Evidence of the non-linear nature of skeletal maturation

Clinical validation of a deep-learning-based bone age software in healthy Korean children

PURPOSE

Bone age (BA) is needed to assess developmental status and growth disorders. We evaluated the clinical performance of a deep-learning-based BA software to estimate the chronological age (CA) of healthy Korean children.

METHODS

This retrospective study included 371 healthy children (217 boys, 154 girls), aged between 4 and 17 years, who visited the Department of Pediatrics for health check-ups between January 2017 and December 2018. A total of 553 left-hand radiographs from 371 healthy Korean children were evaluated using a commercial deep-learning-based BA software (BoneAge, Vuno, Seoul, Korea). The clinical performance of the deep learning (DL) software was determined using the concordance rate and Bland-Altman analysis via comparison with the CA.

RESULTS

A 2-sample t-test (P<0.001) and Fisher exact test (P=0.011) showed a significant difference between the normal CA and the BA estimated by the DL software. There was good correlation between the 2 variables (r=0.96, P<0.001); however, the root mean square error was 15.4 months. With a 12-month cutoff, the concordance rate was 58.8%. The Bland-Altman plot showed that the DL software tended to underestimate the BA compared with the CA, especially in children under the age of 8.3 years.

CONCLUSION

The DL-based BA software showed a low concordance rate and a tendency to underestimate the BA in healthy Korean children.

Maturity-based correction mechanism for talent identification: When is it needed, does it work, and does it help to better predict who will make it to the pros?

When identifying talent, the confounding influence of maturity status on motor performances is an acknowledged problem. To solve this problem, correction mechanisms have been proposed to transform maturity-biased test scores into maturity-unbiased ones. Whether or not such corrections also improve predictive validity remains unclear. To address this question, we calculated correlations between maturity indicators and motor performance variables among a sample of 121 fifteen-year-old elite youth football players in Switzerland. We corrected motor performance scores identified as maturity-biased, and we assessed correction procedure efficacy. Subsequently, we examined whether corrected scores better predicted levels of performance achievement 6 years after data collection (47 professionals vs. 74 non-professional players) compared with raw scores using point biserial correlations, binary logistic regression models, and DeLong tests. Expectedly, maturity indicators correlated with raw scores (0.16 ≤ | r | ≤ 0.72; ps < 0.05), yet not with corrected scores. Contrary to expectations, corrected scores were not associated with an additional predictive benefit (univariate: no significant r-change; multivariate: 0.02 ≤ ΔAUC ≤ 0.03, ps > 0.05). We do not interpret raw and corrected score equivalent predictions as a sign of correction mechanism futility (more work for the same output); rather we view them as an invitation to take corrected scores seriously into account (same output, one fewer problem) and to revise correction-related expectations according to initial predictive validity of motor variables, validity of maturity indicators, initial maturity-bias, and selection systems. Recommending maturity-based corrections is legitimate, yet currently based on theoretical rather than empirical (predictive) arguments.

Maturity-based correction mechanism for talent identification: When is it needed, does it work, and does it help to better predict who will make it to the pros?

When identifying talent, the confounding influence of maturity status on motor performances is an acknowledged problem. To solve this problem, correction mechanisms have been proposed to transform maturity-biased test scores into maturity-unbiased ones. Whether or not such corrections also improve predictive validity remains unclear. To address this question, we calculated correlations between maturity indicators and motor performance variables among a sample of 121 fifteen-year-old elite youth football players in Switzerland. We corrected motor performance scores identified as maturity-biased, and we assessed correction procedure efficacy. Subsequently, we examined whether corrected scores better predicted levels of performance achievement 6 years after data collection (47 professionals vs. 74 non-professional players) compared with raw scores using point biserial correlations, binary logistic regression models, and DeLong tests. Expectedly, maturity indicators correlated with raw scores (0.16 ≤ | r | ≤ 0.72; ps < 0.05), yet not with corrected scores. Contrary to expectations, corrected scores were not associated with an additional predictive benefit (univariate: no significant r-change; multivariate: 0.02 ≤ ΔAUC ≤ 0.03, ps > 0.05). We do not interpret raw and corrected score equivalent predictions as a sign of correction mechanism futility (more work for the same output); rather we view them as an invitation to take corrected scores seriously into account (same output, one fewer problem) and to revise correction-related expectations according to initial predictive validity of motor variables, validity of maturity indicators, initial maturity-bias, and selection systems. Recommending maturity-based corrections is legitimate, yet currently based on theoretical rather than empirical (predictive) arguments.

