Radiological Society Of North America Research Articles

Code and Data Sharing Practices in the Radiology Artificial Intelligence Literature: A Meta-Research Study.

To evaluate code and data sharing practices in original artificial intelligence (AI) scientific manuscripts published in the Radiological Society of North America (RSNA) journals suite from 2017 through 2021. A retrospective meta-research study was conducted of articles published in the RSNA journals suite from January 1, 2017, through December 31, 2021. A total of 218 articles were included and evaluated for code sharing practices, reproducibility of shared code, and data sharing practices. Categorical comparisons were conducted using Fisher exact tests with respect to year and journal of publication, author affiliation(s), and type of algorithm used. Of the 218 included articles, 73 (34%) shared code, with 24 (33% of code sharing articles and 11% of all articles) sharing reproducible code. Radiology and Radiology: Artificial Intelligence published the most code sharing articles (48 [66%] and 21 [29%], respectively). Twenty-nine articles (13%) shared data, and 12 of these articles (41% of data sharing articles) shared complete experimental data by using only public domain datasets. Four of the 218 articles (2%) shared both code and complete experimental data. Code sharing rates were statistically higher in 2020 and 2021 compared with earlier years (P < .01) and were higher in Radiology and Radiology: Artificial Intelligence compared with other journals (P < .01). Original AI scientific articles in the RSNA journals suite had low rates of code and data sharing, emphasizing the need for open-source code and data to achieve transparent and reproducible science.Keywords: Meta-Analysis, AI in Education, Machine LearningSupplemental material is available for this article.© RSNA, 2022.

Standardizing evaluation of patient-specific 3D printed models in surgical planning: development of a cross-disciplinary survey tool for physician and trainee feedback

Background3D printed models are becoming increasingly popular in healthcare as visual and tactile tools to enhance understanding of anatomy and pathology in medical trainee education, provide procedural simulation training, and guide surgical procedures. Patient-specific 3D models are currently being used preoperatively for trainee medical education in planning surgical approaches and intraoperatively to guide decision-making in several specialties. Our study group utilized a modified Delphi process to create a standardized assessment for trainees using patient-specific 3D models as a tool in medical education during pre-surgical planning.MethodsA literature review was conducted to identify survey questions administered to clinicians in published surgical planning studies regarding the use of patient-specific 3D models. A core study team reviewed these questions, removed duplicates, categorized them, mapped them to overarching themes, and, where applicable, modified individual questions into a form generalizable across surgical specialties. The core study panel included a physician, physician-scientist, social scientist, engineer/medical student, and 3D printing lab manager. A modified Delphi process was then used to solicit feedback on the clarity and relevance of the individual questions from an expert panel consisting of 12 physicians from specialties including anesthesiology, emergency medicine, radiology, urology, otolaryngology, and obstetrics/gynecology. When the Radiological Society of North America (RSNA)/American College of Radiology (ACR) 3D Printing Registry Data Dictionary was released, additional survey questions were reviewed. A final cross-disciplinary survey of the utility of 3D printed models in surgical planning medical education was developed.ResultsThe literature review identified 100 questions previously published in surveys assessing patient-specific 3D models for surgical planning. Following the review, generalization, and mapping of survey questions from these studies, a list of 24 questions was generated for review by the expert study team. Five additional questions were identified in the RSNA/ACR 3D Printing Registry Data Dictionary and included for review. A final questionnaire consisting of 20 questions was developed.ConclusionsAs 3D printed models become more common in medical education, the need for standardized assessment is increasingly essential. The standardized questionnaire developed in this study reflects the interests of a variety of stakeholders in patient-specific 3D models across disciplines.

