AIROGS: Artificial Intelligence for Robust Glaucoma Screening Challenge.

The early detection of glaucoma is essential in preventing visual impairment. Artificial intelligence (AI) can be used to analyze color fundus photographs (CFPs) in a cost-effective manner, making glaucoma screening more accessible. While AI models for glaucoma screening from CFPs have shown promising results in laboratory settings, their performance decreases significantly in real-world scenarios due to the presence of out-of-distribution and low-quality images. To address this issue, we propose the Artificial Intelligence for Robust Glaucoma Screening (AIROGS) challenge. This challenge includes a large dataset of around 113,000 images from about 60,000 patients and 500 different screening centers, and encourages the development of algorithms that are robust to ungradable and unexpected input data. We evaluated solutions from 14 teams in this paper and found that the best teams performed similarly to a set of 20 expert ophthalmologists and optometrists. The highest-scoring team achieved an area under the receiver operating characteristic curve of 0.99 (95% CI: 0.98-0.99) for detecting ungradable images on-the-fly. Additionally, many of the algorithms showed robust performance when tested on three other publicly available datasets. These results demonstrate the feasibility of robust AI-enabled glaucoma screening.

Bypassing Heaven's Gate Technique Using Black-Box Testing.

In recent years, the number and sophistication of malware attacks on computer systems have increased significantly. One technique employed by malware authors to evade detection and analysis, known as Heaven's Gate, enables 64-bit code to run within a 32-bit process. Heaven's Gate exploits a feature in the operating system that allows the transition from a 32-bit mode to a 64-bit mode during execution, enabling the malware to evade detection by security software designed to monitor only 32-bit processes. Heaven's Gate poses significant challenges for existing security tools, including dynamic binary instrumentation (DBI) tools, widely used for program analysis, unpacking, and de-virtualization. In this paper, we provide a comprehensive analysis of the Heaven's Gate technique. We also propose a novel approach to bypass the Heaven's Gate technique using black-box testing. Our experimental results show that the proposed approach effectively bypasses and prevents the Heaven's Gate technique and strengthens the capabilities of DBI tools in combating advanced malware threats.

Deep learning and clustering approaches for dental implant size classification based on periapical radiographs.

This study investigated two artificial intelligence (AI) methods for automatically classifying dental implant diameter and length based on periapical radiographs. The first method, deep learning (DL), involved utilizing the pre-trained VGG16 model and adjusting the fine-tuning degree to analyze image data obtained from periapical radiographs. The second method, clustering analysis, was accomplished by analyzing the implant-specific feature vector derived from three key points coordinates of the dental implant using the k-means++ algorithm and adjusting the weight of the feature vector. DL and clustering model classified dental implant size into nine groups. The performance metrics of AI models were accuracy, sensitivity, specificity, F1-score, positive predictive value, negative predictive value, and area under the receiver operating characteristic curve (AUC-ROC). The final DL model yielded performances above 0.994, 0.950, 0.994, 0.974, 0.952, 0.994, and 0.975, respectively, and the final clustering model yielded performances above 0.983, 0.900, 0.988, 0.923, 0.909, 0.988, and 0.947, respectively. When comparing the AI model before tuning and the final AI model, statistically significant performance improvements were observed in six out of nine groups for DL models and four out of nine groups for clustering models based on AUC-ROC. Two AI models showed reliable classification performances. For clinical applications, AI models require validation on various multicenter data.

Deep learning model for distinguishing Mayo endoscopic subscore 0 and 1 in patients with ulcerative colitis.

