Age-related patterns and strong correlation between cranial and ear asymmetry in healthy children: a CT-based cross-sectional study

Abstract

Purpose

Severe cranial deformity has been thought to cause secondary ear misalignment; however, the developmental patterns of cranial and ear asymmetry have not been systematically characterized. We aimed to clarify age-related variations and the interrelationship of cranial and ear asymmetry in healthy Japanese children using head computed tomography (CT).

Methods

We retrospectively analyzed CT data from children imaged for minor head trauma between 2006 and 2023. Cranial asymmetry was assessed using the cranial vault asymmetry index (CVAI), and ear asymmetry was quantified by the absolute ear shift angle (ESA) derived from the alignment of the external auditory canals. Both indices were categorized into five grades, with grade ≥ 3 considered moderate to severe. Age-related trends were evaluated using regression and random-forest models with partial dependence plots (PDPs).

Results

A total of 235 children were included (138 males and 97 females; median age, 12 months). CVAI and absolute ESA showed a significant positive correlation (r = 0.58, p < 0.001). The proportion of grade ≥ 3 deformity declined with age: CVAI peaked at 50% at 4–5 months and decreased to 29% after 5 years, whereas ESA peaked at 56% at 1 month and decreased to 24% after 5 years. PDPs indicated that absolute ESA decreased mainly within the first 6 months, while CVAI showed a more marked decline after 3 years.

Conclusion

Both CVAI and ESA decreased in severity with age and showed a strong positive correlation, suggesting coordinated cranial and ear morphological development. The temporal patterns of improvement differed between the two measures.

Supplementary Information

The online version contains supplementary material available at 10.1007/s00381-026-07160-9.

Introduction

Since prone sleeping positions have been discouraged to prevent sudden infant death syndrome, the prevalence of deformational plagiocephaly has increased [1–3]. The corrective effect of helmet therapy for deformational plagiocephaly has been well demonstrated [4, 5], with earlier interventions generally yielding better outcomes [6–8]. The cranial vault asymmetry index (CVAI) is widely used to evaluate deformational plagiocephaly [9], and previous reports have indicated that CVAI may improve spontaneously with growth [9–12]. These findings have led to a clinical dilemma: while early intervention is necessary to maximize the corrective effect of helmet therapy, waiting for spontaneous improvement risks missing the optimal treatment window.

Severe positional plagiocephaly can cause ear misalignment [13, 14] and facial asymmetry [15]. Although helmet therapy can improve both ear and facial asymmetry [4, 15, 16], its corrective effect on ear shift is smaller than that on cranial asymmetry [8, 17] and is primarily limited to children younger than one year [15]. In contrast to CVAI, the natural developmental course of ear asymmetry has not been systematically characterized, and it remains unclear whether ear misalignment requires correction or merely reflects normal developmental variation.

To address these uncertainties, we aimed to clarify the growth patterns and interrelationship between CVAI and ear shift angle (ESA). We retrospectively analyzed head computed tomography (CT) images from otherwise healthy Japanese children who underwent imaging for minor head trauma. We hypothesized that, because ear asymmetry is generally considered a secondary change arising from severe cranial deformity, the temporal peak of ESA would coincide with or follow that of cranial asymmetry (CVAI).

Methods

Patients and study setting

We retrospectively analyzed head CT data of children evaluated for minor head trauma at Osaka Women’s and Children’s Hospital between March 2006 and May 2023. Exclusion criteria included suspected abuse, severe trauma requiring surgery, ventricular compression, and comorbidities (e.g., craniosynostosis, tumor, epilepsy, autism, intracranial arachnoid cyst, chromosomal abnormalities, cardiovascular disease, or endocrine disorders). Thus, only otherwise healthy children without surgical intervention were included. Part of the study cohort overlapped with the previous reports [18–20].

Data collection

We recorded age (in months) at the time of CT and sex for each participant. On axial CT slices including the frontal horns, bodies, and occipital horns of the lateral ventricles, head circumference (HC), length, width, and right/left diagonals were measured, with the origin defined as the intersection of the length and width. Right and left diagonals were defined at ± 30^◦ from the length axis (Fig. 1a). Cranial asymmetry was quantified by cephalic index (CI) and CVAI [9–11, 21, 22] (Fig. 1a).

