Forensic bone age assessment of hand and wrist joint MRI images in Chinese han male adolescents based on deep convolutional neural networks.

In Chinese criminal law, the ages of 12, 14, 16, and 18 years old play a significant role in the determination of criminal responsibility. In this study, we developed an epiphyseal grading system based on magnetic resonance image (MRI) of the hand and wrist for the Chinese Han population and explored the feasibility of employing deep learning techniques for bone age assessment based on MRI of the hand and wrist. This study selected 282 Chinese Han Chinese males aged 6.0-21.0 years old. In the course of our study, we proposed a novel deep learning model for extracting and enhancing MRI hand and wrist bone features to enhance the prediction of target MRI hand and wrist bone age and achieve precise classification of the target MRI and regression of bone age. The evaluation metric for the classification model including precision, specificity, sensitivity, and accuracy, while the evaluation metrics chosen for the regression model are MAE. The epiphyseal grading was used as a supervised method, which effectively solved the problem of unbalanced sample distribution, and the two experts showed strong consistency in the epiphyseal plate grading process. In the classification results, the accuracy in distinguishing between adults and minors was 91.1%, and the lowest accuracy in the three minor classifications (12, 14, and 16 years of age) was 94.6%, 91.1% and 96.4%, respectively. The MAE of the regression results was 1.24 years. In conclusion, the deep learning model proposed enabled the age assessment of hand and wrist bones based on MRI.

Deep learning-based automated bone age estimation for Saudi patients on hand radiograph images: a retrospective study.

PURPOSE: In pediatric medicine, precise estimation of bone age is essential for skeletal maturity evaluation, growth disorder diagnosis, and therapeutic intervention planning. Conventional techniques for determining bone age depend on radiologists' subjective judgments, which may lead to non-negligible differences in the estimated bone age. This study proposes a deep learning-based model utilizing a fully connected convolutional neural network(CNN) to predict bone age from left-hand radiographs. METHODS: The data set used in this study, consisting of 473 patients, was retrospectively retrieved from the PACS (Picture Achieving and Communication System) of a single institution. We developed a fully connected CNN consisting of four convolutional blocks, three fully connected layers, and a single neuron as output. The model was trained and validated on 80% of the data using the mean-squared error as a cost function to minimize the difference between the predicted and reference bone age values through the Adam optimization algorithm. Data augmentation was applied to the training and validation sets yielded in doubling the data samples. The performance of the trained model was evaluated on a test data set (20%) using various metrics including, the mean absolute error (MAE), median absolute error (MedAE), root-mean-squared error (RMSE), and mean absolute percentage error (MAPE). The code of the developed model for predicting the bone age in this study is available publicly on GitHub at https://github.com/afiosman/deep-learning-based-bone-age-estimation . RESULTS: Experimental results demonstrate the sound capabilities of our model in predicting the bone age on the left-hand radiographs as in the majority of the cases, the predicted bone ages and reference bone ages are nearly close to each other with a calculated MAE of 2.3 [1.9, 2.7; 0.95 confidence level] years, MedAE of 2.1 years, RMAE of 3.0 [1.5, 4.5; 0.95 confidence level] years, and MAPE of 0.29 (29%) on the test data set. CONCLUSION: These findings highlight the usability of estimating the bone age from left-hand radiographs, helping radiologists to verify their own results considering the margin of error on the model. The performance of our proposed model could be improved with additional refining and validation.

Multi-Branch Attention Learning for Bone Age Assessment with Ambiguous Label.

Bone age assessment (BAA) is a typical clinical technique for diagnosing endocrine and metabolic diseases in children's development. Existing deep learning-based automatic BAA models are trained on the Radiological Society of North America dataset (RSNA) from Western populations. However, due to the difference in developmental process and BAA standards between Eastern and Western children, these models cannot be applied to bone age prediction in Eastern populations. To address this issue, this paper collects a bone age dataset based on the East Asian populations for model training. Nevertheless, it is laborious and difficult to obtain enough X-ray images with accurate labels. In this paper, we employ ambiguous labels from radiology reports and transform them into Gaussian distribution labels of different amplitudes. Furthermore, we propose multi-branch attention learning with ambiguous labels network (MAAL-Net). MAAL-Net consists of a hand object location module and an attention part extraction module to discover the informative regions of interest (ROIs) based only on image-level labels. Extensive experiments on both the RSNA dataset and the China Bone Age (CNBA) dataset demonstrate that our method achieves competitive results with the state-of-the-arts, and performs on par with experienced physicians in children's BAA tasks.