The aim of this study was to address the issue of differentiating between Mayo endoscopic subscore (MES) 0 and MES 1 using a deep learning model. A dataset of 492 ulcerative colitis (UC) patients who demonstrated MES improvement between January 2018 and December 2019 at Samsung Medical Center was utilized. Specifically, two representative images of the colon and rectum were selected from each patient, resulting in a total of 984 images for analysis. The deep learning model utilized in this study consisted of a convolutional neural network (CNN)-based encoder, with two auxiliary classifiers for the colon and rectum, as well as a final MES classifier that combined image features from both inputs. In the internal test, the model achieved an F1-score of 0.92, surpassing the performance of seven novice classifiers by an average margin of 0.11, and outperforming their consensus by 0.02. The area under the receiver operating characteristic curve (AUROC) was calculated to be 0.97 when considering MES 1 as positive, with an area under the precision-recall curve (AUPRC) of 0.98. In the external test using the Hyperkvasir dataset, the model achieved an F1-score of 0.89, AUROC of 0.86, and AUPRC of 0.97. The results demonstrate that the proposed CNN-based model, which integrates image features from both the colon and rectum, exhibits superior performance in accurately discriminating between MES 0 and MES 1 in patients with UC.

Clinical relevance of deep learning models in predicting the onset timing of cancer pain exacerbation.

Cancer pain is a challenging clinical problem that is encountered in the management of cancer pain. We aimed to investigate the clinical relevance of deep learning models that predict the onset of cancer pain exacerbation in hospitalized patients. We defined cancer pain exacerbation (CPE) as the pain with a numerical rating scale (NRS) score of ≥ 4. We investigated the performance of the deep learning models using the Matthews correlation coefficient (MCC) with different input lengths and time binning. All the pain records were obtained from the electronic medical records of the hematology-oncology wards in a Samsung Medical Center between July 2016 and February 2020. The model was externally validated using the holdout method with 20% of the datasets. The most common type of cancer was lung cancer (n = 745, 21.7%), and the median CPE per day was 1.01. The NRS pain records showed circadian patterns that correlated with NRS pain patterns of the previous days. The correlation of the NRS scores showed a positive association with the closeness of the NRS pattern of the day with forecast date and size of time binning. The long short-term memory-based model exhibited a good performance by demonstrating 9 times the best performance and 8 times the second-best performance among 21 different settings. The best performance was achieved with 120 h input and 12 h bin lengths (MCC: 0.4927). Our study demonstrated the possibility of predicting CPE using deep learning models, thereby suggesting that preemptive cancer pain management using deep learning could potentially improve patients' daily life.

Fluidic self-assembly for MicroLED displays by controlled viscosity.

Displays in which arrays of microscopic 'particles', or chiplets, of inorganic light-emitting diodes (LEDs) constitute the pixels, termed MicroLED displays, have received considerable attention1,2 because they can potentially outperform commercially available displays based on organic LEDs3,4 in terms of power consumption, colour saturation, brightness and stability and without image burn-in issues1,2,5-7. To manufacture these displays, LED chiplets must be epitaxially grown on separate wafers for maximum device performance and then transferred onto the display substrate. Given that the number of LEDs needed for transfer is tremendous-for example, more than 24 million chiplets smaller than 100 µm are required for a 50-inch, ultra-high-definition display-a technique capable of assembling tens of millions of individual LEDs at low cost and high throughput is needed to commercialize MicroLED displays. Here we demonstrate a MicroLED lighting panel consisting of more than 19,000 disk-shaped GaN chiplets, 45 µm in diameter and 5 µm in thickness, assembled in 60 s by a simple agitation-based, surface-tension-driven fluidic self-assembly (FSA) technique with a yield of 99.88%. The creation of this level of large-scale, high-yield FSA of sub-100-µm chiplets was considered a significant challenge because of the low inertia of the chiplets. Our key finding in overcoming this difficulty is that the addition of a small amount of poloxamer to the assembly solution increases its viscosity which, in turn, increases liquid-to-chiplet momentum transfer. Our results represent significant progress towards the ultimate goal of low-cost, high-throughput manufacture of full-colour MicroLED displays by FSA.

Concurrent self-assembly of RGB microLEDs for next-generation displays.