Fig. 1.

Measurement methods, grading schemes, and age-related associations. (a) Measurement of cephalic index (CI), cranial vault asymmetry index (CVAI), and inter-external auditory canal angle (IEACA) on axial CT images. (b) Grading criteria for scaphocephaly, brachycephaly, plagiocephaly, and ear shift angle (ESA). (c) Cross-tabulation of CVAI and ESA grades with weighted kappa statistics. (d) Logarithmic regression curves of head circumference (HC), length, width, and right/left diagonals with age. (e) Correlation matrix among head-size-related measures. (f, g, h) Associations of age with CI, CVAI, and absolute ESA. Dashed lines in (f) indicate thresholds for brachycephaly (CI = 91%) and scaphocephaly (CI = 84%). (i) Quantile regression analyses for CVAI (upper) and absolute ESA (lower) across 25th, 50th, and 75th percentiles. Shaded areas represent 95% confidence intervals. An asterisk (*) indicates statistical significance (p < 0.05)

Ear asymmetry was assessed on axial CT slices in which the bilateral external auditory canal openings could be simultaneously identified; such slices were available for all cases. To quantitatively evaluate ear positional asymmetry on axial CT images, we employed an original CT-based assessment method. Measurements were performed on a single axial plane, and no correction or normalization was applied for potential vertical asymmetry of the external auditory canals. The inter-external auditory canal angle (IEACA) was defined as the angle between the line connecting the bilateral external auditory canal openings and the cranial midline, which was determined by the line connecting the nasion and inion (Fig. 1a). Because the external auditory canal openings represent bony landmarks, this approach allows assessment of ear positional asymmetry while minimizing the influence of soft-tissue deformation. The ESA was defined as IEACA − 90^◦. Information regarding torticollis or other conditions that could cause vertical misalignment of the external auditory canals was not available in the clinical records and could not be determined from the CT images. Accordingly, no adjustments were made to account for such factors, and all measurements were analyzed as obtained from the axial plane.

The flattened side was determined using two approaches. First, the diagonal-based method defined the flattened side as the side with the longer diagonal length (e.g., leftsided flattening results in a longer left diagonal compared with the right). Second, the ESA-based method defined the flattened side according to the sign of ESA, with negative values indicating right-sided flattening. Concordance between the flattened side determined by diagonal measurements and that determined by ESA was evaluated.

Brachycephaly and scaphocephaly were graded into five levels using CI, and plagiocephaly into five levels using CVAI [23]. Because cranial morphology differs across ethnic groups, with Asian populations generally exhibiting a greater degree of brachycephalization compared with European populations [24], the CVAI grading criteria were slightly modified for application to Japanese infants rather than directly adopting Western reference standards (Fig. 1b).

For ear asymmetry, an original five-grade classification based on absolute ESA was established by referencing CVAI grades and is hereafter referred to as the ESA grade: grade I, < 1.37^◦; grade II, 1.37–2.34^◦; grade III, 2.34–3.85^◦; grade IV, 3.85–4.63^◦; and grade V, ≥ 4.63^◦ (Fig. 1b).

Age clustering

Subjects were stratified into ten age groups: 0, 1, 2–3, 4–5, 6–7, 8–9, 10–15, and 16–23 months, and 2–4 and ≥ 5 years.

Statistical analysis

Continuous variables were summarized as mean ± standard deviation when normally distributed and as median with interquartile range (25th-75th percentile) when non-normally distributed. Group comparisons used t-test, Wilcoxon rank-sum test, or chi-square test, as appropriate. Agreement between CVAI and ESA grades was assessed by weighted kappa. Associations among cranial indices, ESA, and age were examined using linear or logarithmic regression, correlation coefficients, and quantile regression (quantreg function, by Aslak Grinsted). Significance was set at p < 0.05 with Bonferroni correction.