OUP accepted manuscript

Background

The timing of growth is a key factor for correct orthodontic treatment planning. Cervical vertebrae maturation (CVM) is no exception, although the reported chronological ages vary in the literature.

Objective

We aimed to estimate the average chronological age for each Baccetti’s CVM staging.

Search methods

Search on MEDLINE-PubMed, Scopus, LILACS, Google Scholar, Cochrane Central Register of Controlled Trials (CENTRAL) was conducted until July 2021. The review was performed according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.

Selection criteria

Observational or interventional studies reporting chronological age classified through Baccetti’s CVM method were included.

Data collection and analysis

Methodological quality was assessed, and pooled estimates were carried out through random-effects meta-analysis of single means. The impact of sex and continent were also investigated through subgroup analyses.

Results

Forty-one studies were included (9867 participants, 4151 men, and 5716 women). The average chronological age was 9.7 years old (95% confidence interval [CI]: 9.4 to 10.1) in CS1, 10.8 years old (95% CI: 10.5 to 11.1) in CS2, 12.0 years old (95% CI: 11.7 to 12.2) in CS3, 13.4 years old (95% CI: 13.2 to 13.6) in CS4, 14.7 years old (95% CI: 14.4 to 15.1) in CS5, and 15.8 years old (95% CI: 15.3 to 16.3) in CS6. A significant difference was found between the sexes in all CVM stages. We also found significant differences across continents.

Conclusions

For each CVM staging a chronological age range was successfully estimated. Girls presented an earlier skeletal maturation compared to boys. The skeletal maturation differs also according to continents, except for CMV stage 1, pointing to the need for personalized ranges according to each region.

Registration

Registration number: PROSPERO: CRD42021225422

Estimating Craniofacial Growth Cessation: Comparison of Asymptote- and Rate-Based Methods

OBJECTIVE

To identify differences between asymptote- and rate-based methods for estimating age and size at growth cessation in linear craniofacial measurements.

DESIGN

This is a retrospective, longitudinal study. Five linear measurements were collected from lateral cephalograms as part of the Craniofacial Growth Consortium Study (CGCS). Four estimates of growth cessation, including 2 asymptote- (GC_asym, GC_err) and 2 rate-based (GC_abs, GC_10%) methods, from double logistic models of craniofacial growth were compared.

PARTICIPANTS

Cephalometric data from participants in 6 historic longitudinal growth studies were included in the CGCS. At least 1749 individuals (870 females, 879 males), unaffected by craniofacial anomalies, were included in all analyses. Individuals were represented by a median of 11 images between 2.5 and 31.3 years of age.

RESULTS

GC_asym consistently occurred before GC_err and GC_abs consistently occurred before GC_10% within the rate-based approaches. The ordering of the asymptote-based methods compared to the rate-based methods was not consistent across measurements or between males and females. Across the 5 measurements, age at growth cessation ranged from 13.56 (females, nasion-basion, GC_asym) to 24.39 (males, sella-gonion, GC_err).

CONCLUSIONS

Adolescent growth cessation is an important milestone for treatment planning. Based on our findings, we recommend careful consideration of specific definitions of growth cessation in both clinical and research settings since the most appropriate estimation method may differ according to patients' needs. The different methods presented here provide useful estimates of growth cessation that can be applied to raw data and to a variety of statistical models of craniofacial growth.