviation or as percentages where appropriate.

Interrater reliability was calculated using intraclass correlation coefficients (ICCs) between the 2 evaluators’ measurements for each method and viewing condition (standard vs zoom-enhanced). The standard error of measurement (SEM) was calculated as an estimate of the amount of error associated with the technique. The standard detectable difference (SDD) was also computed to determine the difference between 2 observations which can be considered statistically significant. ¹⁴ Agreement between different ultrasound methods and ImageJ were assessed using Bland-Altman plots.

Accuracy of various methods was determined by comparing each ultrasound measurement to the corresponding ImageJ measurement, expressed as a percentage. To compare the measurements across methods, a 1-way repeated measures analysis of variance (ANOVA) was conducted. For specific pairwise comparisons between the 3 ultrasound methods, paired t-tests were performed. Statistical significance was set at P < .05 for all analyses. Measurements were rounded to 2 decimal places, consistent with the output precision of the ultrasound equipment used.

Results

A total of 10 participants were included in this study. The demographic and clinical characteristics of the study population are presented in Table 1. The mean age was 57.2 ± 17.0. 3 (30.0%) participants were male, and the remaining 7 (70.0%) were female.

Table 1.

Demographic and Clinical Characteristics of Study Participants.

Characteristic	Value
Age (years), mean ± SD	57.2 ± 17
Gender, N (%)
Male	3 (30%)
Female	6 (60%)
Transgender female	1 (10%)
Body mass index (kg/m2), mean ± SD	33.73 ± 11.35
Diabetes, N (%)	1 (10%)
Smoker, N (%)	3 (30%)
Hand dominance, N (%)
Right	8 (80%)
Left	1 (10%)
Unknown	1 (10%)
Affected hand (right), N (%)	7 (70%)

The mean CSA measurements for each method were: ImageJ (12.15 ± 5.13 mm²), A-B method (10.71 ± 4.17 mm²), Ellipse method (10.19 ± 4.41 mm²), and Trace method (10.89 ± 4.59 mm²). There were statistically significant differences between ImageJ and the Ellipse method (P = .02) and between ImageJ and the A-B method (P = .05). However, differences between ImageJ and the Trace method did not reach statistical significance (P = .08). Pairwise comparisons between the 3 ultrasound methods showed no statistically significant differences (A-B vs Ellipse: P = .10; A-B vs Trace: P = .62; Ellipse vs Trace: P = .05).

On Bland-Altman plots, the mean differences and 95% limits of agreement for each method were as follows: A-B method (mean difference: −1.46 mm², 95% limits: −4.76 to 1.87 mm²) (Figure 3), Ellipse method (mean difference: −1.96 mm², 95% limits: −5.09 to 1.17 mm²) (Figure 4), and Trace method (mean difference: −1.26 mm², 95% limits: −4.11 to 1.59 mm²) (Figure 5).

Figure 3.

Bland-Altman plot comparing the A-B method with ImageJ measurements of median nerve cross-sectional area. The solid line represents the mean difference (−1.46 mm²), and the dashed lines represent the 95% limits of agreement (−4.76 to 1.87 mm²).

Figure 4.

Bland-Altman plot comparing the Ellipse method with ImageJ measurements of median nerve cross-sectional area. The solid line represents the mean difference (−1.96 mm²), and the dashed lines represent the 95% limits of agreement (−5.09 to 1.17 mm²).

Figure 5.

Bland-Altman plot comparing the Trace method with ImageJ measurements of median nerve cross-sectional area. The solid line represents the mean difference (−1.26 mm²), and the dashed lines represent the 95% limits of agreement (−4.11 to 1.59 mm²).

Interrater reliability was assessed for all methods using ICC. The results are presented in Table 2. Interrater reliability was significantly higher in the zoom condition for all methods (P < .05). The Trace method showed the highest interrater reliability in the zoom condition (ICC = 0.99), followed by the A-B (ICC = 0.98) and Ellipse (ICC = 0.94) methods. In the no-zoom condition, the Trace method also demonstrated the highest interrater reliability (ICC = 0.97), followed by A-B (ICC = 0.95) and Ellipse (ICC = 0.92) methods.

Table 2.

Intraobserver Reliability and Overall Accuracy of Ultrasound Measurement Methods for Median Nerve Cross-Sectional Area.

Method	Inter-observer reliability*		Overall accuracy (%)
	Zoom	No zoom	Zoom	No zoom
A-B	0.983 (0.934-0.996)	0.953 (0.823-0.988)	82.1	83.6
Ellipse	0.940 (0.778-0.985)	0.956 (0.833-0.989)	81.0	82.4
Trace	0.992 (0.967-0.998)	0.977 (0.909-0.994)	84.2	85.5

^*95% Confidence Intervals in brackets.

The SEM and SDD were calculated for each method. For the A-B method, the SEM was 0.74 mm² and the SDD was 2.06 mm². This indicates that a change in measurements greater than 2.06 mm² can be considered a real change with 95% confidence. The Ellipse method showed an SEM of 0.97 mm² and an SDD of 2.69 mm². The Trace method demonstrated the lowest measurement error with an SEM of 0.57 mm² and an SDD of 1.57 mm².

Accuracy was calculated as the percentage agreement with ImageJ measurements (Table 2). All methods showed slightly higher accuracy in the no-zoom condition. The Trace method demonstrated the highest accuracy both with zoom (84.2%) and without zoom (85.5%), followed by the A-B method (zoom: 83.0%, no-zoom: 83.2%) and the Ellipse method (zoom: 78.9%, no-zoom: 80.1%).

Discussion

Our findings demonstrate that while all 3 methods show high reliability, the Trace method consistently demonstrated slightly better performance in terms of both the accuracy and reliability when compared to the A-B and Ellipse methods.