Dimensionality reduction and clustering

Nine variables (sex, age, HC, CI, flattened side, CVAI, CVAI grade, absolute ESA, and ESA grade) were z-scored; sex/flattened side were one-hot encoded. Principal component analysis (PCA) was performed, and k-means clustering (k = 3) was applied to PC1-PC2 scores.

Feature importance and partial dependence plots (PDPs)

Random-forest regression was used to predict CVAI and absolute ESA from sex, age, HC, and flattened side. Feature importance was expressed as normalized mean decrease in mean squared error (MSE). PDPs were generated from the fitted models to illustrate the effects of age or HC on CVAI and absolute ESA, with 95% confidence intervals estimated by bootstrap.

Age-varying effect of CVAI on absolute ESA

Partial R² was computed by comparing a full model (predictors: CVAI, HC, sex, and flattened side) with a reduced model (HC, sex, and flattened side) within each age stratum. Partial R² was defined as:

$$Partial{R}^2=1-\frac{{\mathit{SSE}}_{\mathit{full}}}{{\mathit{SSE}}_{\mathit{reduced}}}$$

where SSE_full and SSE_reduced denote the sums of squared errors for the full and reduced models, respectively. The trajectory of partial R² across age strata was plotted to illustrate growth-related changes. All analyses were performed using MATLAB (MathWorks, Natick, MA).

Results

Baseline characteristics

A total of 235 children were included (138 males, 58.7%; 97 females, 41.3%). The median age at CT was 12.0 months (interquartile range, 5.2–44.8), with no significant sex difference (p = 0.78). HC, width, and the left diagonal were significantly larger in males than in females (HC, p = 0.028; width, p = 0.009; left diagonal, p = 0.022). In contrast, there were no significant sex differences in CI, CI grade, CVAI, CVAI grade, IEACA, absolute ESA, or ESA grade (Table 1).

Table 1.

Baseline characteristics of the study population

	Total (n = 235)	Male (n = 138, 58.7%)	Female (n = 97, 41.3%)	p values
Age (months)	12.0 (5.2–44.8)	11.0 (6.0–55.0)	14.0 (4.8–39.0)	0.78
Head circumference (mm)	466.8 (429.6–501.9)	469.5 (434.2–510.5)	465.5 (422.6–490.4)	0.028*
Length (mm)	155.3 (141.4–169.8)	155.7 (143.1–171.5)	154.5 (139.5–167.3)	0.16
Width (mm)	135.7 (124.8–144.2)	139.1 (126.9–148.0)	133.9 (123.9–141.1)	0.009*
CI (%)	87.0 ± 5.8	87.2 ± 5.3	86.7 ± 6.5	0.49
CI grade
B5 (%)	2 (0.9)	1 (0.7)	1 (1.0)	0.21
B4 (%)	3 (1.3)	2 (1.4)	1 (1.0)
B3 (%)	7 (3.0)	4 (2.9)	3 (3.1)
B2 (%)	15 (6.4)	6 (4.3)	9 (9.3)
B1 (%)	27 (11.5)	15 (10.9)	12 (12.4)
Normal (%)	110 (46.8)	73 (52.9)	37 (38.1)
S1 (%)	34 (14.5)	20 (14.5)	14 (14.4)
S2 (%)	24 (10.2)	14 (10.1)	10 (10.3)
S3 (%)	10 (4.3)	2 (1.4)	8 (8.2)
S4 (%)	3 (1.3)	1 (0.7)	2 (2.1)
S5 (%)	0 (0)	0 (0)	0 (0)
Right-diagonal (mm)	151.9 (138.3–163.0)	152.7 (140.3–165.1)	151.3 (136.9–160.8)	0.12
Left-diagonal (mm)	148.6 (136.9–160.3)	150.7 (138.7–164.1)	147.6 (135.3–156.8)	0.022*
Flattened side
Right (%)	157 (66.8)	86 (62.3)	71 (73.2)	0.08
Left (%)	78 (33.2)	52 (37.7)	26 (26.8)
CVAI (%)	4.6 (1.6–7.0)	4.6 (1.7–6.8)	4.3 (1.6–7.2)	0.73
CVAI Grade
1 (%)	97 (41.3)	58 (42.0)	39 (40.2)	0.98
2 (%)	63 (26.8)	35 (25.4)	28 (28.9)
3 (%)	46 (19.6)	28 (20.3)	18 (18.6)
4 (%)	17 (7.2)	10 (7.2)	7 (7.2)
5 (%)	12 (5.1)	7 (5.1)	5 (5.2)
IEACA (°)	89.8 ± 2.4	90.0 ± 2.5	89.7 ± 2.3	0.33
Absolute ESA (°)	1.5 (0.8–2.6)	1.5 (0.7–2.6)	1.6 (1.1–2.6)	0.38
ESA Grade				0.44
1 (%)	97 (41.3)	63 (45.7)	34 (35.1)
2 (%)	63 (26.8)	33 (23.9)	30 (30.9)
3 (%)	46 (19.6)	24 (17.4)	22 (22.7)
4 (%)	17 (7.2)	10 (7.2)	7 (7.2)
5 (%)	12 5.1)	8 (5.8)	4 (4.1)

