Radiological Society Of North America Research Articles

Ranking the Relative Importance of Image Quality Features in CT by Consensus Survey

ObjectiveThis study sought to determine consensus opinions from subspecialty radiologists and imaging physicists on the relative importance of image quality features in CT. MethodsA prospective survey of subspecialty radiologists and medical physicists was conducted to collect consensus opinions on the relative importance of ten image quality features: axial sharpness, blooming, contrast, longitudinal sharpness, low contrast axial sharpness, metal artifact, motion, noise magnitude, noise texture, and streaking. The survey was first sent to subspecialty radiologists in volunteer leadership roles in the American College of Radiology and Radiological Society of North America, thereafter relying on snowball sampling. Surveyed subspecialties were abdominal, cardiac, emergency, musculoskeletal, neuro, pediatric, and thoracic radiology, and medical physics. Individual respondents’ ratings were normalized for calculation of mean normalized ratings and priority rankings for each feature within subspecialties. Also calculated were intraclass correlation coefficients across image quality features within subspecialties, and analysis of variance across subspecialties within each feature. ResultsMost subspecialties had moderate to excellent intraclass agreement. For every radiology subspecialty except musculoskeletal, motion was the most important image quality feature. There was agreement across subspecialties that axial sharpness and contrast are only moderately important. There was disagreement across subspecialties on the relative importance of noise magnitude. Blooming was highly important to cardiac radiologists, and noise texture was highly important to musculoskeletal radiologists. ConclusionImage quality preferences differ based on clinical tasks and challenges in each anatomical radiology subspecialty. CT image analysis and development of quantitative measures of quality and protocol optimization—and related policy initiatives—should be specific to radiology subspecialty.

-AI-assisted diagnostic potential of CT in bone oncology and its impact on clinical decision-making for intensive care

ObjectiveThis study evaluates the AI-assisted diagnostic potential of computed tomography (CT) for bone cancer and its influence on patient care during the pre- and post-treatment phases. It compares patient management approaches based on CT severity levels and identifies distinct CT phenotypes linked to disease severity. MethodologyWe retrospectively examined 50 patients diagnosed with bone cancer between December 2022 and June 2023. The CT scans were analyzed according to the Radiological Society of North America (RSNA) guidelines. This study was performed using the deep convolutional neutral network (DCNN) model to assist doctors in diagnosing bone tumors through CT scanning. Patients’ management approaches were compared based on the severity levels indicated by CT scans. ResultsFifty patients participated in this study, with a median age of 67.2 years, ranging from 32 to 89 years. Of them, 38 % were female and 62 % were male. In 2022, 19 individuals (13 males and 6 females, ages 32 to 84) were assessed, with a mean age of 59.9 years. In 2023, 31 individuals, aged 54 to 89 with a mean age of 71.6 years, were assessed; among them were 18 men and 13 women. SPECT scans revealed the following key diagnostic features: 85.9 % of patients exhibited bone lesions with ground-glass opacities, 88 % had multipolar involvement, 92.8 % had bilateral involvement, and 92.8 % showed peripheral involvement. The severity scores based on CT scans were significantly higher in patients requiring intensive care, with scores above 14 being more common in this group. ConclusionDistinct CT findings during the AI-assisted diagnosis and treatment of bone cancer provided prompt and sensitive examination capabilities. Notably, two CT phenotypes emerged, associated with large consolidation patterns and high severity scores, offering crucial insights into disease severity and aiding in clinical decision-making for intensive care requirements. The study underscores the importance of CT in the effective monitoring and management of bone cancer pre- and post-treatment.

Doctor simulator: Delta-Age-Sex-AdaIn enhancing bone age assessment through AdaIn style transfer.