An Automated TW3-RUS Bone Age Assessment Method with Ordinal Regression-Based Determination of Skeletal Maturity.

The assessment of bone age is important for evaluating child development, optimizing the treatment for endocrine diseases, etc. And the well-known Tanner-Whitehouse (TW) clinical method improves the quantitative description of skeletal development based on setting up a series of distinguishable stages for each bone individually. However, the assessment is affected by rater variability, which makes the assessment result not reliable enough in clinical practice. The main goal of this work is to achieve a reliable and accurate skeletal maturity determination by proposing an automated bone age assessment method called PEARLS, which is based on the TW3-RUS system (analysis of the radius, ulna, phalanges, and metacarpal bones). The proposed method comprises the point estimation of anchor (PEA) module for accurately localizing specific bones, the ranking learning (RL) module for producing a continuous stage representation of each bone by encoding the ordinal relationship between stage labels into the learning process, and the scoring (S) module for outputting the bone age directly based on two standard transform curves. The development of each module in PEARLS is based on different datasets. Finally, corresponding results are presented to evaluate the system performance in localizing specific bones, determining the skeletal maturity stage, and assessing the bone age. The mean average precision of point estimation is 86.29%, the average stage determination precision is 97.33% overall bones, and the average bone age assessment accuracy is 96.8% within 1 year for the female and male cohorts.

Bone Age Reading by DXA Images should not Replace Bone Age Reading by X-ray Images.

X-ray image of the hand is the most used technique to estimate bone age in children. For the analysis of bone mineral density using DXA in children, bone age may help to adjust such measurement in some cases. During image acquisition in DXA, an anteroposterior image of the hand may be acquired and used to evaluate bone age but few studies have evaluated the agreement between conventional X-ray and DXA images. The aim of the study was to determine bone age estimation agreement between conventional X-ray images and DXA in children and adolescents aged 5 to 16 years of age. We performed an analytical cross-sectional study of 711 healthy subjects. Subject´s bone age, both in conventional X-ray, and DXA images were read independently by two expert evaluators blinded for chronological age. Intraobserver and inter-observer reproducibility were evaluated using Intraclass Correlation Coefficient (ICC), and the agreement between bone age estimations made by both evaluators was analyzed using ICC and Bland-Altman analysis. General agreement between techniques measured through ICC was 0.99 with a mean difference of 6 months between techniques being older the ages obtained by DXA. The agreement limits were around ±2 years, which means that 95% of all differences between techniques were covered within this range. We found a high level of ICC agreement in bone age readings from X-ray and DXA images although we observed overestimation of bone age measurements in DXA. Differences between techniques were greater in women than in men, especially at the ages corresponding to puberty. Bone age measurement in DXA images appears not to be reliable; hence it should be suggested to perform conventional radiography of the hand to assess bone age taking into account that X-ray images have better resolution.

Assessing the Bone Age of Children in an Automatic Manner Newborn to 18 Years Range.