MicroLED displays have been in the spotlight as the next-generation displays owing to their various advantages, including long lifetime and high brightness compared with organic light-emitting diode (OLED) displays. As a result, microLED technology1,2 is being commercialized for large-screen displays such as digital signage and active R&D programmes are being carried out for other applications, such as augmented reality3, flexible displays4 and biological imaging5. However, substantial obstacles in transfer technology, namely, high throughput, high yield and production scalability up to Generation 10+ (2,940 × 3,370 mm2) glass sizes, need to be overcome so that microLEDs can enter mainstream product markets and compete with liquid-crystal displays and OLED displays. Here we present a new transfer method based on fluidic self-assembly (FSA) technology, named magnetic-force-assisted dielectrophoretic self-assembly technology (MDSAT), which combines magnetic and dielectrophoresis (DEP) forces to achieve a simultaneous red, green and blue (RGB) LED transfer yield of 99.99% within 15 min. By embedding nickel, a ferromagnetic material, in the microLEDs, their movements were controlled by using magnets, and by applying localized DEP force centred around the receptor holes, these microLEDs were effectively captured and assembled in the receptor site. Furthermore, concurrent assembly of RGB LEDs were demonstrated through shape matching between microLEDs and receptors. Finally, a light-emitting panel was fabricated, showing damage-free transfer characteristics and uniform RGB electroluminescence emission, demonstrating our MDSAT method to be an excellent transfer technology candidate for high-volume production of mainstream commercial products.

Validation of resonance Raman spectroscopy-measured skin carotenoid status as a biomarker for fruit and vegetable intake in Korean adults.

Blood carotenoid concentration measurement is considered the gold standard for fruit and vegetable (F&V) intake estimation; however, this method is invasive and expensive. Recently, skin carotenoid status (SCS) measured by optical sensors has been evaluated as a promising parameter for F&V intake estimation. In this cross-sectional study, we aimed to validate the utility of resonance Raman spectroscopy (RRS)-assessed SCS as a biomarker of F&V intake in Korean adults. We used data from 108 participants aged 20-69 years who completed SCS measurements, blood collection and 3-d dietary recordings. Serum carotenoid concentrations were quantified using HPLC, and dietary carotenoid and F&V intakes were estimated via 3-d dietary records using a carotenoid database for common Korean foods. The correlations of the SCS with serum carotenoid concentrations, dietary carotenoid intake and F&V intake were examined to assess SCS validity. SCS was positively correlated with total serum carotenoid concentration (r = 0·52, 95 % CI = 0·36, 0·64, P < 0·001), serum ß-carotene concentration (r = 0·60, 95 % CI = 0·47, 0·71, P < 0·001), total carotenoid intake (r = 0·20, 95 % CI = 0·01, 0·37, P = 0·04), ß-carotene intake (r = 0·30, 95 % CI = 0·11, 0·46, P = 0·002) and F&V intake (r = 0·40, 95 % CI = 0·23, 0·55, P < 0·001). These results suggest that SCS can be a valid biomarker of F&V intake in Korean adults.

Deep learning approach to detection of colonoscopic information from unstructured reports.

BACKGROUND: Colorectal cancer is a leading cause of cancer deaths. Several screening tests, such as colonoscopy, can be used to find polyps or colorectal cancer. Colonoscopy reports are often written in unstructured narrative text. The information embedded in the reports can be used for various purposes, including colorectal cancer risk prediction, follow-up recommendation, and quality measurement. However, the availability and accessibility of unstructured text data are still insufficient despite the large amounts of accumulated data. We aimed to develop and apply deep learning-based natural language processing (NLP) methods to detect colonoscopic information. METHODS: This study applied several deep learning-based NLP models to colonoscopy reports. Approximately 280,668 colonoscopy reports were extracted from the clinical data warehouse of Samsung Medical Center. For 5,000 reports, procedural information and colonoscopic findings were manually annotated with 17 labels. We compared the long short-term memory (LSTM) and BioBERT model to select the one with the best performance for colonoscopy reports, which was the bidirectional LSTM with conditional random fields. Then, we applied pre-trained word embedding using large unlabeled data (280,668 reports) to the selected model. RESULTS: The NLP model with pre-trained word embedding performed better for most labels than the model with one-hot encoding. The F1 scores for colonoscopic findings were: 0.9564 for lesions, 0.9722 for locations, 0.9809 for shapes, 0.9720 for colors, 0.9862 for sizes, and 0.9717 for numbers. CONCLUSIONS: This study applied deep learning-based clinical NLP models to extract meaningful information from colonoscopy reports. The method in this study achieved promising results that demonstrate it can be applied to various practical purposes.