Bone age assessment assists physicians in evaluating the growth and development of children. However, deep learning methods for bone age estimation do not currently incorporate differential features obtained through comparisons with other bone atlases. To propose a more accurate method, Delta-Age-Sex-AdaIn (DASA-net), for bone age assessment, this paper combines age and sex distribution through adaptive instance normalization (AdaIN) and style transfer, simulating the process of visually comparing hand images with a standard bone atlas to determine bone age. The proposed Delta-Age-Sex-AdaIn (DASA-net) consists of four modules: BoneEncoder, Binary code distribution, Delta-Age-Sex-AdaIn, and AgeDecoder. It is compared with state-of-the-art methods on both a public Radiological Society of North America (RSNA) pediatric bone age prediction dataset (14,236 hand radiographs, ranging from 1 to 228 months) and a private bone age prediction dataset from Zigong Fourth People's Hospital (474 hand radiographs, ranging from 12 to 218 months, 268 male). Ablation experiments were designed to demonstrate the necessity of incorporating age distribution and sex distribution. The DASA-net model achieved a lower mean absolute deviation (MAD) of 3.52 months on the RSNA dataset, outperforming other methods such as BoneXpert, Deeplasia, BoNet, and other deep learning based methods. On the private dataset, the DASA-net model obtained a MAD of 3.82 months, which is also superior to other methods. The proposed DASA-net model aided the model's learning of the distinctive characteristics of hand bones of various ages and bothsexes by integrating age and sex distribution into style transfer.

A Densely Connected Convolutional Network Towards Efficient Recognition of Intracranial Hemorrhage Using CT Images

Intracranial Bleeding also known as Intracranial Hemorrhage (ICH) is a severe issue for human beings and the most common cause of it is trauma. Majorly ICH occurs due to hypertension and around 2.5 per 10,000 people get affected by one of its subtypes. The human brain consists of lots of soft tissue and nerves, that’s why it is so tough to find affected areas within the shortest period and to apply proper treatment or medication for a radiologist by analyzing the Computed Tomography (CT) images. In addition to ICH treatment being expensive, a densely connected convolutional network (DenseNet-169) model is recommended to accurately detect the damaged region quickly and affordably to facilitate the treatment of the patient. All were private and inaccessible, with the exception of the CQ500 and the Radiological Society of North America (RSNA) datasets about ICH. In our study, we employed stage-2 of the RSNA dataset, which comprises 121,232 test images and 752,803 training images. Phase-1 and phase-2 are the two stages of the dataset. Among the various preprocessing techniques, image type conversion, resizing, and normalization were performed on the dataset. During the learning phase of our model, for hyper-tuning, a portion (30%) of training data was utilized as validation data. The test data was then used to evaluate the model’s efficacy, and it was found that the ICH recognition accuracy of our developed model was 98%. Index Terms— Intracranial Hemorrhage, CT images, DenseNet, Training, Testing, Recognition, RSNA. GUB JOURNAL OF SCIENCE AND ENGINEERING, Vol 9(1), 2022 P 13-22

Optimal Ensemble Transfer Learning Espoused Detection and Classification of Intracranial Haemorrhage

The injury of blood vessels in brain tissues known as intracranial haemorrhage (ICH) is a major cause of stroke. Computed tomography (CT) and X-ray images are frequently used to estimate the location and size of haemorrhages. Radiologists segment CT scan pictures manually using planimetry, which takes a lot of time. To overcome this issue, an optimal ensemble transfer learning espoused detection and classification of intracranial haemorrhage (CIH-V16-V19-R50-DN121-IV3-XN) is proposed in this paper. Initially, the images are gathered via the Radiological Society of North America (RSNA) ICH dataset. Afterwards, the images are fed to pre-processing. The pre-processed output is fed into a classification segment. The model used for classification is the ensemble transfer learning model such as VGG16, VGG19, Resnet50, InceptionV3, Desnet121, and XceptionNet. With the help of ensemble transfer learning techniques, the image is classified into subdural haemorrhage, epidural haemorrhage, subarachnoidal haemorrhage, intraventricular haemorrhage, intraparenchymal haemorrhage, and normal. With the help of voting classifier the best learning model with accuracy is identified. The proposed CIH-V16-V19-R50-DN121-IV3-XN algorithm is implemented in MATLAB using dataset of the RSNA ICH dataset. The performance metrics, such as accuracy, precision, sensitivity, F1-score, specificity, ROC, and error rate, are examined to analyse the efficiency of the proposed method. The performance of the CIH-V16-V19-R50-DN121-IV3-XN approach attains 10.12%, 22.33%, 21.45% high accuracy and 10.09%, 13.65%, 29.87% high specificity compared to the existing methods.