Bone age assessment (BAA) is a radiological process to identify the growth disorders in children. Although this is a frequent task for radiologists, it is cumbersome. The objective of this study is to assess the bone age of children from newborn to 18 years old in an automatic manner through computer vision methods including histogram of oriented gradients (HOG), local binary pattern (LBP), and scale invariant feature transform (SIFT). Here, 442 left-hand radiographs are applied from the University of Southern California (USC) hand atlas. In this experiment, for the first time, HOG-LBP-dense SIFT features with background subtraction are applied to assess the bone age of the subject group. For this purpose, features are extracted from the carpal and epiphyseal regions of interest (ROIs). The SVM and 5-fold cross-validation are used for classification. The accuracy of female radiographs is 73.88% and of the male is 68.63%. The mean absolute error is 0.5 years for both genders' radiographs. The accuracy a within 1-year range is 95.32% for female and 96.51% for male radiographs. The accuracy within a 2-year range is 100% and 99.41% for female and male radiographs, respectively. The Cohen's kappa statistical test reveals that this proposed approach, Cohen's kappa coefficients are 0.71 for female and 0.66 for male radiographs, p value < 0.05, is in substantial agreement with the bone age assessed by experienced radiologists within the USC dataset. This approach is robust and easy to implement, thus, qualified for computer-aided diagnosis (CAD). The reduced processing time and number of ROIs facilitate BAA.

A Deep Automated Skeletal Bone Age Assessment Model with Heterogeneous Features Learning.

Skeletal bone age assessment is a widely used standard procedure in both disease detection and growth prediction for children in endocrinology. Conventional manual assessment methods mainly rely on personal experience in observing X-ray images of left hand and wrist to calculate bone age, which show some intrinsic limitations from low efficiency to unstable accuracy. To address these problems, some automated methods based on image processing or machine learning have been proposed, while their performances are not satisfying enough yet in assessment accuracy. Motivated by the remarkable success of deep learning (DL) techniques in the fields of image classification and speech recognition, we develop a deep automated skeletal bone age assessment model based on convolutional neural networks (CNNs) and support vector regression (SVR) using multiple kernel learning (MKL) algorithm to process heterogeneous features in this paper. This deep framework has been constructed, not only exploring the X-ray images of hand and twist but also some other heterogeneous information like race and gender. The experiment results prove its better performance with higher bone age assessment accuracy on two different data sets compared with the state of the art, indicating that the fused heterogeneous features provide a better description of the degree of bones' maturation.

Region fine-grained attention network for accurate bone age assessment.

Bone age assessment plays a vital role in monitoring the growth and development of adolescents. However, it is still challenging to obtain precise bone age from hand radiography due to these problems: 1) Hand bone varies greatly and is always masked by the background; 2) the hand bone radiographs with successive ages offer high similarity. To solve such issues, a region fine-grained attention network (RFGA-Net) was proposed for bone age assessment, where the region aware attention (RAA) module was developed to distinguish the skeletal regions from the background by modeling global spatial dependency; then the fine-grained feature attention (FFA) module was devised to identify similar bone radiographs by recognizing critical fine-grained feature regions. The experimental results demonstrate that the proposed RFGA-Net shows the best performance on the Radiological Society of North America (RSNA) pediatric bone dataset, achieving the mean absolute error (MAE) of 3.34 and the root mean square error (RMSE) of 4.02, respectively.

Bone age assessment by multi-granularity and multi-attention feature encoding.

Background: Bone age assessment (BAA) is crucial for the diagnosis of growth disorders and the optimization of treatments. However, the random error caused by different observers' experiences and the low consistency of repeated assessments harms the quality of such assessments. Thus, automated assessment methods are needed. Methods: Previous research has sought to design localization modules in a strongly or weakly supervised fashion to aggregate part regions to better recognize subtle differences. Conversely, we sought to efficiently deliver information between multi-granularity regions for fine-grained feature learning and to directly model long-distance relationships for global understanding. The proposed method has been named the "Multi-Granularity and Multi-Attention Net (2M-Net)". Specifically, we first applied the jigsaw method to generate related tasks emphasizing regions with different granularities, and we then trained the model on these tasks using a hierarchical sharing mechanism. In effect, the training signals from the extra tasks created as an inductive bias, enabling 2M-Net to discover task relatedness without the need for annotations. Next, the self-attention mechanism acted as a plug-and-play module to effectively enhance the feature representation capabilities. Finally, multi-scale features were applied for prediction. Results: A public data set of 14,236 hand radiographs, provided by the Radiological Society of North America (RSNA), was used to develop and validate 2M-Net. In the public benchmark testing, the mean absolute error (MAE) between the bone age estimates of the model and of the reviewer was 3.98 months (3.89 months for males and 4.07 months for females). Conclusions: By using the jigsaw method to construct a multi-task learning strategy and inserting the self-attention module for efficient global modeling, we established 2M-Net, which is comparable to the previous best method in terms of performance.

