There was no statistically significant difference in age among the TD group, AIS A/B group, and AIS C/D group (p = 0.902). There was no statistically significant difference in the average time since injury between the AIS A/B group and the AIS C/D group (months; U = 96.5, P = 0.536).

The CSA, APW, and LRW in the AIS A/B group were significantly lower than those in the TD group (CSA: p = 0.002; APW: p = 0.001; LRW: p = 0.029), while CSA and LRW were significantly smaller than AIS C/D group (CSA: p = 0.014; LRW: p = 0.026). However, no statistical differences were found between the AIS C/D group and the TD group (Table 2, Fig. 2). For children with TLSCI, the CSA was positively correlated with PPSS (r = 0.430, p = 0.018). APW (Fig. 3)was positively correlated with PPSS (r = 0.470, p = 0.009) and LTSS (r = 0.382, p = 0.037). Neither sensory nor motor scores were associated with LRW.

The AIS A/B group was significantly lower than those in the TD group (CSA: p = 0.002; APW: p = 0.001; LRW: p = 0.029), while CSA and LRW were significantly smaller than AIS C/D (CSA: p = 0.014; LRW: p = 0.026). ANOVA One-way Analysis of variance, CSA cross-sectional area, APW anterior-posterior width, LRW left-right width, TD typically developing, AIS American Spinal Injury Association Impairment Scale. *p ≤ 0.05; **p ≤ 0.01.

CSA and PPSS were positively correlated (r = 0.430, p = 0.018). APW was positively correlated with PPSS (r = 0.470, p = 0.009) and LTSS (r = 0.382, p = 0.037). CSA cross-sectional area, APW anterior-posterior width, LRW left-right width, LTSS light touch sensory scores, PPSS pinprick sensory scores, ISNCSCI International Standard for the Neurological Classification of Spinal Cord Injury.

In children with TLSCI, the results of ROC analysis for morphologic spinal cord parameters in AIS A/B and AIS C/D groups are shown in Table 3. Figure 4 shows the sensitivity and specificity of LRW, with a sensitivity of 92.3%, a specificity of 70.6%, and an area under the curve (AUC) of 0.77 (95% CI, 0.594–0.952).

The area under the curve (AUC) for the ROC was 0.774 with a sensitivity of 92.3% and a specificity of 70.6%. LRW left-right width, AUC area under the curve, ROC receiver operator characteristic, AIS American Spinal Injury Association Impairment Scale.

The impact of age at inclusion, time since injury (months), and clinical grading at inclusion on CSA, APW, and LRW were investigated using multiple linear regression. Referring to the previous article, scoring of the extent of the clinical grading was done based on the AIS grades-A five points system (4–0; AIS A to AIS E in order of decreasing sensorimotor deficit) [22, 23].

Multiple linear regression analysis was performed on CSA, and the multiple linear regression model was statistically significant (F = 10.812, P < 0.001, R2 = 0.555). The regression equation: CSA = 56.473 mm2 + 0.682 (age) − 0.157 (time since injury) − 2.147 (clinical grading). Where age (β = 0.299, p = 0.038) significantly and positively influenced CSA. Time since injury (β = −0.525, p = 0.001) and AIS grade (β = −0.549, p < 0.001) significantly and negatively influenced CSA.

When multiple linear regression analysis was performed on APW, it was found that the model was statistically significant (F = 9.068, P < 0.001, R2 = 0.511). The regression equation: APW = 6.717 mm + 0.051 (age) − 0.014 (time since injury) − 0.127 (clinical grading). Where age (β = 0.30, p < 0.047) significantly and positively affected APW, time since injury (β = −0.014, p < 0.001), and clinical grading (β = −0.431, p = 0.004) significantly and negatively affected APW.

Multiple linear regression analysis was performed on LRW, and the multiple linear regression model was statistically significant (F = 3.253, P = 0.038, R2 = 0.273). The regression equation: LRW = 10.624 mm + 0.043 (age) − 0.007 (time since injury) − 0.238 (clinical grading). Where only clinical grading (β = −0.489, p = 0.007) significantly and negatively affected LRW.