• Journal of Gastric Cancer
• /
• v.24 no.3
• /
• pp.327-340
• /
• 2024

Purpose: Results of initial endoscopic biopsy of gastric lesions often differ from those of the final pathological diagnosis. We evaluated whether an artificial intelligence-based gastric lesion detection and diagnostic system, ENdoscopy as AI-powered Device Computer Aided Diagnosis for Gastroscopy (ENAD CAD-G), could reduce this discrepancy. Materials and Methods: We retrospectively collected 24,948 endoscopic images of early gastric cancers (EGCs), dysplasia, and benign lesions from 9,892 patients who underwent esophagogastroduodenoscopy between 2011 and 2021. The diagnostic performance of ENAD CAD-G was evaluated using the following real-world datasets: patients referred from community clinics with initial biopsy results of atypia (n=154), participants who underwent endoscopic resection for neoplasms (Internal video set, n=140), and participants who underwent endoscopy for screening or suspicion of gastric neoplasm referred from community clinics (External video set, n=296). Results: ENAD CAD-G classified the referred gastric lesions of atypia into EGC (accuracy, 82.47%; 95% confidence interval [CI], 76.46%-88.47%), dysplasia (88.31%; 83.24%-93.39%), and benign lesions (83.12%; 77.20%-89.03%). In the Internal video set, ENAD CAD-G identified dysplasia and EGC with diagnostic accuracies of 88.57% (95% CI, 83.30%-93.84%) and 91.43% (86.79%-96.07%), respectively, compared with an accuracy of 60.71% (52.62%-68.80%) for the initial biopsy results (P<0.001). In the External video set, ENAD CAD-G classified EGC, dysplasia, and benign lesions with diagnostic accuracies of 87.50% (83.73%-91.27%), 90.54% (87.21%-93.87%), and 88.85% (85.27%-92.44%), respectively. Conclusions: ENAD CAD-G is superior to initial biopsy for the detection and diagnosis of gastric lesions that require endoscopic resection. ENAD CAD-G can assist community endoscopists in identifying gastric lesions that require endoscopic resection.

• Korean Journal of Computational Design and Engineering
• /
• v.15 no.4
• /
• pp.297-305
• /
• 2010

Medical image acquisition techniques such as CT and MRI have disadvantages in that the numerous time and efforts are needed. Furthermore, a great amount of radiation exposure is an inherent proberty of the CT imaging technique, a number of side-effects are expected from such method. To improve such conventional methods, a number of novel methods that can obtain 3D medical images from a few X-ray images, such as algebraic reconstruction technique (ART), have been developed. Such methods deform a generic model of the internal body part and fit them into the X-ray images to obtain the 3D model; the initial shape, therefore, affects the entire fitting process in a great deal. From this fact, we propose a novel method that can generate a 3D vertebraic generic model based on the statistical database of CT scans in this study. Moreover, we also discuss a method to generate patient-tailored generic model using the facts obtained from the statistical analysis. To do so, the mesh topologies of CT-scanned 3D vertebra models are modified to be identical to each other, and the database is constructed based on them. Furthermore, from the results of a statistical analysis on the database, the tendency of shape distribution is characterized, and the modeling parameters are extracted. By using these modeling parameters for generating the patient-tailored generic model, the computational speed and accuracy of ART can greatly be improved. Furthermore, although this study only includes an application to the C1 (Atlas) vertebra, the entire framework of our method can be applied to other body parts generally. Therefore, it is expected that the proposed method can benefit the various medical imaging applications.

• Journal of radiological science and technology
• /
• v.26 no.3
• /
• pp.13-17
• /
• 2003

By converting movable indirect mass chest X-ray devices for vehicles into digital systems and upgrading it to share information with the hospital's medical image information system, excellencies have been confirmed as a result of installing and running this type of system and are listed hereinafter. 1. Upgrading analog systems, such as indirect mass chest X-ray devices dependent on printed film, to digital systems allows them to be run and managed much more efficiently, contributing to the increase in the stability and the efficiency of the system. 2. Unlike existing images, communication based on DICOM standards allow images to be compatible with the hospital's outer and inner network PACS systems, extending the scope of the radiation departments information system. 3. Assuming chest-exclusive indirect mass chest X-rays, a linked development of CAD (Computer Aided Diagnosis, Detector) becomes possible. 4. By applying wireless Internet, Web-PACS for movable indirect mass chest X-ray devices for vehicles will become possible. Research in these fields must continue and if the superior image quality and convenience of digital systems are confirmed, I believe that the conversion of systems still dependent on analog images to modernized digital systems is a must.

• Journal of Korea Multimedia Society
• /
• v.9 no.12
• /
• pp.1588-1595
• /
• 2006

In this paper, we proposed a liver extracting procedure for computer aided liver diagnosis system. Extraction of liver region in an abdominal CT image is difficult due to interferences of other organs. For this reason, liver region is extracted in a region of interest(ROI). ROI is selected by the window which can measure the distribution of Hounsfield Unit(HU) value of liver region in an abdominal CT image. The distribution is measured by an existential probability of HU value of lever region in the window. If the probability of any window is over 50%, the center point of the window would be assigned to ROI. Actually, liver region is not clearly discerned from the adjacent organs like muscle, spleen, and pancreas in an abdominal CT image. Liver region is extracted by the watershed segmentation algorithm which is effective in this situation. Because it is very sensitive to the slight valiance of contrast, it generally produces over segmentation regions. Therefore these regions are required to merge into the significant regions for optimal segmentation. Finally, a liver region can be selected and extracted by prier information based on anatomic information.