Supervised representation learning based on various levels of pediatric radiographic views for transfer learning.

Transfer learning plays a pivotal role in addressing the paucity of data, expediting training processes, and enhancing model performance. Nonetheless, the prevailing practice of transfer learning predominantly relies on pre-trained models designed for the natural image domain, which may not be well-suited for the medical image domain in grayscale. Recognizing the significance of leveraging transfer learning in medical research, we undertook the construction of class-balanced pediatric radiograph datasets collectively referred to as PedXnets, grounded in radiographic views using the pediatric radiographs collected over 24 years at Asan Medical Center. For PedXnets pre-training, approximately 70,000 X-ray images were utilized. Three different pre-training weights of PedXnet were constructed using Inception V3 for various radiation perspective classifications: Model-PedXnet-7C, Model-PedXnet-30C, and Model-PedXnet-68C. We validated the transferability and positive effects of transfer learning of PedXnets through pediatric downstream tasks including fracture classification and bone age assessment (BAA). The evaluation of transfer learning effects through classification and regression metrics showed superior performance of Model-PedXnets in quantitative assessments. Additionally, visual analyses confirmed that the Model-PedXnets were more focused on meaningful regions of interest.

Multi-scale multi-reception attention network for bone age assessment in X-ray images.

Bone age assessment plays a significant role in estimating bone maturity. However, radiograph/X-ray images of hand bones contain a large amount of redundant information. Some detection or segmentation based methods have recently been proposed to solve this issue. These network structures are often of high complexity and might require extra annotations, which make them less applicable in practice. In this paper, we present a Multi-scale Multi-reception Attention Net (MMANet), which combines a novel Multi-scale Multi-reception Complement Attention (MMCA) network and a graph attention module with a ResNet backbone to enhance the feature representation of key regions and suppress the influence of background regions to achieve significant performance improvement. Experimental results show our MMANet is able to accurately detect key regions and achieves 3.88 mean absolute error (MAE) on the RSNA 2017 Paediatric Bone Age Challenge dataset. Our method, without explicit modelling of anatomical information, outperforms the current state-of-the-art method (MAE=3.91) by 0.03 (months) which requires extra annotations. Code is available at https://github.com/yzc1122333/BoneAgeAss.

Automated Skeletal Bone Age Assessment with Two-Stage Convolutional Transformer Network Based on X-ray Images.

Human skeletal development is continuous and staged, and different stages have various morphological characteristics. Therefore, bone age assessment (BAA) can accurately reflect the individual's growth and development level and maturity. Clinical BAA is time consuming, highly subjective, and lacks consistency. Deep learning has made considerable progress in BAA in recent years by effectively extracting deep features. Most studies use neural networks to extract global information from input images. However, clinical radiologists are highly concerned about the ossification degree in some specific regions of the hand bones. This paper proposes a two-stage convolutional transformer network to improve the accuracy of BAA. Combined with object detection and transformer, the first stage mimics the bone age reading process of the pediatrician, extracts the hand bone region of interest (ROI) in real time using YOLOv5, and proposes hand bone posture alignment. In addition, the previous information encoding of biological sex is integrated into the feature map to replace the position token in the transformer. The second stage extracts features within the ROI by window attention, interacts between different ROIs by shifting the window attention to extract hidden feature information, and penalizes the evaluation results using a hybrid loss function to ensure its stability and accuracy. The proposed method is evaluated on the data from the Pediatric Bone Age Challenge organized by the Radiological Society of North America (RSNA). The experimental results show that the proposed method achieves a mean absolute error (MAE) of 6.22 and 4.585 months on the validation and testing sets, respectively, and the cumulative accuracy within 6 and 12 months reach 71% and 96%, respectively, which is comparable to the state of the art, markedly reducing the clinical workload and realizing rapid, automatic, and high-precision assessment.

