Background: This study aimed to evaluate the applicability of deep learning technology to thyroid ultrasound images for classification of thyroid nodules. Methods: This retrospective analysis included ultrasound images of patients with thyroid nodules investigated by fine-needle aspiration at the thyroid clinic of a single center from April 2010 to September 2012. Thyroid nodules with cytopathologic results of Bethesda category V (suspicious for malignancy) or VI (malignant) were defined as thyroid cancer. Multiple deep learning algorithms based on convolutional neural networks (CNNs) —ResNet, DenseNet, and EfficientNet—were utilized, and Siamese neural networks facilitated multi-view analysis of paired transverse and longitudinal ultrasound images. Results: Among 1,048 analyzed thyroid nodules from 943 patients, 306 (29%) were identified as thyroid cancer. In a subgroup analysis of transverse and longitudinal images, longitudinal images showed superior prediction ability. Multi-view modeling, based on paired transverse and longitudinal images, significantly improved the model performance; with an accuracy of 0.82 (95% confidence intervals [CI], 0.80 to 0.86) with ResNet50, 0.83 (95% CI, 0.83 to 0.88) with DenseNet201, and 0.81 (95% CI, 0.79 to 0.84) with EfficientNetv2_ s. Training with high-resolution images obtained using the latest equipment tended to improve model performance in association with increased sensitivity. Conclusion CNN algorithms applied to ultrasound images demonstrated substantial accuracy in thyroid nodule classification, indicating their potential as valuable tools for diagnosing thyroid cancer. However, in real-world clinical settings, it is important to aware that model performance may vary depending on the quality of images acquired by different physicians and imaging devices.

Background and Objectives: The Fractional Flow Reserve and Intravascular UltrasoundGuided Intervention Strategy for Clinical Outcomes in Patients with Intermediate Stenosis (FLAVOUR) trial demonstrated non-inferiority of fractional flow reserve (FFR)-guided percutaneous coronary intervention (PCI) compared with intravascular ultrasound (IVUS)-guided PCI. We sought to investigate the cost-effectiveness of FFR-guided PCI compared to IVUS-guided PCI in Korea. Methods: A 2-part cost-effectiveness model, composed of a short-term decision tree model and a long-term Markov model, was developed for patients who underwent PCI to treat intermediate stenosis (40% to 70% stenosis by visual estimation on coronary angiography).The lifetime healthcare costs and quality-adjusted life-years (QALYs) were estimated from the healthcare system perspective. Transition probabilities were mainly referred from the FLAVOUR trial, and healthcare costs were mainly obtained through analysis of Korean National Health Insurance claims data. Health utilities were mainly obtained from the Seattle Angina Questionnaire responses of FLAVOUR trial participants mapped to EQ-5D. Results: From the Korean healthcare system perspective, the base-case analysis showed that FFR-guided PCI was 2,451 U.S. dollar lower in lifetime healthcare costs and 0.178 higher in QALYs compared to IVUS-guided PCI. FFR-guided PCI remained more likely to be cost-effective over a wide range of willingness-to-pay thresholds in the probabilistic sensitivity analysis. Conclusions Based on the results from the FLAVOUR trial, FFR-guided PCI is projected to decrease lifetime healthcare costs and increase QALYs compared with IVUS-guided PCI in intermediate coronary lesion, and it is a dominant strategy in Korea.

Background: This study aimed to evaluate the applicability of deep learning technology to thyroid ultrasound images for classification of thyroid nodules. Methods: This retrospective analysis included ultrasound images of patients with thyroid nodules investigated by fine-needle aspiration at the thyroid clinic of a single center from April 2010 to September 2012. Thyroid nodules with cytopathologic results of Bethesda category V (suspicious for malignancy) or VI (malignant) were defined as thyroid cancer. Multiple deep learning algorithms based on convolutional neural networks (CNNs) —ResNet, DenseNet, and EfficientNet—were utilized, and Siamese neural networks facilitated multi-view analysis of paired transverse and longitudinal ultrasound images. Results: Among 1,048 analyzed thyroid nodules from 943 patients, 306 (29%) were identified as thyroid cancer. In a subgroup analysis of transverse and longitudinal images, longitudinal images showed superior prediction ability. Multi-view modeling, based on paired transverse and longitudinal images, significantly improved the model performance; with an accuracy of 0.82 (95% confidence intervals [CI], 0.80 to 0.86) with ResNet50, 0.83 (95% CI, 0.83 to 0.88) with DenseNet201, and 0.81 (95% CI, 0.79 to 0.84) with EfficientNetv2_ s. Training with high-resolution images obtained using the latest equipment tended to improve model performance in association with increased sensitivity. Conclusion CNN algorithms applied to ultrasound images demonstrated substantial accuracy in thyroid nodule classification, indicating their potential as valuable tools for diagnosing thyroid cancer. However, in real-world clinical settings, it is important to aware that model performance may vary depending on the quality of images acquired by different physicians and imaging devices.