Attention-based multiple-instance learning for Pediatric bone age assessment with efficient and interpretable

Pediatric bone age assessment (BAA) is a common clinical technique for evaluating children's endocrine, genetic, and growth disorders. However, the deep learning BAA method based on global images neglects fine-grained concerns, and regions of interest (ROIs) need additional annotation and complex processing. To overcome these shortcomings, we proposed an interpretable deep-learning architecture based on multiple-instance learning to address BAA efficiently without additional annotations. We cropped the entire image into small patches and got patch features by feature extraction network. Then, an attention backbone ranked feature vectors of the entire image and aggregates its information according to its relative importance. Finally, each image's features and gender were aggregated to predict bone age. The proposed method can identify ROIs by attention-based multi-instance aggregation without additional labels and produce interpretable heatmaps. Moreover, by cropping the complete image into patches and reducing the dimensionality, the proposed model can notice the fine-grained information of the image and improve the model training speed. We validated the proposed method in the Radiological Society of North America 2017 dataset. The results showed that the proposed model achieved an advanced performance of MAE 4.17 months. Furthermore, the visualization results indicated that the proposed model was highly interpretable, which can localize the ROIs without spatial labeling. In conclusion, a novel method for high performance and interpretable bone age prediction without additional manual annotations has been developed, which can be used to effectively assess the pediatric's bone age.

Intracranial hemorrhage subtype classification using learned fully connected separable convolutional network

SummaryIn recent decades, intracranial hemorrhage detection from computed tomography (CT) scans has gained considerable attention among researchers in the medical community. The major problem in dealing with the Radiological Society of North America (RSNA) dataset is a three dimensional representation of CT scan, where the labeled data are scarce and hard to obtain. To highlight this problem, a novel learned fully connected separable convolutional network is proposed in this research article. After collecting the CT scans, data augmentation is used to generate multiple image variations to improve the capacity of the proposed model generalization. Based on the albumentations library, the transformations are selected for data augmentation such as brightness adjustment, horizontal flipping, shifting, rotation, and scaling. The intracranial hemorrhage subtype classification is accomplished utilizing a learned fully connected separable convolutional network which significantly classifies six classes as any, intraparenchymal, subarachnoid, epidural, intraventricular, and subdural. In the resulting phase, the learned fully connected separable convolutional network obtained an average accuracy of 98.63%, sensitivity of 73.32%, specificity of 99.49%, and area under the curve of 98.98%, where the obtained results are effective compared with ResNet‐50, SE‐ResNeXt‐50, ResNeXt‐101, and ResNeXt‐101 with bidirectional long short term memory network.

GAN augmentation for multiclass image classification using hemorrhage detection as a case-study.

Purpose: In recent years, the development and exploration of deeper and more complex deep learning models has been on the rise. However, the availability of large heterogeneous datasets to support efficient training of deep learning models is lacking. While linear image transformations for augmentation have been used traditionally, the recent development of generative adversarial networks (GANs) could theoretically allow us to generate an infinite amount of data from the real distribution to support deep learning model training. Recently, the Radiological Society of North America (RSNA) curated a multiclass hemorrhage detection challenge dataset that includes over 800,000 images for hemorrhage detection, but all high-performing models were trained using traditional data augmentation techniques. Given a wide variety of selections, the augmentation for image classification often follows a trial-and-error policy. Approach: We designed conditional DCGAN (cDCGAN) and in parallel trained multiple popular GAN models to use as online augmentations and compared them to traditional augmentation methods for the hemorrhage case study. Results: Our experimentations show that the super-minority, epidural hemorrhages with cDCGAN augmentation presented a minimum of improvement in their performance against the traditionally augmented model using the same classifier configuration. Conclusion: This shows that for complex and imbalanced datasets, traditional data imbalancing solutions may not be sufficient and require more complex and diverse data augmentation methods such as GANs to solve.

Comparison of the RSNA chest CT classification system and CO-RADS system in reporting COVID-19 pneumonia in symptomatic and asymptomatic patients

BackgroundCoronavirus disease (COVID-19) is a new infection with three pandemic waves up till now. CT plays an important role in diagnosis with multiple reporting systems that can be used during CT analysis. We aimed to compare reporting using the recommendations of the radiological society of North America (RSNA) versus the coronavirus disease reporting and data system (CO-RADS) and to assess the performance of CT if used in asymptomatic patients as a screening. Two hundred and fifty-one patients who underwent chest CT scanning either due to clinical suspicion or as screening before hospital admission were included in this retrospective observational cross-sectional study. This was followed by RT-PCR for confirmation. Three radiologists with different years of experience interpreted the CT findings using the RSNA recommendations and the CO-RADS reporting. The data were collected and compared.ResultsThere was no statistically significant difference noted in the diagnostic accuracy obtained while using the RSNA recommendations and the CO-RADS reporting system. Also, a good inter-rater agreement was noticed while using the two reporting systems. The CT showed a highly significant value while used in the assessment of symptomatic patients in controversy to the screening of asymptomatic patients.ConclusionBoth reporting systems show similar diagnostic accuracy with a good almost similar inter-rater agreement. Both can be used while interpreting the CT images of cases with suspected COVID-19 infection. CT can be used effectively in the detection of COVID-19 infection between symptomatic patients while it is of a lower value in the screening of asymptomatic patients.

