Evaluation of height prediction models: from traditional methods to artificial intelligence.

BACKGROUND: Traditional methods for predicting adult height (AHP) rely on manual readings of bone age (BA). However, the incorporation of artificial intelligence has recently improved the accuracy of BA readings and their incorporation into AHP models. METHODS: This study aimed to identify the AHP model that fits the current average height for adults in Mexico. Using a cross-sectional design, the study included 1173 participants (5-18 yr). BA readings were done by two experts (manually) and with an automated method (BoneXpert®). AHP was carried out using both traditional and automated methods. The best AHP model was the one that was closest to the population mean. RESULTS: All models overestimated the population mean (males: 0.7-6.7 cm, females: 0.9-3.7 cm). The AHP models with the smallest difference were BoneXpert for males and Bayley & Pinneau for females. However, the manual readings of BA showed significant interobserver variability (up to 43% of predictions between observers exceeded 5 cm using the Bayley & Pinneau method). CONCLUSION: Traditional AHP models relying on manual BA readings have high interobserver variability. Therefore, BoneXpert is the most reliable option, reducing such variability and providing AHP models that remain close to the mean population height. IMPACT: Traditional models for predicting adult height often result in overestimated height predictions. The manual reading of bone age is prone to interobserver variability, which can introduce significant biases in the prediction of adult height. The BoneXpert method minimizes the variability associated with traditional methods and demonstrates consistent results in relation to the average height of the population. This study is the first to assess adult height prediction models specifically in the current generations of Mexican children.

Automated bone age assessment in a German pediatric cohort: agreement between an artificial intelligence software and the manual Greulich and Pyle method.

OBJECTIVES: This study aimed to evaluate the performance of artificial intelligence (AI) software in bone age (BA) assessment, according to the Greulich and Pyle (G&P) method in a German pediatric cohort. MATERIALS AND METHODS: Hand radiographs of 306 pediatric patients aged 1-18 years (153 boys, 153 girls, 18 patients per year of life)-including a subgroup of patients in the age group for which the software is declared (243 patients)-were analyzed retrospectively. Two pediatric radiologists and one endocrinologist made independent blinded BA reads. Subsequently, AI software estimated BA from the same images. Both agreements, accuracy, and interchangeability between AI and expert readers were assessed. RESULTS: The mean difference between the average of three expert readers and AI software was 0.39 months with a mean absolute difference (MAD) of 6.8 months (1.73 months for the mean difference and 6.0 months for MAD in the intended use subgroup). Performance in boys was slightly worse than in girls (MAD 6.3 months vs. 5.6 months). Regression analyses showed constant bias (slope of 1.01 with a 95% CI 0.99-1.02). The estimated equivalence index for interchangeability was - 14.3 (95% CI -27.6 to - 1.1). CONCLUSION: In terms of BA assessment, the new AI software was interchangeable with expert readers using the G&P method. CLINICAL RELEVANCE STATEMENT: The use of AI software enables every physician to provide expert reader quality in bone age assessment. KEY POINTS: • A novel artificial intelligence-based software for bone age estimation has not yet been clinically validated. • Artificial intelligence showed a good agreement and high accuracy with expert radiologists performing bone age assessment. • Artificial intelligence showed to be interchangeable with expert readers.

Machine learning and deep learning enabled age estimation on medial clavicle CT images.

The medial clavicle epiphysis is a crucial indicator for bone age estimation (BAE) after hand maturation. This study aimed to develop machine learning (ML) and deep learning (DL) models for BAE based on medial clavicle CT images and evaluate the performance on normal and variant clavicles. This study retrospectively collected 1049 patients (mean± SD: 22.50±4.34 years) and split them into normal training and test sets, and variant training and test sets. An additional 53 variant clavicles were incorporated into the variant test set. The development stages of normal MCE were used to build a linear model and support vector machine (SVM) for BAE. The CT slices of MCE were automatically segmented and used to train DL models for automated BAE. Comparisons were performed by linear versus ML versus DL, and normal versus variant clavicles. Mean absolute error (MAE) and classification accuracy was the primary parameter of comparison. For BAE, the SVM had the best MAE of 1.73 years, followed by the commonly-used CNNs (1.77-1.93 years), the linear model (1.94 years), and the hybrid neural network CoAt Net (2.01 years). In DL models, SE Net 18 was the best-performing DL model with similar results to SVM in the normal test set and achieved an MAE of 2.08 years in the external variant test. For age classification, all the models exhibit superior performance in the classification of 18-, 20-, 21-, and 22-year thresholds with limited value in the 16-year threshold. Both ML and DL models produce desirable performance in BAE based on medial clavicle CT.

