Data Imbalance in Immunity Bone Age Assessment System Using Independent Autoencoders

Bone age can reflect the true growth and development status of a child; thus, it plays a critical role in evaluating growth and endocrine disorders. This study established and validated an optimized Tanner-Whitehouse 3 artificial intelligence (TW3-AI) bone age assessment (BAA) system based on a convolutional neural network (CNN). A data set of 9,059 clinical radiographs of the left hand was obtained from the picture archives and communication systems (PACS) between January 2012 and December 2016. Among these, 8,005/9,059 (88%) samples were treated as the training set for model implementation, 804/9,059 (9%) samples as the validation set for parameters optimization, and the remaining 250/9,059 (3%) samples were used to verify the accuracy and reliability of the model compared to that of 4 experienced endocrinologists and 2 experienced radiologists. The overall variation of TW3-metacarpophalangeal, radius, ulna and short bones (RUS) and TW3-Carpal bone score, as well as each bone (13 RUS + 7 Carpal) between reviewers and the AI, were compared by Bland-Altman (BA) chart and Kappa test, respectively. Furthermore, the time consumption between the model and reviewers was also compared. The performance of TW3-AI model was highly consistent with the reviewers' overall estimation, and the root mean square (RMS) was 0.50 years. The accuracy of the BAA of the TW3-AI model was better than the estimate of the reviewers. Further analysis revealed that human interpretations of the male capitate, hamate, the first distal and fifth middle phalanx and female capitate, the trapezoid, and the third and fifth middle phalanx, were most inconsistent. The average image processing time was 1.5±0.2 s in the TW3-AI model, which was significantly shorter than manual interpretation. The diagnostic performance of CNN-based TW3 BAA was accurate and timesaving, and possesses better stability compared to diagnostics made by experienced experts.

Bone age cluster assessment and feature clustering analysis based on phalangeal image rough segmentation

A quantitative study was conducted to direct the design and development of a computerized system for bone age assessment (BAA) in University of Malaya Medical Center (UMMC). Bone age assessment is a clinical procedure performed in pediatric radiology for evaluation the stage of skeletal maturation. It is usually performed by comparing an x-ray of a child’s left hand with a standard of known samples. The current methods utilized in clinical environment to estimate bone age are time consuming and prone to observer variability. This is motivation for developing a computerized method for BAA. A primary analysis shows the current method used by UMMC radiologists for bone age assessment, their feedbacks, problems encountered and their opinions about new approach for BAA. Our study also extracts user requirements for designing and developing a computerized method for BAA.KeywordsComputerized bone age assessmentQuantitative studyBone ageRadiographyBone age assessment

To explore the integration method and technical realization of artificial intelligence bone age assessment system with the hospital RIS-PACS network and workflow. Two sets of artificial intelligence based on bone age assessment systems (CHBoneAI 1.0/2.0) were developed. The intelligent system was further integrated with RIS-PACS based on the http protocol in Python flask web framework. The two sets of systems were successfully integrated into the local network and RIS-PACS in hospital. The deployment has been smoothly running for nearly 3 years. Within the current network setting, it takes less than 3 s to complete bone age assessment for a single patient. The artificial intelligence based bone age assessment system has been deployed in clinical RIS-PACS platform and the "running in parallel", which is marking a success of Stage-I and paving the way to Stage-II where the intelligent systems can evolve to become more powerful in particular of the system self-evolution and the "running alternatively".

Bone age is an indicator of bone maturity and is useful for the treatment of different pediatric conditions as well as for legal issues. Bone age can be assessed by the analysis of different skeletal segments and teeth and through several methods; however, traditional bone age assessment is a complicated and time-consuming process, prone to inter- and intra-observer variability. There is a high demand for fully automated systems, but creating an accurate and reliable solution has proven difficult. Deep learning technology, machine learning, and Convolutional Neural Networks-based systems, which are rapidly evolving, have shown promising results in automated bone age assessment. We provide the background of bone age estimation, its usefulness and traditional methods of assessment, and review the currently artificial-intelligence-based solutions for bone age assessment and the future perspectives of these applications.