Clinical significance of chest CT for the exclusion of COVID-19 in pre-admission screening: Is it worthwhile using chest CT with reverse-transcription polymerase chain reaction test?

BackgroundA single reverse-transcription polymerase chain reaction (RT-PCR) test is not sufficient to exclude COVID-19 in hospital pre-admission screening. However, repeated RT-PCR tests are time-consuming. This study investigates the utility of chest computed tomography (CT) for COVID-19 screening in asymptomatic patients.MethodsBetween April 2020 and March 2021, RT-PCR testing and chest CT were performed to screen COVID-19 in 10 823 asymptomatic patients prior to admission. Chest CT findings were retrospectively evaluated using the reporting system of the Radiological Society of North America. Using RT-PCR results as a reference, we assessed the diagnostic efficacy of chest CT during both the low- and high-prevalence periods of the COVID-19 pandemic.ResultsFollowing a positive RT-PCR test, 20 asymptomatic patients (0.18%) were diagnosed with COVID-19; in the low-prevalence period, 5 of 6556 patients (0.076%) were positive; and in the high-prevalence period, 15 of 4267 patients (0.35%) were positive. Of the 20 asymptomatic COVID-19 positive patients, chest CT results were positive for COVID-19 pneumonia in 8 patients. Chest CT results were false-positive in 185 patients (1.7% false-positive rate, and 60% false-negative rate). Pneumonia that was classified as a “typical appearance” of COVID-19 reported as false-positives in 36 of 39 patients (92.3%). Across the study period, the diagnostic efficacy of “typical appearance” on chest CT were characterized by a sensitivity, specificity, accuracy, positive predictive value (PPV), and negative predictive value (NPV) of 15%, 99.7%, 99.7%, 7.7%, and 99.8%; 20%, 99.6%, 99.6%, 4%, and 99.9%; and 13.3%, 99.7%, 99.7%, 14.3%, and 99.7%, in the entire study, low-, and high-prevalence periods, respectively.ConclusionsAddition of chest CT to RT-PCR testing provides no benefit to the detection of COVID-19 in asymptomatic patients.

Accuracy and self-validation of automated bone age determination

The BoneXpert method for automated determination of bone age from hand X-rays was introduced in 2009 and is currently running in over 200 hospitals. The aim of this work is to present version 3 of the method and validate its accuracy and self-validation mechanism that automatically rejects an image if it is at risk of being analysed incorrectly. The training set included 14,036 images from the 2017 Radiological Society of North America (RSNA) Bone Age Challenge, 1642 images of normal Dutch and Californian children, and 8250 images from Tübingen from patients with Short Stature, Congenital Adrenal Hyperplasia and Precocious Puberty. The study resulted in a cross-validated root mean square (RMS) error in the Tübingen images of 0.62 y, compared to 0.72 y in the previous version. The RMS error on the RSNA test set of 200 images was 0.45 y relative to the average of six manual ratings. The self-validation mechanism rejected 0.4% of the RSNA images. 121 outliers among the self-validated images of the Tübingen study were rerated, resulting in 6 cases where BoneXpert deviated more than 1.5 years from the average of the three re-ratings, compared to 72 such cases for the original manual ratings. The accuracy of BoneXpert is clearly better than the accuracy of a single manual rating. The self-validation mechanism rejected very few images, typically with abnormal anatomy, and among the accepted images, there were 12 times fewer severe bone age errors than in manual ratings, suggesting that BoneXpert could be safer than manual rating.

Applicability of CO-RADS in an Anonymized Cohort Including Earlyand Advanced Stages of COVID-19 in Comparison to the Recommendations of the German Radiological Society and Radiological Society of North America.