Validation of age estimation methods through the scapula in a Mediterranean non-adult population.

Age estimation methods through the metric study of bones in non-adult individuals are a frequently used tool in Biological and Forensic Anthropology; however, few methodological validation studies are available, given the limited number of adequate samples for their study. In this context, the aim of this work is to test the effectiveness of the methods of Fazekas and Kósa (1978), Saunders et al. (1993), Rissech and Black (2007) and Cardoso et al. (2017), for age estimation through the measurements of the scapula, in the non-adult osteological collection of the Anthropology Laboratory of the University of Granada (Spain), one of the most representative identified non-adult collections internationally. For this purpose, nine different measurements of the scapula were taken from a total of 149 non-adult individuals, applying the regression formulae proposed by the different authors and verifying their accuracy. Results have shown the existence of significant differences between the real age and the estimated age in almost all the tested methods, with degrees of error that differ from those obtained in the original study populations. Based on the obtained results, it is recommended to use the method of Cardoso et al. in postnatal individuals, always adjusting the degrees of error with those obtained in this work.

Radiological age assessment based on clavicle ossification in CT: enhanced accuracy through deep learning.

BACKGROUND: Radiological age assessment using reference studies is inherently limited in accuracy due to a finite number of assignable skeletal maturation stages. To overcome this limitation, we present a deep learning approach for continuous age assessment based on clavicle ossification in computed tomography (CT). METHODS: Thoracic CT scans were retrospectively collected from the picture archiving and communication system. Individuals aged 15.0 to 30.0 years examined in routine clinical practice were included. All scans were automatically cropped around the medial clavicular epiphyseal cartilages. A deep learning model was trained to predict a person's chronological age based on these scans. Performance was evaluated using mean absolute error (MAE). Model performance was compared to an optimistic human reader performance estimate for an established reference study method. RESULTS: The deep learning model was trained on 4,400 scans of 1,935 patients (training set: mean age = 24.2 years ± 4.0, 1132 female) and evaluated on 300 scans of 300 patients with a balanced age and sex distribution (test set: mean age = 22.5 years ± 4.4, 150 female). Model MAE was 1.65 years, and the highest absolute error was 6.40 years for females and 7.32 years for males. However, performance could be attributed to norm-variants or pathologic disorders. Human reader estimate MAE was 1.84 years and the highest absolute error was 3.40 years for females and 3.78 years for males. CONCLUSIONS: We present a deep learning approach for continuous age predictions using CT volumes highlighting the medial clavicular epiphyseal cartilage with performance comparable to the human reader estimate.

Bone age assessment based on different MRI modalities of the proximal humerus epiphysis: the comparisons of T1WI, T2WI, and PDWI.

Bone age assessment (BAA) is crucial in various fields, including legal proceedings, athletic competitions, and clinical medicine. However, the use of X-ray methods for age estimation without medical indication is subject to ethical debate, especially in forensic and athletic fields. The application of magnetic resonance imaging (MRI) with non-ionizing radiation can overcome this limitation in BAA. This study aimed to compare the application value of several MRI modalities of proximal humeral in BAA. A total of 468 patients with shoulder MRIs were retrospectively collected from a Chinese Han population aged 12-30 years (259 males and 209 females) for training and testing, including T1 weighted MRI (T1WI), T2 weighted MRI (T2WI), and Proton density weighted MRI (PDWI). Optimal regression models were established for age estimation, yielding mean absolute error (MAE) values below 2.0 years. The MAE values of T1WI were the lowest, with 1.700 years in males and 1.798 years in females. The area under the curve (AUC) and accuracy values of different MRI modalities of 16-year and 18-year thresholds were all around 0.9. For the 18-year threshold, T1WI outperformed T2WI and PDWI. In conclusion, the three MRI modalities of the proximal humerus can serve as reliable indicators for age assessment, while the T1WI performed better in age assessment and classification.

