• Korean Journal of Radiology
• /
• 제25권4호
• /
• pp.363-373
• /
• 2024

Objective: To develop and evaluate a deep learning model for automated segmentation and detection of bone metastasis on spinal MRI. Materials and Methods: We included whole spine MRI scans of adult patients with bone metastasis: 662 MRI series from 302 patients (63.5 ± 11.5 years; male:female, 151:151) from three study centers obtained between January 2015 and August 2021 for training and internal testing (random split into 536 and 126 series, respectively) and 49 MRI series from 20 patients (65.9 ± 11.5 years; male:female, 11:9) from another center obtained between January 2018 and August 2020 for external testing. Three sagittal MRI sequences, including non-contrast T1-weighted image (T1), contrast-enhanced T1-weighted Dixon fat-only image (FO), and contrast-enhanced fat-suppressed T1-weighted image (CE), were used. Seven models trained using the 2D and 3D U-Nets were developed with different combinations (T1, FO, CE, T1 + FO, T1 + CE, FO + CE, and T1 + FO + CE). The segmentation performance was evaluated using Dice coefficient, pixel-wise recall, and pixel-wise precision. The detection performance was analyzed using per-lesion sensitivity and a free-response receiver operating characteristic curve. The performance of the model was compared with that of five radiologists using the external test set. Results: The 2D U-Net T1 + CE model exhibited superior segmentation performance in the external test compared to the other models, with a Dice coefficient of 0.699 and pixel-wise recall of 0.653. The T1 + CE model achieved per-lesion sensitivities of 0.828 (497/600) and 0.857 (150/175) for metastases in the internal and external tests, respectively. The radiologists demonstrated a mean per-lesion sensitivity of 0.746 and a mean per-lesion positive predictive value of 0.701 in the external test. Conclusion: The deep learning models proposed for automated segmentation and detection of bone metastases on spinal MRI demonstrated high diagnostic performance.

• Korean Journal of Radiology
• /
• 제24권6호
• /
• pp.498-511
• /
• 2023

Objective: To evaluate the diagnostic performance of chest computed tomography (CT)-based qualitative and radiomics models for predicting residual axillary nodal metastasis after neoadjuvant chemotherapy (NAC) for patients with clinically node-positive breast cancer. Materials and Methods: This retrospective study included 226 women (mean age, 51.4 years) with clinically node-positive breast cancer treated with NAC followed by surgery between January 2015 and July 2021. Patients were randomly divided into the training and test sets (4:1 ratio). The following predictive models were built: a qualitative CT feature model using logistic regression based on qualitative imaging features of axillary nodes from the pooled data obtained using the visual interpretations of three radiologists; three radiomics models using radiomics features from three (intranodal, perinodal, and combined) different regions of interest (ROIs) delineated on pre-NAC CT and post-NAC CT using a gradient-boosting classifier; and fusion models integrating clinicopathologic factors with the qualitative CT feature model (referred to as clinical-qualitative CT feature models) or with the combined ROI radiomics model (referred to as clinical-radiomics models). The area under the curve (AUC) was used to assess and compare the model performance. Results: Clinical N stage, biological subtype, and primary tumor response indicated by imaging were associated with residual nodal metastasis during the multivariable analysis (all P < 0.05). The AUCs of the qualitative CT feature model and radiomics models (intranodal, perinodal, and combined ROI models) according to post-NAC CT were 0.642, 0.812, 0.762, and 0.832, respectively. The AUCs of the clinical-qualitative CT feature model and clinical-radiomics model according to post-NAC CT were 0.740 and 0.866, respectively. Conclusion: CT-based predictive models showed good diagnostic performance for predicting residual nodal metastasis after NAC. Quantitative radiomics analysis may provide a higher level of performance than qualitative CT features models. Larger multicenter studies should be conducted to confirm their performance.

