This study was approved by the Institutional Review Board of the First Hospital of Jilin University. Written informed consent to participate in this study was obtained from all participants prior to enrollment. The study was conducted in accordance with the Declaration of Helsinki.

Between May 2024 and May 2025, we retrospectively identified 32 patients with carpal tunnel syndrome (CTS) from the Department of Hand and Foot Surgery, Lequn Campus, First Hospital of Jilin University. In total, 47 wrists were included (8 wrists from 5 men and 39 wrists from 27 women; age, 42–74 years). Fifteen patients had bilateral CTS and 17 had unilateral CTS. CTS was diagnosed based on typical clinical symptoms and electrophysiological examinations. Clinical manifestations included unilateral or bilateral numbness in the median nerve distribution (thumb to the radial half of the ring finger), weakness, wrist discomfort, and nocturnal paresthesia. Physical examination demonstrated positive Tinel's sign and Phalen's test on the affected side; thenar atrophy was noted in advanced cases. All CTS patients underwent electrophysiological examinations.

The control group comprised 94 volunteers recruited from individuals undergoing routine physical examinations at the Ultrasound Clinic, Lequn Campus, First Hospital of Jilin University during the same period. One wrist was randomly selected from each participant (20 wrists from 20 men and 74 wrists from 74 women; age, 39–75 years). Controls had no symptoms suggestive of CTS.

Exclusion criteria for both groups were: (1) age <18 years; (2) history of wrist trauma or surgery; (3) anatomical variants, including a persistent median artery or bifid median nerve; (4) systemic diseases (e.g., diabetes mellitus, gout, acromegaly, rheumatoid arthritis, or hyperparathyroidism); (5) history or clinical evidence of polyneuropathy or hereditary neuropathy with liability to pressure palsies; (6) other neurological disorders such as cervical radiculopathy or brachial plexus injury; and (7) incomplete clinical data or inability to cooperate with the examination procedures.

All ultrasound examinations were performed using a Canon Aplio i800 system (Canon Medical Systems, Japan) with i24LX8 and i18LX5 linear-array transducers, which support Superb Microvascular Imaging (SMI) and Shear Wave Elastography (SWE). Musculoskeletal/nerve presets were used, and imaging settings were adjusted to optimize visualization of the median nerve and maintained as consistently as possible across participants.

Participants were seated comfortably with the forearm resting on the examination table, the palm facing upward, fingers slightly flexed, and the wrist in a relaxed neutral position. The transducer was first aligned with the longitudinal axis of the median nerve to identify the nerve course and key landmarks; cross-sectional images were then obtained by rotating the probe to the transverse plane. Throughout scanning, the probe was kept perpendicular to the skin to minimize obliquity, and excessive pressure was avoided to prevent artificial deformation of the nerve. Adequate coupling gel was applied to ensure optimal acoustic contact.

The sonographer who performed the ultrasound examinations (10 years of experience in musculoskeletal ultrasound) was blinded to the electrophysiological results at the time of image acquisition. Measurements were performed using a standardized protocol, and quantitative parameters were recorded at the time of examination.

The median nerve was systematically scanned from the distal arm to the wrist to evaluate morphology, course, and surrounding tissue abnormalities. On transverse images, the median nerve cross-sectional area (CSA) was measured at four predefined levels: the pisiform level (carpal tunnel inlet), the hook of hamate level (carpal tunnel outlet), the distal wrist crease, and the pronator quadratus level (distal forearm). CSA was obtained by manual tracing along the inner margin of the hyperechoic epineurial rim (Figure 1a). At each level, the flattening ratio (FR) was calculated as the transverse diameter divided by the anteroposterior diameter. ΔCSA was defined as the difference in CSA between the pisiform and pronator quadratus levels. To improve delineation of the transverse carpal ligament (TCL), participants were asked to gently move the fingertips to help distinguish the TCL from the palmar superficial fascia; TCL thickness was measured directly superficial to the median nerve (Figure 1b). Each measurement was repeated three times by the same examiner, and the mean value was used for analysis. The BSA-based Z-score model was derived from CSA measured at the pisiform level.

A 24-MHz high-frequency linear transducer was gently placed on the wrist with sufficient coupling gel to avoid excessive compression. After obtaining routine transverse 2D images at the pisiform and hook of hamate levels, the probe was rotated by 90° to obtain a longitudinal (long-axis) view of the median nerve and SMI was activated. The region of interest (sampling frame) was centered on the median nerve, and the frame size, position, and flow velocity scale were set identically for all participants and kept constant during acquisition. Intraneural flow was quantified automatically as the pixel ratio, defined as the percentage of color pixels within the sampling frame relative to the total pixel count. Care was taken to avoid contamination from adjacent non-neural vascular signals (Figure 1c). Measurements were repeated twice, and the mean value was used for analysis.

