Abstract
Background
Ultrasonography offers several advantages in medical diagnostics, including the absence of radiation exposure, convenience, and real-time dynamic imaging. In this study, radiographic bone age assessment based on left-hand wrist X-rays was used as the reference standard to evaluate the feasibility and association with radiographic bone age of ultrasound for bone age estimation.
Methods
This study was designed as a prospective cross-sectional diagnostic feasibility study. Children who underwent radiographic bone age assessment using left-hand wrist X-rays at a large tertiary hospital were enrolled. 139 children (63 boys and 76 girls) aged 3–16 years were recruited. Within 7 days, ultrasound examinations were performed on the metacarpals and phalanges of the first, third, and fifth fingers, as well as the ulna, radius, and femur. The longitudinal diameters of the ossification centers and epiphyses were measured, and the ossification ratio [OsR = ossification center length/epiphyseal length (d/D)] and skeletal maturity score (SMS) were calculated for each bone.
Results
Ultrasound imaging clearly showed the ossification centers and epiphyseal cartilage. Radiographic bone age was correlated with the SMS of the metacarpals, phalanges, ulna, radius, and femur, with the strongest correlation observed between radiographic bone age and the combined skeletal maturity score (CMS) of the ulna, radius, and femur [boys: r=0.961, 95 % confidence interval (CI) 0.936–0.977; P<0.001; girls: r=0.962, 95 % CI: 0.940–0.976; P<0.001]. The URF (ulna + radius + femur) combination demonstrated the highest intraclass correlation coefficient (ICC) values and 95 % CIs in both the inter-observer and the intra-observer assessments, with an ICC of 0.99 and a 95 % CI of 0.98–1.00.
Conclusions
In this study, the CMS of the URF were strongly correlated with Tanner-Whitehouse 3 radius-ulna-short bone (TW3-RUS) radiographic bone age, suggesting that ultrasound has potential research value in pediatric bone age assessment. Because agreement and prediction-error analysis was not performed, feasibility and association with radiographic bone age of this examination method for pediatric bone age assessment requires further validation through subsequent large-sample agreement studies.
Keywords: Bone age, skeletal maturity score (SMS), ossification ratio (OsR), ultrasound
Highlight box.
Key findings
• In this study, the combined skeletal maturity score of the URF (ulna + radius + femur) were strongly correlated with Tanner-Whitehouse 3 radius ulna-short bone radiographic bone age.
What is known and what is new?
• Left‑wrist X‑ray is the bone‑age gold standard but involves ionizing radiation. Radiation‑free alternatives such as magnetic resonance imaging, dual-energy X-ray absorptiometry and earlier ultrasound approaches are limited by high cost, poor reliability or lack of standard protocols, without widespread clinical adoption.
• A streamlined standardized ultrasound protocol targeting the ulna, radius and femur.
What is the implication, and what should change now?
• This radiation‑free ultrasound method is a promising adjunct for bone‑age assessment, not yet a replacement for X‑ray. Multi‑centre large‑sample studies with agreement analysis and external validation are required before clinical implementation.
Introduction
Bone age, which is distinct from chronological age, serves as an indicator of skeletal and biological maturity and represents the degree of osseous maturation (1,2). Assessment of bone age serves as a valuable diagnostic tool for endocrinologists and pediatricians in evaluating growth and developmental disorders in children. This method not only facilitates the prediction of final adult height but also has important implications for monitoring children’s health development. Furthermore, bone age evaluation plays important roles in both athletic screening (3) and forensic age estimation (4). Common methods for assessing bone age in athletes include X-ray and magnetic resonance imaging (MRI). However, to obtain more reliable bone age information for athletes, some researchers recommend the adoption of supplementary assessment methods (5). One study used ultrasound to determine the bone age of soccer players, aiming to evaluate whether it could serve as a radiation-free alternative to X-ray. The results did not demonstrate significant effectiveness (6). However, the ultrasound method employed in that study was proposed in 2019 and has not gained international recognition. Therefore, whether improved ultrasound methods exist requires further research. The literature indicates that quantitative ultrasound technology is highly promising. If an ultrasound method capable of accurately assessing bone age is developed in the future, it will provide additional options for identifying youth talent in sports (7).
Currently, the primary methods for assessing bone age are the Greulich-Pyle (GP) (8) atlas method and the Tanner-Whitehouse (TW) (8) scoring system. The GP method provides a visually intuitive and convenient approach for assessing bone age, with high accuracy and broad age coverage. However, as the reference population of the GP method primarily consists of European and American children, some countries consider this method not universally applicable across all ethnic groups (9). Compared with the GP method, the TW method is more accurate for assessing bone age. However, the TW method involves complex procedures and is more time-consuming. In recent years, artificial intelligence (AI) has been rapidly developed for automated bone age assessment. Compared with manual assessment, AI-based evaluation offers advantages, including rapid processing and reduced subjectivity (10,11), significantly improving the efficiency of radiologists in determining bone age. However, the accuracy of AI-based bone age assessment depends heavily on data quality, diversity, and annotation consistency rather than on dataset size alone. In addition, AI model development involves substantial costs, and bone age assessment outcomes may vary among different algorithms.
