• Korean Journal of Radiology
• /
• v.24 no.11
• /
• pp.1151-1163
• /
• 2023

Objective: To develop a deep-learning-based bone age prediction model optimized for Korean children and adolescents and evaluate its feasibility by comparing it with a Greulich-Pyle-based deep-learning model. Materials and Methods: A convolutional neural network was trained to predict age according to the bone development shown on a hand radiograph (bone age) using 21036 hand radiographs of Korean children and adolescents without known bone development-affecting diseases/conditions obtained between 1998 and 2019 (median age [interquartile range {IQR}], 9 [7-12] years; male:female, 11794:9242) and their chronological ages as labels (Korean model). We constructed 2 separate external datasets consisting of Korean children and adolescents with healthy bone development (Institution 1: n = 343; median age [IQR], 10 [4-15] years; male: female, 183:160; Institution 2: n = 321; median age [IQR], 9 [5-14] years; male: female, 164:157) to test the model performance. The mean absolute error (MAE), root mean square error (RMSE), and proportions of bone age predictions within 6, 12, 18, and 24 months of the reference age (chronological age) were compared between the Korean model and a commercial model (VUNO Med-BoneAge version 1.1; VUNO) trained with Greulich-Pyle-based age as the label (GP-based model). Results: Compared with the GP-based model, the Korean model showed a lower RMSE (11.2 vs. 13.8 months; P = 0.004) and MAE (8.2 vs. 10.5 months; P = 0.002), a higher proportion of bone age predictions within 18 months of chronological age (88.3% vs. 82.2%; P = 0.031) for Institution 1, and a lower MAE (9.5 vs. 11.0 months; P = 0.022) and higher proportion of bone age predictions within 6 months (44.5% vs. 36.4%; P = 0.044) for Institution 2. Conclusion: The Korean model trained using the chronological ages of Korean children and adolescents without known bone development-affecting diseases/conditions as labels performed better in bone age assessment than the GP-based model in the Korean pediatric population. Further validation is required to confirm its accuracy.

• Korean Journal of Radiology
• /
• v.23 no.4
• /
• pp.466-478
• /
• 2022

Objective: ¹⁸F-fluorodeoxyglucose (FDG) PET/CT is often used for detecting malignancy in patients with newly diagnosed hemophagocytic lymphohistiocytosis (HLH), with acceptable sensitivity but relatively low specificity. The aim of this study was to improve the diagnostic ability of ¹⁸F-FDG PET/CT in identifying malignancy in patients with HLH by combining ¹⁸F-FDG PET/CT and clinical parameters. Materials and Methods: Ninety-seven patients (age ≥ 14 years) with secondary HLH were retrospectively reviewed and divided into the derivation (n = 71) and validation (n = 26) cohorts according to admission time. In the derivation cohort, 22 patients had malignancy-associated HLH (M-HLH) and 49 patients had non-malignancy-associated HLH (NM-HLH). Data on pretreatment ¹⁸F-FDG PET/CT and laboratory results were collected. The variables were analyzed using the Mann-Whitney U test or Pearson's chi-square test, and a nomogram for predicting M-HLH was constructed using multivariable binary logistic regression. The predictors were also ranked using decision-tree analysis. The nomogram and decision tree were validated in the validation cohort (10 patients with M-HLH and 16 patients with NM-HLH). Results: The ratio of the maximal standardized uptake value (SUVmax) of the lymph nodes to that of the mediastinum, the ratio of the SUVmax of bone lesions or bone marrow to that of the mediastinum, and age were selected for constructing the model. The nomogram showed good performance in predicting M-HLH in the validation cohort, with an area under the receiver operating characteristic curve of 0.875 (95% confidence interval, 0.686-0.971). At an appropriate cutoff value, the sensitivity and specificity for identifying M-HLH were 90% (9/10) and 68.8% (11/16), respectively. The decision tree integrating the same variables showed 70% (7/10) sensitivity and 93.8% (15/16) specificity for identifying M-HLH. In comparison, visual analysis of ¹⁸F-FDG PET/CT images demonstrated 100% (10/10) sensitivity and 12.5% (2/16) specificity. Conclusion: ¹⁸F-FDG PET/CT may be a practical technique for identifying M-HLH. The model constructed using ¹⁸F-FDG PET/CT features and age was able to detect malignancy with better accuracy than visual analysis of ¹⁸F-FDG PET/CT images.