Bone age estimation is very useful in children undergoing epiphysiodesis or guided growth surgery especially during the years of accelerated growth. It may be noted that no data are available on bone age estimation for Indian children of this age group. Sauvegrain (French) method is a very useful and simple method for bone age assessment during the years of accelerated growth. We decided to check the usefulness and the accuracy of the Sauvegrain method in Indian children. A team of two pediatric orthopaedic surgeons and a radiologist scored elbow X-rays of 80 healthy children (40 boys and 40 girls), using the Sauvegrain method twice. Interobserver reliability and intraobserver reproducibility of the Sauvegrain scoring were assessed. There was a very strong correlation between all observers in both rounds (r = > 0.8) and an excellent reproducibility by the same observer in both rounds (r = 0.955). Chronological and bone age are considered the same if the difference between them is less than 6 months. With this criterion bone and chronological ages matched in >37% of boys and girls, similar to the study done in French children. In the nonmatching group, more children had delayed bone age compared to their chronological age. The Sauvegrain method of bone age assessment described for French children was found to be useful in estimating bone age in Indian children. It is especially helpful in the clinical practice for detecting mismatch between the chronological and the radiological age before undertaking guided growth or epiphysiodesis.

In 104 normal boys, aged 7 to 14 years (bone ages 5 to 15 years), plasma dehydroepiandrosterone (DHEA) rose from 52.7 at 7 years, to 112.0 ng/100 ml at 10 years. A further rise occurred at 12 years (188 ng/100 ml). In relation to the bone age, DHEA increased from a mean plasma level of 31.1 at a bone age of 5 years to 77.1 ng/100 ml at one of 7 years. Further increases were observed with mean values of 163.2 at a bone age of 11 years, and of 221.2 at a bone age of 12 years, with a maximum of 333.4 ng/100 ml at bone ages of 14-15 years. The first significant increase of plasma testosterone (T) was noted at a bone age of 12 years (54.8 ng/100 ml). The major rise of T was preceded by the rise of plasma LH and was accompanied by the rise of plasma FSH. Plasma DHEA and T were also measured in 123 normal girls, ages 6 to 13 years (bone ages 5 to 15 years). DHEA rose significantly from a mean level of 44.7 at 6 years, to 80.9 ng/100 ml at 8 years, with further increases between 9 and 10 years and between 10 and 11 years. In relation to bone age, DHEA increased significantly from a mean plasma concentration of 30.9 at a bone age of 5 years, to that of 58.6 ng/100 ml at 7 years. Further increases were observed with values of 191.1 at a bone age of 10 years and 485.6 ng/100 ml at a bone age of 13 years. The first significant rise of testosterone (T) occurred at 10 years of both chronological and bone age. DHEA rose before the increase of gonadotropins. The major rise of T at a bone age of 10 years occurred concurrently with increases in plasma FSH and LH. Low levels of DHEA were observed in Addison's disease. In hypogonadotropin hypogonadism and in anorchia, DHEA levels were normal, suggesting that DHEA is produced primarily in the adrenal gland. In seven girls with early adrenarche, plasma concentrations of DHEA were in the upper range of normal values, whereas T levels were within the normal range. Conversely in girls with late adrenarche, plasms DHEA was lower than normal but T was within the normal limits. The elevation of DHEA prior to the first signs of puberty suggests that DHEA may play a role in the maturation of the hypothalamic-hypophysealgonadal axis. However, the mechanism that triggers the secretion of DHEA is not known.

This study aimed to evaluate the clinical accuracy of automated BoneXpert (BX) vs. clinician-based Greulich-Pyle (GP) bone age (BA) assessments by comparing their respective predicted adult heights (PAH) with near-final height (NFH) in patients with congenital adrenal hyperplasia (CAH). Forty-seven patients with classic salt-wasting CAH diagnosed in the neonatal period who had reached NFH were included. Left hand-wrist radiographs obtained during the prepubertal period (6-9 years) were evaluated manually by two clinicians using the GP atlas and by the BX-GP method. PAH was calculated for both sexes using the Bayley-Pinneau (BP) method based on clinician- and BX-derived BA. PAH standard deviation scores (SDS) were compared with NFH-SDS and target height (TH)-SDS. Agreement was assessed using Bland-Altman analysis. Thirty patients (63.8 %) were female. Mean ages at bone age assessment and NFH were 7.3 and 16.1 years, respectively. Mean prepubertal height-SDS was 0.0 in girls and-0.25 in boys, while NFH-SDS was-0.79 and-1.14, respectively. Mean bone age according to clinician-GP was 7.80 years in girls and 8.16 in boys, and 7.69 and 8.14 years according to BX-GP. Clinician-based PAH-SDS was-1.0 in girls and-1.18 in boys, whereas BX-based PAH-SDS was-0.86 and-1.17, respectively. BX-based PAH was closer to NFH in girls (p<0.001), with no significant difference in boys. In both sexes, PAH-SDS and NFH-SDS were significantly lower than TH-SDS (p<0.001). PAH calculations using the BP method based on BX-derived GP bone age readings more accurately predicted NFH in patients with CAH.