Besides traditional X-ray skeletal age evaluation, novel imaging alternatives include dual-energy X-ray absorptiometry (DXA), MRI and ultrasound. All three modalities eliminate or lower ionizing radiation relative to radiography, yet each carries notable drawbacks. DXA reduces radiation exposure but suffers from high equipment cost, lengthy scanning, and systematic bone age overestimation as previously documented (12). MRI visualizes epiphyseal cartilage clearly with zero radiation, though its clinical use is restricted by high costs, long acquisition time, susceptibility to imaging artifacts and absent unified assessment standards. Ultrasound is portable and radiation-free, with good visualization of epiphyseal ossification centers and cartilage. However, relevant research stays preliminary (13,14), which restricts its large-scale clinical adoption for bone age measurement.
Previous studies have employed various ultrasound-based methods for assessing bone age, including the measurement of femoral head cartilage thickness, the use of ultrasound versions of left wrist atlases, and the detection of changes in sound velocity in the left wrist. However, these methods have not been widely adopted in clinical practice because of their limitations. For example, the use of ultrasound to measure the thickness of the femoral head cartilage has low sensitivity, whereas the method for measuring changes in the sound velocity of the left wrist has poor positive and negative predictive values and is easily affected by surrounding soft tissues (15). Currently, there is no standardized protocol for ultrasound evaluation of bone age. Therefore, this study uses radiographic bone age as the diagnostic reference standard to evaluate the feasibility of ultrasound-determined bone age. We hypothesized that a simplified ultrasound-derived skeletal maturity score (SMS) based on the ulna, radius, and femur would show strong correlation with Tanner-Whitehouse 3 radius-ulna-short bone (TW3-RUS) radiographic bone age and high inter- and intra-observer reliability. We present this article in accordance with the STROBE reporting checklist (available at https://tp.amegroups.com/article/view/10.21037/tp-2026-0620/rc).
Methods
Participants and instruments
This study protocol was reviewed and approved by the Ethics Committee of Affiliated Hospital of North Sichuan Medical College (approval No. 2023ER438-1). Written informed consent was obtained from the legal guardians of all pediatric participants before enrollment. All study procedures were conducted in accordance with the ethical principles laid down in the World Medical Association Declaration of Helsinki and its subsequent amendments. The inclusion criteria required that the participants had no pathological changes in the left wrist or knee joint or any systemic diseases that might affect normal bone development (such as infection, deformity, trauma, rickets, or the use of hormones that influence normal bone development), as skeletal dysplasia can result in inaccurate bone age assessment. Additionally, clinical data, including medical history, height, and weight, were collected from the medical records of each participant.
Sample size calculation was performed using PASS software for Pearson correlation analysis. Based on previous pediatric bone age studies, we hypothesized the correlation coefficient between ultrasonic combined skeletal maturity score (CMS) and radiographic bone age was r=0.90. We set a two-sided α of 0.05 and statistical power of 0.80, with an acceptable margin of error of 0.06 for correlation estimation. The calculation showed that a minimum of 110 participants was required to guarantee sufficient statistical precision.
Subjects in this study were children aged 3 to 16 years who underwent left wrist radiographic bone age assessment at a tertiary hospital from December 2023 to November 2024. A total of 142 children were initially screened. Two participants were excluded due to failed image acquisition caused by poor cooperation during ultrasound scanning, and one participant was excluded because the interval between radiographic bone age evaluation and ultrasound examination exceeded 1 week. Finally, 139 children were enrolled, including 63 boys and 76 girls. All subjects were consecutively recruited, and no missing data were observed. All 139 participants were included in every statistical analysis. The participants included healthy children (29 girls, 37 boys), children with short stature (16 girls, 18 boys), precocious puberty (21 girls, 1 boy), obesity (6 girls, 4 boys), delayed puberty (3 boys, 3 girls), and a chest mass (1 girl). All participants’ bone ages were independently evaluated by two pediatricians and one radiologist using the TW3-RUS method (16). In cases of discordant assessments, the three physicians performed a joint re-evaluation to reach a consensus. Radiographic skeletal age was independently evaluated by three pediatric physicians using the Chinese population-adapted TW3-RUS method, with 12, 13 and 9 years of bone age reading experience respectively. All readers were fully blinded to children’s chronological age, clinical information and ultrasonic data during assessment. Cases with inconsistent readings were discussed by all three physicians to reach a unified skeletal age value as the reference standard. Figure 1 presents the flowchart of study participants.
Figure 1.
Participant flowchart. CMS, combined skeletal maturity score; SMS, skeletal maturity scores.
Ultrasonic assessments were performed using two ultrasound systems, including the Aplio 500 (Canon Medical Systems, Tokyo, Japan) and the Aixplorer (Supersonic Imagine, Aix-en-Provence, France). A 5–13 MHz high-frequency linear array probe operating in MSU imaging mode was used for all participants. The scanning depth was standardized to 2–3 cm with minor individualized adjustments according to pediatric bone size. No fixed unified values were predefined for the focal zone, gain, dynamic range, tissue harmonic imaging, and compound imaging. All imaging parameters were flexibly adjusted by operators to ensure clear visualization of bony and cartilaginous contours, with consistent imaging settings maintained on the two ultrasound devices as far as possible.
All measurements were acquired in real time during live scanning rather than via offline reanalysis of stored static images. A total of 14 images were acquired per participant, with image eligibility defined as clear visualization of the targeted anatomical structures required for measurement. All images were saved in JPG format. Each ossification center was measured only once, and the single realtime measurement values were directly included in the final statistical analyses. All measurements were completed on the original ultrasound scanners.