Lessons Learned in Building Expertly Annotated Multi-Institution Datasets and Hosting the RSNA AI Challenges.

The Radiological Society of North America (RSNA) has held artificial intelligence competitions to tackle real-world medical imaging problems at least annually since 2017. This article examines the challenges and processes involved in organizing these competitions, with a specific emphasis on the creation and curation of high-quality datasets. The collection of diverse and representative medical imaging data involves dealing with issues of patient privacy and data security. Furthermore, ensuring quality and consistency in data, which includes expert labeling and accounting for various patient and imaging characteristics, necessitates substantial planning and resources. Overcoming these obstacles requires meticulous project management and adherence to strict timelines. The article also highlights the potential of crowdsourced annotation to progress medical imaging research. Through the RSNA competitions, an effective global engagement has been realized, resulting in innovative solutions to complex medical imaging problems, thus potentially transforming health care by enhancing diagnostic accuracy and patient outcomes. Keywords: Use of AI in Education, Artificial Intelligence © RSNA, 2024.

ChatGPT may free time needed by the interventional radiologist for administration / documentation

Motive: Documentation and administration, unpleasant necessities, take a substantial part of the working time in the subspecialty of interventional radiology. With increasing future demand for clinical radiology predicted, time savings from use of text drafting technologies could be a valuable contribution towards our field. Method: Three cases of peripherally inserted central catheter (PICC) line insertion were defined for the present study. The current version of ChatGPT was tasked with drafting reports, following the Radiological Society of North America (RSNA) template. Key results: Score card evaluation by human radiologists indicates that time savings in documentation / administration can be expected without loss of quality from using ChatGPT. Further, automatically generated texts were not assessed to be clearly identifiable as AI-produced. Conclusions: Patients, doctors, and hospital administrators would welcome a reduction of the time that interventional radiologists need for documentation and administration these days. If AI-tools as tested in the present study are brought into clinical application, questions about trust into those systems eg with regard to medical complications will have to be addressed.

The Relation of Polymerase Chain Reaction and Computed Tomography Infiltrations of Suspected Covid-19 Patients in Emergency Department

Aims: In the context of the persistent global challenge posed by COVID-19, effective diagnostic methods are crucial. Considering the limitations of Polymerase Chain Reaction (PCR) in the Emergency Department (ED), alternative confirmation methods for assessing Covid-19 are needed. This study aimed to evaluate the performance of Thorax-Computed Tomography (T-CT), following Radiological Society of North America (RSNA) recommendations, for suspected Covid-19 patients with pulmonary infiltrations. Materials and Methods: From March to August 2020, 324 ED patients with suspected Covid-19 underwent T-CT scans. Blinded radiologists independently assessed T-CT scans based on RSNA guidelines. Reverse-transcriptase polymerase chain reaction (RT-PCR) served as the reference test. Results: Of 324 patients, 35% tested positive via RT-PCR. T-CT categories were typical (35.2%), indeterminate (47.5%), and atypical (11.1%). Using a typical T-CT category threshold resulted in 66% sensitivity, 81% specificity, 65% positive predictive value (PPV), 82% negative predictive value (NPV), and 76% accuracy. Subgroup analysis of repeat RT-PCR tests improved performance: 80% sensitivity, 79% specificity, 76% PPV, 83% NPV, and 79% accuracy. Combining RT-PCR and T-CT in the ED achieved 95.5% sensitivity, 79% PPV, and 86.4% accuracy. Conclusion: Following RSNA guidelines, T-CT exhibits moderate sensitivity and high specificity for detecting Covid-19. In ED settings with suspected cases, T-CT aids in recommending retesting after an initial negative RT-PCR result, facilitating early management and enabling timely isolation measures. The combined use of RT-PCR and T-CT enhances diagnostic performance, emphasizing the potential benefits of integrating these methods.