^*P < 0.05 was considered statistically significant. Continuous variables are presented as mean ± standard deviation or median (interquartile range), as appropriate. Continuous variables were compared using unpaired t-tests or Wilcoxon rank-sum tests, and categorical variables were compared using chi-square tests. CI cephalic index, CVAI cranial vault asymmetry index, ESA ear shift angle, IEACA inter-external auditory canal angle

Age-related changes in cranial and ear asymmetry

Because the ESA grading was newly defined in this study, its concordance with the CVAI grading was evaluated using weighted kappa statistics. The weighted agreement was moderate (weighted κ = 0.55, p < 0.001; Fig. 1c), supporting the validity of the ESA classification. Head-size measures (HC, length, width, and diagonals) showed excellent fits to logarithmic growth models (R² = 0.79 − 0.89; Fig. 1d) and strong positive intercorrelations (Fig. 1e).

CI shifted toward normal with age, showing a significant negative slope in the brachycephaly range (CI > 91%; p = 0.04) and a non-significant positive slope in the scaphocephaly range (CI < 84%) (Fig. 1f). Both CVAI and absolute ESA declined with age; the association was not statistically significant for CVAI (p = 0.06; Fig. 1g) or for absolute ESA (p = 0.10; Fig. 1h).

Quantile regression revealed that CVAI values decreased across all quantiles, with the steepest slope at the 50th percentile, indicating a greater reduction in moderate cases. Absolute ESA showed the largest decline at the 75th percentile, suggesting more pronounced reduction in severe cases (Fig. 1i).

Associations between cranial and ear asymmetry

CVAI correlated positively with absolute ESA (r = 0.58, p < 0.001; Fig. 2a), while ESA correlated negatively with the right-left diagonal difference (r =  − 0.82, p < 0.001; Fig. 2b), indicating a close link between cranial and ear asymmetry. The concordance between diagonal- and ear shift-based flattened sides was 80.4% (chi-square test, p < 0.001; Fig. 2c).

Fig. 2.

Associations between cranial and ear asymmetry. (a) Correlation between CVAI and absolute ESA. (b) Correlation between ESA (IEACA − 90.^◦) and right–left diagonal difference. (c) Concordance between diagonal-based and ear shift–based flattened sides. (d) Concordance rates across 10 age groups (mo, month; y, year). (e, f) Comparisons of CVAI and absolute ESA between right- and left-flattened groups defined by diagonals (e) or ear shift (f). (g, h) Comparisons of CVAI and absolute ESA between Match and Non-match groups (i.e., concordant and discordant classifications based on diagonal- and ear shift-derived flattened sides). An asterisk (*) indicates statistical significance (p < 0.05)

In age stratified clusters (0 months, n = 8; 1 month, n = 16; 2–3 months, n = 15; 4–5 months, n = 20; 6–7 months, n = 24; 8–9 months, n = 24; 10–15 months, n = 21; 16–23 months, n = 21; 2–4 years, n = 45; and ≥ 5 years, n = 41), concordance rates ranged from 76 to 88%, peaking at birth and remaining stable thereafter (Fig. 2d).