Image acquisition and measurement were performed by three radiologists with different levels of clinical experience, including one senior radiologist and two trained junior radiologists. Before formal enrollment, the two junior radiologists completed a pre-experimental training session consisting of 20 cases. Qualification was determined by the senior radiologist, who assessed the standardization of image quality and measurement positioning; only qualified operators participated in the official data collection. The full 14-ossification-center scanning protocol required 7–10 minutes per child. All ultrasound scans were performed within 1 week after radiography (98 on the same day, 37 on days 2, 3 on day 3, and 1 on day 4). All scans were conducted by a radiologist who was blinded to the radiologic bone age assessment results. The representative images were selected prior to knowledge of the study results.
Ultrasound
In accordance with the TW3-RUS method for bone age assessment, our study initially selected the ulna and radius, along with the first, third, and fifth metacarpals and phalanges. On the basis of Wan et al.’s (17) research demonstrating the strong correlation between distal femoral growth and bone age, we included the distal femur, resulting in a total of 14 bones for evaluation. Ultrasound imaging was used to measure both the longitudinal diameter of the ossification center and the epiphyseal length for each individual bone. When the metacarpals and phalanges were scanned, we innovatively incorporated the flexor digitorum tendons into the standard cross-section. Furthermore, horizontal measuring lines parallel to the horizontal axis of the ultrasound image were drawn to measure the maximum longitudinal length of ossification centers and epiphyseal cartilages.
Participants were positioned supine on the examination table. For metacarpophalangeal joint scanning, the palm faces upward, with the fingers extended and slightly abducted. During ulnar examination, the palm was positioned downward in a naturally relaxed state. Radial scans were performed with the hand in the sagittal position while natural relaxation was maintained. For knee joint evaluation, the knee was kept in a naturally relaxed position with slight external rotation (Figure 2).
Figure 2.
Schematic diagram of ultrasound scanning positions and the correct positioning of the participant: (A-C) sagittal scanning of the 1st, 3rd and 5th metacarpals and phalanges with the ultrasound probe. (D-F) Coronal scanning of the ulna, radius and femur. All scans were performed on the left side. The probe marker was oriented toward the subject’s feet in panels (A-E). The probe marker was oriented toward the subject’s head in panel (F).
When ultrasound scanning of the metacarpal and phalangeal bones was performed, the flexor digitorum tendons served as anatomical markers for obtaining sagittal plane images. Ultrasound scanning of the ulna, radius, and femur was performed in the coronal plane. The standardized scanning protocol for the distal ulna required clear visualization of the ulnar styloid process, triangular fibrocartilage complex, and diaphysis. For the distal radius, the protocol included the radial styloid process and diaphysis, along with the tendons of the abductor pollicis longus and extensor pollicis brevis (Figure 3A). The distal femur scanning protocol revealed the medial collateral ligament, meniscus, and diaphysis. All the scanning planes were used to clearly visualize the contours of both the ossification center and the epiphysis.
Figure 3.
Radius (A), femur (B). The red lines denote ossification centers, the blue lines represent epiphyseal cartilage, and the yellow lines indicate cortical bone. d = the longitudinal diameter of the ossification center, D = the longitudinal diameter of epiphyseal cartilage. For parameter d, the measurement starts at the proximal margin and terminates at the distal margin of the ossification center. For parameter D, the measurement originates at the junction of epiphyseal cartilage and diaphysis and ends at the distal edge of epiphyseal cartilage.
The ossification center and epiphysis of each bone were measured. For the distal femur, a reference point was established at the meniscal apex (Figure 3B). The ossification ratio (OsR) was calculated for each bone, where OsR = ossification center length/epiphyseal length (d/D), with d representing the longitudinal diameter of the ossification center and D denoting the longitudinal diameter of the epiphysis. The ratio d/D equals 0 if the ossification center has not appeared, and equals 1 when complete fusion occurs.
Figure 4 shown representative ultrasound images of female children of different ages for measuring the ossification center and the longitudinal diameter of the epiphysis. The SMS for an individual bone was calculated as SMS = OsR × 100. The composite CMS for multiple bones were determined by CMS = Sermi × Wi, where W represents the weighting coefficient and i denotes the 14 examined bones. In accordance with the TW3-RUS method, distinct weighting coefficients were assigned to different bones based on their varying contributions to BA (bone age) assessment: 0.67 for the first metacarpal and phalanges, 0.5 for the third and fifth metacarpals and phalanges, and 2 for the ulna and radius. Because TW3-RUS radiographic assessment and the ultrasound measurements in this study both quantify the ossification progression of bone. The TW3-RUS scoring system inherently defines the relative developmental contribution of each skeletal site to overall skeletal maturation. Since the ultrasonic OsR captures the same temporal sequence of epiphyseal maturation within these identical anatomical structures, the relative weight assigned to individual skeletal sites is biologically plausible across imaging modalities. In our study, based on previous research (17), we empirically assigned the same weighting coefficient to the ulna and radius of the femur.
Figure 4.
Representative ultrasound images of female children of different ages for measuring the ossification center and longitudinal diameter of the epiphysis, showing that the OsR changes with age. d/D, ossification center length/epiphyseal length; DP1, first distal phalanx; MC5, fifth metacarpal bone; MP3, third middle phalanx; OsR, ossification ratio; PP3, third proximal phalanx; Y, years.
