• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제14권6호
• /
• pp.2480-2496
• /
• 2020

Considering that high-dose X-ray radiation during CT scans may bring potential risks to patients, in the medical imaging industry there has been increasing emphasis on low-dose CT. Due to complex statistical characteristics of noise found in low-dose CT images, many traditional methods are difficult to preserve structural details effectively while suppressing noise and artifacts. Inspired by the deep learning techniques, we propose a densely connected residual network (DCRN) for low-dose CT image noise cancelation, which combines the ideas of dense connection with residual learning. On one hand, dense connection maximizes information flow between layers in the network, which is beneficial to maintain structural details when denoising images. On the other hand, residual learning paired with batch normalization would allow for decreased training speed and better noise reduction performance in images. The experiments are performed on the 100 CT images selected from a public medical dataset-TCIA(The Cancer Imaging Archive). Compared with the other three competitive denoising algorithms, both subjective visual effect and objective evaluation indexes which include PSNR, RMSE, MAE and SSIM show that the proposed network can improve LDCT images quality more effectively while maintaining a low computational cost. In the objective evaluation indexes, the highest PSNR 33.67, RMSE 5.659, MAE 1.965 and SSIM 0.9434 are achieved by the proposed method. Especially for RMSE, compare with the best performing algorithm in the comparison algorithms, the proposed network increases it by 7 percentage points.

• 한국방사선학회논문지
• /
• 제16권3호
• /
• pp.295-302
• /
• 2022

희박뷰 전산화단층촬영(computed tomography; CT) 영상화 기술은 피폭 방사선량을 감소시킬 수 있을 뿐만 아니라 획득한 투영상의 균일성을 유지하고 잡음을 감소시킬 수 있는 장점이 있다. 하지만 재구성 영상 내 인공물 발생으로 인하여 화질 및 피사체 구조가 왜곡되는 단점이 있다. 본 연구에서는 희박뷰 CT 영상의 인공물 감소를 위해 wavelet 변환과 잔차 학습(residual learning)을 적용한 콘볼루션 신경망(convolutional neural network; CNN) 기반 영상화 모델을 개발하고, 개발한 모델을 통한 희박뷰 CT 영상의 인공물 감소 정도를 정량적으로 분석하였다. CNN은 wavelet 변환 층, 콘볼루션 층 및 역 wavelet 변환 층으로 구성하였으며, 희박뷰 CT 영상과 잔차 영상을 각각 입출력 영상으로 설정하여 영상화 모델 학습을 진행하였다. 영상화 모델 학습을 위해 평균제곱오차(mean squared error; MSE)를 손실함수로, Adam 함수를 최적화 함수로 사용하였다. 학습된 모델을 통해 입력 희박뷰 CT 영상에 대한 예측 잔차 영상을 획득하고, 두 영상간의 감산을 통해 최종 결과 영상을 획득하였다. 또한 최종 결과 영상에 대한 시각적 특성, 최대신호대잡음비(peak signal-to- noise ratio; PSNR) 및 구조적유사성지수(structural similarity; SSIM)를 측정하였다. 연구결과 본 연구에서 개발한 영상화 모델을 통해 희박뷰 CT 영상의 인공물이 효과적으로 제거되며, 공간분해능이 향상되는 결과를 확인하였다. 또한 wavelet 변환과 잔차 학습을 미적용한 영상화 모델에 비해 본 연구에서 개발한 영상화 모델은 결과 영상의 PSNR 및 SSIM을 각각 8.18% 및 19.71% 향상시킬 수 있음을 확인하였다. 따라서 본 연구에서 개발한 영상화 모델을 이용하여 희박뷰 CT 영상의 인공물 제거는 물론 공간분해능 향상 및 정량적 정확도 향상 효과를 획득할 수 있다.