• Korean Journal of Radiology
• /
• 제24권3호
• /
• pp.190-203
• /
• 2023

Objective: We aimed to assess and validate the radiologic and clinical factors that were associated with recurrence and survival after curative surgery for heterogeneous targetoid primary liver malignancies in patients with chronic liver disease and to develop scoring systems for risk stratification. Materials and Methods: This multicenter retrospective study included 197 consecutive patients with chronic liver disease who had a single targetoid primary liver malignancy (142 hepatocellular carcinomas, 37 cholangiocarcinomas, 17 combined hepatocellular carcinoma-cholangiocarcinomas, and one neuroendocrine carcinoma) identified on preoperative gadoxetic acid-enhanced MRI and subsequently surgically removed between 2010 and 2017. Of these, 120 patients constituted the development cohort, and 77 patients from separate institution served as an external validation cohort. Factors associated with recurrence-free survival (RFS) and overall survival (OS) were identified using a Cox proportional hazards analysis, and risk scores were developed. The discriminatory power of the risk scores in the external validation cohort was evaluated using the Harrell C-index. The Kaplan-Meier curves were used to estimate RFS and OS for the different risk-score groups. Results: In RFS model 1, which eliminated features exclusively accessible on the hepatobiliary phase (HBP), tumor size of 2-5 cm or > 5 cm, and thin-rim arterial phase hyperenhancement (APHE) were included. In RFS model 2, tumors with a size of > 5 cm, tumor in vein (TIV), and HBP hypointense nodules without APHE were included. The OS model included a tumor size of > 5 cm, thin-rim APHE, TIV, and tumor vascular involvement other than TIV. The risk scores of the models showed good discriminatory performance in the external validation set (C-index, 0.62-0.76). The scoring system categorized the patients into three risk groups: favorable, intermediate, and poor, each with a distinct survival outcome (all log-rank p < 0.05). Conclusion: Risk scores based on rim arterial enhancement pattern, tumor size, HBP findings, and radiologic vascular invasion status may help predict postoperative RFS and OS in patients with targetoid primary liver malignancies.

• Korean Journal of Radiology
• /
• 제24권2호
• /
• pp.133-144
• /
• 2023

Objective: Cyclin-dependent kinase inhibitor (CDKN)2A/B homozygous deletion is a key molecular marker of isocitrate dehydrogenase (IDH)-mutant astrocytomas in the 2021 World Health Organization. We aimed to investigate whether qualitative and quantitative MRI parameters can predict CDKN2A/B homozygous deletion status in IDH-mutant astrocytomas. Materials and Methods: Preoperative MRI data of 88 patients (mean age ± standard deviation, 42.0 ± 11.9 years; 40 females and 48 males) with IDH-mutant astrocytomas (76 without and 12 with CDKN2A/B homozygous deletion) from two institutions were included. A qualitative imaging assessment was performed. Mean apparent diffusion coefficient (ADC), 5th percentile of ADC, mean normalized cerebral blood volume (nCBV), and 95th percentile of nCBV were assessed via automatic tumor segmentation. Logistic regression was performed to determine the factors associated with CDKN2A/B homozygous deletion in all 88 patients and a subgroup of 47 patients with histological grades 3 and 4. The discrimination performance of the logistic regression models was evaluated using the area under the receiver operating characteristic curve (AUC). Results: In multivariable analysis of all patients, infiltrative pattern (odds ratio [OR] = 4.25, p = 0.034), maximal diameter (OR = 1.07, p = 0.013), and 95th percentile of nCBV (OR = 1.34, p = 0.049) were independent predictors of CDKN2A/B homozygous deletion. The AUC, accuracy, sensitivity, and specificity of the corresponding model were 0.83 (95% confidence interval [CI], 0.72-0.91), 90.4%, 83.3%, and 75.0%, respectively. On multivariable analysis of the subgroup with histological grades 3 and 4, infiltrative pattern (OR = 10.39, p = 0.012) and 95th percentile of nCBV (OR = 1.24, p = 0.047) were independent predictors of CDKN2A/B homozygous deletion, with an AUC accuracy, sensitivity, and specificity of the corresponding model of 0.76 (95% CI, 0.60-0.88), 87.8%, 80.0%, and 58.1%, respectively. Conclusion: The presence of an infiltrative pattern, larger maximal diameter, and higher 95th percentile of the nCBV may be useful MRI biomarkers for CDKN2A/B homozygous deletion in IDH-mutant astrocytomas.

• Korean Journal of Radiology
• /
• 제23권10호
• /
• pp.1009-1018
• /
• 2022

Objective: This study aimed to investigate the feasibility of using artificial intelligence (AI) to identify normal chest radiography (CXR) from the worklist of radiologists in a health-screening environment. Materials and Methods: This retrospective simulation study was conducted using the CXRs of 5887 adults (mean age ± standard deviation, 55.4 ± 11.8 years; male, 4329) from three health screening centers in South Korea using a commercial AI (Lunit INSIGHT CXR3, version 3.5.8.8). Three board-certified thoracic radiologists reviewed CXR images for referable thoracic abnormalities and grouped the images into those with visible referable abnormalities (identified as abnormal by at least one reader) and those with clearly visible referable abnormalities (identified as abnormal by at least two readers). With AI-based simulated exclusion of normal CXR images, the percentages of normal images sorted and abnormal images erroneously removed were analyzed. Additionally, in a random subsample of 480 patients, the ability to identify visible referable abnormalities was compared among AI-unassisted reading (i.e., all images read by human readers without AI), AI-assisted reading (i.e., all images read by human readers with AI assistance as concurrent readers), and reading with AI triage (i.e., human reading of only those rendered abnormal by AI). Results: Of 5887 CXR images, 405 (6.9%) and 227 (3.9%) contained visible and clearly visible abnormalities, respectively. With AI-based triage, 42.9% (2354/5482) of normal CXR images were removed at the cost of erroneous removal of 3.5% (14/405) and 1.8% (4/227) of CXR images with visible and clearly visible abnormalities, respectively. In the diagnostic performance study, AI triage removed 41.6% (188/452) of normal images from the worklist without missing visible abnormalities and increased the specificity for some readers without decreasing sensitivity. Conclusion: This study suggests the feasibility of sorting and removing normal CXRs using AI with a tailored cut-off to increase efficiency and reduce the workload of radiologists.