Changes in CT-Based Morphological Features of the Kidney with Declining Glomerular Filtration Rate in Chronic Kidney Disease.

Chronic kidney disease (CKD) progression involves morphological changes in the kidney, such as decreased length and thickness, with associated histopathological alterations. However, the relationship between morphological changes in the kidneys and glomerular filtration rate (GFR) has not been quantitatively and comprehensively evaluated. We evaluated the three-dimensional size and shape of the kidney using computed tomography (CT)-derived features in relation to kidney function. We included 257 patients aged ≥18 years who underwent non-contrast abdominal CT at the Inha University Hospital. The features were quantified using predefined algorithms in the pyRadiomics package after kidney segmentation. All features, except for flatness, significantly correlated with estimated GFR (eGFR). The surface-area-to-volume ratio (SVR) showed the strongest negative correlation (r = -0.75, p < 0.0001). Kidney size features, such as volume and diameter, showed moderate to high positive correlations; other morphological features showed low to moderate correlations. The calculated area under the receiver operating characteristic (ROC) curve (AUC) for different features ranged from 0.51 (for elongation) to 0.86 (for SVR) for different eGFR thresholds. Diabetes patients had weaker correlations between the studied features and eGFR and showed less bumpy surfaces in three-dimensional visualization. We identified alterations in the CKD kidney based on various three-dimensional shape and size features, with their potential diagnostic value.

Task-Specific Adaptive Differential Privacy Method for Structured Data.

Data are needed to train machine learning (ML) algorithms, and in many cases often include private datasets that contain sensitive information. To preserve the privacy of data used while training ML algorithms, computer scientists have widely deployed anonymization techniques. These anonymization techniques have been widely used but are not foolproof. Many studies showed that ML models using anonymization techniques are vulnerable to various privacy attacks willing to expose sensitive information. As a privacy-preserving machine learning (PPML) technique that protects private data with sensitive information in ML, we propose a new task-specific adaptive differential privacy (DP) technique for structured data. The main idea of the proposed DP method is to adaptively calibrate the amount and distribution of random noise applied to each attribute according to the feature importance for the specific tasks of ML models and different types of data. From experimental results under various datasets, tasks of ML models, different DP mechanisms, and so on, we evaluate the effectiveness of the proposed task-specific adaptive DP method. Thus, we show that the proposed task-specific adaptive DP technique satisfies the model-agnostic property to be applied to a wide range of ML tasks and various types of data while resolving the privacy-utility trade-off problem.

Associations of Cardiorespiratory Fitness With Estimated Remnant Cholesterol and Non-High-Density Lipoprotein Cholesterol in Healthy Men.

Remnant cholesterol (RC) and non-high-density lipoprotein cholesterol (non-HDL-C) may contribute to the residual risk for atherosclerotic cardiovascular disease. High cardiorespiratory fitness (CRF) is associated with favorable traditional lipid profiles, but its relation with RC and non-HDL-C remains unclear. We analyzed cross-sectional data on 4,613 healthy men (mean age 49 years). CRF was measured using peak oxygen uptake during incremental exercise testing and categorized into quartiles. RC was estimated as total cholesterol minus HDL-C and low-density lipoprotein cholesterol, and elevated RC was defined as ≥38 mg/100 ml (90 percentile). Non-HDL-C was calculated as total cholesterol minus HDL-C, and high non-HLD-C was defined as ≥190 mg/100 ml. CRF was inversely associated with RC (ß -0.31, 95% confidence interval [CI] -0.39 to -0.24) and non-HDL-C (ß -0.34, 95% CI -0.57 to -0.11) after adjustment for several risk factors. Each metabolic equivalent increment in CRF was associated with lower odds of having elevated RC (odds ratio [OR] 0.85, 95% CI 0.77 to 0.93) and non-HDL-C (OR 0.93, 95% CI 0.85 to 1.00) in multivariable analysis. Compared with the bottom quartile, the top quartile of CRF had significantly lower odds of elevated RC (OR 0.63, 95% CI 0.45 to 0.88) and non-HDL-C (OR 0.68, 95% CI 0.51 to 0.91). In conclusion, higher CRF was independently associated with lower levels of RC and non-HDL-C and lower odds of the prevalence of elevated RC and non-HDL-C in healthy men.