Statistical analysis
The primary analysis of this study focused on the correlation between CMS and radiographic bone age, while intra- and interobserver repeatability were assessed as secondary analyses. All data in this study were verified to follow a normal distribution. Accordingly, Pearson’s correlation coefficient r was used to analyze the correlations between ultrasound derived skeletal parameters and radiographic bone age. Prior to Pearson correlation analysis, the linear trend was verified via scatter plots, and the normality of variables was assessed using the Shapiro-Wilk test. Boxplots were used to screen outliers, and Cook’s distance was applied to identify influential points. The 95% confidence intervals (CIs) of correlation coefficients were calculated via Fisher’s Z transformation, and the percentile method was adopted for bootstrap CIs. Intraclass correlation coefficient (ICC) was applied to evaluate interobserver and intraobserver measurement repeatability. Single-measurement ICC was calculated using a two-way random-effects model with absolute agreement. Absolute agreement ICC evaluates the degree of concordance between the absolute values of two sets of measurements; the ICC approaches 1 only when the true values of the two measurements are highly consistent. 30 children (14 boys and 16 girls, aged 3–16 years) were stratified randomly sampled covering the full age range of this study. Two radiologists independently interpreted the images to assess interobserver reliability; the second radiologist performed measurements on stored images. For intraobserver reliability assessment, the same radiologist remeasured these 30 image sets after a 3-month interval. Each original measurement was performed once, and intraobserver repeatability was assessed by repeated measurements of identical images. Both readers were radiologists with 2 and 3 years of image-reading experience, respectively. The assessors were blinded to their previous measurements and radiographic findings. Finally, Bland-Altman analysis was performed to evaluate outcome agreement.
Results
Participants
The baseline characteristics of the 139 children are summarized in Table 1. The mean age was 8.3±3.3 (3.0–15.5) years for boys and 8.6±2.6 (3.1–14.2) years for girls.
Table 1. Baseline characteristics of study participants.
| Variable | Boys | Girls | ||||
|---|---|---|---|---|---|---|
| 3–<6 years | 6–<12 years | 12–15 years | 3–<6 years | 6–<12 years | 12–15 years | |
| Chronological age (years) | 5.0±1.1 | 10.5±1.5 | 13.7±1.0 | 5.1±1.0 | 9.6±1.6 | 14.1±0.2 |
| Radiographic bone age (years) | 5.1±1.4 | 10.3±2.2 | 14.2±2.0 | 5.1±1.3 | 10.2±1.9 | 15.2±0.6 |
| Height (cm) | 106.2±8.3 | 137.3±13.3 | 154.1±10.0 | 107.0±9.1 | 136.8±9.8 | 151.0±6.9 |
| Weight (kg) | 17.2±3.2 | 31.1±11.0 | 48.5±9.2 | 17.0±5.9 | 33.6±9.9 | 43.8±8.8 |
| BMI (kg/m2) | 15.1±1.7 | 16.3±3.8 | 20.2±2.0 | 14.5±2.3 | 18.2±3.2 | 19.1±2.1 |
Data are presented as mean ± standard deviation. BMI, body mass index.
Correlation analysis of radiographic bone age and ultrasound bone parameters
Correlation analyses were conducted between radiologic bone age and ultrasound-measured bone parameters from individual bones, including the 11 metacarpals and phalanges, ulna, radius, and femur. The primary analysis focused on CMS, whereas the exploratory analysis examined the ossification center and epiphysis. The results are presented in Table 2,3. Most individual bone parameters were weakly correlated with radiographic bone age. Only the ossification center of the radius (boys), femoral SMS (boys), and ulnar SMS (both sexes) demonstrated strong correlations (r>0.9). Correlation analysis was conducted between radiologic bone age and grouped ultrasound bone parameters from the 11 metacarpals and phalanges, ulna, radius, and femur, and the results are presented in Table 4. The analysis revealed consistently strong correlations between radiographic bone age and ultrasound parameters across all the groups. In particular, the CMS of the URF (ulna + radius + femur) showed the strongest correlation with radiographic bone age (boys: r=0.961, 95 % CI: 0.936–0.977; P<0.001; girls: r=0.962, 95 % CI: 0.940–0.976; P<0.001) (Table 4). CMS (URF) was constructed using weighted coefficients without normalization. This study did not prospectively establish separate training and validation subsets. Therefore, this empirically weighted CMS (URF) serves only as an exploratory indicator and requires external validation in independent cohorts prior to clinical implementation. The composite CMS (URF) score in this exploratory analysis was constructed using predefined empirical weights rather than weights fitted from the present sample, which reduces the risk of overfitting to this cohort. We additionally constructed a simple unweighted model for comparison (Table 5). The results from the unweighted and empirically weighted models were similar.
Table 2. Pearson correlation coefficients (r) between radiographic bone age and ultrasound measurements of ossification center length, epiphyseal length, and SMS (boys).
| Variable | DP1 | PP1 | MC1 | DP3 | MP3 | PP3 | MC3 | DP5 | MP5 | PP5 | MC5 | Ulna | Radius | Femur |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Ossi | 0.602 | 0.755 | 0.470 | 0.811 | 0.783 | 0.674 | 0.878 | 0.855 | 0.669 | 0.619 | 0.813 | 0.898 | 0.936 | 0.924 |
| Epi | 0.264 | 0.359 | 0.143 | 0.571 | 0.559 | 0.301 | 0.726 | 0.567 | 0.321 | 0.155 | 0.499 | 0.133 | 0.806 | 0.668 |
| SMS | 0.705 | 0.774 | 0.618 | 0.629 | 0.697 | 0.623 | 0.856 | 0.775 | 0.599 | 0.581 | 0.841 | 0.944 | 0.896 | 0.931 |
DP1, first distal phalanx; DP3, third distal phalanx; DP5, fifth distal phalanx; Epi, epiphysis; MC1, first metacarpal; MC3, third metacarpal; MC5, fifth metacarpal; MP3, third middle phalanx; MP5, fifth middle phalanx; Ossi, ossification center; PP1, first proximal phalanx; PP3, third proximal phalanx; PP5, fifth proximal phalanx; SMS, skeletal maturity scores.