Classifications were created to facilitate radiological evaluation of the novel coronavirus disease 2019 (COVID-19) on computed tomography (CT) images. The categorical CT assessment scheme (CO-RADS) categorizes lung parenchymal changes according to their likelihood of being caused by SARS-CoV-2 infection. This study investigates the diagnostic accuracy of diagnosing COVID-19 with CO-RADS compared to the Thoracic Imaging Section of the German Radiological Society (DRG) classification and Radiological Society of North America (RSNA) classification in an anonymized patient cohort. To mimic advanced disease stages, follow-up examinations were included as well. This study includes all patients undergoing chest CT in the case of a suspected SARS-CoV-2 infection or an already confirmed infection between March 13 and November 30, 2020. During the study period, two regional lockdowns occurred due to high incidence values, increasing the pre-test probability of COVID-19. Anonymized CT images were reviewed retrospectively and in consensus by two radiologists applying CO-RADS, DRG, and RSNA classification. Afterwards, CT findings were compared to results of sequential real-time reverse transcriptase polymerase chain reaction (qPCR) test performed during hospitalization to determine statistical analysis for diagnosing COVID-19. 536 CT examinations were included. CO-RADS, DRG and RSNA achieved an NPV of 96 %/94 %/95 % (CO-RADS/DRG/RSNA), PPV of 83 %/80 %/88 %, sensitivity of 86 %/76 %/80 %, and specificity of 96 %/95 %/97 %. The disease prevalence was 20 %. All applied classifications can reliably exclude a SARS-CoV-2 infection even in an anonymous setting. Nevertheless, pre-test probability was high in our study setting and has a great influence on the classifications. Therefore, the applicability of the individual classifications will become apparent in the future with lower prevalence and incidence of COVID-19. · CO-RADS, DRG, and RSNA classifications help to reliably detect infected patients in an anonymized setting. · Pre-test probability has a great influence on the individual classifications. · Difficulties in an anonymized study setting are severe pulmonary changes and residuals.. · Valentin B, Steuwe A, Wienemann T et al. Applicability of CO-RADS in an Anonymized Cohort Including Early and Advanced Stages of COVID-19 in Comparison to the Recommendations of the German Radiological Society andRadiological Society of North America. Fortschr Röntgenstr 2022; 194: 862 - 872.

MEDUSA: Multi-Scale Encoder-Decoder Self-Attention Deep Neural Network Architecture for Medical Image Analysis.

Medical image analysis continues to hold interesting challenges given the subtle characteristics of certain diseases and the significant overlap in appearance between diseases. In this study, we explore the concept of self-attention for tackling such subtleties in and between diseases. To this end, we introduce, a multi-scale encoder-decoder self-attention (MEDUSA) mechanism tailored for medical image analysis. While self-attention deep convolutional neural network architectures in existing literature center around the notion of multiple isolated lightweight attention mechanisms with limited individual capacities being incorporated at different points in the network architecture, MEDUSA takes a significant departure from this notion by possessing a single, unified self-attention mechanism with significantly higher capacity with multiple attention heads feeding into different scales in the network architecture. To the best of the authors' knowledge, this is the first “single body, multi-scale heads” realization of self-attention and enables explicit global context among selective attention at different levels of representational abstractions while still enabling differing local attention context at individual levels of abstractions. With MEDUSA, we obtain state-of-the-art performance on multiple challenging medical image analysis benchmarks including COVIDx, Radiological Society of North America (RSNA) RICORD, and RSNA Pneumonia Challenge when compared to previous work. Our MEDUSA model is publicly available.

Disparate participation by gender of conference attendants in scientific discussions.