Background and Objectives: The Fractional Flow Reserve and Intravascular UltrasoundGuided Intervention Strategy for Clinical Outcomes in Patients with Intermediate Stenosis (FLAVOUR) trial demonstrated non-inferiority of fractional flow reserve (FFR)-guided percutaneous coronary intervention (PCI) compared with intravascular ultrasound (IVUS)-guided PCI. We sought to investigate the cost-effectiveness of FFR-guided PCI compared to IVUS-guided PCI in Korea. Methods: A 2-part cost-effectiveness model, composed of a short-term decision tree model and a long-term Markov model, was developed for patients who underwent PCI to treat intermediate stenosis (40% to 70% stenosis by visual estimation on coronary angiography).The lifetime healthcare costs and quality-adjusted life-years (QALYs) were estimated from the healthcare system perspective. Transition probabilities were mainly referred from the FLAVOUR trial, and healthcare costs were mainly obtained through analysis of Korean National Health Insurance claims data. Health utilities were mainly obtained from the Seattle Angina Questionnaire responses of FLAVOUR trial participants mapped to EQ-5D. Results: From the Korean healthcare system perspective, the base-case analysis showed that FFR-guided PCI was 2,451 U.S. dollar lower in lifetime healthcare costs and 0.178 higher in QALYs compared to IVUS-guided PCI. FFR-guided PCI remained more likely to be cost-effective over a wide range of willingness-to-pay thresholds in the probabilistic sensitivity analysis. Conclusions Based on the results from the FLAVOUR trial, FFR-guided PCI is projected to decrease lifetime healthcare costs and increase QALYs compared with IVUS-guided PCI in intermediate coronary lesion, and it is a dominant strategy in Korea.

Background: and Purpose To determine the clinical phenotypes, relapse timing, treatment responses, and outcomes of children with relapsing myelin oligodendrocyte glycoprotein antibody-associated disease (MOGAD). Methods: We collected the demographic, clinical, laboratory, and radiological data of patients aged <18 years who had been diagnosed with MOGAD at Seoul National University Children’s Hospital between January 2010 and January 2022; 100 were identified as positive for MOG antibodies, 43 of whom experienced relapse. Results: The median age at onset was 7 years (range 2–16 years). The median number of relapses was 2 (range 1–8), and patients were followed up for a median of 65 months (range 5–214 months). The first relapse was experienced before 3 months from onset by 15 patients (34.9%). The most-common initial phenotypes were acute disseminated encephalomyelitis (n=17, 39.5%) and optic neuritis (ON; n=11, 25.6%). The most-common relapse phenotypes were neuromyelitis optica spectrum disorder (n=9, 20.9%), relapsing ON (n=6, 14.0%), and multiphasic disseminated encephalomyelitis (n=6, 14.0%). Many of the patients (n=18, 41.9%) were not specifically categorized. A high proportion of these patients had non-acute disseminated encephalomyelitis encephalitis. Atypical phenotypes such as prolonged fever or hemiplegic migraine-like episodes were also noted. Mycophenolate mofetil and cyclic immunoglobulin treatment significantly reduced the annual relapse rates. Conclusions Our 43 pediatric patients with relapsing MOGAD showed a tendency toward early relapse and various relapse phenotypes. The overall prognoses of these patients were good regardless of phenotype or response to second-line immunosuppressant treatment.

Background: and Purpose Treatments for neuromyelitis optica spectrum disorder (NMOSD) such as eculizumab, ravulizumab, satralizumab, and inebilizumab have significantly advanced relapse prevention, but they remain expensive. Rituximab is an off-label yet popular alternative that offers a cost-effective solution, but its real-world efficacy needs better quantification for guiding the application of newer approved NMOSD treatments (ANTs). This study aimed to determine real-world rituximab failure rates to anticipate the demand for ANTs and aid in resource allocation. Methods: We conducted a nationwide retrospective study involving 605 aquaporin-4-antibody-positive NMOSD patients from 22 centers in South Korea that assessed the efficacy and safety of rituximab over a median follow-up of 47 months. Results: The 605 patients treated with rituximab included 525 (87%) who received continuous therapy throughout the follow-up period (median=47 months, interquartile range=15–87 months). During this period, 117 patients (19%) experienced at least 1 relapse. Notably, 68 of these patients (11% of the total cohort) experienced multiple relapses or at least 1 severe relapse.Additionally, 2% of the patients discontinued rituximab due to adverse events, which included severe infusion reactions, neutropenia, and infections. Conclusions This study has confirmed the efficacy of rituximab in treating NMOSD, as evidenced by an 87% continuation rate among patients over a 4-year follow-up period. Nevertheless, the occurrence of at least one relapse in 19% of the cohort, including 11% who experienced multiple or severe relapses, and a 2% discontinuation rate due to adverse events highlight the urgent need for alternative therapeutic options.

Background: This study aimed to evaluate the applicability of deep learning technology to thyroid ultrasound images for classification of thyroid nodules. Methods: This retrospective analysis included ultrasound images of patients with thyroid nodules investigated by fine-needle aspiration at the thyroid clinic of a single center from April 2010 to September 2012. Thyroid nodules with cytopathologic results of Bethesda category V (suspicious for malignancy) or VI (malignant) were defined as thyroid cancer. Multiple deep learning algorithms based on convolutional neural networks (CNNs) —ResNet, DenseNet, and EfficientNet—were utilized, and Siamese neural networks facilitated multi-view analysis of paired transverse and longitudinal ultrasound images. Results: Among 1,048 analyzed thyroid nodules from 943 patients, 306 (29%) were identified as thyroid cancer. In a subgroup analysis of transverse and longitudinal images, longitudinal images showed superior prediction ability. Multi-view modeling, based on paired transverse and longitudinal images, significantly improved the model performance; with an accuracy of 0.82 (95% confidence intervals [CI], 0.80 to 0.86) with ResNet50, 0.83 (95% CI, 0.83 to 0.88) with DenseNet201, and 0.81 (95% CI, 0.79 to 0.84) with EfficientNetv2_ s. Training with high-resolution images obtained using the latest equipment tended to improve model performance in association with increased sensitivity. Conclusion CNN algorithms applied to ultrasound images demonstrated substantial accuracy in thyroid nodule classification, indicating their potential as valuable tools for diagnosing thyroid cancer. However, in real-world clinical settings, it is important to aware that model performance may vary depending on the quality of images acquired by different physicians and imaging devices.