Time-dependent CT score-based model for identifying severe/critical COVID-19 at a fever clinic after the emergence of Omicron variant

Rationale and objectivesThe computed tomography (CT) score has been used to evaluate the severity of COVID-19 during the pandemic; however, most studies have overlooked the impact of infection duration on the CT score. This study aimed to determine the optimal cutoff CT score value for identifying severe/critical COVID-19 during different stages of infection and to construct corresponding predictive models using radiological characteristics and clinical factors. Materials and methodsThis retrospective study collected consecutive baseline chest CT images of confirmed COVID-19 patients from a fever clinic at a tertiary referral hospital from November 28, 2022, to January 8, 2023. Cohorts were divided into three subcohorts according to the time interval from symptom onset to CT examination at the hospital: early phase (0–3 days), intermediate phase (4–7 days), and late phase (8–14 days). The binary endpoints were mild/moderate and severe/critical infection. The CT scores and qualitative CT features were manually evaluated. A logistic regression analysis was performed on the CT score as determined by a visual assessment to predict severe/critical infection. Receiver operating characteristic analysis was performed and the area under the curve (AUC) was calculated. The optimal cutoff value was determined by maximizing the Youden index in each subcohort. A radiology score and integrated models were then constructed by combining the qualitative CT features and clinical features, respectively, using multivariate logistic regression with stepwise elimination. ResultsA total of 962 patients (aged, 61.7 ± 19.6 years; 490 men) were included; 179 (18.6%) were classified as severe/critical COVID-19, while 344 (35.8%) had a typical Radiological Society of North America (RSNA) COVID-19 appearance. The AUCs of the CT score models reached 0.91 (95% confidence interval (CI) 0.88–0.94), 0.82 (95% CI 0.76–0.87), and 0.83 (95% CI 0.77–0.89) during the early, intermediate, and late phases, respectively. The best cutoff values of the CT scores during each phase were 1.5, 4.5, and 5.5. The predictive accuracies associated with the time-dependent cutoff values reached 88% (vs.78%), 73% (vs. 63%), and 87% (vs. 57%), which were greater than those associated with universal cutoff value (all P < 0.001). The radiology score models reached AUCs of 0.96 (95% CI 0.94–0.98), 0.90 (95% CI 0.87–0.94), and 0.89 (95% CI 0.84–0.94) during the early, intermediate, and late phases, respectively. The integrated models including demographic and clinical risk factors greatly enhanced the AUC during the intermediate and late phases compared with the values obtained with the radiology score models; however, an improvement in accuracy was not observed. ConclusionThe time interval between symptom onset and CT examination should be tracked to determine the cutoff value for the CT score for identifying severe/critical COVID-19. The radiology score combining qualitative CT features and the CT score complements clinical factors for identifying severe/critical COVID-19 patients and facilitates timely hierarchical diagnoses and treatment.

Evaluating the Robustness of a Deep Learning Bone Age Algorithm to Clinical Image Variation Using Computational Stress Testing.

Purpose To evaluate the robustness of an award-winning bone age deep learning (DL) model to extensive variations in image appearance. Materials and Methods In December 2021, the DL bone age model that won the 2017 RSNA Pediatric Bone Age Challenge was retrospectively evaluated using the RSNA validation set (1425 pediatric hand radiographs; internal test set in this study) and the Digital Hand Atlas (DHA) (1202 pediatric hand radiographs; external test set). Each test image underwent seven types of transformations (rotations, flips, brightness, contrast, inversion, laterality marker, and resolution) to represent a range of image appearances, many of which simulate real-world variations. Computational "stress tests" were performed by comparing the model's predictions on baseline and transformed images. Mean absolute differences (MADs) of predicted bone ages compared with radiologist-determined ground truth on baseline versus transformed images were compared using Wilcoxon signed rank tests. The proportion of clinically significant errors (CSEs) was compared using McNemar tests. Results There was no evidence of a difference in MAD of the model on the two baseline test sets (RSNA = 6.8 months, DHA = 6.9 months; P = .05), indicating good model generalization to external data. Except for the RSNA dataset images with an appended radiologic laterality marker (P = .86), there were significant differences in MAD for both the DHA and RSNA datasets among other transformation groups (rotations, flips, brightness, contrast, inversion, and resolution). There were significant differences in proportion of CSEs for 57% of the image transformations (19 of 33) performed on the DHA dataset. Conclusion Although an award-winning pediatric bone age DL model generalized well to curated external images, it had inconsistent predictions on images that had undergone simple transformations reflective of several real-world variations in image appearance. Keywords: Pediatrics, Hand, Convolutional Neural Network, Radiography Supplemental material is available for this article. © RSNA, 2024 See also commentary by Faghani and Erickson in this issue.