Bone age assessment from articular surface and epiphysis using deep neural networks.

Bone age assessment is of great significance to genetic diagnosis and endocrine diseases. Traditional bone age diagnosis mainly relies on experienced radiologists to examine the regions of interest in hand radiography, but it is time-consuming and may even lead to a vast error between the diagnosis result and the reference. The existing computer-aided methods predict bone age based on general regions of interest but do not explore specific regions of interest in hand radiography. This paper aims to solve such problems by performing bone age prediction on the articular surface and epiphysis from hand radiography using deep convolutional neural networks. The articular surface and epiphysis datasets are established from the Radiological Society of North America (RSNA) pediatric bone age challenge, where the specific feature regions of the articular surface and epiphysis are manually segmented from hand radiography. Five convolutional neural networks, i.e., ResNet50, SENet, DenseNet-121, EfficientNet-b4, and CSPNet, are employed to improve the accuracy and efficiency of bone age diagnosis in clinical applications. Experiments show that the best-performing model can yield a mean absolute error (MAE) of 7.34 months on the proposed articular surface and epiphysis datasets, which is more accurate and fast than the radiologists. The project is available at https://github.com/YameiDeng/BAANet/, and the annotated dataset is also published at https://doi.org/10.5281/zenodo.7947923.

A real-time automated bone age assessment system based on the RUS-CHN method.

Background: Bone age is the age of skeletal development and is a direct indicator of physical growth and development in children. Most bone age assessment (BAA) systems use direct regression with the entire hand bone map or first segmenting the region of interest (ROI) using the clinical a priori method and then deriving the bone age based on the characteristics of the ROI, which takes more time and requires more computation. Materials and methods: Key bone grades and locations were determined using three real-time target detection models and Key Bone Search (KBS) post-processing using the RUS-CHN approach, and then the age of the bones was predicted using a Lightgbm regression model. Intersection over Union (IOU) was used to evaluate the precision of the key bone locations, while the mean absolute error (MAE), the root mean square error (RMSE), and the root mean squared percentage error (RMSPE) were used to evaluate the discrepancy between predicted and true bone age. The model was finally transformed into an Open Neural Network Exchange (ONNX) model and tested for inference speed on the GPU (RTX 3060). Results: The three real-time models achieved good results with an average (IOU) of no less than 0.9 in all key bones. The most accurate outcomes for the inference results utilizing KBS were a MAE of 0.35 years, a RMSE of 0.46 years, and a RMSPE of 0.11. Using the GPU RTX3060 for inference, the critical bone level and position inference time was 26 ms. The bone age inference time was 2 ms. Conclusions: We developed an automated end-to-end BAA system that is based on real-time target detection, obtaining key bone developmental grade and location in a single pass with the aid of KBS, and using Lightgbm to obtain bone age, capable of outputting results in real-time with good accuracy and stability, and able to be used without hand-shaped segmentation. The BAA system automatically implements the entire process of the RUS-CHN method and outputs information on the location and developmental grade of the 13 key bones of the RUS-CHN method along with the bone age to assist the physician in making judgments, making full use of clinical a priori knowledge.

Bone age recognition based on mask R-CNN using xception regression model.

Background and Objective: Bone age detection plays an important role in medical care, sports, judicial expertise and other fields. Traditional bone age identification and detection is according to manual interpretation of X-ray images of hand bone by doctors. This method is subjective and requires experience, and has certain errors. Computer-aided detection can effectually enhance the validity of medical diagnosis, especially with the fast development of machine learning and neural network, the method of bone age recognition using machine learning has gradually become the focus of research, which has the advantages of simple data pretreatment, good robustness and high recognition accuracy. Methods: In this paper, the hand bone segmentation network based on Mask R-CNN was proposed to segment the hand bone area, and the segmented hand bone region was directly input into the regression network for bone age evaluation. The regression network is using an enhancd network Xception of InceptionV3. After the output of Xception, the convolutional block attention module is connected to refine the feature mapping from channel and space to obtain more effective features. Results: According to the experimental results, the hand bone segmentation network model based on Mask R-CNN can segment the hand bone region and eliminate the interference of redundant background information. The average Dice coefficient on the verification set is 0.976. The mean absolute error of predicting bone age on our data set was only 4.97 months, which exceeded the accuracy of most other bone age assessment methods. Conclusion: Experiments show that the accuracy of bone age assessment can be enhancd by using the Mask R-CNN-based hand bone segmentation network and the Xception bone age regression network to form a model, which can be well applied to actual clinical bone age assessment.