One important metric of a radiologist’s visibility and influence is their ability to participate in discussion within their community. The goal of our study was to compare the participation level of men and women in scientific discussions at the annual meeting of the Radiological Society of North America (RSNA). Eleven volunteers collected participation data by gender in 59 sessions (286 presentations) at the 2018 RSNA meeting. Data was analyzed using a combination of Chi-squared, paired Wilcoxon signed-rank and T-test. Of all RSNA professional attendees at the RSNA, 68% were men and 32% were women. Of the 2869 presentations listed in the program, 65% were presented by men and 35% were presented by women. Of the 286 presentations in our sample, 177 (61.8%) were presented by men and 109 (38.1%) were presented by women. Of these 286 presentations, 81 (63%) were moderated by men and 47 (37%) were moderated by women. From the audience, 190 male attendees participated in 134 question-and-answer (Q&A) sessions following presentations and 58 female attendees participated in 52 Q&A sessions (P<0.001). Female attendees who did participate in Q&A sessions talked for a significantly shorter period of time (mean 7.14 ± 17.7 seconds, median 0) compared to male attendees (28.7 ± 29.6 seconds, median 16; P<0.001). Overall, our findings demonstrate that women participated less than men in the Q&A sessions at RSNA 2018, and talked for a shorter period of time. The fact that women were outnumbered among their male peers may explain the difference in behavior by gender.

Comparison of chest CT scan findings between COVID-19 and pulmonary contusion in trauma patients based on RSNA criteria: Established novel criteria for trauma victims

ProposeIn this study, we re-assessed the criteria defined by the radiological society of North America (RSNA) to determine novel radiological findings helping the physicians differentiating COVID-19 from pulmonary contusion. MethodsAll trauma patients with blunt chest wall trauma and subsequent pulmonary contusion, COVID-19-related signs and symptoms before the trauma were enrolled in this retrospective study from February to May 2020. Included patients (Group P) were then classified into two groups based on polymerase chain reaction tests (Group Pa for positive patients and Pb for negative ones). Moreover, 44 patients from the pre-pandemic period (Group PP) were enrolled. They were matched to Group P regarding age, sex, and trauma-related scores. Two radiologists blindly reviewed the CT images of all enrolled patients according to criteria defined by the RSNA criteria. The radiological findings were compared between Group P and Group PP; statistically significant ones were re-evaluated between Group Pa and Group Pb thereafter. Finally, the sensitivity and specificity of each significant findings were calculated. The Chi-square test was used to compare the radiological findings between Group P and Group PP. ResultsIn the Group PP, 73.7% of all ground-glass opacities (GGOs) and 80% of all multiple bilateral GGOs were detected (p < 0.001 and p = 0.25, respectively). Single bilateral GGOs were only seen among the Group PP. The Chi-square tests showed that the prevalence of diffused GGOs, multiple unilateral GGOs, multiple consolidations, and multiple bilateral consolidations were significantly higher in the Group P (p = 0.001, 0.01, 0.003, and 0.003, respectively). However, GGOs with irregular borders and single consolidations were more significant among the Group PP (p = 0.01 and 0.003, respectively). Of note, reticular distortions and subpleural spares were exclusively detected in the Group PP. ConclusionWe concluded that the criteria set by RSNA for the diagnosis of COVID-19 are not appropriate in trauma patients. The clinical signs and symptoms are not always useful either. The presence of multiple unilateral GGOs, diffused GGOs, and multiple bilateral consolidations favor COVID-19 with 88%, 97.62%, and 77.7% diagnostic accuracy.

Bruce G. Haffty, MD, President, Radiological Society of North America, 2022.

Bruce G. Haffty, MD, President, Radiological Society of North America, 2022.

Validation of the North America expert consensus statement on reporting CT findings for COVID-19 in individuals with lung cancer.

The aim of our study was to validate the use of the standardized Radiological Society of North America (RSNA) reporting system in individuals with known lung cancer who presented to the emergency department with suspected COVID-19. We included patients aged 18 years or older from the Cancer Institute of the State of São Paulo (ICESP) with a confirmed diagnosis of lung cancer, admitted to the emergency department and undergoing chest computed tomography (CT) for suspicion of COVID-19. Comparison between SARS-CoV2 RT-PCR across RSNA categories was performed in all patients and further stratified by diagnosis of lung cancer progression. Among 58 individuals included in the analysis (65±9 years, 43% men), 20 had positive RT-PCR. Less than a half (43%) had no new lung findings in the CT. Positive RT-PCR was present in 75% of those with typical findings according to RSNA and in only 9% when these findings were classified as atypical or negative (P<0.001). Diagnostic accuracy was even higher when stratified by the presence or absence of progressive disease (PD). Extent of pulmonary inflammatory changes was strongly associated with higher mortality, reaching a lethality of 83% in patients with >25% of lung involvement and 100% when there was >50% of lung involvement. The lung involvement score was also highly predictive of prognosis in this population as was reported for non-lung cancer individuals. Collectively, our results demonstrated that diagnostic and prognostic values of chest CT findings in COVID-19 are robust to the presence of lung abnormalities related to lung cancer.

