• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제16권7호
• /
• pp.2359-2376
• /
• 2022

With the exponential growth of medical image big data represented by high-resolution CT images(CTI), the high-resolution CTI data is of great importance for clinical research and diagnosis. The paper takes lung CTI as an example to study. Retrieving answer CTIs similar to the input one from the large-scale lung CTI database can effectively assist physicians to diagnose. Compared with the conventional content-based image retrieval(CBIR) methods, the CBIR for lung CTIs demands higher retrieval accuracy in both the contour shape and the internal details of the organ. In traditional supervised deep learning networks, the learning of the network relies on the labeling of CTIs which is a very time-consuming task. To address this issue, the paper proposes a Weakly Supervised Similarity Evaluation Network (WSSENet) for efficiently support similarity analysis of lung CTIs. We conducted extensive experiments to verify the effectiveness of the WSSENet based on which the CBIR is performed.

• 대한임베디드공학회논문지
• /
• 제17권5호
• /
• pp.265-272
• /
• 2022

As the risk of lung cancer has increased, early-stage detection and treatment of cancers have received a lot of attention. Among various medical imaging approaches, computer tomography (CT) has been widely utilized to examine the size and growth rate of lung nodules. However, the process of manual examination is a time-consuming task, and it causes physical and mental fatigue for medical professionals. Recently, many computer-aided diagnostic methods have been proposed to reduce the workload of medical professionals. In recent studies, encoder-decoder architectures have shown reliable performances in medical image segmentation, and it is adopted to predict lesion candidates. However, localizing nodules in lung CT images is a challenging problem due to the extremely small sizes and unstructured shapes of nodules. To solve these problems, we utilize atrous spatial pyramid pooling (ASPP) to minimize the loss of information for a general U-Net baseline model to extract rich representations from various receptive fields. Moreover, we propose mixed-up attention mechanism of reverse, boundary and convolutional block attention module (CBAM) to improve the accuracy of segmentation small scale of various shapes. The performance of the proposed model is compared with several previous attention mechanisms on the LIDC-IDRI dataset, and experimental results demonstrate that reverse, boundary, and CBAM (RB-CBAM) are effective in the segmentation of small nodules.

• 핵의학기술
• /
• 제21권1호
• /
• pp.54-59
• /
• 2017

폐암(Lung cancer) 환자의 경우 PET/CT 검사에서 호흡으로 인하여 영상의 정합오차가 발생하게 되는데 이로 인해 정확한 SUV 와 Tumor volume측정을 방해하는 요인으로 작용된다. $SUV_{max}$를 이용하여 폐암 환자의 수술 후 예측 및 항암화학요법의 효과를 평가하고 있으며, 방사선치료의 예후 예측 및 평가를 위해 현재 Tumor volume과 SUV를 이용한 지표가 사용되고 있다. 그렇기 때문에 정합오차를 줄이기 위해 본원에서는 Respiratory gating method를 적용하여 검사를 시행하고 있다. 본 연구는 Step and Go 방식이 아닌 Flow mode를 적용하여 Non-gating 영상과 첫 번째 Respiratory Gating영상, 그리고 추가로 부분 Respiratory gating 촬영하여 Respiratory gating method의 유용성에대해 알아보았다. 2016년 6월부터 2016년 9월까지 분당서울대학교병원에서 PET/CT 검사를 한 폐암 환자 20명(남:12명, 여:8명)을 대상으로 amplitude rang 15% 미만인 호흡이 안정한 환자군 10명 15%초과한 호흡이 불안정한 환자군 10명으로 나누어 비교분석하였다. 전체 환자에서 Non-gating 영상의 $SUV_{max}$는 $9.43{\pm}3.93$, $SUV_{mean}$은 $1.77{\pm}0.89$, Tumor Volume은 $4.17{\pm}2.41$로 측정되었고 기존 Gating 영상에서 $SUV_{max}$는 $10.08{\pm}4.07$, $SUV_{mean}$은 $1.75{\pm}0.81$, Tumor Volume은 $3.56{\pm}2.11$로 측정되었다. 그리고 추가 Lung gating 영상에서 $SUV_{max}$는 $10.86{\pm}4.36$, $SUV_{mean}$은 $1.77{\pm}0.85$, Tumor volume은 $3.36{\pm}1.98$을 얻었다. Non-gating 영상과 기존 Gating 영상, 그리고 기존 Gating 영상과 추가 Lung gating 영상을 비교했을 때 둘 다 $SUV_{mean}$ 값에서 통계적으로 유의한 차이를 보이지 않았으나(P>0.05) $SUV_{max}$와 Tumor volume에서 유의한 차이를 보였다(P<0.05). 그중 호흡이 안정한 환자군보다 호흡이 불안정한 환자군에서의 증감률이 더 크게 나타났다. Amplitude range 폭은 전체 20명 중 12명(Signal이 안정된 환자 3명 불안정한 환자 9명)이 추가 Lung gating을 했을 때 기존 Gating 영상보다 더 낮게 나타났다. 본 연구에 의하면 Flow mode를 적용하여 Respiration Gating Method로 촬영한 결과 추가적인 CT 촬영 없이 호흡으로 인해 발생하는 병변의 움직임을 보정해 주어 $SUV_{max}$, Tumor volume을 Non-gating 영상보다 더 정확하게 측정할 수 있었다. 그리고 처음 Gating 할 때보다 추가 촬영 시 호흡의 안정에 따른 Amplitude range 폭의 낮아짐을 알 수 있었다. 따라서 Gating 영상이 Non-gating 영상보다 진단에 더 유용한 정보를 제공함을 알 수 있었고, Signal이 불규칙적인 환자에게 시간적 여유가 있다면 추가로 부분 촬영을 하는 것이 도움이 될 것이라고 사료된다.