• 방송공학회논문지
• /
• 제28권3호
• /
• pp.293-301
• /
• 2023

표준화되지 않은 의료 데이터 수집 및 관리는 여전히 수동으로 진행되고 있어, 이 문제를 해결하기 위해 딥 러닝을 사용해 CT 데이터를 분류하는 연구들이 진행되고 있다. 하지만 대부분 연구에서는 기본적인 CT slice인 axial 평면만을 기반으로 모델을 개발하고 있다. CT 영상은 일반 이미지와 다르게 인체 구조만 묘사하기 때문에 CT scan을 재구성하는 것만으로도 더 풍부한 신체적 특징을 나타낼 수 있다. 이 연구는 axial 평면뿐만 아니라 CT 데이터를 2D로 변환하는 여러가지 방법들을 통해 보다 높은 성능을 달성할 수 있는 방법을 찾고자 한다. 훈련은 5가지 부위의 CT 스캔 1042개를 사용했고, 모델 평가를 위해 테스트셋 179개, 외부 데이터셋으로 448개를 수집했다. 딥러닝 모델 개발을 위해 ImageNet으로 사전 학습된 InceptionResNetV2를 백본으로 사용하였으며, 모델의 전체 레이어를 재 학습했다. 실험결과 신체 부위 분류에서는 재구성 데이터 모델이 99.33%를 달성하며 axial 모델보다 1.12% 더 높았고, 조영제 분류에서는 brain과 neck에서만 axial모델이 높았다. 결론적으로 axial slice로만 훈련했을 때 보다 해부학적 특징이 잘 나타나는 데이터로 학습했을 때 더 정확한 성능 달성이 가능했다.

• Korean Journal of Radiology
• /
• 제21권8호
• /
• pp.1007-1017
• /
• 2020

Objective: The purpose of our study was to investigate the predictive abilities of clinical and computed tomography (CT) features for outcome prediction in patients with coronavirus disease (COVID-19). Materials and Methods: The clinical and CT data of 238 patients with laboratory-confirmed COVID-19 in our two hospitals were retrospectively analyzed. One hundred sixty-six patients (103 males; age 43.8 ± 12.3 years) were allocated in the training cohort and 72 patients (38 males; age 45.1 ± 15.8 years) from another independent hospital were assigned in the validation cohort. The primary composite endpoint was admission to an intensive care unit, use of mechanical ventilation, or death. Univariate and multivariate Cox proportional hazard analyses were performed to identify independent predictors. A nomogram was constructed based on the combination of clinical and CT features, and its prognostic performance was externally tested in the validation group. The predictive value of the combined model was compared with models built on the clinical and radiological attributes alone. Results: Overall, 35 infected patients (21.1%) in the training cohort and 10 patients (13.9%) in the validation cohort experienced adverse outcomes. Underlying comorbidity (hazard ratio [HR], 3.35; 95% confidence interval [CI], 1.67-6.71; p < 0.001), lymphocyte count (HR, 0.12; 95% CI, 0.04-0.38; p < 0.001) and crazy-paving sign (HR, 2.15; 95% CI, 1.03-4.48; p = 0.042) were the independent factors. The nomogram displayed a concordance index (C-index) of 0.82 (95% CI, 0.76-0.88), and its prognostic value was confirmed in the validation cohort with a C-index of 0.89 (95% CI, 0.82-0.96). The combined model provided the best performance over the clinical or radiological model (p < 0.050). Conclusion: Underlying comorbidity, lymphocyte count and crazy-paving sign were independent predictors of adverse outcomes. The prognostic nomogram based on the combination of clinical and CT features could be a useful tool for predicting adverse outcomes of patients with COVID-19.

• 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.

• 정보처리학회논문지B
• /
• 제10B권5호
• /
• pp.509-514
• /
• 2003

본 논문은 흉부를 촬영한 CT(Computed Tomography) 영상을 이용하여 기관지 내부를 시각화(visualization)하기 위해 가상기관지경(virtual bronchoscopy)을 구현하였다. 실제 광학내시경은 침습(invasive)검사로 환자가 불편함을 감수해야하며 검사 전에 사전 준비 작업이 필요하고 절개, 감염, 출혈 등의 부작용을 수반하는데 비해 가상내시경은 CT 나 MRI 등과 같은 스캐너의 단면 영상을 사용하여 구현한다. 가상기관지경의 항해경로 결정을 위해 CT 단면 원천영상에서 기관지(trachea wall)만을 분리하였고, 분리된 기관지 영상을 이용하여 중앙축 변환(MAT : .Medial Axis Transformation)을 통해 구해진 좌표값을 가상 카메라의 운행 경로로 사용하였다. 원근투영(perspective projection) 및 볼륨 데이터의 표면을 렌더링하기 위해 마칭큐브(marching cube) 알고리즘을 사용하였다. 기관지질환은 기관지 내부의 염증(inflammation) 이나 폐암(lung cancer)으로 기인하여 기관지 통로가 좁아지는 기관지 협착증, 기관지 확장증 및 기관지암으로 분류된다. 가상기관지경은 기관지 내부의 질환 위치와 질환 정도를 정성적으로 파악 할 수 있으며 기관지질환의 진단과 교육에 사용될 수 있다.