Semi-supervised Learning for Generalizable Intracranial Hemorrhage Detection and Segmentation.

Purpose To develop and evaluate a semi-supervised learning model for intracranial hemorrhage detection and segmentation on an out-of-distribution head CT evaluation set. Materials and Methods This retrospective study used semi-supervised learning to bootstrap performance. An initial "teacher" deep learning model was trained on 457 pixel-labeled head CT scans collected from one U.S. institution from 2010 to 2017 and used to generate pseudo labels on a separate unlabeled corpus of 25 000 examinations from the Radiological Society of North America and American Society of Neuroradiology. A second "student" model was trained on this combined pixel- and pseudo-labeled dataset. Hyperparameter tuning was performed on a validation set of 93 scans. Testing for both classification (n = 481 examinations) and segmentation (n = 23 examinations, or 529 images) was performed on CQ500, a dataset of 481 scans performed in India, to evaluate out-of-distribution generalizability. The semi-supervised model was compared with a baseline model trained on only labeled data using area under the receiver operating characteristic curve, Dice similarity coefficient, and average precision metrics. Results The semi-supervised model achieved a statistically significant higher examination area under the receiver operating characteristic curve on CQ500 compared with the baseline (0.939 [95% CI: 0.938, 0.940] vs 0.907 [95% CI: 0.906, 0.908]; P = .009). It also achieved a higher Dice similarity coefficient (0.829 [95% CI: 0.825, 0.833] vs 0.809 [95% CI: 0.803, 0.812]; P = .012) and pixel average precision (0.848 [95% CI: 0.843, 0.853]) vs 0.828 [95% CI: 0.817, 0.828]) compared with the baseline. Conclusion The addition of unlabeled data in a semi-supervised learning framework demonstrates stronger generalizability potential for intracranial hemorrhage detection and segmentation compared with a supervised baseline. Keywords: Semi-supervised Learning, Traumatic Brain Injury, CT, Machine Learning Supplemental material is available for this article. Published under a CC BY 4.0 license. See also the commentary by Swimburne in this issue.

Deep learning convolutional neural network ResNet101 and radiomic features accurately analyzes mpMRI imaging to predict MGMT promoter methylation status with transfer learning approach

AbstractAccurate brain tumor classification is crucial for enhancing the diagnosis, prognosis, and treatment of glioblastoma patients. We employed the ResNet101 deep learning method with transfer learning to analyze the 2021 Radiological Society of North America (RSNA) Brain Tumor challenge dataset. This dataset comprises four structural magnetic resonance imaging (MRI) sequences: fluid‐attenuated inversion‐recovery (FLAIR), T1‐weighted pre‐contrast (T1w), T1‐weighted post‐contrast (T1Gd), and T2‐weighted (T2). We assessed the model's performance using standard evaluation metrics. The highest performance to detect MGMT methylation status for patients suffering glioblastoma was an accuracy (85.48%), sensitivity (80.64%), specificity (90.32%). Whereas classification performance with no tumor was yielded with accuracy (85.48%), sensitivity (90.32%), specificity (80.64%). The radiomic features (74) computed with ensembled Bagged Tree and relief feature selection method (30/74) improved the validation accuracy of 84.3% and AUC of 0.9038 to detect. O6‐methylguanine‐DNA methyltransferase (MGMT) promoter methylation status in glioblastoma patients holds promise for optimizing treatment planning and prognosis. By understanding MGMT methylation status, clinicians can make informed decisions about treatment strategies, potentially leading to improved clinical outcomes.