ObjectivesThis 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 methodsHand 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.ResultsThe 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).ConclusionIn terms of BA assessment, the new AI software was interchangeable with expert readers using the G&P method.Clinical relevance statementThe 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.

This research has three parts addressing classification, especially the imbalanced data problem, which is one of the most popular and essential issues in the domain of classification. The first part is to study the Adaptive Boosting (AdaBoost) algorithm. AdaBoost is an effective solution for classification, but it still needs improvement in the imbalanced data problem. This part proposes a method to improve the AdaBoost algorithm using the new weighted vote parameters for the weak classifiers. Our proposed weighted vote parameters are determined not only by the global error rate but also by the classification accuracy rate of the positive class, which is our primary interest. The imbalanced index of the data is also a factor in constructing our algorithms. The numeric studies show that our proposed algorithms outperform the traditional ones, especially regarding the evaluation criterion of the F--1 Measure. Theoretic proofs of the advantages of our proposed algorithms are presented. The second part treats the Relevance Vector Machine (RVM), which is a supervised learning algorithm extended from the Support Vector Machine (SVM) based on the Bayesian sparsity model. Compared with the regression problem, RVM classification is challenging to conduct because there is no closed-form solution for the weight parameter posterior. The original RVM classification algorithm uses Newton's method in optimization to obtain the mode of weight parameter posterior, then approximates it by a Gaussian distribution in Laplace's method. This original model would work, but it just applies the frequency methods in a Bayesian framework. This part first proposes a Generic Bayesian RVM classification, which is a pure Bayesian model. We conjecture that our algorithm achieves convergent estimates of the quantities of interest compared with the nonconvergent estimates of the original RVM classification algorithm. Furthermore, a fully Bayesian approach with the hierarchical hyperprior structure for RVM classification is proposed, which improves the classification performance, especially in the imbalanced data problem. The third part is an extended work of the second one. The original RVM classification model uses the logistic link function to build the likelihood, which makes the model hard to conduct since the posterior of the weight parameter has no closed-form solution. This part proposes the probit link function approach instead of the logistic one for the likelihood function in RVM classification, namely PRVM (RVM with the Probit link function). We show that the posterior of the weight parameter in our model follows the multivariate normal distribution and achieves a closed-form solution. A latent variable is needed in our algorithm to simplify the Bayesian computation greatly, and its conditional posterior follows a truncated normal distribution. Compared with the original RVM classification model, our proposed one is another pure Bayesian approach and it has a more efficient computation process. For the prior structure, we first consider the Normal-Gamma independent prior to propose a Generic Bayesian PRVM algorithm. Furthermore, the Fully Bayesian PRVM algorithm with a hierarchical hyperprior structure is proposed, which improves the classification performance, especially in the imbalanced data problem.

The integration of industrialization and informatization has exposed industrial control systems (ICSs) to increasingly serious security challenges. Currently, the mainstream method to protect the security of ICSs is intrusion detection system (IDS) based on deep-learning. However, these methods depend on a massive amount of high-quality data. Owing to the characteristics and protocol limitations, ICSs data usually experience low-quality and data imbalance problems, which significantly affects the accuracy of IDS.In this study, an IDS for ICS that combines data expansion algorithm and CNN was proposed. A novel normalized neighborhood weighted convex combined random sample (NNW-CCRS) oversampling algorithm was designed, which automatically attenuates the effects of noise and expanding imbalanced data to produce balanced ICS datasets. By reducing the impact of imbalanced ICS data on IDSs, our system effectively protects the security of ICS. Secure Water Treatment dataset (SWaT) was used for experimental validation. The experimental results confirmed that the accuracy of the proposed system improved by approximately 20%, compared to the ICS without data expansion.