All 3 methods consistently underestimated CSA compared to ImageJ measurements. This systematic underestimation may be attributed to the inherent limitations of 2-dimensional ultrasound in capturing the complex 3-dimensional structure of the median nerve. Bland-Altman analysis revealed that the Trace method showed the smallest mean difference (−1.26 mm²) and narrowest limits of agreement (−4.11 to 1.59 mm²) compared to ImageJ, followed by the A-B method (−1.46 mm², −4.76 to 1.87 mm²) and the Ellipse method (−1.96 mm², −5.09 to 1.17 mm²). These results suggest that while all ultrasound methods tend to produce lower measurements than ImageJ, the difference is most pronounced and consistent for the Ellipse method, less so for the A-B method, and least pronounced for the Trace method.

The high interrater reliability observed across all methods (ICC ranging from 0.92 to 0.99) aligns with previous studies that have reported good repeatability of ultrasound measurements of the median nerve.^16-18 Our study extends these observations by directly comparing 3 different measurement techniques against a reference standard using ImageJ tracing methodology, providing valuable insights into their relative performance.

Our finding that the Trace method demonstrated marginally superior accuracy may be attributed to its ability to precisely follow the actual contours of the nerve. Unlike the A-B and Ellipse methods, which rely on geometric approximations that assume an elliptical shape, the Trace method allows the operator to follow the exact perimeter of the nerve, capturing irregular contours that deviate from a perfect ellipse. This advantage is particularly relevant for the median nerve, which often exhibits shape irregularities at the carpal tunnel inlet in patients with CTS.

Our study revealed interesting findings regarding the use of zoom features during measurements. While all methods showed significantly higher interrater reliability in the zoom condition compared to the no-zoom condition, measurements across all methods were slightly more accurate in the no-zoom condition. This may be explained by considering that zooming might improve the visibility of nerve boundaries, leading to more consistent measurements between different operators (higher reliability), but might also introduce subtle distortions or emphasize artifacts that reduce overall accuracy compared to ImageJ measurements. Similar findings were reported by Jelsing et al, ¹⁹ who found that zoomed images tended to yield larger CSA measurements for the median nerve.

All methods showed good agreement with the gold standard ImageJ measurements, but the Trace method consistently demonstrated the highest accuracy both with zoom (84.2%) and without zoom (85.5%). Our findings on the median nerve align with similar studies on other peripheral nerves. Alshami et al ¹⁴ reported excellent intrarater and interrater reliability (ICC ≥ 0.86) for tibial nerve CSA measurements at the tarsal tunnel, comparable to our results for the median nerve. They similarly found that the direct tracing method was slightly more reliable than the ellipsoid formula method. This finding is likely because the Trace method used in the ultrasound system to measure CSA closely resembles the approach ImageJ uses to calculate the median nerve’s CSA, as both methods employ manual tracing of the nerve boundary rather than geometric approximations. In addition, the superior accuracy and reliability of the Trace method may be attributed to the straightforward nature of the tracing process with modern ultrasound equipment versus the more precise placement requirements of the A-B and Ellipse methods, where operators must carefully position measurement points or elliptical shapes. The complete control afforded by the Trace method allows operators to follow the exact nerve contours without needing repeated adjustments to fit predetermined shapes to irregular nerve boundaries.

It is important to acknowledge that all 3 methods for CSA calculation have their own strengths and limitations. The A-B method is simple to perform, requires minimal training, and is less affected by image quality. However, it assumes the nerve is perfectly elliptical, which is not always the case, potentially underestimating CSA for irregular shapes. The Ellipse method takes slight irregularities into consideration but still assumes a generally elliptical shape and requires slightly more skill to place the best-fit ellipse optimally. The Trace method accounts for irregular shapes but requires care and may be affected slightly by image quality. Equally important is to focus on providing timely training, as research has demonstrated that ultrasound measurement of median nerve CSA can be efficiently taught in just a 5-minute session, with novice operators achieving results consistent with experienced examiners. ²⁰ It is important to note that while statistically significant in some cases, the absolute differences between measurement methods (approximately 0.5-1.5 mm²) represent only 5% to 15% of the typical diagnostic threshold for CTS (10-12 mm²), suggesting these variations may not substantially alter clinical decision-making in most cases but could be relevant when monitoring subtle changes during longitudinal assessment.

Our study has several limitations. First, our sample size of 10 patients is relatively small, which may limit the generalizability of our findings. Future studies with larger and more diverse populations would provide more robust results. Second, our use of only 2 evaluators (1 experienced surgeon and 1 resident) may not fully capture the performance characteristics of these methods across operators with varying levels of experience. Third, while we used ImageJ as a reference standard, there is no true gold standard for nerve CSA measurement, and these measurements may themselves contain errors or biases. Fourth, the export resolution and image quality can influence ImageJ measurements, which is a limitation of using it as a reference standard. Fifth, it is important to emphasize that the absolute differences in measurements between methods (approximately 0.5-1.5 mm²) are unlikely to significantly alter clinical decision-making in most cases. Finally, we did not evaluate the impact of any demographic variables on measurement accuracy, which could be important considerations for future research.

Conclusion

All 3 ultrasound methods demonstrated good interrater reliability in measuring median nerve CSA, with the Trace method showing modest advantages in precision and accuracy. While zoom-enhanced views improved reliability, they slightly reduced measurement accuracy across all methods. Based on our findings, we recommend the nonzoom Trace method for consistent longitudinal measurements, though maintaining proficiency in a single chosen method is more important than which specific technique is selected. The superior performance of the Trace method likely reflects its methodological similarity to the ImageJ reference standard. Our small sample size necessitates further validation with larger studies exploring factors such as body mass index, gender, and operator experience on measurement accuracy.