Table 3. Pearson correlation coefficients (r) between radiographic bone age and ultrasound measurements of ossification center length, epiphyseal length, and SMS (girls).
| Variable | DP1 | PP1 | MC1 | DP3 | MP3 | PP3 | MC3 | DP5 | MP5 | PP5 | MC5 | Ulna | Radius | Femur |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Ossi | 0.617 | 0.663 | 0.545 | 0.692 | 0.754 | 0.630 | 0.594 | 0.822 | 0.365 | 0.613 | 0.701 | 0.868 | 0.892 | 0.811 |
| Epi | 0.374 | 0.322 | 0.268 | 0.447 | 0.458 | 0.272 | 0.299 | 0.619 | 0.257 | 0.407 | 0.526 | 0.484 | 0.618 | 0.687 |
| SMS | 0.549 | 0.630 | 0.672 | 0.512 | 0.290 | 0.657 | 0.669 | 0.590 | 0.539 | 0.328 | 0.762 | 0.950 | 0.598 | 0.741 |
DP1, first distal phalanx; DP3, third distal phalanx; DP5, fifth distal phalanx; Epi, epiphysis; MC1, first metacarpal; MC3, third metacarpal; MC5, fifth metacarpal; MP3, third middle phalanx; MP5, fifth middle phalanx; Ossi, ossification center; PP1, first proximal phalanx; PP3, third proximal phalanx; PP5, fifth proximal phalanx; SMS, skeletal maturity scores.
Table 4. Pearson correlation coefficients (r) between radiographic bone age and ultrasound measurements of ossification center length, epiphyseal length and SMS.
| Variable | Ossi (SUM) | Epi (SUM) | CMS (SUM) | Ossi (SUM + F) | Epi (SUM + F) | CMS (SUM + F) | Ossi (UR) | Epi (UR) | CMS (UR) | Ossi (URF) | Epi (URF) | CMS (URF) |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Boys | 0.919 | 0.914 | 0.927 | 0.955 | 0.889 | 0.945 | 0.940 | 0.864 | 0.940 | 0.958 | 0.841 | 0.961 |
| Girls | 0.922 | 0.889 | 0.916 | 0.930 | 0.881 | 0.927 | 0.936 | 0.811 | 0.870 | 0.908 | 0.812 | 0.962 |
CMS, combined skeletal maturity score; Epi, epiphysis; F, femur; Ossi, ossification center; SMS, skeletal maturity scores; SUM, ulna + radius + 11 metacarpals and phalanges; UR, ulna + radius; URF, ulna + radius + femur.
Table 5. Pearson correlation coefficients (r) between the CMS (URF) and radiographic bone age when different weighting coefficients are assigned to the femur.
| Variable | URF (Wf =1) | URF (Wf =2) |
|---|---|---|
| Boys | 0.953 | 0.961 |
| Girls | 0.965 | 0.962 |
Wf, weighting coefficient for the femur; Wf = 1 represents the unweighted scheme; Wf = 2 represents the empirically weighted scheme. CMS, combined skeletal maturity score; URF, ulna + radius + femur.
Linear regression analysis was performed between radiographic bone age and CMS (URF), revealing strong correlations for both sexes. Notably, the simple linear regression plot in this study was only used to visually illustrate the linear trend and was not intended for constructing a predictive model. The analysis yielded high coefficients of determination, with R2=0.92 for boys and R2=0.92 for girls (Figure 5). Internal validation was performed using 1,000 bootstrap resampling iterations. The 95 % bootstrap CIs for the correlation coefficient were 0.94–0.98 in the boys’ group and 0.92–0.97 in the girls’ group.
Figure 5.
Linear regression plots of CMS (URF) and RBA for boys (A, n=63) and girls (B, n=76). Narrow shaded area: 95% confidence band for the population mean of radiographic bone age at a given CMS (URF) value; wide shaded area: 95% prediction band for the radiographic bone age of an individual new subject with a fixed CMS (URF) score. CMS, combined skeletal maturity score; RBA, radiographic bone age; URF, ulna + radius + femur.
Analysis of the mean CMS values across bone age groups aged 3–12 years revealed consistent trends in both boys and girls. SMS values increased progressively with advancing bone age. Within each identical bone age group, CMS measurements followed a stable hierarchical order: femur > radius > ulna, with females consistently demonstrating higher values than males (Table 6). For the remaining bone age subgroups, the small sample sizes led to unstable means and standard deviations; therefore, no inferences were drawn from the stratified data.