• 대한핵의학회:학술대회논문집
• /
• 대한핵의학회 2002년도 춘계학술대회 및 총회
• /
• pp.25-27
• /
• 2002

With the advent of multidetector-row CT, lung cancer imaging is much more promising than before. However, the effectiveness of multidetector-row CT in making an initial diagnosis, staging, and evaluating post-treatment changes of lung cancer still remains to be proved. Fast imaging along with volumetric data set and attendant multi-planar imaging provide much more details on the anatomic changes and pathology associated with lung cancer. However, with images showing anatomic and pathologic changes only, radiologists confront with several questions the answers of which may help evaluate lung cancer more thoroughly. The frequent questions that I have in dally practice of chest CT interpretation are as follows.

• 대한방사선기술학회지:방사선기술과학
• /
• 제47권1호
• /
• pp.39-48
• /
• 2024

In this study, we optimized the FNLM algorithm through a simulation study and applied it to a phantom scanned by low-dose CT to evaluate whether the FNLM algorithm can be used to obtain improved image quality images. We optimized the FNLM algorithm with MASH phantom and FASH phantom, which the algorithm was applied with MATLAB, increasing the smoothing factor from 0.01 to 0.05 with increments of 0.001 and measuring COV, RMSE, and PSNR values of the phantoms. For both phantom, COV and RMSE decreased, and PSNR increased as the smoothing factor increased. Based on the above results, we optimized a smoothing factor value of 0.043 for the FNLM algorithm. Then we applied the optimized FNLM algorithm to low dose lung CT and lung CT under normal conditions. In both images, the COV decreased by 55.33 times and 5.08 times respectively, and we confirmed that the quality of the image of low dose CT applying the optimized FNLM algorithm was 5.08 times better than the image of lung CT under normal conditions. In conclusion, we found that the smoothing factor of 0.043 among the factors of the FNLM algorithm showed the best results and validated the performance by reducing the noise in the low-quality CT images due to low dose with the optimized FNLM algorithm.

• 한국멀티미디어학회논문지
• /
• 제24권3호
• /
• pp.373-381
• /
• 2021

In this paper, we propose a classification model based on convolutional neural network(CNN) for predicting 2-year recurrence in non-small cell lung cancer(NSCLC) patients using preoperative chest CT images. Based on the region of interest(ROI) defined as the tumor internal and external area, the input images consist of an intratumoral patch, a peritumoral patch and a peritumoral texture patch focusing on the texture information of the peritumoral patch. Each patch is trained through AlexNet pretrained on ImageNet to explore the usefulness and performance of various patches. Additionally, ensemble learning of network trained with each patch analyzes the performance of different patch combination. Compared with all results, the ensemble model with intratumoral and peritumoral patches achieved the best performance (ACC=98.28%, Sensitivity=100%, NPV=100%).

• 한국콘텐츠학회논문지
• /
• 제11권10호
• /
• pp.333-341
• /
• 2011

분진에 노출 되었던 집단을 대상으로 폐질환 선별검사를 위해 시행한 저선량 CT영상의 관상동맥 석회화 소견과 석회화수치 CT검사 결과를 비교 분석하였다. 연구윤리심의위원회의 승인과 연구대상자로부터 동의서를 받은 후, 과거 분진에 노출된 직업력을 갖고 있는 61명의 남자를 대상으로 폐질환의 선별검사를 위해 저선량 CT촬영과 석회화수치 CT검사를 동시에 실시하였다. 저선량 CT영상(Axial image)은 영상의학과 전문의로부터 진폐 소견 및 폐질환과 관상동맥 석회화소견에 대해 판독하였고, 석회화수치 CT검사로부터 얻은 기초 영상(Row image)은 별도의 워크스테이션으로 보내져 상업적인 라피디아 소프트웨어(ver 2.8)를 이용하여 관상동맥 석회화수치를 구하였다. 저선량 CT영상에서 석회화소견을 보이지 않은 그룹(42명, 68.9%)과 보인 그룹(19명, 31.1%)사이에 총 석회화(13.68 vs. 582.93, p=.009), 좌전하행관상동맥 (3.15 vs. 248.95, p=.006)의 석회화수치가 통계학적으로 유의한 차이를 보였고, 좌주관상동맥, 좌회선관상동맥, 우관상동맥에서 석회화소견을 보인 그룹에서 높게 나타났다(p>0.05). 저선량 CT영상의 석회화소견은 석회화수치 CT검사의 석회화수치 100에서 가장 높은 일치도(K-value=0.80, 95% 신뢰구간=0.69-0.91)를 보였다. 폐질환을 조기 발견하기 위해 시행된 저선량 CT영상에서 보여진 석회화소견은 석회화수치 검사결과와 높은 관련성을 보임으로써 폐쇄성 관상동맥질환을 예측할 수 있는 것으로 사료된다.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제13권1호
• /
• pp.347-370
• /
• 2019