• Korean Journal of Radiology
• /
• 제24권6호
• /
• pp.529-540
• /
• 2023

Objective: To compare the diagnostic performance and clinical outcomes of 2-mSv computed tomography (CT) and conventional-dose CT (CDCT), following radiology residents' interpretation of CT examinations for suspected appendicitis. Materials and Methods: Altogether, 3074 patients with suspected appendicitis aged 15-44 years (28 ± 9 years, 1672 females) from 20 hospitals were randomly assigned to the 2-mSv CT (n = 1535) or CDCT (n = 1539) groups in a pragmatic trial from December 2013 and August 2016. Overall, 107 radiology residents participated in the trial as readers in the form of daily practice after online training for 2-mSv CT. They made preliminary CT reports, which were later finalized by attending radiologists via addendum reports, for 640 and 657 patients in the 2-mSv CT and CDCT groups, respectively. We compared the diagnostic performance of the residents, discrepancies between preliminary and addendum reports, and clinical outcomes between the two groups. Results: Patient characteristics were similar between the 640 and 657 patients. Residents' diagnostic performance was not significantly different between the 2-mSv CT and CDCT groups, with a sensitivity of 96.0% and 97.1%, respectively (difference [95% confidence interval {CI}], -1.1% [-4.9%, 2.6%]; P = 0.69) and specificity of 93.2% and 93.1%, respectively (0.1% [-3.6%, 3.7%]; P > 0.99). The 2-mSv CT and CDCT groups did not significantly differ in discrepancies between the preliminary and addendum reports regarding the presence of appendicitis (3.3% vs. 5.2%; -1.9% [-4.2%, 0.4%]; P = 0.12) and alternative diagnosis (5.5% vs. 6.4%; -0.9% [-3.6%, 1.8%]; P = 0.56). The rates of perforated appendicitis (12.0% vs. 12.6%; -0.6% [-4.3%, 3.1%]; P = 0.81) and negative appendectomies (1.9% vs. 1.1%; 0.8% [-0.7%, 2.3%]; P = 0.33) were not significantly different between the two groups. Conclusion: Diagnostic performance and clinical outcomes were not significantly different between the 2-mSv CT and CDCT groups following radiology residents' CT readings for suspected appendicitis.

• Korean Journal of Radiology
• /
• 제22권7호
• /
• pp.1213-1224
• /
• 2021

Objective: To develop a machine learning (ML) pipeline based on radiomics to predict Coronavirus Disease 2019 (COVID-19) severity and the future deterioration to critical illness using CT and clinical variables. Materials and Methods: Clinical data were collected from 981 patients from a multi-institutional international cohort with real-time polymerase chain reaction-confirmed COVID-19. Radiomics features were extracted from chest CT of the patients. The data of the cohort were randomly divided into training, validation, and test sets using a 7:1:2 ratio. A ML pipeline consisting of a model to predict severity and time-to-event model to predict progression to critical illness were trained on radiomics features and clinical variables. The receiver operating characteristic area under the curve (ROC-AUC), concordance index (C-index), and time-dependent ROC-AUC were calculated to determine model performance, which was compared with consensus CT severity scores obtained by visual interpretation by radiologists. Results: Among 981 patients with confirmed COVID-19, 274 patients developed critical illness. Radiomics features and clinical variables resulted in the best performance for the prediction of disease severity with a highest test ROC-AUC of 0.76 compared with 0.70 (0.76 vs. 0.70, p = 0.023) for visual CT severity score and clinical variables. The progression prediction model achieved a test C-index of 0.868 when it was based on the combination of CT radiomics and clinical variables compared with 0.767 when based on CT radiomics features alone (p < 0.001), 0.847 when based on clinical variables alone (p = 0.110), and 0.860 when based on the combination of visual CT severity scores and clinical variables (p = 0.549). Furthermore, the model based on the combination of CT radiomics and clinical variables achieved time-dependent ROC-AUCs of 0.897, 0.933, and 0.927 for the prediction of progression risks at 3, 5 and 7 days, respectively. Conclusion: CT radiomics features combined with clinical variables were predictive of COVID-19 severity and progression to critical illness with fairly high accuracy.