Forensic age estimation of the knee by post-mortem DR, CT, and MR imaging: a comparative study.

It is believed by many that reference data for age estimation purposes must be imaging-modality specific. A study from our department has however proven otherwise. We therefore found it interesting to investigate this further by looking at the level of agreement between different imaging modalities. The aim of this study was to investigate the level of agreement between the three radiological modalities, computed tomography (CT), magnetic resonance imaging (MRI), and digital radiography (DR), in assessing the ossification of the epiphyses of the knee. A total of 34 deceased individuals of 10-25 years of age, brought in for a medicolegal autopsy at our department, were scanned by CT, MRI, and DR. The ossification stages of the three bones of the right knee, distal femoral, proximal tibial, and proximal fibular epiphysis were assessed using the established combined staging method by Schmeling et al. and Kellinghaus et al. Analysis of the results by Cohen's weighted kappa showed a good agreement between CT and DR (K = 0.61-0.70), and MRI and DR (K = 0.68-0.79) but only moderate agreement between CT and MRI (K = 0.55-0.57). This leads us to conclude that different radiological images cannot be used interchangeably for age estimation purposes, so reference material needs to be imaging-modality specific. However, to make a more general conclusion research on a larger population is needed.

Body size and three estimates of skeletal age: Relationships with strength and motor performance among male soccer players 9-12 and 13-16 years.

OBJECTIVE: To estimate the associations between height, weight, and three estimates of skeletal age (SA) and the strength and motor performance of male soccer players in two chronological age (CA) groups, 9-12 (n = 60) and 13-16 (n = 52) years. METHODS: Height, weight, strength (grip), speed (5 m, 20 m sprints), acceleration (time at crossing 10 m in 20 m sprint), agility (figure-of-eight run), power (vertical jump), and endurance (intermittent shuttle run) were measured. SA was assessed with the TW2 RUS, TW3 RUS, and Fels methods; each SA was expressed as the standardized residual of the regression of SA on CA (SAsr). Hierarchical multiple regression was used. RESULTS: Body size accounted for ≥50% of the variance in grip strength in both CA groups; the body size × SAsr for each method and SAsr alone added little to the explained variance. Body size, body size × SAsr interactions, and SAsr per se with each method accounted for small percentages of variance in motor tasks among players 9-12 years, while body size explained a larger proportion of variance in motor tasks (except the endurance run) among players 13-16 years; body size × SAsr interactions for TW2 and TW3 more so than Fels added to the explained variances. For the endurance run, only SAsr per se with each method accounted for significant portions of the variance. CONCLUSION: Body size and the three estimates of SA significantly influenced strength and motor performance, but the explained variance varied between CA groups and among SA methods and performance tasks.

Deeplasia: deep learning for bone age assessment validated on skeletal dysplasias.