A Deep Learning Approach to Classification Pneumonia in Thorax Images

An algorithm of automatic learning was developed. That is abke to identify radiographic images with a pneumonia and not pneumonia diagnose based on a data set published in the “Kaggle” platform by the Radiological Society of North America, from which we obtained a set of specific images with their labels that were divided by 70% to train a convolutional neuronal network model consisting of two convolutional layers for the extraction of characteristics in each image, and ends with a classification stage in the training of the model, to conclude with 74% in metrics of accuracy given by test tests with 30% of the data set.

The Role of Imaging for GI Bleeding: ACG and SAR Consensus Recommendations.

Gastrointestinal (GI) bleeding is the most common GI diagnosis leading to hospitalization within the United States. Prompt diagnosis and treatment of GI bleeding is critical to improving patient outcomes and reducing high health care utilization and costs. Radiologic techniques including CT angiography, catheter angiography, CT enterography, MR enterography, nuclear medicine red blood cell scan, and technetium-99m pertechnetate scintigraphy (Meckel scan) are frequently used to evaluate patients with GI bleeding and are complementary to GI endoscopy. However, multiple management guidelines exist, which differ in the recommended utilization of these radiologic examinations. This variability can lead to confusion as to how these tests should be used in the evaluation of GI bleeding. In this document, a panel of experts from the American College of Gastroenterology and Society of Abdominal Radiology provide a review of the radiologic examinations used to evaluate for GI bleeding including nomenclature, technique, performance, advantages, and limitations. A comparison of advantages and limitations relative to endoscopic examinations is also included. Finally, consensus statements and recommendations on technical parameters and utilization of radiologic techniques for GI bleeding are provided. © Radiological Society of North America and the American College of Gastroenterology, 2024. Supplemental material is available for this article. This article is being published concurrently in American Journal of Gastroenterology and Radiology. The articles are identical except for minor stylistic and spelling differences in keeping with each journal's style. Citations from either journal can be used when citing this article. See also the editorial by Lockhart in this issue.

Standardization of Weightage Assigned to Different Segments of the Hand X-ray for Assessment of Bone Age by the Greulich-Pyle Method.

Bone age (BA) assessment is important in evaluating disorders of growth and puberty; the Greulich and Pyle atlas method (GP) is most used. We aimed to determine the weightage to be attributed by raters to various segments of the hand x-ray, namely, distal end of radius-ulna (RU), carpals, and short bones for rating bone age using the GP atlas method. 692 deidentified x-rays from a previous study (PUNE-dataset) and 400 from the Radiological Society of North America (RSNA-dataset) were included in the study. Mean of BA assessed by experienced raters was termed reference rating. Linear regression was used to model reference age as function of age ratings of the three segments. The root-mean-square-error (RMSE) of segmental arithmetic mean and weighted mean with respect to reference rating were computed for both datasets. Short bones were assigned the highest weightage. Carpals were assigned higher weightage in pre-pubertal PUNE participants as compared to RSNA, vice-versa in RU segment of post-pubertal participants. The RMSE of weighted mean ratings was significantly lower than for the arithmetic mean in the PUNE dataset. We thus determined weightage to be attributed by raters to segments of the hand x-ray for assessment of bone age by the GP method.

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.

Strengthening deep-learning models for intracranial hemorrhage detection: strongly annotated computed tomography images and model ensembles.