Outcomes of Endoscopic Resection for Early Gastric Cancer in Very Elderly Patients: A Nationwide Population-Based Study.

Background/Aims: Few studies have investigated the long-term outcomes of endoscopic resection for early gastric cancer (EGC) in very elderly patients. The aim of this study was to determine the appropriate treatment strategy and identify the risk factors for mortality in these patients. Methods: Patients with EGC who underwent endoscopic resection from 2006 to 2017 were identified using National Health Insurance Data and divided into three age groups: very elderly (≥85 years), elderly (65 to 84 years), and non-elderly (≤64 years). Their long- and short-term outcomes were compared in the three age groups, and the survival in the groups was compared with that in the control group, matched by age and sex. We also evaluated the risk factors for long- and short-term outcomes. Results: A total of 8,426 patients were included in our study: 118 very elderly, 4,583 elderly, and 3,725 non-elderly. The overall survival and cancer-specific survival rates were significantly lower in the very elderly group than in the elderly and the non-elderly groups. Congestive heart failure was negatively associated with cancer-specific survival. A significantly decreased risk for mortality was observed in all groups (p<0.001). The very elderly group had significantly higher readmission and mortality rates within 3 months of endoscopic resection than the non-elderly and elderly groups. Furthermore, the cerebrovascular disease was associated with mortality within 3 months after endoscopic resection. Conclusions: Endoscopic resection for EGC can be helpful for very elderly patients, and it may play a role in achieving overall survival comparable to that of the control group.

Clustering Approach for Detecting Multiple Types of Adversarial Examples.

With intentional feature perturbations to a deep learning model, the adversary generates an adversarial example to deceive the deep learning model. As an adversarial example has recently been considered in the most severe problem of deep learning technology, its defense methods have been actively studied. Such effective defense methods against adversarial examples are categorized into one of the three architectures: (1) model retraining architecture; (2) input transformation architecture; and (3) adversarial example detection architecture. Especially, defense methods using adversarial example detection architecture have been actively studied. This is because defense methods using adversarial example detection architecture do not make wrong decisions for the legitimate input data while others do. In this paper, we note that current defense methods using adversarial example detection architecture can classify the input data into only either a legitimate one or an adversarial one. That is, the current defense methods using adversarial example detection architecture can only detect the adversarial examples and cannot classify the input data into multiple classes of data, i.e., legitimate input data and various types of adversarial examples. To classify the input data into multiple classes of data while increasing the accuracy of the clustering model, we propose an advanced defense method using adversarial example detection architecture, which extracts the key features from the input data and feeds the extracted features into a clustering model. From the experimental results under various application datasets, we show that the proposed method can detect the adversarial examples while classifying the types of adversarial examples. We also show that the accuracy of the proposed method outperforms the accuracy of recent defense methods using adversarial example detection architecture.

Joint Associations of Obesity and Cardiorespiratory Fitness With Coronary Artery Calcium Composition: IS THERE EVIDENCE FOR FAT-BUT-FIT?