• Korean Journal of Radiology
• /
• v.20 no.5
• /
• pp.719-728
• /
• 2019

Objective: To investigate the diagnostic validity of coronary computed tomography angiography (cCTA) in vasospastic angina (VA) and factors associated with discrepant results between invasive coronary angiography with the ergonovine provocation test (iCAG-EPT) and cCTA. Materials and Methods: Of the 1397 patients diagnosed with VA from 2006 to 2016, 33 patients (75 lesions) with available cCTA data from within 6 months before iCAG-EPT were included. The severity of spasm (% diameter stenosis [%DS]) on iCAGEPT and cCTA was assessed, and the difference in %DS (Δ%DS) was calculated. Δ%DS was compared after classifying the lesions according to pre-cCTA-administered sublingual nitroglycerin (SL-NG) or beta-blockers. The lesions were further categorized with %DS ≥ 50% on iCAG-EPT or cCTA defined as a significant spasm, and the diagnostic performance of cCTA on identifying significant spasm relative to iCAG-EPT was assessed. Results: Compared to lesions without SL-NG treatment, those with SL-NG treatment showed a higher Δ%DS (39.2% vs. 22.1%, p = 0.002). However, there was no difference in Δ%DS with or without beta-blocker treatment (35.1% vs. 32.6%, p = 0.643). The significant difference in Δ%DS associated with SL-NG was more prominent in patients who were aged < 60 years, were male, had body mass index < 25 kg/m², and had no history of hypertension, diabetes, or dyslipidemia. Based on iCAG-EPT as the reference, the per-lesion-based sensitivity, specificity, positive predictive value, negative predictive value, and accuracy of cCTA for VA diagnosis were 7.5%, 94.0%, 60.0%, 47.1%, and 48.0%, respectively. Conclusion: For patients with clinically suspected VA, confirmation with iCAG-EPT needs to be considered without completely excluding the diagnosis of VA simply based on cCTA results, although further prospective studies are required for confirmation.

• Korean Journal of Radiology
• /
• v.20 no.5
• /
• pp.791-800
• /
• 2019

Objective: To compare various models of diffusion-weighted imaging including monoexponential apparent diffusion coefficient (ADC), biexponential (fast diffusion coefficient [D_f], slow diffusion coefficient [D_s], and fraction of fast diffusion), stretched-exponential (distributed diffusion coefficient and anomalous exponent term [α]), and kurtosis (mean diffusivity and mean kurtosis [MK]) models in the differentiation of renal solid masses. Materials and Methods: A total of 81 patients (56 men and 25 women; mean age, 57 years; age range, 30-69 years) with 18 benign and 63 malignant lesions were imaged using 3T diffusion-weighted MRI. Diffusion model selection was investigated in each lesion using the Akaike information criteria. Mann-Whitney U test and receiver operating characteristic (ROC) analysis were used for statistical evaluations. Results: Goodness-of-fit analysis showed that the stretched-exponential model had the highest voxel percentages in benign and malignant lesions (90.7% and 51.4%, respectively). ADC, D_s, and MK showed significant differences between benign and malignant lesions (p < 0.05) and between low- and high-grade clear cell renal cell carcinoma (ccRCC) (p < 0.05). α was significantly lower in the benign group than in the malignant group (p < 0.05). All diffusion measures showed significant differences between ccRCC and non-ccRCC (p < 0.05) except D_f and α (p = 0.143 and 0.112, respectively). α showed the highest diagnostic accuracy in differentiating benign and malignant lesions with an area under the ROC curve of 0.923, but none of the parameters from these advanced models revealed significantly better performance over ADC in discriminating subtypes or grades of renal cell carcinoma (RCC) (p > 0.05). Conclusion: Compared with conventional diffusion parameters, α may provide additional information for differentiating benign and malignant renal masses, while ADC remains the most valuable parameter for differentiation of RCC subtypes and for ccRCC grading.