In the computer-aided detection (CAD) system of pulmonary nodules, a high false positive rate is common because the density and the computed tomography (CT) values of the vessel and the nodule in the CT images are similar, which affects the detection accuracy of pulmonary nodules. In this paper, a method of automatic detection of pulmonary nodules based on multi-scale enhancement filters and 3D shape features is proposed. The method uses an iterative threshold and a region growing algorithm to segment lung parenchyma. Two types of multi-scale enhancement filters are constructed to enhance the images of nodules and blood vessels in 3D lung images, and most of the blood vessel images in the nodular images are removed to obtain a suspected nodule image. An 18 neighborhood region growing algorithm is then used to extract the lung nodules. A new pulmonary nodules feature descriptor is proposed, and the features of the suspected nodules are extracted. A support vector machine (SVM) classifier is used to classify the pulmonary nodules. The experimental results show that our method can effectively detect pulmonary nodules and reduce false positive rates, and the feature descriptor proposed in this paper is valid which can be used to distinguish between nodules and blood vessels.

• 한국멀티미디어학회논문지
• /
• 제16권6호
• /
• pp.700-707
• /
• 2013

본 논문에서는 시간차 흉부 CT 영상의 폐 실질 비강체 정합을 위하여 영역 이진화 모델링과 지역적 변형 모델을 이용한 정합 기법을 제안한다. 제안 기법은 먼저 폐 혈관과 실질을 분할하고, 영역 이진화 모델링을 수행하여 두 영상 사이의 밝기값의 차이에 따른 정합 오차를 최소화 한다. 다음으로 초기 정합 기법으로 두 폐 표면을 전역적으로 정렬하고, 지역적 변형 변환 모델을 제안하여 비강체 정합을 수행한다. 또한, 정합 후 감산된 시간에 따른 밝기값 차이가 미리 정의된 칼라 맵을 이용하여 가시화 된다. 실험 결과는 제안기법이 10명의 환자에 대하여 최대호흡과 최소호흡 CT 영상에서 폐 실질을 정확하게 정합하였음을 보여주었다. 제안된 비강체 정합 기법은 폐 실질에 대한 정량적 분석 결과의 직관적인 칼라 매핑을 통하여 다양한 폐 질환의 정량적 분석에 유용하게 사용될 수 있다.

• International Journal of Vascular Biomedical Engineering
• /
• 제4권1호
• /
• pp.27-30
• /
• 2006

Background; Pleural micro-metastasis of lung cancer is detected by touch print cytology or pleural lavage cytology, but its prognostic impact has not elucidated yet. We hypothesize that recurrence may depend on the amount of tumor cells disseminated in pleural cavity, if the invasiveness of all cancer is the same. To predict the amount of tumor cells disseminated in pleural cavity, we need pleural surface area, distributed pattern of cells and concentration of cells per unit area. Human pleural surface area has not reported yet. In this report, we calculate the normal human pleural surface area using CT image data processing. Methods; Twenty persons were checked CT scan, and we obtained the data from each image. In order to calculate the pleural surface, the outline of lung was firstly extruded from CT image data using home-made Digitizer program. And the distance between CT images was calculated from the extruded outline. Finally a normal human pleural surface was calculated from function between the distance of consecutive CT images and the calculated length. Results; Their mean age is $65{\pm}12$ years old (range $26{\sim}77$), body weight is $62{\pm}9\;kg\;(48{\sim}80)$, and height is $167{\pm}6\;cm\;(156{\sim}176)$. The number of images used is $36{\pm}7\;(24{\sim}51)$. Pleural surface area is $211,888{\pm}35,756\;mm^2\;(143,880{\sim}279,576)$. Right-side pleural surface area is $107,932\;mm^2$ and Lt is $103,955\;mm^2$. Costal, mediastinal and diaphragmatic surfaces of right-side pleura are $77,483\;mm^2,\;39,057\;mm^2,\;and\;8,608\;mm^2$ respectively, and left-side are $72,497\;mm^2,\;35,578\;mm^2,\;and\;4,120\;mm^2$ respectively. Conclusion; Normal human pleural surface area is calculated using CT image data at first and the result is about $0.212\;m^2$.