BACKGROUND: Skeletal dysplasias collectively affect a large number of patients worldwide. Most of these disorders cause growth anomalies. Hence, evaluating skeletal maturity via the determination of bone age (BA) is a useful tool. Moreover, consecutive BA measurements are crucial for monitoring the growth of patients with such disorders, especially for timing hormonal treatment or orthopedic interventions. However, manual BA assessment is time-consuming and suffers from high intra- and inter-rater variability. This is further exacerbated by genetic disorders causing severe skeletal malformations. While numerous approaches to automate BA assessment have been proposed, few are validated for BA assessment on children with skeletal dysplasias. OBJECTIVE: We present Deeplasia, an open-source prior-free deep-learning approach designed for BA assessment specifically validated on patients with skeletal dysplasias. MATERIALS AND METHODS: We trained multiple convolutional neural network models under various conditions and selected three to build a precise model ensemble. We utilized the public BA dataset from the Radiological Society of North America (RSNA) consisting of training, validation, and test subsets containing 12,611, 1,425, and 200 hand and wrist radiographs, respectively. For testing the performance of our model ensemble on dysplastic hands, we retrospectively collected 568 radiographs from 189 patients with molecularly confirmed diagnoses of seven different genetic bone disorders including achondroplasia and hypochondroplasia. A subset of the dysplastic cohort (149 images) was used to estimate the test-retest precision of our model ensemble on longitudinal data. RESULTS: The mean absolute difference of Deeplasia for the RSNA test set (based on the average of six different reference ratings) and dysplastic set (based on the average of two different reference ratings) were 3.87 and 5.84 months, respectively. The test-retest precision of Deeplasia on longitudinal data (2.74 months) is estimated to be similar to a human expert. CONCLUSION: We demonstrated that Deeplasia is competent in assessing the age and monitoring the development of both normal and dysplastic bones.

The Reliability of the Modified Fels Knee Skeletal Maturity System.

BACKGROUND: The recently described Modified Fels knee skeletal maturity system (mFels) has proven utility in prediction of ultimate lower extremity length in modern pediatric patients. mFels users evaluate chronological age, sex, and 7 anteroposterior knee radiographic parameters to produce a skeletal age estimate. We developed a free mobile application to minimize the learning curve of mFels radiographic parameter evaluation. We sought to identify the reliability of mFels for new users. METHODS: Five pediatric orthopaedic surgeons, 5 orthopaedic surgery residents, 3 pediatric orthopaedic nurse practitioners, and 5 medical students completely naïve to mFels each evaluated a set of 20 pediatric anteroposterior knee radiographs with the assistance of the (What's the Skeletal Maturity?) mobile application. They were not provided any guidance beyond the instructions and examples embedded in the app. The results of their radiographic evaluations and skeletal age estimates were compared with those of the mFels app developers. RESULTS: Averaging across participant groups, inter-rater reliability for each mFels parameter ranged from 0.73 to 0.91. Inter-rater reliability of skeletal age estimates was 0.98. Regardless of group, steady proficiency was reached by the seventh radiograph measured. CONCLUSIONS: mFels is a reliable means of skeletal maturity evaluation. No special instruction is necessary for first time users at any level to utilize the (What's the Skeletal Maturity?) mobile application, and proficiency in skeletal age estimation is obtained by the seventh radiograph. LEVEL OF EVIDENCE: Level II.

Skeletal maturation evaluation: which is the reliability of dental calcification Demirjian method versus hand-wrist X-ray in growing subjects? A systematic review.

OBJECTIVES: This systematic review aimed at evaluating the reliability of dental maturation (DM) according to Demirjian method compared to hand and wrist maturation (HWM) to assess skeletal maturity (SM) in growing subjects, to identify the teeth and the corresponding mineralisation stages related to the pubertal growth spurt (PGS). MATERIALS AND METHODS: PubMed, Scopus, and Web of Science were systematically searched until January 5th, 2024, to identify observational cross-sectional studies that assessed the reliability of Demirjian method compared to the HWM methods (i.e., Grave and Brown and Fishman) in growing subjects. The quality assessment was evaluated using the Joanna Briggs Institute (JBI) Critical Appraisal Checklist. RESULTS: Out of 136 papers suitable for title/abstract screening, 19 included studies. Of them, 17 papers showed the reliability of Demirjian DM method compared to HWM Fishman and Grave and Brown methods to assess SM in growing subjects. According to JBI Critical Appraisal Checklist, 12 papers were high-quality studies and 7 papers were medium-quality studies.  Conclusions: The mandibular second molar might be considered as the best indicator compared to other teeth and that the peak of growth occurs no earlier than stage F in females and stage G in males according to Demirjian method. Also, the mandibular canine might be analysed as indicator of SM in males, and results suggest that the peak of growth occurs no earlier than maturation stage F according to Demirjian method, only in male subjects. Further studies are needed to confirm these findings.