Bone age assessment based on deep neural networks with annotation-free cascaded critical bone region extraction.

Bone age assessment (BAA) from hand radiographs is crucial for diagnosing endocrinology disorders in adolescents and supplying therapeutic investigation. In practice, due to the conventional clinical assessment being a subjective estimation, the accuracy of BAA relies highly on the pediatrician's professionalism and experience. Recently, many deep learning methods have been proposed for the automatic estimation of bone age and had good results. However, these methods do not exploit sufficient discriminative information or require additional manual annotations of critical bone regions that are important biological identifiers in skeletal maturity, which may restrict the clinical application of these approaches. In this research, we propose a novel two-stage deep learning method for BAA without any manual region annotation, which consists of a cascaded critical bone region extraction network and a gender-assisted bone age estimation network. First, the cascaded critical bone region extraction network automatically and sequentially locates two discriminative bone regions via the visual heat maps. Second, in order to obtain an accurate BAA, the extracted critical bone regions are fed into the gender-assisted bone age estimation network. The results showed that the proposed method achieved a mean absolute error (MAE) of 5.45 months on the public dataset Radiological Society of North America (RSNA) and 3.34 months on our private dataset.

Bone age assessment from lateral cephalograms using deep learning algorithms in the Indian population.

Purpose: The assessment of bone age has applications in a wide variety of fields: from orthodontics to immigration. The traditional non-automated methods are time-consuming and subject to inter- and intra-observer variability. This is the first study of its kind done on the Indian population. In this study, we analyse different pre-processing techniques and architectures to determine the degree of maturation (i.e. cervical vertebral maturation [CVM]) from cephalometric radiographs using machine learning algorithms. Methods: Cephalometric radiographs-labelled with the correct CVM stage using Baccetti et al. method-from 383 individuals aged between 10 and 36 years were used in the study. Data expansion and in-place data augmentation were used to handle high data imbalances. Different pre-processing techniques like Sobel filters and canny edge detectors were employed. Several deep learning convolutional neural network (CNN) architectures along with numerous pre-trained models like ResNet-50 and VGG-19 were analysed for their efficacy on the dataset. Results: Models with 6 and 8 convolutional layers trained on 64 × 64-size grayscale images trained the fastest and achieved the highest accuracy of 94%. Pre-trained ResNet-50 with the first 49 layers frozen and VGG-19 with 10 layers frozen to training had remarkable performances on the dataset with accuracies of 91% and 89%, respectively. Conclusions: Custom deep CNN models with 6-8 layers on 64 × 64-sized greyscale images were successfully used to achieve high accuracies to classify the majority classes. This study is a launchpad in the development of an automated method for bone age assessment from lateral cephalograms for clinical purposes.

SMANet: multi-region ensemble of convolutional neural network model for skeletal maturity assessment.

Background: Bone age assessment (BAA) is a crucial research topic in pediatric radiology. Interest in the development of automated methods for BAA is increasing. The current BAA algorithms based on deep learning have displayed the following deficiencies: (I) most methods involve end-to-end prediction, lacking integration with clinically interpretable methods; (II) BAA methods exhibit racial and geographical differences. Methods: A novel, automatic skeletal maturity assessment (SMA) method with clinically interpretable methods was proposed based on a multi-region ensemble of convolutional neural networks (CNNs). This method predicted skeletal maturity scores and thus assessed bone age by utilizing left-hand radiographs and key regional patches of clinical concern. Results: Experiments included 4,861 left-hand radiographs from the database of Beijing Jishuitan Hospital and revealed that the mean absolute error (MAE) was 31.4±0.19 points (skeletal maturity scores) and 0.45±0.13 years (bone age) for the carpal bones-series and 29.9±0.21 points and 0.43±0.17 years, respectively, for the radius, ulna, and short (RUS) bones series based on the Tanner-Whitehouse 3 (TW3) method. Conclusions: The proposed automatic SMA method, which was without racial and geographical influence, is a novel, automatic method for assessing childhood bone development by utilizing skeletal maturity. Furthermore, it provides a comparable performance to endocrinologists, with greater stability and efficiency.