Table 6. Mean values and standard deviations of CMS for the ulna, radius, femur, and their combined total.
| BA | Boys | Girls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| N | Ulna | Radius | Femur | URF | N | Ulna | Radius | Femur | URF | |
| 1–1.9 | 1 | 0 | 27.0 | 37.0 | 64.0 | – | – | – | – | – |
| 2–2.9 | 1 | 0 | 24.0 | 46.0 | 70.0 | – | – | – | – | – |
| 3–3.9 | 3 | 0 | 35.3±10.0 | 43.3±6.4 | 78.3±16.3 | 4 | 0 | 42.5±6.5 | 50.8±8.8 | 93.3±12.6 |
| 4–4.9 | 9 | 0 | 40.9±6.3 | 46.3±2.6 | 87.1±6.9 | 5 | 0 | 38.8±5.4 | 57.6±4.0 | 96.2±7.5 |
| 5–5.9 | 8 | 0 | 43.0±4.1 | 53.5±8.1 | 96.4±5.7 | 7 | 1.4±3.8 | 47±7.9 | 66.4±6.1 | 115.0±10.4 |
| 6–6.9 | 4 | 0 | 52.0±7.7 | 58.3±7.8 | 110.0±13.4 | 4 | 15.8±13.6 | 50.8±6.0 | 67.8±2.9 | 133.5±13.8 |
| 7–7.9 | 3 | 2.3±4.0 | 46.7±6.4 | 55.3±11.2 | 104.0±20.8 | 6 | 19.5±10.3 | 49.7±6.6 | 71.0±4.4 | 140.3±17.3 |
| 8–8.9 | 7 | 9.4±12.3 | 51.0±2.3 | 64.9±5.9 | 125.3±13.9 | 5 | 30.0±13.5 | 54.4±3.9 | 73.4±6.4 | 157.6±21.1 |
| 9–9.9 | 5 | 23.8±15.1 | 61.6±3.5 | 67.6±5.7 | 153.0±20.6 | 8 | 42.9±5.2 | 58.8±5.4 | 77.1±4.5 | 178.6±6.7 |
| 10–10.9 | 4 | 38.0±5.8 | 58.0±5.3 | 75.0±6.7 | 170.8±10.0 | 12 | 52.3±8.5 | 62.4±6.5 | 74.3±19.4 | 189.2±22.3 |
| 11–11.9 | 6 | 40.0±9.4 | 64.7±11.4 | 73.3±3.8 | 177.8±7.5 | 11 | 57.3±8.6 | 74.4±7.5 | 82.6±5.3 | 214.4±13.2 |
| 12–12.9 | 2 | 45.5±7.1 | 60.5±0.7 | 84.0±5.7 | 189.5±6.4 | 5 | 72.0±9.9 | 82.0±7.0 | 85.8±4.5 | 240.0±16.9 |
| 13–13.9 | 2 | 59.0±15.6 | 64.5±9.2 | 77.0±1.4 | 201.0±25.5 | 4 | 77.5±8.6 | 84.0±4.2 | 89.5±3.1 | 251.3±11.9 |
| 14–14.9 | 3 | 71.7±3.2 | 84.3±6.7 | 84.0±5.0 | 240.0±6.1 | 1 | 80.0 | 92.0 | 96.0 | 267.0 |
| 15–15.9 | 1 | 70.0 | 96.0 | 94.0 | 259.0 | 1 | 82.0 | 85.0 | 93.0 | 260.0 |
| 16–16.9 | 2 | 93.5±0.7 | 90.5±4.9 | 95.0±2.8 | 278.5±7.8 | – | – | – | – | – |
Data are presented as mean ± standard deviation. BA, bone age; CMS, combined skeletal maturity score; URF, ulna + radius + femur.
The interobserver ICC indicated low repeatability for single bone measurements of metacarpals and phalanges, whereas higher ICC values were obtained for the ulna, radius, and femur. The interobserver and intraobserver ICCs for the combined URF three-bone score both reached 0.99 (95% CI: 0.98–1.00). The intraobserver ICC obtained from repeated measurements by the same radiologist at a 3-month interval was comparable to the interobserver reliability (Table 7). For CMS (URF), the interobserver SEM (standard error of measurement) was 5.26, the CV (coefficient of variation) was 3.34%, and the CR (coefficient of repeatability) was 14.58. The intraobserver SEM, CV and CR of CMS (URF) were 4.68, 3.00%, and 12.98, respectively. Bland-Altman analysis demonstrated that the 95% limits of agreement were [11.76, 14.19] for intraobserver comparisons and [11.32, 16.23] for interobserver comparisons (Figure 6).
Table 7. Inter-observer and intra-observer ICC values for the 14 examined bones.
| Variable | Intra-observer (95 % CI) | Inter-observer (95 % CI) |
|---|---|---|
| DP1 | 0.83 (0.68–0.92) | 0.87 (0.64–0.94) |
| PP1 | 0.85 (0.70–0.92) | 0.84 (0.67–0.92) |
| MC1 | 0.79 (0.58–0.90) | 0.86 (0.73–0.93) |
| DP3 | 0.69 (0.43–0.84) | 0.75 (0.50–0.88) |
| MP3 | 0.85 (0.71–0.93) | 0.82 (0.54–0.92) |
| PP3 | 0.81 (0.64–0.91) | 0.90 (0.77–0.95) |
| MC3 | 0.61 (0.33–0.79) | 0.89 (0.78–0.95) |
| DP5 | 0.85 (0.72–0.93) | 0.88 (0.73–0.95) |
| MP5 | 0.87 (0.71–0.94) | 0.75 (0.54–0.87) |
| PP5 | 0.73 (0.55–0.88) | 0.85 (0.71–0.93) |
| MC5 | 0.66 (0.38–0.83) | 0.48 (0.15–0.71) |
| U | 0.96 (0.90–0.98) | 0.97 (0.93–0.99) |
| R | 0.96 (0.93–0.98) | 0.98 (0.95–0.99) |
| F | 0.99 (0.97–0.99) | 0.97 (0.93–0.98) |
| URF | 0.99 (0.98–1.00) | 0.99 (0.98–1.00) |
DP, distal phalanx; F, femur; ICC, intraclass correlation coefficient; MC, metacarpal; MP, middle phalanx; PP, proximal phalanx; R, radius; U, ulna; URF, ulna + radius + femur.
