• 한국정보전자통신기술학회논문지
• /
• 제6권1호
• /
• pp.37-44
• /
• 2013

본 논문에서는 voxel classification을 이용한 폐 결절 자동 검출 시스템을 제안한다. 제안하는 폐 영상 분석 방법은 크게 세 단계로 구성된다. 첫 번째 단계에서는 분석 대상 폐 영역을 분할한다. 그리고 두 번째 단계는 분할된 폐 영역 내에서 폐 구조물을 분할한다. 마지막으로 두 번째 과정에서 분할된 폐결절후보와 폐혈관 voxel을 대상으로 log-polar sampling을 이용한 특징 벡터를 만들고, 특징벡터를 입력 값으로 하여 support vector machine classifier를 이용하여 분석대상 voxel을 폐 결절 voxel과 비결절 voxel로 구분하여 폐 결절을 검출한다.

• 전자공학회논문지CI
• /
• 제46권6호
• /
• pp.35-43
• /
• 2009

환자의 흉부 CT 영상을 입력으로 하여 폐 영역의 결절(nodule)을 효과적으로 분류하는 것이 목적인 컴퓨터 조력 진단(Computer Aided Diagnosis, CAD) 시스템에서는 대부분 폐 영역의 혈관 추출 단계가 우선적으로 진행된다. 혈관 조영제를 투여한 환자 영상에서 폐 혈관과 폐 결절은 흉부 CT 영상에서 비슷한 감쇄를 보이므로 혈관 추출 단계에서 혈관과 결절이 함께 추출될 수 있다. 때문에 폐 혈관이 제거된 나머지 영역에서 폐 결절을 탐지 및 분류하는 방식의 폐 결절 분류기는 혈관 추출 기법의 성능에 큰 영향을 받는다. 본 논문에서는 폐 결절이 혈관으로 오인식 되어 혈관과 함께 추출될 수 있는 문제를 극복하기 위해 혈관의 두께 패턴을 분석하여 폐 결절을 재분류하는 기법을 제안한다. 제안된 방법은 폐 영역 결정 단계, 폐 혈관 영역 추출 및 골격 형성 단계, 혈관 토폴로지 구성 및 보정 단계, 그리고 혈관 후보 내 결절 재분류 단계 등의 네 단계로 이루어져 있다. 제안한 방법의 정확도를 분석하기 위해 폐 결절이 혈관을 침투하여 분포하는 환자들의 입력 영상을 이용하여 판독 전문의의 도움을 받아 분류기의 정확도 평가를 진행하였다. 실험 결과에서는 제안한 혈관 추출 및 결절 재분류 기법을 사용한 CAD 시스템의 폐 결절 분류기 성능과 재분류 절차를 진행하지 않은 분류기 성능에 대해 비교하고 제안한 방법이 혈관 영역으로 잘못 포함된 결절을 자동으로 정확하게 재분류 할 수 있음을 보여준다.

• 한국멀티미디어학회논문지
• /
• 제20권7호
• /
• pp.994-1003
• /
• 2017

Ground-glass opacity nodules(GGNs) in chest CT images are associated with lung cancer, and have a different malignant rate depending on existence of solid component in the nodules. In this paper, we propose a method to classify pure GGNs and part-solid GGNs using multiview images and texture analysis in pulmonary GGNs with solid components of 5mm or smaller. We extracted 1521 features from the GGNs segmented from the chest CT images and classified the GGNs using a SVM classification model with selected features that classify pure GGNs and part-solid GGNs through a feature selection method. Our method showed 85% accuracy using the SVM classifier with the top 10 features selected in the multiview images.

• 대한의용생체공학회:의공학회지
• /
• 제38권5호
• /
• pp.237-241
• /
• 2017

Early detection of the pulmonary nodule is important for diagnosis and treatment of lung cancer. Recently, CT has been used as a screening tool for lung nodule detection. And, it has been reported that computer aided detection(CAD) systems can improve the accuracy of the radiologist in detection nodules on CT scan. The previous study has been proposed a method using Convolutional Neural Network(CNN) in Lung CAD system. But the proposed model has a limitation in accuracy due to its sparse layer structure. Therefore, we propose a Deep Convolutional Neural Network to overcome this limitation. The model proposed in this work is consist of 14 layers including 8 convolutional layers and 4 fully connected layers. The CNN model is trained and tested with 61,404 regions-of-interest (ROIs) patches of lung image including 39,760 nodules and 21,644 non-nodules extracted from the Lung Image Database Consortium(LIDC) dataset. We could obtain the classification accuracy of 91.79% with the CNN model presented in this work. To prevent overfitting, we trained the model with Augmented Dataset and regularization term in the cost function. With L1, L2 regularization at Training process, we obtained 92.39%, 92.52% of accuracy respectively. And we obtained 93.52% with data augmentation. In conclusion, we could obtain the accuracy of 93.75% with L2 Regularization and Data Augmentation.