• Korean Journal of Radiology
• /
• 제24권10호
• /
• pp.974-982
• /
• 2023

Objective: Recent studies have highlighted the active and potential role of perivascular adipose tissue (PVAT) in atherosclerosis and aneurysm progression, respectively. This study explored the link between PVAT attenuation and abdominal aortic aneurysm (AAA) progression using computed tomography angiography (CTA). Materials and Methods: This multicenter retrospective study analyzed patients with AAA who underwent CTA at baseline and follow-up between March 2015 and July 2022. The following parameters were obtained: maximum diameter and total volume of the AAA, presence or absence of intraluminal thrombus (ILT), maximum diameter and volume of the ILT, and PVAT attenuation of the aortic aneurysm at baseline CTA. PVAT attenuation was divided into high (> -73.4 Hounsfield units [HU]) and low (≤ -73.4 HU). Patients who had or did not have AAA progression during the follow-up, defined as an increase in the aneurysm volume > 10 mL from baseline, were identified. Kaplan-Meier and multivariable Cox regression analyses were used to investigate the association between PVAT attenuation and AAA progression. Results: Our study included 167 participants (148 males; median age: 70.0 years; interquartile range: 63.0-76.0 years), of which 145 (86.8%) were diagnosed with AAA accompanied by ILT. Over a median period of 11.3 months (range: 6.0-85.0 months), AAA progression was observed in 67 patients (40.1%). Multivariable Cox regression analysis indicated that high baseline PVAT attenuation (adjusted hazard ratio [aHR] = 2.23; 95% confidence interval [CI], 1.16-4.32; P = 0.017) was independently associated with AAA progression. This association was demonstrated within the patients of AAA with ILT subcohort, where a high baseline PVAT attenuation (aHR = 2.23; 95% CI, 1.08-4.60; P = 0.030) was consistently independently associated with AAA progression. Conclusion: Elevated PVAT attenuation is independently associated with AAA progression, including patients of AAA with ILT, suggesting the potential of PVAT attenuation as a predictive imaging marker for AAA expansion.

• Korean Journal of Radiology
• /
• 제22권9호
• /
• pp.1579-1586
• /
• 2021

Objective: To evaluate the diagnostic performance of the modified Korean Thyroid Imaging Reporting and Data System (K-TIRADS), and compare it with the 2016 version of K-TIRADS using the Thyroid Imaging Network of Korea. Materials and Methods: Between June and September 2015, 5708 thyroid nodules (≥ 1.0 cm) from 5081 consecutive patients who had undergone thyroid ultrasonography at 26 institutions were retrospectively evaluated. We used a biopsy size threshold of 2 cm for K-TIRADS 3 and 1 cm for K-TIRADS 4 (modified K-TIRADS 1) or 1.5 cm for K-TIRADS 4 (modified K-TIRADS 3). The modified K-TIRADS 2 subcategorized the K-TIRADS 4 into 4A and 4B, and the cutoff sizes for the biopsies were defined as 1 cm for K-TIRADS 4B and 1.5 cm for K-TIRADS 4A. The diagnostic performance and the rate of unnecessary biopsies of the modified K-TIRADS for detecting malignancy were compared with those of the 2016 K-TIRAD, which were stratified by nodule size (with a threshold of 2 cm). Results: A total of 1111 malignant nodules and 4597 benign nodules were included. The sensitivity, specificity, and unnecessary biopsy rate of the benign nodules were 94.9%, 24.4%, and 60.9% for the 2016 K-TIRADS; 91.0%, 39.7%, and 48.6% for the modified K-TIRADS 1; 84.9%, 45.9%, and 43.5% for the modified K-TIRADS 2; and 76.1%, 50.2%, and 40.1% for the modified K-TIRADS 3. For small nodules (1-2 cm), the diagnostic sensitivity of the modified K-TIRADS decreased by 5.2-25.6% and the rate of unnecessary biopsies reduced by 19.2-32.8% compared with those of the 2016 K-TIRADS (p < 0.001). For large nodules (> 2 cm), the modified K-TIRADSs maintained a very high sensitivity for detecting malignancy (98%). Conclusion: The modified K-TIRADSs significantly reduced the rate of unnecessary biopsies for small (1-2 cm) nodules while maintaining a very high sensitivity for malignancy for large (> 2 cm) nodules.