Evaluation of an objective staging system for assessment of cervical vertebral maturation.

BACKGROUND: The aim of this study was to evaluate an objective method for Cervical Vertebral Maturation (CVM) staging. METHODS: An initial sample of 647 Lateral Cephalometric Radiographs (LCR) were staged according to the CVM (Baccetti et al.) by 4 examiners. The final sample (n = 394) included LCR on which the staging of the 4 investigators matched. The objective staging was performed by a single operator. The sample was divided according to the maturational stages into pre-pubertal, pubertal and post-pubertal groups. Measurements were performed on the cervical vertebrae (C2, C3 and C4). The angle between posterior and superior borders for C3 and C4 was the Superior Wall Inclination Angle (SWIA). Concavity Depth (CD) for C2, C3 and C4, and Body Shape (BS) (ratio of width to height of C3 and C4). Measurements of the 3 groups were compared. RESULTS: Reliability of subjective staging was high (intra-observer reliability, 0.948; inter-observer reliability, 0.967). Good agreement was observed for the outcomes measured. Intra-observer reliability was good (0.918, 0.885 and 0.722 for CD, BS and SWIA, respectively). The same was for the inter-observer reliability results (0.902, 0.889 and 0.728 for CD, BS and SWIA, respectively). Significant differences were observed for mean values of SWIA and BS and median values of CD within maturational stage. Similar findings were observed when the outcomes were compared at different phases (P < 0.001). CONCLUSIONS: A standardized, objective staging system using linear, angular measurements and ratios was applied for the determination of cervical vertebral maturation.

Is hand-wrist radiography still necessary in orthodontic treatment planning?

OBJECTIVES: The aim of our study is to compare the relationship between hand-wrist and cervical vertebra maturation stages with chronological age and to investigate the effect of malocclusion type on the relationship between these methods. MATERIALS AND METHODS: Hand-wrist and cephalometric radiographs of 1000 patients (526 females, 474 males) with a mean age of 13.41 ± 1.83 were analyzed. The methods of Bacetti et al. were used for the cervical vertebra maturation stage, and Björk, Grave and Brown's methods were used for the hand-wrist maturation stage. One-way ANOVA test was applied to compare skeletal classes between them. Tukey post hoc test was used to determine the differences. The relationship between the malocclusion type, cervical vertebra and hand-wrist maturation stages was evaluated with the Spearman correlation test. RESULTS: Spearman's correlation coefficient was 0.831, 0.831 and 0.760 in Class I, II and III females, respectively. In males, it was calculated as 0.844, 0.889 and 0.906, respectively. When sex and malocclusion were not differentiated, the correlation was found to be 0.887. All were statistically significant (P < 0.001). The highest correlation was observed in class III males, while the lowest was found in class III females. CONCLUSION: Cervical vertebrae can be used safely to assess pubertal spurt without hand-wrist radiography. Diagnosing growth and development stages from cephalometric images is important in reducing additional workload and preventing radiation risk.

Greulich and Pyle atlas: a non-reliable skeletal maturity assessment method in the North Indian population.

Forensic age assessments are crucial in the evaluation of criminal responsibility and preventing false age claims. Of all the methods available, the Greulich and Pyle (GP) atlas is most commonly used for age estimation purposes. Therefore, the current study sought to analyze the reliability and applicability of the GP standard and, additionally, to determine any possible association between the socioeconomic status (SES), food habits, and estimated skeletal maturity in the North Indian population. The study included 627 (334 males and 293 females) healthy children up to 19 years of age with varying SES and food habits. The skeletal age (SA) was estimated by three different evaluators using the GP atlas. The chronological mean age (CA) and SA were compared in different age cohorts. A paired t-test and a Pearson chi-square test were applied to show the difference between CA and estimated SA and the association of skeletal maturity with SES and food habits. The estimated skeletal age in males was retarded by 0.142 years or 1.72 months (p ≤ 0.05), whereas in females, it was retarded by 0.259 years or 3.12 months (p ≤ 0.05). In males, the GP method has significantly underestimated SA in age cohorts 3-4, 4-5, 6-7, 7-8, 8-9, and 12-13, whereas it overestimated in 10-11 and 18-19 years. However, in females, the SA was significantly underestimated in age groups 10-11, 12-13, and 14-15, respectively. Estimated skeletal maturity had no significant association with SES and food habits. The current study concludes that the GP atlas may not be applicable to North India's population. The observed difference in assessed skeletal maturity may be due to geographical region, genetics, hormonal effects, etc., which require further investigation. Hence, population-specific standards are necessary to determine the bone age of Indian children accurately.