Shear wave elastography was performed using an 18-MHz linear transducer with minimal probe pressure. The SWE display scale was set to 0.5–12 m/s. Images were acquired only when the shear wave propagation lines were smooth and parallel. A 1-mm circular region of interest (ROI) was placed within the median nerve, avoiding the hyperechoic epineurium and adjacent structures. For each acquisition, five consecutive measurements with stable propagation and without obvious artifacts were recorded (Figure 1d), and their mean value was calculated. The acquisition was repeated twice, and the average of the two mean values was used for analysis. All ultrasound settings were kept consistent, and all examinations were performed by the same experienced operator.

Demographic and anthropometric variables were collected, including sex, age, height, weight, wrist circumference, body mass index (BMI), and body surface area (BSA). Height and weight were measured using standardized equipment with participants standing upright barefoot and without heavy clothing. BMI was calculated as weight (kg)/height² (m²). BSA was calculated using the Du Bois formula. Wrist circumference was measured circumferentially at the distal wrist crease using a flexible tape measure with the wrist in a relaxed neutral position. All measurements were standardized and obtained by a single examiner to ensure data accuracy and consistency.

Individualized Z-scores adjusted by BSA were calculated using the formula: Z = (X−µ)/σ = (measured CSA−BSA-predicted CSA)/standard deviation. X represented the mean of three actual CSA measurements, µ represented the predicted CSA based on BSA, and σ was the standard deviation. Thus, the formula can be expressed as: Z-score × σ = measured CSA−predicted CSA, enabling calculation of individualized upper limits of normal (ULN) CSA values based on BSA.

All statistical analyses were performed using SPSS Statistics 29.0 (IBM, Armonk, NY, USA). Normality was assessed using the Shapiro–Wilk test. Continuous variables were compared between groups using the independent-samples t-test for normally distributed data or the Mann–Whitney U test for non-normally distributed data. Categorical variables were compared using the chi-square test. Normally distributed data are presented as mean ± standard deviation (SD), and non-normally distributed data as median [interquartile range (IQR)]. Spearman's rank correlation was used for correlation analyses when variables were not normally distributed. McNemar's test was used for paired categorical comparisons when applicable. Linear regression analysis was performed to establish a BSA-based prediction equation for median nerve CSA at the pisiform level and to derive individualized Z-scores. Receiver operating characteristic (ROC) analyses were conducted to evaluate the diagnostic performance of 2D ultrasound parameters, SMI, SWE, the BSA-based Z-score, and the combined SMI + SWE logistic regression model. Areas under the ROC curve (AUCs), optimal cutoff values (determined by the Youden index), sensitivity, and specificity were calculated. AUCs were reported with 95% confidence intervals (CIs) from ROC analyses. In addition, 95% CIs for sensitivity and specificity at the selected cutoffs were calculated using exact binomial (Clopper–Pearson) methods. Logistic regression analysis was performed with CTS status as the dependent variable and SMI and SWE parameters as independent variables. In addition, a sensitivity analysis was performed by further adjusting the multivariable logistic regression model for age and sex to assess the robustness of the associations. A two-sided P value < 0.05 was considered statistically significant.

A total of 126 participants (141 wrists) were included in the analysis. The CTS group comprised 32 patients (47 wrists; 8 wrists from 5 men and 39 wrists from 27 women), and the control group comprised 94 healthy subjects (94 wrists; 20 wrists from 20 men and 74 wrists from 74 women). The mean age was 53.98 ± 8.63 years in the CTS group and 53.52 ± 8.28 years in the control group (P = 0.86). There were no significant between-group differences in sex distribution, body weight, or body surface area (BSA) (all P > 0.05). In contrast, the CTS group had a lower height (158.85 ± 9.92 vs. 163.11 ± 9.68 cm, P = 0.02), a higher BMI (27.00 ± 5.16 vs. 24.95 ± 3.94 kg/m², P = 0.04), and a larger wrist circumference (16.58 ± 1.37 vs. 16.02 ± 1.45 cm, P = 0.03) (Table 1).

Two-dimensional (2D) ultrasound parameters, Superb Microvascular Imaging pixel ratio (SMI PR), and Shear Wave Elastography (SWE)–derived metrics are summarized in Table 2. Compared with controls, the CTS group showed significantly higher CSA at the pisiform (CSA1), hook of hamate (CSA2), and distal wrist crease (CSA3) levels, a higher flattening ratio at the hook of hamate level (FR2), a larger ΔCSA, and increased transverse carpal ligament (TCL) thickness (all P < 0.05). In addition, SMI PR and SWE parameters [elastic modulus [E] and shear wave velocity [SWV]] were significantly higher in the CTS group than in the control group (all P < 0.05). No significant between-group differences were observed for FR1, FR3, CSA4, or FR4 (all P > 0.05).