Figure 6.
Bland-Altman plots of CMS measurements. (A) Intra-observer repeated measurements; (B) inter-observer measurements. X axis: mean of paired measurements; Y axis: measurement difference. Dashed horizontal lines = 95% limits of agreement. CMS, combined skeletal maturity score.
Discussion
As individuals age, their skeletal structures undergo predictable growth patterns. Based on these developmental changes, researchers have established various methods for assessing bone age. Currently, radiographic bone age evaluation using left wrist X-rays remains the reference standard in clinical practice.
Emerging methods for assessing bone age primarily include DXA, MRI, and ultrasound. While compared with conventional methods, DXA offers reduced radiation exposure, its clinical application is limited by high equipment costs, time-consuming procedures, and questionable reliability. MRI provides radiation-free imaging with excellent cartilage visualization; however, its prohibitive examination costs and lengthy scan times have prevented widespread clinical adoption. Ultrasound combines portability with a radiation-free operation, making it particularly suitable for pediatric patients who require long-term serial bone age monitoring (18). Although ultrasound bone age assessment has been implemented in limited regions, these methods have not gained widespread clinical acceptance. For instance, the BonAge system (19) has been reported to demonstrate insufficient sensitivity in bone age evaluation and cannot replace conventional radiographic assessment using X-ray imaging as the reference standard.
In our study, ultrasound was used to measure the maximum longitudinal diameters of the ossification centers and epiphyses in the left first, third, and fifth metacarpals and phalanges, as well as the ulna, radius, and femur. The OsR was defined as the ratio of the ossification center length to the epiphyseal length. The SMS was calculated by multiplying the OsR by 100. Systematic grouping and combination of bones from different anatomical regions were subsequently conducted for comprehensive analysis. The results demonstrated that the CMS of the ulna, radius, and femur were most strongly correlated with radiographic bone age, with Pearson correlation coefficients (r) of 0.961 for boys and 0.962 for girls. Compared with previous ultrasound approaches (e.g., wrist sound velocity methods), the multiparameter weighted scoring system (CMS = ΣWi × SMSi) reduces individual bone variability and enhances scanning stability through comprehensive parameter integration.
Inter-observer and intra-observer reliability analyses revealed relatively low ICC values for the metacarpals and phalanges, whereas higher ICC values were observed for the ulna, radius, and femur. Notably, the URF combination demonstrated the highest ICC (0.99) for both the inter- and the intra-observer assessments. This discrepancy may be attributed to the smaller size of the metacarpals and phalanges, which are less clearly visualized on ultrasound than the more distinct anatomical structures of the ulna, radius, and femur.
The strengths of this study include the well-defined standard sectional views (including ossification centers and epiphyseal cartilage) that are clearly visible on ultrasound imaging. The simple operation and measurement protocols minimize subjective errors. Furthermore, the URF combination requires the scanning of only three anatomical sites, resulting in a shorter examination time (2–4 min), thereby establishing a practical foundation for implementation in primary healthcare settings.
Beyond clinical endocrinology and primary care, the proposed ultrasound method may also be useful in youth sports, where serial maturity monitoring is needed to interpret performance, address selection biases, and manage injury risk without cumulative radiation exposure (5).
This study has several limitations. First, it was a single center investigation with a limited sample size and uneven age distribution across subgroups, and external cohort validation was lacking, which may introduce selection bias. Second, pubertal staging was not fully stratified in the present study. Third, clinically interpretable predictive error indicators were not provided, and formal consistency analysis was not performed. In addition, the extremely wide age range of the enrolled participants may artificially elevate the overall correlation coefficient. Furthermore, the weight coefficient assigned to the femur was empirically determined and has not been formally validated. Future multicenter, large sample clinical studies are warranted to optimize and validate our findings.
Conclusions
In summary, our proposed method is rapid and straightforward to perform, with an average scanning time of 2–4 min for the left ulna, radius, and femur. The ultrasound-derived ulna-radius-femur SMS showed a strong cross-sectional association with TW3-RUS radiographic bone age and excellent measurement reliability. These findings support further development of the method; however, they do not yet establish individual-level agreement, predictive accuracy, or diagnostic interchangeability with radiographic bone-age assessment.
Supplementary
The article’s supplementary files as
Acknowledgments
The authors express sincere gratitude to the staff of the Affiliated Hospital of North Sichuan Medical College for contributing to this research, in addition to providing excellent care to the pediatric patients.
Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. This study was approved by the Ethics Committee of Affiliated Hospital of North Sichuan Medical College (approval No. 2023ER438-1). Written informed consent was obtained from the legal guardians of all pediatric participants before enrollment.
Footnotes
Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://tp.amegroups.com/article/view/10.21037/tp-2026-0620/rc
Funding: This work was supported by the Doctoral Initiation Fund of Affiliated Hospital of North Sichuan Medical College (Nos. 2023-2GC013 and 210930).
Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tp.amegroups.com/article/view/10.21037/tp-2026-0620/coif). All authors report that this work was supported by the Doctoral Initiation Fund of Affiliated Hospital of North Sichuan Medical College (Nos. 2023-2GC013 and 210930). The authors have no other conflicts of interest to declare.
Data Sharing Statement
Available at https://tp.amegroups.com/article/view/10.21037/tp-2026-0620/dss
References
- 1. Cavallo F, Mohn A, Chiarelli F, et al. Evaluation of Bone Age in Children: A Mini-Review. Front Pediatr 2021;9:580314.. 10.3389/fped.2021.580314 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Zhang Y, Yang X, Zhang L, et al. Ultrasound-based bone age assessment in children and adolescents: a mini-review. Front Pediatr 2025;13:1615923.. 10.3389/fped.2025.1615923 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Fransen J, Skorski S, Baxter-Jones ADG. Estimating is not measuring: the use of non-invasive estimations of somatic maturity in youth football. Sci Med Footb 2021;5:261-2.. 10.1080/24733938.2021.1975808 [DOI] [PubMed] [Google Scholar]
- 4. Schmidt S, Koch B, Schulz R, et al. Studies in use of the Greulich-Pyle skeletal age method to assess criminal liability. Leg Med (Tokyo) 2008;10:190-5.. 10.1016/j.legalmed.2008.01.003 [DOI] [PubMed] [Google Scholar]
- 5. Leyhr D, Murr D, Basten L, et al. Biological Maturity Status in Elite Youth Soccer Players: A Comparison of Pragmatic Diagnostics With Magnetic Resonance Imaging. Front Sports Act Living 2020;2:587861.. 10.3389/fspor.2020.587861 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6. Widmer C, Busch JD, Born DP, et al. Assessment of Biological Age With Conventional Ultrasound Imaging as an Alternative to X-Ray-A Pilot Study in Youth Soccer. Eur J Sport Sci 2025;25:e12264.. 10.1002/ejsc.12264 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Rüeger E, Hutmacher N, Eichelberger P, et al. Ultrasound Imaging-Based Methods for Assessing Biological Maturity during Adolescence and Possible Application in Youth Sport: A Scoping Review. Children (Basel) 2022;9:1985.. 10.3390/children9121985 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Garn SM. Radiographic Atlas of Skeletal Development of the Hand and Wrist. Am J Hum Genet 1959;11:282-3. [Google Scholar]
- 9. Prokop-Piotrkowska M, Marszałek-Dziuba K, Moszczyńska E, et al. Traditional and New Methods of Bone Age Assessment-An Overview. J Clin Res Pediatr Endocrinol 2021;13:251-62.. 10.4274/jcrpe.galenos.2020.2020.0091 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Martin DD, Calder AD, Ranke MB, et al. Accuracy and self-validation of automated bone age determination. Sci Rep 2022;12:6388.. 10.1038/s41598-022-10292-y [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Nguyen T, Hermann AL, Ventre J, et al. High performance for bone age estimation with an artificial intelligence solution. Diagn Interv Imaging 2023;104:330-6.. 10.1016/j.diii.2023.04.003 [DOI] [PubMed] [Google Scholar]
- 12. Cruz-Priego GA, Guagnelli MA, Miranda-Lora AL, et al. Bone Age Reading by DXA Images should not Replace Bone Age Reading by X-ray Images. J Clin Densitom 2022;25:456-63.. 10.1016/j.jocd.2022.08.004 [DOI] [PubMed] [Google Scholar]
- 13. Chen S, Xiong Y, Liu J, et al. Assessment Model of Pediatric Bone Age Based on High-Frequency Ultrasound: A Prospective Cohort Study. Ultrasound Med Biol 2025;51:2067-74.. 10.1016/j.ultrasmedbio.2025.07.020 [DOI] [PubMed] [Google Scholar]
- 14. Lin M, Yan L, He H, et al. Preliminary study on ultrasound-based quantitative assessment of epiphyseal ossification angle for bone age evaluation in children. Quant Imaging Med Surg 2026;16:136.. 10.21037/qims-2025-2009 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Rachmiel M, Naugolni L, Mazor-Aronovitch K, et al. Bone Age Assessments by Quantitative Ultrasound (SonicBone) and Hand X-ray Based Methods are Comparable. Isr Med Assoc J 2017;19:533-8. [PubMed] [Google Scholar]
- 16. Zhang SY, Liu LJ, Wu ZL, et al. Standards of TW3 skeletal maturity for Chinese children. Ann Hum Biol 2008;35:349-54.. 10.1080/03014460801953781 [DOI] [PubMed] [Google Scholar]
- 17. Wan J, Zhao Y, Feng Q, et al. Summation of Ossification Ratios of Radius, Ulna and Femur: A New Parameter to Evaluate Bone Age by Ultrasound. Ultrasound Med Biol 2020;46:1761-8.. 10.1016/j.ultrasmedbio.2020.03.021 [DOI] [PubMed] [Google Scholar]
- 18. Daneff M, Casalis C, Bruno CH, et al. Bone age assessment with conventional ultrasonography in healthy infants from 1 to 24 months of age. Pediatr Radiol 2015;45:1007-15.. 10.1007/s00247-014-3253-0 [DOI] [PubMed] [Google Scholar]
- 19. Khan KM, Miller BS, Hoggard E, et al. Application of ultrasound for bone age estimation in clinical practice. J Pediatr 2009;154:243-7.. 10.1016/j.jpeds.2008.08.018 [DOI] [PubMed] [Google Scholar]