Multiple attempts at intracranial hemorrhage (ICH) detection using deep-learning techniques have been plagued by clinical failures. We aimed to compare the performance of a deep-learning algorithm for ICH detection trained on strongly and weakly annotated datasets, and to assess whether a weighted ensemble model that integrates separate models trained using datasets with different ICH improves performance. We used brain CT scans from the Radiological Society of North America (27,861 CT scans, 3,528 ICHs) and AI-Hub (53,045 CT scans, 7,013 ICHs) for training. DenseNet121, InceptionResNetV2, MobileNetV2, and VGG19 were trained on strongly and weakly annotated datasets and compared using independent external test datasets. We then developed a weighted ensemble model combining separate models trained on all ICH, subdural hemorrhage (SDH), subarachnoid hemorrhage (SAH), and small-lesion ICH cases. The final weighted ensemble model was compared to four well-known deep-learning models. After external testing, six neurologists reviewed 91 ICH cases difficult for AI and humans. InceptionResNetV2, MobileNetV2, and VGG19 models outperformed when trained on strongly annotated datasets. A weighted ensemble model combining models trained on SDH, SAH, and small-lesion ICH had a higher AUC, compared with a model trained on all ICH cases only. This model outperformed four deep-learning models (AUC [95% C.I.]: Ensemble model, 0.953[0.938-0.965]; InceptionResNetV2, 0.852[0.828-0.873]; DenseNet121, 0.875[0.852-0.895]; VGG19, 0.796[0.770-0.821]; MobileNetV2, 0.650[0.620-0.680]; p < 0.0001). In addition, the case review showed that a better understanding and management of difficult cases may facilitate clinical use of ICH detection algorithms. We propose a weighted ensemble model for ICH detection, trained on large-scale, strongly annotated CT scans, as no model can capture all aspects of complex tasks.

Creating the freedom to thrive: Honoring the legacy of RSNA gold medalist, Deborah Levine, MD

Awarded the Radiological Society of North America (RSNA) Gold Medal in 2018, Dr. Deborah Levine's research, journalism, and mentorship have left an indelible mark on the radiology field. Her work in ultrasound led to its use as the standard for monitoring benign adnexal cysts. She helped popularize obstetric magnetic resonance imaging (MRI) through her research on its use in placental accreta and fetal abnormalities, which led to the development of the ‘Compendium of Fetal MRI’ website. This work in research led naturally to a career in journalism, where she eventually became Senior Deputy Editor of Radiology and founded Radiology Select. Concurrently with her personal achievements, Dr. Levine has dedicated herself to the mentorship of her female trainees. She sought various leadership positions to learn more about and advocate for the promotion and support of female leadership in radiology departments. In many ways, Dr. Levine's career and work have transformed radiology as we know it today for both patients and physicians.

Skeletal age evaluation using hand X-rays to determine growth problems

A common clinical method for identifying anomalies in bone growth in infants and newborns is skeletal age estimation with X-ray images. Children’s bone abnormalities can result from several conditions including wounds, infections, or tumors. One of the most frequent reasons for bone issues is that most youngsters are affected by the slow displacement of bones caused by pressure applied to the growth plates as youngsters develop. The growth plate can be harmed by a lack of blood supply, separation from other parts of the bone, or slight misalignment. Problems with the growth plate prevent bones from developing, cause joint distortion, and may cause permanent joint injury. A significant discrepancy between the chronological and assessed ages may indicate a growth problem because determining bone age represents the real level of growth. Therefore, skeletal age estimation is performed to look for endocrine disorders, genetic problems, and growth anomalies. To address the bone age assessment challenge, this study uses the Radiological Society of North America’s Pediatric Bone Age Challenge dataset which contains 12,600 radiological images of the left hand of a patient that includes the gender and bone age information. A bone age evaluation system based on the hand skeleton guidelines is proposed in this study for the detection of hand bone maturation. The proposed approach is based on a customized convolutional neural network. For the calculation of the skeletal age, different data augmentation techniques are used; these techniques not only increase the dataset size but also impact the training of the model. The performance of the model is assessed against the Visual Geometry Group (VGG) model. Results demonstrate that the customized convolutional neural network (CNN) model outperforms the VGG model with 97% accuracy.

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

BackgroundSkeletal 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.ObjectiveWe present Deeplasia, an open-source prior-free deep-learning approach designed for BA assessment specifically validated on patients with skeletal dysplasias.Materials and methodsWe 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.ResultsThe 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.ConclusionWe demonstrated that Deeplasia is competent in assessing the age and monitoring the development of both normal and dysplastic bones.Graphical