• Korean Journal of Radiology
• /
• v.20 no.1
• /
• pp.114-125
• /
• 2019

Objective: Segmented cardiac cine magnetic resonance imaging (MRI) is the gold standard for cardiac ventricular volumetric assessment. In patients with difficulty in breath-holding or arrhythmia, this technique may generate images with inadequate quality for diagnosis. Real-time cardiac cine MRI has been developed to address this limitation. We aimed to assess the performance of retrospective electrocardiography-gated real-time cine MRI at 3T for left ventricular (LV) volume and mass measurement. Materials and Methods: Fifty-one patients were consecutively enrolled. A series of short-axis cine images covering the entire left ventricle using both segmented and real-time balanced steady-state free precession cardiac cine MRI were obtained. End-diastolic volume (EDV), end-systolic volume (ESV), stroke volume (SV), ejection fraction (EF), and LV mass were measured. The agreement and correlation of the parameters were assessed. Additionally, image quality was evaluated using European CMR Registry (Euro-CMR) score and structure visibility rating. Results: In patients without difficulty in breath-holding or arrhythmia, no significant difference was found in Euro-CMR score between the two techniques (0.3 ± 0.7 vs. 0.3 ± 0.5, p > 0.05). Good agreements and correlations were found between the techniques for measuring EDV, ESV, EF, SV, and LV mass. In patients with difficulty in breath-holding or arrhythmia, segmented cine MRI had a significant higher Euro-CMR score (2.3 ± 1.2 vs. 0.4 ± 0.5, p < 0.001). Conclusion: Real-time cine MRI at 3T allowed the assessment of LV volume with high accuracy and showed a significantly better image quality compared to that of segmented cine MRI in patients with difficulty in breath-holding and arrhythmia.

• Journal of the Korean Society of Radiology
• /
• v.18 no.3
• /
• pp.249-256
• /
• 2024

The study aimed to evaluate the efficacy of treatment plans using full Arc and Partial Arc Coplanar volumetric modulated arc therapy and Non-Coplanar volumetric modulated arc therapy to minimize radiation treatment side effects, such as pneumonia, and protect normal organs in esophageal cancer radiotherapy. 30 patients who underwent Concurrent Chemoradiotherapy for esophageal cancer were included. Compared planning target volume, lung, heart, spinal cord and total monitor units among three treatment plans: fVMAT(2 Full Arc), pVMAT(4 Partial Arc), and ncVMAT(2 Partial Arc + 2 Non-Coplanar Arc). All plans met the PTV criteria, showing uniform distribution. The average dose to the heart was 5.8 Gy for fVMAT, 6.97 Gy for pVMAT, and 7.6 Gy for ncVMAT, with the lowest value in fVMAT, which was statistically significant. However, the average lung dose was 9.01 Gy for fVMAT, 7.71 Gy for pVMAT, and 7.12 Gy for ncVMAT, with V_5Gy(%) values of 52.22%, 38.61%, 36.35% and V_10Gy(%) values of 37.8%, 27.33%, 24.15% respectively. ncVMAT showed the lowest values, while fVMAT had the highest, with statistical significance. In conclusion, ncVMAT effectively reduces lung radiation exposure in esophageal cancer radiotherapy, potentially reducing the incidence of side effects such as pneumonia. However, considering factors like setup accuracy and treatment time, applying an appropriate treatment plan may lead to better outcomes.

• Korean Journal of Radiology
• /
• v.21 no.8
• /
• pp.998-1006
• /
• 2020

Objective: To compare the accuracies of quantitative computed tomography (CT) parameters and semiquantitative visual score in evaluating clinical classification of severity of coronavirus disease (COVID-19). Materials and Methods: We retrospectively enrolled 187 patients with COVID-19 treated at Tongji Hospital of Tongji Medical College from February 15, 2020, to February 29, 2020. Demographic data, imaging characteristics, and clinical data were collected, and based on the clinical classification of severity, patients were divided into groups 1 (mild) and 2 (severe/critical). A semiquantitative visual score was used to estimate the lesion extent. A three-dimensional slicer was used to precisely quantify the volume and CT value of the lung and lesions. Correlation coefficients of the quantitative CT parameters, semiquantitative visual score, and clinical classification were calculated using Spearman's correlation. A receiver operating characteristic curve was used to compare the accuracies of quantitative and semi-quantitative methods. Results: There were 59 patients in group 1 and 128 patients in group 2. The mean age and sex distribution of the two groups were not significantly different. The lesions were primarily located in the subpleural area. Compared to group 1, group 2 had larger values for all volume-dependent parameters (p < 0.001). The percentage of lesions had the strongest correlation with disease severity with a correlation coefficient of 0.495. In comparison, the correlation coefficient of semiquantitative score was 0.349. To classify the severity of COVID-19, area under the curve of the percentage of lesions was the highest (0.807; 95% confidence interval, 0.744-0.861: p < 0.001) and that of the quantitative CT parameters was significantly higher than that of the semiquantitative visual score (p = 0.001). Conclusion: The classification accuracy of quantitative CT parameters was significantly superior to that of semiquantitative visual score in terms of evaluating the severity of COVID-19.