• 대한치매학회지
• /
• 제22권3호
• /
• pp.100-108
• /
• 2023

Background and Purpose: The efficacy and safety of GV1001 have been demonstrated in patients with moderate-to-severe Alzheimer's disease (AD). In this study, we aimed to further demonstrate the effectiveness of GV1001 using subscales of the Severe Impairment Battery (SIB), which is a validated measure to assess cognitive function in patients with moderate-to-severe AD. Methods: We performed a post hoc analysis of data from a 6 month, multicenter, phase 2, randomized, double-blind, placebo-controlled trial with GV1001 (ClinicalTrials.gov, NCT03184467). Patients were randomized to receive either GV1001 or a placebo for 24 weeks. In the current study, nine subscales of SIB-social interaction, memory, orientation, language, attention, praxis, visuospatial ability, construction, and orientation to name-were compared between the treatment (GV1001 1.12 mg) and placebo groups at weeks 12 and 24. The safety endpoints for these patients were also determined based on adverse events. Results: In addition to the considerable beneficial effect of GV1001 on the SIB total score, GV1001 1.12 mg showed the most significant effect on language function at 24 weeks compared to placebo in both the full analysis set (FAS) and per-protocol set (PPS) (p=0.017 and p=0.011, respectively). The rate of adverse events did not differ significantly between the 2 groups. Conclusions: Patients with moderate-to-severe AD receiving GV1001 had greater language benefits than those receiving placebo, as measured using the SIB language subscale.

• Clinical Endoscopy
• /
• 제56권6호
• /
• pp.744-753
• /
• 2023

Background/Aims: This study aimed to evaluate the prevalence and natural progression of subepithelial lesions (SELs) in the upper gastrointestinal (UGI) tract. Methods: The medical records of patients with UGI SELs who underwent endoscopic screening at eight university hospitals between January and December 2010 were retrospectively investigated. The follow-up evaluations were performed until December 2016. Results: UGI SELs were found in 1,044 of the 65,233 participants screened (endoscopic prevalence, 1.60%; the total number of lesions, 1,062; mean age, 55.1±11.2 years; men, 53.6%). The median follow-up period was 48 (range, 8-74) months. SELs were most frequently found in the stomach (63.8%) and had a mean size of 9.9±6.1 mm. Endoscopic ultrasonography (EUS) was performed in 293 patients (28.1%). The most common lesions were leiomyomas, followed by gastrointestinal stromal tumors (GISTs), and ectopic pancreas. The proportions of SELs with malignant potential according to size were 3% (<1 cm), 22% (1-2 cm), 27% (2-3 cm), and 38% (≥3 cm). In gastric SELs larger than 1 cm, resections were performed in 20 patients because of an increase in size, of which 12 were found to be GISTs. Conclusions: The prevalence of UGI SELs was 1.60%. Further, 23% of gastric SELs ≥1 cm were precancerous lesions, most followed by EUS and clinical decisions without initial pathological confirmation.

• Korean Circulation Journal
• /
• 제53권10호
• /
• pp.693-707
• /
• 2023

Background and Objectives: Inherited arrhythmia (IA) is a more common cause of sudden cardiac death in Asian population, but little is known about the genetic background of Asian IA probands. We aimed to investigate the clinical characteristics and analyze the genetic underpinnings of IA in a Korean cohort. Methods: This study was conducted in a multicenter cohort of the Korean IA Registry from 2014 to 2017. Genetic testing was performed using a next-generation sequencing panel including 174 causative genes of cardiovascular disease. Results: Among the 265 IA probands, idiopathic ventricular fibrillation (IVF) and Brugada Syndrome (BrS) was the most prevalent diseases (96 and 95 cases respectively), followed by long QT syndrome (LQTS, n=54). Two-hundred-sixteen probands underwent genetic testing, and 69 probands (31.9%) were detected with genetic variant, with yield of pathogenic or likely pathogenic variant as 6.4%. Left ventricular ejection fraction was significantly lower in genotype positive probands (54.7±11.3 vs. 59.3±9.2%, p=0.005). IVF probands showed highest yield of positive genotype (54.0%), followed by LQTS (23.8%), and BrS (19.5%). Conclusions: There were significant differences in clinical characteristics and genetic yields among BrS, LQTS, and IVF. Genetic testing did not provide better yield for BrS and LQTS. On the other hand, in IVF, genetic testing using multiple gene panel might enable the molecular diagnosis of concealed genotype, which may alter future clinical diagnosis and management strategies.