• Journal of Korea Multimedia Society
• /
• v.13 no.11
• /
• pp.1698-1705
• /
• 2010

In this paper, we propose a fast liver segmentation method from magnetic resonance(MR) images. Our method efficiently divides a MR image into a set of discrete objects, and boundaries based on the normalized gradient magnitude information. Then, the objects belonging to the liver are detected by using 2D seeded region growing with seed points, which are extracted from the segmented liver region of the slice immediately above or below the current slice. Finally, rolling ball algorithm, and connected component analysis minimizes false positive error near the liver boundaries. Our method was validated by twenty data sets and the results were compared with the manually segmented result. The average volumetric overlap error was 5.2%, and average absolute volumetric measurement error was 1.9%. The average processing time for segmenting one data set was about three seconds. Our method could be used for computer-aided liver diagnosis, which requires a fast and accurate segmentation of liver.

• Asian Pacific Journal of Cancer Prevention
• /
• v.17 no.2
• /
• pp.835-838
• /
• 2016

Background: Lung cancer remains one of the most common cancers in the world, both in terms of new cases (about 13% of total per year) and deaths (nearly one cancer death in five), because of the high case fatality. Errors in lung cancer type or malignant growth determination lead to degraded treatment efficacy, because anticancer strategy depends on tumor morphology. Materials and Methods: We have made an attempt to evaluate effectiveness of machine learning algorithms in the task of lung cancer classification based on gene expression levels. We processed four publicly available data sets. The Dana-Farber Cancer Institute data set contains 203 samples and the task was to classify four cancer types and sound tissue samples. With the University of Michigan data set of 96 samples, the task was to execute a binary classification of adenocarcinoma and non-neoplastic tissues. The University of Toronto data set contains 39 samples and the task was to detect recurrence, while with the Brigham and Women's Hospital data set of 181 samples it was to make a binary classification of malignant pleural mesothelioma and adenocarcinoma. We used the k-nearest neighbor algorithm (k=1, k=5, k=10), naive Bayes classifier with assumption of both a normal distribution of attributes and a distribution through histograms, support vector machine and C4.5 decision tree. Effectiveness of machine learning algorithms was evaluated with the Matthews correlation coefficient. Results: The support vector machine method showed best results among data sets from the Dana-Farber Cancer Institute and Brigham and Women's Hospital. All algorithms with the exception of the C4.5 decision tree showed maximum potential effectiveness in the University of Michigan data set. However, the C4.5 decision tree showed best results for the University of Toronto data set. Conclusions: Machine learning algorithms can be used for lung cancer morphology classification and similar tasks based on gene expression level evaluation.

• Journal of Biomedical Engineering Research
• /
• v.38 no.5
• /
• 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.

• Journal of KIISE:Software and Applications
• /
• v.29 no.11
• /
• pp.803-808
• /
• 2002

This paper describes automatic methods for detection of kidney and kidney tumor on abdominal CT scans. The abdominal CT images were digitalized using a film digitizer and a gray-level threshold method was used to segment the kidney. Based on texture analysis results, which were perform on sample images of kidney tumors, SEED region of kidney tumor was selected as result of homogeneity test. The average and standard deviation, which are representative statistical moments, were used to as an acceptance criteria for homogeneous test. Region growing method was used to segment the kidney tumor from the center pixel of selected SEED region using a gray-level value as an acceptance criteria for homogeneity test. These method were applied to 113 images of 9 cases, which were scanned by GE Hispeed Advantage CT scanner and digitalized by Lumisvs LS-40 film digitizer. The sensitivity was 85% and there was no false-positive results.

• Journal of Korea Multimedia Society
• /
• v.14 no.3
• /
• pp.348-355
• /
• 2011

In this paper, we propose an automatic prostate segmentation method from dynamic magnetic resonance (MR) images. Our method detects contrast-enhanced images among the dynamic MR images using an average intensity analysis. Then, the candidate regions of prostate are detected by the B-spline non-rigid registration and subtraction between the pre-contrast and contrast-enhanced MR images. Finally, the prostate is segmented by performing a dilation operation outward, and sequential shape propagation inward. Our method was validated by ten data sets and the results were compared with the manually segmented results. The average volumetric overlap error was 6.8%, and average absolute volumetric measurement error was 2.5%. Our method could be used for the computer-aided prostate diagnosis, which requires an accurate prostate segmentation.

• Journal of Broadcast Engineering
• /
• v.18 no.5
• /
• pp.702-712
• /
• 2013

In this paper, we proposed an algorithm that extracts the airway and lung without loss of information in spite of the pulmonary vessel and nodules of the chest wall in the chest CT images. We use a mask image in order to improve the performance and to save processing time of airway and lung segmentation. In the second step, by converting left and right lungs to binary image using the morphological information, we have removed the solitary pulmonary nodule to identify the value of the threshold lung and the chest wall. The last step is to connect the outer shell of the lung with cubic Spline interpolation by adding the perfect pixel and computing the distance of the removed part. Experimental results using Matlab verified that the proposed method could overcome the drawbacks of the conventional methods.