Comparison of Demographic Profile and CT Findings in COVID-19 Pneumonia between First and Second Waves of COVID-19 Pandemic: A Retrospective Study

Introduction: The Coronavirus Disease 2019 (COVID-19) pandemic has affected the world in two waves. The human civilisation took a heavy toll in the two waves of pandemic in terms of human deaths, loss of livelihood and socioeconomic regression. Aim: To present the Computed Tomography (CT) thorax findings of COVID-19 pneumonia during peak three months of first and second waves of COVID-19 pandemic and compare the demographic profile, radiological patterns and CT severity scores to document if significant difference exists between the two waves. Materials and Methods: The present observational, retrospective and single-centric study included all Real-time Reverse Transcription-polymerase Chain Reaction (RT-PCR) tests proven COVID-19 patients who underwent CT scan during the peak three months of first (August to October 2020) and second (April to June 2021) waves of COVID-19 pandemic at TRIHMS, Naharlagun, Arunachal Pradesh, India. A total of 93 and 120 patients were included from the first and second waves, respectively. The demographic profile, radiological pattern based on the Radiological Society of North America (RSNA) consensus statement and CT severity index in COVID- 19 pneumonia were compared during the peak three months of first and second waves of the pandemic. Chi-squared and t-test were used for statistical analysis. Results: The patients affected during the second wave were younger compared to the first wave (mean age of 43 vs 49, p=0.001). The patients having moderate or severe COVID-19 pneumonia during the second wave were significantly younger compared to the first wave (mean age of 42 vs 55, p&lt;0.001). More females were affected with moderate or severe COVID-19 pneumonia during the second wave compared to the first wave (p=0.017). However, the overall sex difference in COVID-19 pneumonia (including mild category) was not significant with p-value of 0.127. There was same prevalence of radiological patterns in the two waves (p=0.981). The difference in the prevalence of CT severity score was not significant in the two waves (p=0.256). Conclusion: Similar radiological patterns were observed in the two waves of the pandemic. Younger patients were significantly more affected during the second wave of the pandemic.

Diagnostic performance of the RSNA-proposed classification for COVID-19 pneumonia versus pre-pandemic controls

ObjectiveTo evaluate the diagnostic accuracy of the Radiological Society of North America (RSNA) classification system for coronavirus disease 2019 (COVID-19) pneumonia compared to pre-pandemic chest computed tomography (CT) scan images to mitigate the risk of bias regarding the reference standard. Materials and methodsThis was a retrospective, cross-sectional, diagnostic test accuracy study. Chest CT scans, carried out from May 1 to June 30, 2020, and from May 1 to July 17, 2017, were consecutively selected for the COVID-19 (positive reverse transcription-polymerase chain reaction [RT-PCR] for severe acute respiratory syndrome coronavirus 2 result) and control (pre-pandemic) groups, respectively. Four expert thoracic radiologists blindly interpreted each CT scan image. Sensitivity and specificity were calculated. ResultsA total of 160 chest CT scan images were included: 79 in the COVID-19 group (56 [43.5–67] years old, 41 men) and 81 in the control group (62 [52–72] years old, 44 men). Typically, an estimated specificity of 98.5% (95% confidence interval [CI] 98.1%–98.4%) was obtained. For the indeterminate classification as a diagnostic threshold, an estimated sensitivity of 88.3% (95% CI 84.7%–91.7%) and a specificity of 79.0% (95% CI 74.5%–83.4%), with an area under the curve of 0.865 (95% CI 0.838–0.895), were obtained. ConclusionThe RSNA classification system shows strong diagnostic accuracy for COVID-19 pneumonia, even against pre-pandemic controls. It can be an important aid in clinical decision-making, especially when a typical or indeterminate pattern is found, possibly advising retesting following an initial negative RT-PCR result and streamlining early management and isolation.