Between right- and left-flattened groups, CVA was significantly higher in the right-sided group (p = 0.039; Fig. 2e), whereas absolute ESA showed no difference (p = 0.70; Fig. 2f). When divided into match versus non-match groups, both CVAI and absolute ESA were significantly higher in match cases (p < 0.001; Fig. 2g, h), suggesting that non-match cases generally represented milder deformities.

Prevalence of higher asymmetry grades across age groups

For CVAI, the proportion of children with grade ≥ 3 (helmet therapy threshold at our hospital) peaked at 50% at the 4–5 month group and then declined to 29% at the ≥ 5-year group (Fig. 3a). For ESA grade, the proportion was highest in the 1 month group (56%) and decreased steadily to 24% in the ≥ 5-year group (Fig. 3b). Age-clustered analyses showed that CVAI was significantly lower in the 2–4 year and ≥ 5-year groups than in the 2–3 month and 10–15 month groups (p < 0.05; Fig. 3c). Absolute ESA was significantly lower in the 2–4 year and ≥ 5-year groups than in the 1-month group (p < 0.05; Fig. 3d). These differences, however, lost significance after Bonferroni correction (Supplemental Digital Content Table 1, 2).

Fig. 3.

Age-related prevalence and values of cranial and ear asymmetry. (a) Proportion of children with CVAI grades 1–5 across 10 age groups; the combined percentage of grade ≥ 3 (helmet therapy threshold) is also shown. (b) Proportion of children with ESA grades 1–5 across 10 age groups; the combined percentage of grade ≥ 3 is also shown. (c) CVAI values by age cluster. (d) Absolute ESA values by age cluster. Error bars indicate interquartile ranges (25th-75th percentile). An asterisk (*) indicates statistical significance at p < 0.05; however, the difference was not significant after Bonferroni correction

Cluster analysis and predictive modeling

PCA with k-means clustering (k = 3) identified three distinct clusters (Fig. 4a). Cluster 3 (65; 50–102 months) was significantly older than cluster 1(7; 4–13 months) and 2 (9; 4–23 months) (adjusted p < 0.001; Fig. 4b). Both CVAI and absolute ESA were significantly higher in cluster 2 compared with cluster 1 and 3 (adjusted p < 0.001; Fig. 4c, d).

Fig. 4.

Cluster analysis and predictive modeling of cranial and ear asymmetry. (a) Principal component analysis (PCA) with k-means clustering (k = 3) plotted in PC1–PC2 space. (b, c, d) Comparisons of age (b), CVAI (c), and absolute ESA (d) among clusters. Adjusted p-values (Bonferroni correction) are shown. (e) Feature importance for predicting CVAI and absolute ESA from random-forest regression. (f, g) Partial dependence plots (PDPs) of age for CVAI (f) and absolute ESA (g). Small ticks along the x-axis indicate the distribution of individual subjects (rug plots). The vertical dashed lines mark 6 and 36 months. (h, i) PDPs for head circumference for CVAI (h) and absolute ESA (i), with rug plots showing the distribution of subjects along the x-axis. The vertical dashed line indicates a head circumference of 470 mm. (j) Unique contribution of CVAI to predicting absolute ESA, with age-specific partial R.² (upper) and patient distribution (lower). An asterisk (*) indicates statistical significance (p < 0.05)

Random-forest regression indicated that HC was the most important predictor, followed by age, for both CVAI and absolute ESA (Fig. 4e). PDPs for age indicated that CVAI showed a slight decline before 6 months, followed by a gradual increase until 36 months, and then a marked decrease at 36 months (Fig. 4f). By contrast, absolute ESA exhibited a steep decline within the first 6 months, remained stable thereafter, and showed only a modest decrease at 36 months (Fig. 4g). PDPs for HC demonstrated a trough around 470 mm for both CVAI and absolute ESA (Fig. 4h, i). PDPs for age demonstrated significant negative correlations with both CVAI and absolute ESA (p = 0.014; Fig. 4f and p < 0.001; Fig. 4g). The association between CVAI and absolute ESA peaked at 36 months and then declined (Fig. 4j).