INTRODUCTION: The purpose of this study was to examine the individual and joint associations of obesity and cardiorespiratory fitness (CRF) with indices of coronary artery calcification (CAC) in 2090 middle-aged men. METHODS: Obesity was defined as a body mass index (BMI) ≥25 kg/m2 and a waist circumference (WC) ≥90 cm. Cardiorespiratory fitness was operationally defined as peak oxygen uptake (V˙o2peak) directly measured using gas analysis. Participants were then divided into unfit and fit categories based on age-specific V˙o2peak percentiles. Agatston scores >100 and volume and density scores >75th percentile were defined as indices of CAC, signifying advanced subclinical atherosclerosis. RESULTS: Obese men had increased CAC Agatston, volume, and density scores, while higher CRF was associated with lower Agatston and volume scores after adjusting for potential confounders. In the joint analysis, unfit-obese men had higher CAC Agatston and CAC volume. The fit-obesity category was not associated with CAC Agatston (OR = 0.91: 95% CI, 0.66-1.25, for BMI and OR = 1.21: 95% CI, 0.86-1.70, for WC) and CAC volume (OR = 1.14: 95% CI, 0.85-1.53, for BMI and OR = 1.23: 95% CI, 0.90-1.69, for WC), which were similar to estimates for the fit-normal weight category. CONCLUSIONS: These findings demonstrate that while obesity is positively associated with the prevalence of moderate to severe CAC scores, CRF is inversely associated with the prevalence of moderate to severe CAC scores. Additionally, the combination of being fit and obese was not associated with CAC scores, which could potentially reinforce the fat-but-fit paradigm.

Prediction Algorithms for Blood Pressure Based on Pulse Wave Velocity Using Health Checkup Data in Healthy Korean Men: Algorithm Development and Validation.

BACKGROUND: Pulse transit time and pulse wave velocity (PWV) are related to blood pressure (BP), and there were continuous attempts to use these to predict BP through wearable devices. However, previous studies were conducted on a small scale and could not confirm the relative importance of each variable in predicting BP. OBJECTIVE: This study aims to predict systolic blood pressure and diastolic blood pressure based on PWV and to evaluate the relative importance of each clinical variable used in BP prediction models. METHODS: This study was conducted on 1362 healthy men older than 18 years who visited the Samsung Medical Center. The systolic blood pressure and diastolic blood pressure were estimated using the multiple linear regression method. Models were divided into two groups based on age: younger than 60 years and 60 years or older; 200 seeds were repeated in consideration of partition bias. Mean of error, absolute error, and root mean square error were used as performance metrics. RESULTS: The model divided into two age groups (younger than 60 years and 60 years and older) performed better than the model without division. The performance difference between the model using only three variables (PWV, BMI, age) and the model using 17 variables was not significant. Our final model using PWV, BMI, and age met the criteria presented by the American Association for the Advancement of Medical Instrumentation. The prediction errors were within the range of about 9 to 12 mmHg that can occur with a gold standard mercury sphygmomanometer. CONCLUSIONS: Dividing age based on the age of 60 years showed better BP prediction performance, and it could show good performance even if only PWV, BMI, and age variables were included. Our final model with the minimal number of variables (PWB, BMI, age) would be efficient and feasible for predicting BP.

Is Homomorphic Encryption-Based Deep Learning Secure Enough?

As the amount of data collected and analyzed by machine learning technology increases, data that can identify individuals is also being collected in large quantities. In particular, as deep learning technology-which requires a large amount of analysis data-is activated in various service fields, the possibility of exposing sensitive information of users increases, and the user privacy problem is growing more than ever. As a solution to this user's data privacy problem, homomorphic encryption technology, which is an encryption technology that supports arithmetic operations using encrypted data, has been applied to various field including finance and health care in recent years. If so, is it possible to use the deep learning service while preserving the data privacy of users by using the data to which homomorphic encryption is applied? In this paper, we propose three attack methods to infringe user's data privacy by exploiting possible security vulnerabilities in the process of using homomorphic encryption-based deep learning services for the first time. To specify and verify the feasibility of exploiting possible security vulnerabilities, we propose three attacks: (1) an adversarial attack exploiting communication link between client and trusted party; (2) a reconstruction attack using the paired input and output data; and (3) a membership inference attack by malicious insider. In addition, we describe real-world exploit scenarios for financial and medical services. From the experimental evaluation results, we show that the adversarial example and reconstruction attacks are a practical threat to homomorphic encryption-based deep learning models. The adversarial attack decreased average classification accuracy from 0.927 to 0.043, and the reconstruction attack showed average reclassification accuracy of 0.888, respectively.

The Relationship Between Breast Density Change During Menopause and the Risk of Breast Cancer in Korean Women.