Age Estimation Based on Computed Tomography Analysis of the Scapula.

Background and Objectives: Age estimation from skeletal remains and in living individuals is an important issue for human identification, and also plays a critical role in judicial proceedings for migrants. Forensic analysis of ossification centers is the main evaluation method for age estimation, and ossification degree can be determined using computed tomography analysis. The purpose of this study is to investigate the applicability of CT (computed tomography) in the analysis of left scapula ossification centers, for forensic age estimation in Turkish society. Materials and Methods: We analyzed six ossification centers of the left scapula and these ossification centers are the coracoid, subcoracoid, coracoid apex, acromial, glenoid, and inferior angle ossification centers. A pediatric radiologist analyzed these six ossification centers of the scapula by using a staging method defined by Schmeling et al. in 2004. Two months after the first assessment, 20 randomly selected cases was reanalyzed by the first observer and by another pediatric radiologist. Correlation between the age and ossification stage was assessed using Spearman's nonparametric correlation test. Linear regression analysis was performed using a backwards model. Cohen's kappa coefficient was used for evaluating interobserver and intraobserver variability. Results: In this retrospective study, 397 (248 male and 149 female) cases were evaluated. Ages ranged between 7.1 and 30.9. The mean age was 19.83 ± 6.49. We determined a positive significant correlation between the age and the ossification stages of ossification centers analyzed in both sexes. In each ossification center, except inferior angle, all of the stage 1 and 2 cases in both sexes were under 18 years old. Intraobserver and interobserver evaluations showed that reproducibility and consistency of the method was relatively good. Conclusions: The present study indicated that CT analysis of scapula ossification centers might be helpful in forensic age assessment of living individuals and dry bones.

Bone Age Determination of Epiphyseal Fusion at Knee Joint and Its Correlation with Chronological Age.

Background and Objectives: Bone age determination is a valuable method for forensic and disaster identifications of unknown human remains, as well as for medical and surgical procedural purposes. This retrospective research study aimed to determine the age based on epiphyseal fusion stages and investigate differences related to gender. Materials and Methods: X-rays of the knee were collected from medical imaging centers in hospitals in the south of Jordan and examined by two observers who determined the bone epiphyseal phase of closure for the femur, tibia, and fibula bone ends close to the knee based on a three-stage classification. Results: The main results revealed that females showed earlier epiphyseal union (Stage II) at the lower end of the femur and the upper ends of the tibia and fibula compared to males. In males, the start of complete union (Stage III) at knee bones was seen at the age of 17-18 years, while in females, it was seen at the age of 16-17 years. Additionally, knee bones showed complete union in 100% of males and females in the age groups 21-22 years and 20-21 years, respectively. Although females showed an earlier start and end of epiphyseal complete union than males, analysis of collected data showed no significant age differences between males and females at the three stages of epiphyseal union of the knee bones. Conclusions: Findings of the radiographic analysis of bone epiphyseal fusion at the knee joint are a helpful method for chronological age determination. This study supports the gender and ethnicity variation among different geographical locations. Studies with a high sample number would be needed to validate our findings.

The utility of sternum in creating a biological profile: A review and future directions.