Discussion

We initially hypothesized that ear asymmetry would coincide with or follow cranial asymmetry, as ear asymmetry is generally considered a secondary change associated with severe cranial deformity. However, our results from a population-based cohort of neurologically normal Japanese children who underwent head CT for minor head trauma contradicted this expectation. Both CVAI and ESA decreased in severity with age. Notably, the peak of ear asymmetry preceded that of cranial asymmetry: ESA grade ≥ 3 was observed most frequently at 1 month of age, whereas CVAI peaked at 4–5 months. Moreover, PDP analyses revealed that ear asymmetry improved mainly within the first 6 months of life, while cranial asymmetry showed a more prominent decline after 3 years of age.

Feature importance analysis suggested that HC was the strongest predictor, followed by age, although these variables were highly correlated. To clarify their distinct roles, PDPs were generated. PDPs show the effect of varying one predictor on the model’s prediction, while averaging over the other variables [25]. If cranial and ear asymmetry simply accumulated with growth, without spontaneous improvement, both PDPs would be expected to show monotonic increases. Instead, CVAI exhibited a modest rise until 36 months, followed by a sharp decline, indicating that growth-related improvement becomes evident after three years. In contrast, absolute ESA showed a steep decline during the first 6 months, suggesting that growth-related improvement of ear asymmetry occurs primarily in infancy. Both HC-PDPs of CVAI and absolute ESA demonstrated troughs around 470 mm, corresponding to 18 months by the ageHC logarithmic regression. These troughs suggest a potential transitional range, in which increases in head circumference were associated with more pronounced improvement below this range, whereas the magnitude of improvement appeared to attenuate beyond it. Taken together, these findings suggest that ear asymmetry undergoes most of its growth-related improvement during early infancy, whereas cranial deformity shows a more gradual improvement with a delayed but pronounced phase after three years. When considered in terms of HC, 470 mm may represent a threshold for growth-related normalization.

Age-clustered analysis revealed that cranial asymmetry peaked at 4–5 months, consistent with previous reports indicating that cranial deformation worsens after birth [26, 27] and reaches its maximum at 4 months [12]. Absolute ESA was also greatest at 1 month but subsequently decreased with growth, reflecting growth-related improvement. Quantile regression further suggested that greater ear asymmetry was more likely to improve with age, as indicated by the steeper negative slope at the 75th percentile. PCA further supported this trajectory: children in the intermediate-age cluster showed higher CVAI and absolute ESA compared with both the youngest and oldest clusters, indicating that cranial and ear asymmetry worsen after birth, peak during infancy, and then improve thereafter. Importantly, the 4–5 month period appears to represent a physiologically vulnerable stage when cranial deformation is most pronounced. This highlights the importance of preventive strategies during this window, particularly repositioning therapy and physiotherapy [28], which are generally recommended before 6 months of age [8].

Traditionally, ear asymmetry has been regarded as a secondary consequence of cranial deformation [3, 8]. If this were the case, the peak of ear asymmetry would be expected to follow that of CVAI. However, we observed the opposite: ESA peaked as early as 1 month, earlier than the 4–5 month peak observed for CVAI. This finding suggests that ear asymmetry may arise, at least partly, independently of cranial deformation. We observed a significant correlation between CVAI and ESA in an otherwise healthy population with minor head trauma, with high concordance between diagonal- and ear shift-based assessments. In contrast, a previous study of helmettreated cohorts reported no such association [4]. Moreover, our partial R² analysis showed that the relationship between cranial and ear asymmetry remained stable from birth to 18 months, increased thereafter, and peaked at 36 months. This indicates that cranial deformation contributes most strongly to ear shift later in development, rather than during the typical therapeutic window of 6–12 months. Collectively, these results suggest that early ear asymmetry may be influenced by factors other than cranial deformation, but such factors remain unclear based on both our findings and prior literature, warranting further investigation.