Fully automatic deep convolutional approaches for the analysis of COVID-19 using chest X-ray images

Covid-19 is a new infectious disease caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). Given the seriousness of the situation, the World Health Organization declared a global pandemic as the Covid-19 rapidly around the world. Among its applications, chest X-ray images are frequently used for an early diagnostic/screening of Covid-19 disease, given the frequent pulmonary impact in the patients, critical issue to prevent further complications caused by this highly infectious disease.In this work, we propose 4 fully automatic approaches for the classification of chest X-ray images under the analysis of 3 different categories: Covid-19, pneumonia and healthy cases. Given the similarity between the pathological impact in the lungs between Covid-19 and pneumonia, mainly during the initial stages of both lung diseases, we performed an exhaustive study of differentiation considering different pathological scenarios. To address these classification tasks, we evaluated 6 representative state-of-the-art deep network architectures on 3 different public datasets: (I) Chest X-ray dataset of the Radiological Society of North America (RSNA); (II) Covid-19 Image Data Collection; (III) SIRM dataset of the Italian Society of Medical Radiology. To validate the designed approaches, several representative experiments were performed using 6,070 chest X-ray radiographs. In general, satisfactory results were obtained from the designed approaches, reaching a global accuracy values of 0.9706 ± 0.0044, 0.9839 ± 0.0102, 0.9744 ± 0.0104 and 0.9744 ± 0.0104, respectively, thus helping the work of clinicians in the diagnosis and consequently in the early treatment of this relevant pandemic pathology.

Deep Learning Framework for Automatic Bone Age Assessment.

Bone age Assessment or the skeletal age is a general clinical practice to detect endocrine and metabolic disarrangement in child development. The bone age indicates the level of structural and biological growth better than chronological age calculated from the birth date. The X-Ray of the wrist and hand is used in common to estimate the bone age of a person. The degree of agreement among the automated methods used to evaluate the X-rays is more than any other manual method. In this work, we propose a fully automated deep learning approach for bone age assessment. The dataset used is from the 2017 Pediatric Bone Age Challenge released by the Radiological Society of North America. Each X-Ray image in this dataset is an image of a left hand tagged with the age and gender of the patient. Transfer learning is employed by using pre-trained neural network architecture. InceptionV3 architecture is used in the present work, and the difference between the actual and predicted age obtained is 5.921 months.Clinical Relevance- This provides an AI-based computer assistance system as a supplement tool to help clinicians make bone age predictions.

Imaging AI in Practice: A Demonstration of Future Workflow Using Integration Standards

Artificial intelligence (AI) tools are rapidly being developed for radiology and other clinical areas. These tools have the potential to dramatically change clinical practice; however, for these tools to be usable and function as intended, they must be integrated into existing radiology systems. In a collaborative effort between the Radiological Society of North America, radiologists, and imaging-focused vendors, the Imaging AI in Practice (IAIP) demonstrations were developed to show how AI tools can generate, consume, and present results throughout the radiology workflow in a simulated clinical environment. The IAIP demonstrations highlight the critical importance of semantic and interoperability standards, as well as orchestration profiles for successful clinical integration of radiology AI tools. Keywords: Computer Applications-General (Informatics), Technology Assessment © RSNA, 2021.

Chest computed tomography in the diagnosis of COVID-19 in patients with false negative RT-PCR.

ABSTRACTObjective To evaluate the role of chest computed tomography in patients with COVID-19 who presented initial negative result in reverse transcriptase-polymerase chain reaction (RT-PCR).Methods A single-center, retrospective study that evaluated 39 patients with negative RT-PCR for COVID-19, who underwent chest computed tomography and had a final clinical or serological diagnosis of COVID-19. The visual tomographic classification was evaluated according to the Consensus of the Radiological Society of North America and software developed with artificial intelligence for automatic detection of findings and chance estimation of COVID-19.Results In the visual tomographic analysis, only one of them (3%) presented computed tomography classified as negative, 69% were classified as typical and 28% as indeterminate. In the evaluation using the software, only four (about 10%) had a probability of COVID-19 <25%.Conclusion Computed tomography can play an important role in management of suspected cases of COVID-19 with initial negative results in RT-PCR, especially considering those patients outside the ideal window for sample collection for RT-PCR.