This review delves into the forensic utility of the sternum in creating a biological profile, focusing on sex, stature, and age estimation. Emphasizing the sternum's significance in challenging scenarios, the study supports the combined length of the manubrium and sternal body as a crucial indicator in sex and stature estimation. However, it highlights the need for caution in applying findings across diverse populations and questions the reliability of Hyrtl's law. Age estimation, primarily based on morphological changes and ossification ages, is explored, with one study showing promise but requiring further validation. While acknowledging the sternum's advantages, the review underscores potential limitations and the absence of specific studies on ancestry estimation, leaving this aspect open for future research. In conclusion, the review provides a comprehensive overview of the sternum's forensic applications, urging continued research to enhance accuracy and applicability.

Application of Medical Statistical and Machine Learning Methods in the Age Estimation of Living Individuals.

In the study of age estimation in living individuals, a lot of data needs to be analyzed by mathematical statistics, and reasonable medical statistical methods play an important role in data design and analysis. The selection of accurate and appropriate statistical methods is one of the key factors affecting the quality of research results. This paper reviews the principles and applicable principles of the commonly used medical statistical methods such as descriptive statistics, difference analysis, consistency test and multivariate statistical analysis, as well as machine learning methods such as shallow learning and deep learning in the age estimation research of living individuals, and summarizes the relevance and application prospects between medical statistical methods and machine learning methods. This paper aims to provide technical guidance for the age estimation research of living individuals to obtain more scientific and accurate results.

Adults Ischium Age Estimation Based on Deep Learning and 3D CT Reconstruction.

OBJECTIVES: To develop a deep learning model for automated age estimation based on 3D CT reconstructed images of Han population in western China, and evaluate its feasibility and reliability. METHODS: The retrospective pelvic CT imaging data of 1 200 samples (600 males and 600 females) aged 20.0 to 80.0 years in western China were collected and reconstructed into 3D virtual bone models. The images of the ischial tuberosity feature region were extracted to create sex-specific and left/right site-specific sample libraries. Using the ResNet34 model, 500 samples of different sexes were randomly selected as training and verification set, the remaining samples were used as testing set. Initialization and transfer learning were used to train images that distinguish sex and left/right site. Mean absolute error (MAE) and root mean square error (RMSE) were used as primary indicators to evaluate the model. RESULTS: Prediction results varied between sexes, with bilateral models outperformed left/right unilateral ones, and transfer learning models showed superior performance over initial models. In the prediction results of bilateral transfer learning models, the male MAE was 7.74 years and RMSE was 9.73 years, the female MAE was 6.27 years and RMSE was 7.82 years, and the mixed sexes MAE was 6.64 years and RMSE was 8.43 years. CONCLUSIONS: The skeletal age estimation model, utilizing ischial tuberosity images of Han population in western China and employing the ResNet34 combined with transfer learning, can effectively estimate adult ischium age.

Age Estimation by Machine Learning and CT-Multiplanar Reformation of Cranial Sutures in Northern Chinese Han Adults.

OBJECTIVES: To establish age estimation models of northern Chinese Han adults using cranial suture images obtained by CT and multiplanar reformation (MPR), and to explore the applicability of cranial suture closure rule in age estimation of northern Chinese Han population. METHODS: The head CT samples of 132 northern Chinese Han adults aged 29-80 years were retrospectively collected. Volume reconstruction (VR) and MPR were performed on the skull, and 160 cranial suture tomography images were generated for each sample. Then the MPR images of cranial sutures were scored according to the closure grading criteria, and the mean closure grades of sagittal suture, coronal sutures (both left and right) and lambdoid sutures (both left and right) were calculated respectively. Finally taking the above grades as independent variables, the linear regression model and four machine learning models for age estimation (gradient boosting regression, support vector regression, decision tree regression and Bayesian ridge regression) were established for northern Chinese Han adults age estimation. The accuracy of each model was evaluated. RESULTS: Each cranial suture closure grade was positively correlated with age and the correlation of sagittal suture was the highest. All four machine learning models had higher age estimation accuracy than linear regression model. The support vector regression model had the highest accuracy among the machine learning models with a mean absolute error of 9.542 years. CONCLUSIONS: The combination of skull CT-MPR and machine learning model can be used for age estimation in northern Chinese Han adults, but it is still necessary to combine with other adult age estimation indicators in forensic practice.