Our additional findings were consistent with previous reports. Our cohort, which consisted of children who underwent CT for minor head trauma, also showed a predominance of male subjects, in line with previous studies [10, 11]. Growth curves of HC followed previously reported patterns [29]. Furthermore, cranial deformation is known to occur more frequently on the right side [1], and in our analysis, the diagonal-based flattened side was on the right in 66.8% of cases.

This study has several limitations that should be acknowledged. First, this study was based on a population-based cohort of neurologically normal Japanese children who underwent head CT for minor head trauma. As such, the present findings primarily reflect the natural course of mild cranial and ear asymmetry and do not directly address whether severe deformities would show a similar degree of spontaneous improvement. In particular, the extent to which markedly deformed cases—typically referred for early intervention—would follow the same trajectory remains uncertain. Second, because both the study population and the grading system were specific to Japanese children, the generalizability of these findings to other ethnic or racial populations has not been established. Third, information on helmet therapy was not available in this cohort. The first cranial remolding orthosis was formally approved as a medical device in Japan in 2018. Based on the timing of CT acquisition and subjects’ age, 97 cases in our cohort could have been potential candidates for helmet therapy. Although the overall impact is expected to be limited, it remains possible that some post-treatment cases were included. Fourth, because the cohort consisted of children undergoing CT for minor head trauma, there was a predominance of boys, although no sex-related differences were observed in CVAI or ESA. Fifth, although predictive modeling suggested growth-related improvement, these results should be interpreted as associative rather than causal. In addition, the cross-sectional design precludes direct longitudinal tracking of individual subjects, and the use of CT images obtained for trauma may not fully reflect normal development. Finally, the measurement of ear shift is not yet standardized. In this study, we proposed a quantitative metric and grading system for ear asymmetry (ESA) derived from head CT images, which enabled precise and reproducible measurements. However, this grading system was developed within a single population and imaging modality, and external validation in independent cohorts is warranted. In addition, although CT allowed precise quantification, it is invasive and not feasible for routine assessment, underscoring the need for future studies to evaluate the applicability of ESA-based assessment using noninvasive imaging modalities.

Conclusion

This study demonstrates that both cranial vault asymmetry and ear asymmetry undergo age-related improvement in a Japanese pediatric cohort, although their developmental trajectories differ. Cranial asymmetry showed the greatest improvement after 3 years of age, whereas ear asymmetry improved earlier than that, most prominently during the first 6 months of life. Together, these findings refine the current understanding of the natural course of cranial and ear asymmetry in early childhood.

Supplementary Information

Below is the link to the electronic supplementary material.

Abbreviations

CI

Cephalic index

CT

Computed tomography

CVAI

Cranial vault asymmetry index

ESA

Ear shift angle

HC

Head circumference

IEACA

Inter-external auditory canal angle

MSE

Mean squared error

PCA

Principal component analysis

PDP

Partial dependence plot

SSE

Sum of squared errors

Author contributions

H.H. conceived the study, collected and analyzed the data, developed the MATLAB program, created all figures and tables, and was primarily responsible for writing the manuscript. All authors provided clinical care and evaluated the patients. O.T. and Y.C. supervised the study. All authors have reviewed the manuscript.

Funding

Open Access funding provided by The University of Osaka. The Japan Society for the Promotion of Science (JSPS) KAKENHI [JP21K16629 (Hiroaki Hashimoto)] supported this work.

Data availability

The data used in this study are available from the corresponding authors upon reasonable request and after additional ethics approval.

Declarations

Ethical approval and consent to participate

Ethical approval was obtained from the Ethics Committee of Osaka Women’s and Children’s Hospital (Izumi, Japan, approval no. 1634), adhering to the Declaration of Helsinki. Informed consent was obtained through the opt-out method on our center’s website, considering the retrospective and noninvasive nature of the study.

Consent for publication

All authors reviewed the complete manuscript and consented to publication.

Clinical trial number

Not applicable.

Conflict interest

The authors declare no competing interests.