• Korean Journal of Radiology
• /
• 제22권3호
• /
• pp.395-404
• /
• 2021

Objective: To compare the utility of computed tomography perfusion (CTP) and three different 4-point scoring systems in computed tomography angiography (CTA) in confirming brain death (BD) in patients with and without skull defects. Materials and Methods: Ninety-two patients clinically diagnosed as BD using CTA and/or CTP for confirmation were retrospectively reviewed. For the final analysis, 86 patients were included in this study. Images were re-evaluated by three radiologists according to the 4-point scoring systems that consider the vessel opacification on 1) the venous phase for both M4 segments of the middle cerebral arteries (MCAs-M4) and internal cerebral veins (ICVs) (A60-V60), 2) the arterial phase for the MCA-M4 and venous phase for the ICVs (A20-V60), 3) the venous phase for the ICVs and superior petrosal veins (ICV-SPV). The CTP images were independently reviewed. The presence of an open skull defect and stasis filling was noted. Results: Sensitivities of the ICV-SPV, A20-V60, A60-V60 scoring systems, and CTP in the diagnosis of BD were 89.5%, 82.6%, 67.4%, and 93.3%, respectively. The sensitivity of A20-V60 scoring was higher than that of A60-V60 in BD patients (p < 0.001). CTP was found to be the most sensitive method (86.5%) in patients with open skull defect (p = 0.019). Interobserver agreement was excellent in the diagnosis of BD, in assessing A20-V60, A60-V60, ICV-SPV, CTP, and good in stasis filling (κ: 0.84, 0.83, 0.83, 0.83, and 0.67, respectively). Conclusion: The sensitivity of CTA confirming brain death differs between various proposed 4-point scoring systems. Although the ICV-SPV is the most sensitive, evaluation of the SPV is challenging. Adding CTP to the routine BD CTA protocol, especially in cases with open skull defect, could increase sensitivity as a useful adjunct.

• 한국방사선학회논문지
• /
• 제14권3호
• /
• pp.211-220
• /
• 2020

고화질의 전산화단층촬영상을 통해 정확한 병변 검출과 진단을 할 수 있다. 이와 같은 장점 때문에 전산화단층촬영 시 방사선량을 줄이면서 영상 화질을 개선하기 위해 많은 연구가 수행되었다. 최근 전산화단층촬영상 화질을 향상시키기 위한 딥러닝 기반 기술이 개발되었고, 기존의 기술에 비해 우수한 성능을 보이고 있다. 본 연구에서는 전산화단층촬영상의 공간분해능을 향상시키기 위해 초고해상도 합성곱 신경망 모델을 사용하였으며, 초고해상도 합성곱 신경망 모델의 성능을 결정하는 초 매개 변수 변화에 따른 영상 화질을 평가하여 초고해상도 합성곱 신경망 모델에 대한 초 매개 변수의 효과를 검증하였다. Profile, 구조적 유사성 지수, 최대신호 대 잡음비 및 반치폭을 측정하여 초 매개 변수 변화에 따른 초고해상도 합성곱 신경망 모델의 성능을 평가하였다. 연구결과, 초고해상도 합성곱 신경망 모델의 성능은 epoch와 training set이 증가함에 따라 향상되었으며, 전산화단층촬영상 화질을 향상시키기 위해 learning rate 최적화가 필요하다는 사실을 확인하였다. 따라서 최적의 초 매개 변수와 함께 구현된 초고해상도 합성곱 신경망 모델은 전산화단층촬영상의 품질을 향상시킬 수 있다.