Receiver operating characteristic (ROC) analyses were performed to evaluate the ability of each ultrasound parameter to differentiate CTS wrists from control wrists (Figure 2). Table 3 summarizes the area under the ROC curve (AUC), optimal cutoff values, sensitivity, and specificity for each parameter. For each parameter, AUCs as well as sensitivities and specificities are reported with 95% confidence intervals (Table 3). TCL thickness yielded the highest AUC (0.945); at an optimal cutoff of 0.80 mm, sensitivity and specificity were 87% and 86%, respectively. CSA at the pisiform level (CSA1) showed the second-highest AUC (0.917), with 94% sensitivity and 75% specificity at a cutoff of 0.11 cm². ΔCSA demonstrated a similar AUC (0.916), with both sensitivity and specificity of 85% at a cutoff of 0.05 cm². SMI PR achieved an AUC of 0.893, corresponding to 79% sensitivity and 92% specificity at a cutoff of 2.05. SWE-derived elastic modulus (E) had an AUC of 0.887 (sensitivity 91%, specificity 75%) with a cutoff of 94.55 kPa, and shear wave velocity (SWV) showed a comparable AUC of 0.886 (sensitivity 87%, specificity 79%) with a cutoff of 5.64 m/s. CSA3 yielded an AUC of 0.839 (sensitivity 66%, specificity 86%) with a cutoff of 0.12 cm², and FR2 yielded an AUC of 0.735 (sensitivity 70%, specificity 72%) with a cutoff of 3.97. CSA2 showed the lowest AUC (0.664), with 57% sensitivity and 76% specificity at a cutoff of 0.10 cm².

A significant positive correlation was observed between body surface area (BSA) and CSA at the pisiform level (CSA1) (R = 0.230, P = 0.026). Linear regression analysis yielded a BSA-based prediction equation: CSA1 = 0.022 × BSA + 0.057 (Figure 3). This equation was used to derive individualized expected CSA1 values based on BSA.

Individualized Z-scores were calculated as: Z = (X−μ)/σ, where X denotes the mean of three repeated CSA1 measurements, μ is the BSA-predicted CSA1 from the regression equation, and σ is the standard deviation used for standardization. ROC analysis of the Z-score (Figure 4) demonstrated an AUC of 0.924 (95% CI 0.879–0.969), with an optimal cutoff of 0.7023, yielding 92% sensitivity (95% CI 0.796–0.976) and 83% specificity (95% CI 0.738–0.899).

To assess the clinical utility of this approach, McNemar's test was used to compare the classification performance of the BSA-based Z-score with the conventional fixed CSA1 threshold of 0.11 cm². A significant difference was observed (P = 0.039). Compared with the fixed cutoff, the Z-score approach improved specificity (83% vs. 75%) while achieving comparable sensitivity (92% vs. 94%). These findings suggest that BSA-based individualized Z-scores may provide a more tailored diagnostic criterion for CTS, particularly by improving specificity.

The diagnostic utility of Superb Microvascular Imaging pixel ratio (SMI PR) and shear wave elastography (SWE)–derived elastic modulus (E) was further evaluated using logistic regression analysis. CTS status was entered as the dependent variable, with SMI PR and E as independent predictors. Based on the estimated regression coefficients, a predictive model was established. Both SMI PR (β = 1.302) and E (β = 0.069) were independently associated with CTS (both P < 0.001; Table 4).

In the primary multivariable model, SMI PR (OR 3.678, P < 0.001) and elastic modulus (OR 1.072, P < 0.001) were independently associated with CTS. After additional adjustment for age and sex, SMI (OR 3.793, 95% CI 2.186–6.583; P < 0.001) and elastic modulus (OR 1.074, 95% CI 1.038–1.111; P < 0.001) remained significant, whereas age (OR 1.020, 95% CI 0.945–1.100; P = 0.620) and sex (male vs. female; OR 1.616, 95% CI 0.435–5.996; P = 0.473) were not statistically significant. The effect estimates changed minimally after adjustment (SMI: 3.678 vs. 3.793; elastic modulus: 1.072 vs. 1.074), indicating robustness to demographic confounding (Supplementary Table S1).

ROC analysis of the combined logistic regression model (Figure 5) showed an AUC of 0.944 (95% CI 0.906–0.982). At the optimal cutoff determined by the maximum Youden index, the model achieved a sensitivity of 83% (95% CI 0.692–0.924) and a specificity of 90% (95% CI 0.826–0.955), indicating high diagnostic performance for differentiating CTS wrists from controls.

These results suggest that integrating SMI PR and SWE-derived elastic modulus provides improved diagnostic accuracy compared with single-parameter assessment and may serve as a useful adjunct to conventional ultrasound evaluation for CTS.