Bone age assessment is challenging in clinical practice due to the complicated bone age assessment process. Current automatic bone age assessment methods were designed with rare consideration of the diagnostic logistics and thus may yield certain uninterpretable hidden states and outputs. Consequently, doctors can find it hard to cooperate with such models harmoniously because it is difficult to check the correctness of the model predictions. In this work, we propose a new graph-based deep learning framework for bone age assessment with hand radiographs, called Doctor Imitator (DI). The architecture of DI is designed to learn the diagnostic logistics of doctors using the scoring methods (e.g., the Tanner-Whitehouse method) for bone age assessment. Specifically, the convolutions of DI capture the local features of the anatomical regions of interest (ROIs) on hand radiographs and predict the ROI scores by our proposed Anatomy-based Group Convolution, summing up for bone age prediction. Besides, we develop a novel Dual Graph-based Attention module to compute patient-specific attention for ROI features and context attention for ROI scores. As far as we know, DI is the first automatic bone age assessment framework following the scoring methods without fully supervised hand radiographs. Experiments on hand radiographs with only bone age supervision verify that DI can achieve excellent performance with sparse parameters and provide more interpretability.

Legg-Calve-Perthes disease (LCPD) outcomes are largely determined by their age and maturity at onset. In LCPD, there is a known association with delayed skeletal maturity of up to 1.9 years in affected children based on maturity assessment with the Greulich and Pyle (GP) maturity atlas. The GP atlas is the standard for assessing bone age but requires obtaining a separate radiograph of the hand. A new methodology for assessing skeletal maturation in comparison to the GP bone age in children with LCPD was sought. A retrospective review of a prospective, multicenter study of patients with LCPD treated from 1984 to 1991 and followed to skeletal maturity was performed. Patients were included if they had LCPD diagnosed on anteroposterior pelvis radiographs that included the contralateral hip who had bone age radiographs obtained at the time of presentation. Patients were excluded if they presented with bilateral LCPD, the contralateral hip was not visualized on their presenting radiographs, they lacked bone age radiographs at the time of presentation, or they presented outside the range for the Optimized Oxford system. A formula using the greater trochanteric height to femoral head diameter ratio and patient sex (GT+ Sex) for predicting GP bone age was developed. The GP and GT+ Sex bone ages were compared with the chronologic age (CA) to determine the mean discrepancy. Seventy-one patients were included (mean 9.5 ± 1.2y at presentation, 42.2% females). Skeletal maturity assessment by the GP bone age method demonstrated a mean discrepancy of 1.4 years younger than CA (95% CI: 1.01-1.76y). GT+ Sex bone age assessment demonstrated a mean discrepancy of 1.4 years younger than CA (95% CI: 1.03-1.75y). The GP bone age was a mean of 0.00 years different than the GT+ Sex assessment bone age (95% CI: -0.3 to 0.3y). The GT+ Sex assessment bone age correlated significantly with GP bone age ( R =0.89, P <0.0001). Male patients had a significantly younger GP bone age relative to CA compared with female patients (1.8 vs. 0.86y, P = 0.02); however, there were fewer sex differences in the GT+ Sex assessment bone age relative to CA (male, 1.77y younger vs. 1.12y younger; P = 0.01). The GT+ Sex skeletal maturity assessment system in children with LCPD significantly correlated with the GP bone age system when compared. The GT+ Sex assessment may provide an assessment of the true bone age in LCPD patients without the need for hand radiographic imaging studies. Level III-retrospective review and analysis.

Bone age assessment (BAA) evaluates individual skeletal maturity by comparing the characteristics of skeletal development to the standard in a specific population. The X-ray image examination for bone age is tedious and subjective, and it requires high professional skills. Therefore, AI techniques are desired to innovate and improve BAA methods. Most of the BAA method use the whole X-ray image in an end-to-end model directly. Such whole-image-based approaches fail to characterize local changes and provide limited aid for diagnosis and understanding disease progress. To address these issues, we collected and curated a dataset of 2129 cases for the study of BAA with fine-grained skeletal maturity level labels of the 13 ROIs in hand bone based on the expert knowledge from TW method. We designed a four-stage automatic BAA model based on recursive feature pyramid network. Firstly, the palm region was segmented using U-Net, followed by the extraction of multi-target ROIs of hand bone using a recursive feature pyramid network. Given the extracted ROIs, we employed a transfer learning model with attention mechanism to predict the skeletal maturity level of each ROI. Finally, the bone age is assessed based on the percentile curve of bone maturity. The proposed BAA model can automate the BAA. In addition, it provides the detection result of the 13 ROIs and their ROI-level skeletal maturity. The MAE can reach 0.61 years on the dataset with the labeling precision of one year. All the data and annotations used in this paper are released publicly.

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.