BACKGROUND: The aim of this study was to investigate the relationship between changes in breast density during menopause and breast cancer risk. METHODS: This study was a retrospective, longitudinal cohort study for women over 30 years of age who had undergone breast mammography serially at baseline and postmenopause during regular health checkups at Samsung Medical Center. None of the participants had been diagnosed with breast cancer at baseline. Mammographic breast density was measured using the American College of Radiology Breast Imaging Reporting and Data System. RESULTS: During 18,615 person-years of follow-up (median follow-up 4.8 years; interquartile range 2.8-7.5 years), 45 participants were diagnosed with breast cancer. The prevalence of dense breasts was higher in those who were younger, underweight, had low parity or using contraceptives. The cumulative incidence of breast cancer increased 4 years after menopause in participants, and the consistently extremely dense group had a significantly higher cumulative incidence (CI) of breast cancer compared with other groups [CI of extremely dense vs. others (incidence rate per 100,000 person-years): 375 vs. 203, P < 0.01]. CONCLUSION: Korean women whose breast density was extremely dense before menopause and who maintained this density after menopause were at two-fold greater risk of breast cancer. PREVENTION RELEVANCE: Extremely dense breast density that is maintained persistently from premenopause to postmenopause increases risk of breast cancer two fold in Korean women. Therefore, women having risk factors should receive mammography frequently and if persistently extremely dense breast had been detected, additional modalities of BC screening could be considered.

COVID-19 Vaccination-Associated Lymphadenopathy on FDG PET/CT: Distinctive Features in Adenovirus-Vectored Vaccine.

PURPOSE: This study investigated 18F-FDG PET/CT features of adenovirus-vectored vaccination against COVID-19 in healthy subjects. PATIENTS AND METHODS: Thirty-one health care workers had been vaccinated Vaxzevria and underwent FDG PET/CT as an optional test for a cancer screening program. Size and FDG uptake of the hypermetabolic lymph nodes were measured. Uptake value of spleen was also measured with liver for comparison. RESULTS: All examinees who underwent FDG PET/CT within 14 days' interval showed hypermetabolic lymphadenopathies ipsilateral to vaccine injection. All examinees with hypermetabolic lymphadenopathy had simultaneous muscular uptakes until 23 days' interval. Among 12 examinees who underwent FDG PET/CT more than 15 days after vaccination, only 3 male examinees did not show hypermetabolism in the axillary lymph nodes. There was no female examinee with negative hypermetabolic lymphadenopathy until 29 days after vaccination. CONCLUSIONS: Hypermetabolic reactive lymphadenopathy in the ipsilateral axillary area with or without supraclavicular area is most likely to occur in a healthy person with recent adenovirus-vectored COVID-19 vaccination on FDG PET/CT.

Deep learning for detection and segmentation of artefact and disease instances in gastrointestinal endoscopy.

The Endoscopy Computer Vision Challenge (EndoCV) is a crowd-sourcing initiative to address eminent problems in developing reliable computer aided detection and diagnosis endoscopy systems and suggest a pathway for clinical translation of technologies. Whilst endoscopy is a widely used diagnostic and treatment tool for hollow-organs, there are several core challenges often faced by endoscopists, mainly: 1) presence of multi-class artefacts that hinder their visual interpretation, and 2) difficulty in identifying subtle precancerous precursors and cancer abnormalities. Artefacts often affect the robustness of deep learning methods applied to the gastrointestinal tract organs as they can be confused with tissue of interest. EndoCV2020 challenges are designed to address research questions in these remits. In this paper, we present a summary of methods developed by the top 17 teams and provide an objective comparison of state-of-the-art methods and methods designed by the participants for two sub-challenges: i) artefact detection and segmentation (EAD2020), and ii) disease detection and segmentation (EDD2020). Multi-center, multi-organ, multi-class, and multi-modal clinical endoscopy datasets were compiled for both EAD2020 and EDD2020 sub-challenges. The out-of-sample generalization ability of detection algorithms was also evaluated. Whilst most teams focused on accuracy improvements, only a few methods hold credibility for clinical usability. The best performing teams provided solutions to tackle class imbalance, and variabilities in size, origin, modality and occurrences by exploring data augmentation, data fusion, and optimal class thresholding techniques.