• Tuberculosis and Respiratory Diseases
• /
• 제69권5호
• /
• pp.354-360
• /
• 2010

Background: Primary non-Hodgkin's lymphoma of the lung is a rare entity. It is represented commonly as marginal zone B-cell lymphoma of mucosa-associated lymphoid tissue (MALT) type. Although there have been a few reviews of this lymphoma, clinical features, radiologic findings, management and prognosis have not been well defined. Methods: We reviewed the medical records of 24 patients with primary pulmonary lymphoma between January 1995 and September 2008; all diagnoses had been confirmed based on pathology. Results: The median follow-up time was 42.3 months (range, 0.1~131.2 months). Five (20.8%) patients were asymptomatic, 17 (70.8%) patients had pulmonary symptoms, and the remaining 2 (8.3%) patients presented with constitutional symptoms. There were 16 (66.7%) patients with MALT lymphoma, 4 (16.7%) patients with diffuse large B-cell lymphoma and 4 (16.7%) patients with lymphoma that had not received a WHO classification. Radiologic findings of primary pulmonary lymphoma were diverse and multiple nodule or consolidation was the most common finding regardless of pathologic lymphoma type. PET scan was carried out in 13 (54.2%) patients and all lesions showed notable FDG uptake. MALT lymphoma showed a trend of better prognosis (3-year survival, 78.8% vs. 70.0%; 5-year survival, 78.8% vs. 52.5%; p=0.310) than non-MALT lymphoma. Conclusion: Primary non-Hodgkin's lymphoma of the lung occurs with nonspecific clinical features and radiologic findings. MALT lymphoma is the most common pathologic type of primary pulmonary lymphoma. This entity of lymphoma appears to have a good prognosis and in this study, there was a trend of better outcome than non-MALT lymphoma.

• 재활복지공학회논문지
• /
• 제10권1호
• /
• pp.87-92
• /
• 2016

본 논문에는 관심 영역의 폐실질 영역을 양성과 악성 결절의 분류를 위한 특징인자에 포함으로써 분류성능을 개선하였다. CT를 통해 확인되는 매우 작은 폐결절(4~10mm)은 고형 종양 내에 CT 데이터 복셀 수가 제한되어 기존 컴퓨터보조 진단도구를 통해 처리하기가 어렵다. 이러한 아주 작은 폐 결절의 경우 분석을 위해 주변의 실질을 포함하여 특징인자를 추출하는 것이 CT 복셀 세트를 증가시킬 수 있으며, CT 스캐너와 매개 변수에 대한 컴퓨터 보조진단도구의 유연성을 확보함으로써 진단 성능을 개선할 수 있다. 나이브 베이스와 SVM 약분류기를 이용하는 아다부스트 학습을 통해 304개의 특징인자로부터 유효한 특징인자를 결정하였으며, 제안한 방법을 COPDGene 데이터에 적용한 결과 100%의 정확도, 민감도 및 특이도의 결과를 획득하여 컴퓨터 보조진단에 유용하게 사용될 수 있음을 보였다.

• Journal of Preventive Medicine and Public Health
• /
• 제29권3호
• /
• pp.471-482
• /
• 1996

Early recognition of coalescence in pneumoconiotic lesions is important because such coalescence is associated with the respiratory symptoms and deterioration of lung function. This complicated form of pneumoconiosis also has worse prognosis than does simple pneumoconiosis. High resolution computerized tomography(HRCT) provides significant additional information on the stage of the pneumoconiosis because it easily detects coalescence of nodules and emphysema that may not be apparent on the simple radiograph. The purpose of this study is to clarify the role of HRCT in detection of large opacity and the relationship of change between the coalescence of nodules or emphysema and lung function in dust exposed workers. 1. There was good correlation between the HRCT grade of pneumoconiosis and ILO category of profusion. 5(9.09%) in 55 study population had confluent nodule extending eve, two o, more cuts on HRCT. HRCT could identify the pneumoconiotic nodules which was not found by simple radiogrphy in 6 workers with category 0/0. 2. No significant difference was observed coalescence of nodules and emphysema by dust type. 3. There was no significant difference in pulmonary function according to ILO and HRCT classification. 4. HRCT could detect the significant reduction in $FEV_1,\;FEV_1/FVC$, PEFR, $FEF_{25},\;FEF_{50},\;and\;FEF_{75}$ and remarkable increase in RV and TLC in study persons with emphysema compared with non-emphysema group. 5. Emphysema was found more often in nodules-coalescence group than small opacity group by HRCT. We found that HRCT could easily detect areas of coalescence and complicated emphysema compared to plain chest X-ray. Also our data suggest that it is primarily the degree of emphysema rather than the degree of pneumoconiosis that determines the level of pulmonary function.