Clinical application of artificial intelligence in longitudinal image analysis of bone age among GHD patients.

Objective: This study aims to explore the clinical value of artificial intelligence (AI)-assisted bone age assessment (BAA) among children with growth hormone deficiency (GHD). Methods: A total of 290 bone age (BA) radiographs were collected from 52 children who participated in the study at Sun Yat-sen Memorial Hospital between January 2016 and August 2017. Senior pediatric endocrinologists independently evaluated BA according to the China 05 (CH05) method, and their consistent results were regarded as the gold standard (GS). Meanwhile, two junior pediatric endocrinologists were asked to assessed BA both with and without assistance from the AI-based BA evaluation system. Six months later, around 20% of the images assessed by the junior pediatric endocrinologists were randomly selected to be re-evaluated with the same procedure half a year later. Root mean square error (RMSE), mean absolute error (MAE), accuracy, and Bland-Altman plots were used to compare differences in BA. The intra-class correlation coefficient (ICC) and one-way repeated ANOVA were used to assess inter- and intra-observer variabilities in BAA. A boxplot of BA evaluated by different raters during the course of treatment and a mixed linear model were used to illustrate inter-rater effect over time. Results: A total of 52 children with GHD were included, with mean chronological age and BA by GS of 6.64 ± 2.49 and 5.85 ± 2.30 years at baseline, respectively. After incorporating AI assistance, the performance of the junior pediatric endocrinologists improved (P < 0.001), with MAE and RMSE both decreased by more than 1.65 years (Rater 1: ΔMAE = 1.780, ΔRMSE = 1.655; Rater 2: ΔMAE = 1.794, ΔRMSE = 1.719), and accuracy increasing from approximately 10% to over 91%. The ICC also increased from 0.951 to 0.990. During GHD treatment (at baseline, 6-, 12-, 18-, and 24-months), the difference decreased sharply when AI was applied. Furthermore, a significant inter-rater effect (P = 0.002) also vanished upon AI involvement. Conclusion: AI-assisted interpretation of BA can improve accuracy and decrease variability in results among junior pediatric endocrinologists in longitudinal cohort studies, which shows potential for further clinical application.

Automated Bone Age Assessment: A New Three-Stage Assessment Method from Coarse to Fine.

Bone age assessment (BAA) based on X-ray imaging of the left hand and wrist can accurately reflect the degree of the body's physiological development and physical condition. However, the traditional manual evaluation method relies too much on inefficient specialist labor. In this paper, to propose automatic BAA, we introduce a hierarchical convolutional neural network to detect the regions of interest (ROI) and classify the bone grade. Firstly, we establish a dataset of children's BAA containing 2518 left hand X-rays. Then, we use the fine-grained classification to obtain the grade of the region of interest via object detection. Specifically, fine-grained classifiers are based on context-aware attention pooling (CAP). Finally, we perform the model assessment of bone age using the third version of the Tanner-Whitehouse (TW3) methodology. The end-to-end BAA system provides bone age values, the detection results of 13 ROIs, and the bone maturity of the ROIs, which are convenient for doctors to obtain information for operation. Experimental results on the public dataset and clinical dataset show that the performance of the proposed method is competitive. The accuracy of bone grading is 86.93%, and the mean absolute error (MAE) of bone age is 7.68 months on the clinical dataset. On public dataset, the MAE is 6.53 months. The proposed method achieves good performance in bone age assessment and is superior to existing fine-grained image classification methods.