• Journal of Korea Water Resources Association
• /
• v.57 no.2
• /
• pp.73-85
• /
• 2024

Predicting water quality of rivers and reservoirs is necessary for the management of water resources. Artificial Neural Networks (ANNs) have been used in many studies to predict water quality with high accuracy. Previous studies have used Gradient Descent (GD)-based optimizers as an optimizer, an operator of ANN that searches parameters. However, GD-based optimizers have the disadvantages of the possibility of local optimal convergence and absence of a solution storage and comparison structure. This study developed improved optimizers to overcome the disadvantages of GD-based optimizers. Proposed optimizers are optimizers that combine adaptive moments (Adam) and Nesterov-accelerated adaptive moments (Nadam), which have low learning errors among GD-based optimizers, with Harmony Search (HS) or Novel Self-adaptive Harmony Search (NSHS). To evaluate the performance of Long Short-Term Memory (LSTM) using improved optimizers, the water quality data from the Dasan water quality monitoring station were used for training and prediction. Comparing the learning results, Mean Squared Error (MSE) of LSTM using Nadam combined with NSHS (NadamNSHS) was the lowest at 0.002921. In addition, the prediction rankings according to MSE and R² for the four water quality indices for each optimizer were compared. Comparing the average of ranking for each optimizer, it was confirmed that LSTM using NadamNSHS was the highest at 2.25.

• Journal of the Korea institute for structural maintenance and inspection
• /
• v.28 no.4
• /
• pp.55-61
• /
• 2024

Recently, the number of aging concrete structures is steadily increasing. This is because many of these structures are reaching their expected lifespan. Such structures require accurate inspections and persistent maintenance. Otherwise, their original functions and performance may degrade, potentially leading to safety accidents. Therefore, research on objective inspection technologies using deep learning and computer vision is actively being conducted. High-resolution images can accurately observe not only micro cracks but also spalling and exposed rebar, and deep learning enables automated detection. High detection performance in deep learning is only guaranteed with diverse and numerous training datasets. However, surface damage to concrete is not commonly captured in images, resulting in a lack of training data. To overcome this limitation, this study proposed a method for generating concrete surface damage images, including cracks, spalling, and exposed rebar, using stable diffusion. This method synthesizes new damage images by paired text and image data. For this purpose, a training dataset of 678 images was secured, and fine-tuning was performed through low-rank adaptation. The quality of the generated images was compared according to three base models of stable diffusion. As a result, a method to synthesize the most diverse and high-quality concrete damage images was developed. This research is expected to address the issue of data scarcity and contribute to improving the accuracy of deep learning-based damage detection algorithms in the future.

• Applied Chemistry for Engineering
• /
• v.35 no.5
• /
• pp.423-428
• /
• 2024

Microneedles with micron-sized needle arrays are an emerging technology for the transdermal administration of active pharmaceutical ingredients with minimally invasive pain. Over the past decade, although various additive manufacturing technologies have been employed for precise fabrication of microneedles, these methods are often limited by material compatibility and bioavailability, in addition to being time-consuming and costly. In here, we compare the resolution of Polyjet and DLP-SLA 3D printing methods for the precise fabrication of biodegradable PCLDA/PEGDA microneedles. To enhance the structural accuracy of the microneedles from both printing methods, we evaluate the 3D printing conditions, including 3D printing angle and needle height and diameter. Molds for microneedles are fabricated using optimized 3D printing methods, and subsequent replica molding processes are employed to fabricate the polymeric microneedles with sharp need tips. Finally, we use photocurable PCLDA and PEGDA for biodegradable and biocompatible microneedles, and their mechanical properties as PCLDA concentrations are analyzed to assess the strength required for skin insertion. This study has demonstrated the efficient and low-cost fabrication of high-resolution microneedles for transdermal drug delivery.