http://chineseinput.net/에서 pinyin(병음)방식으로 중국어를 변환할 수 있습니다.
변환된 중국어를 복사하여 사용하시면 됩니다.
Identification of abnormal tissue from CT images using improved ResNet34
Naoya Honda,Tohru Kamiya,Shoji Kido 제어로봇시스템학회 2022 제어로봇시스템학회 국제학술대회 논문집 Vol.2022 No.11
In recent years, CT examinations have been widely used as a screening method to detect lung cancer. However, reading enormous CT images become a heavy burden to the physician. To avoid this problem, computer-aided diagnosis systems have been introduced on CT screening. In general, physicians consider patient information in addition to image information when they make a diagnosis, new efforts are being made to improve the accuracy of diagnosis by mimicking this information with a machine. In this paper, we propose a method for identifying pulmonary nodules by adding medical record information to images to improve the accuracy of diagnosis. We classify nodules from unknown data by assigning branching information of vascular opacities, straight vascular shadows, and nodular shadows as labeled image, which are a cause of misrecognition based on image features in machine learning. In the experiment, the classification accuracy of the nodule class was improved by adding clinical information to 644 images including 161 nodal images.
ROI-based Fully Automated Liver Registration in Multi-phase CT Images
Kentaro SAITO,Huimin LU,Hyoungseop KIM,Shoji KIDO,Masahiro TANABE 제어로봇시스템학회 2018 제어로봇시스템학회 국제학술대회 논문집 Vol.2018 No.10
In this paper, we propose a registration method for fully automated liver tumor detection. Multiple phases CT is used for the detection of the liver tumor because multiple phase CT can give different characteristic features of lesions for each time phases. Registration accuracy is important when obtaining image features from multiple time phases. However, since each time phases have different image density characteristics, therefore registration of multi-phase CT is a challenging task. In this paper, we propose a robust initial alignment method independent of changing image density features in each time phase, and deformable registration method with region of interests (ROI) as liver region extracted by U-Net. Our proposed method is evaluated on 15 patient image sets. This method is applied to the early arterial phase and the equilibrium phase to registries. Experimental results show that segmentation of early arterial phase is 83% and registration is 93% accuracy.
Naoki Asatani,Tohru Kamiya,Shingo Mabu,Shoji Kido 제어로봇시스템학회 2020 제어로봇시스템학회 국제학술대회 논문집 Vol.2020 No.10
According to the 2016 World Health Organization (WHO) survey, respiratory diseases are serious diseases that account for four of the top ten causes of death in the world, accounting for more than 8 million deaths worldwide. Currently, the diagnosis of respiratory disease is made by auscultation, but in order to make an accurate diagnosis, a number of abnormal patterns of respiratory sounds need to be memorized, and the results of the diagnosis are dependent on the proficiency of the physician. Therefore, a computer aided diagnosis (CAD) system is needed to quantitatively classify the respiratory sounds and output the results as a "second opinion". In this paper, a short-time Fourier transformed spectrogram, a Constant-Q transformed logarithmic frequency spectrogram, and a continuous wavelet transformed scalogram are simultaneously input to VGG16 which is one of the network models of CNN(Convolutional Neural Network) and classified by LSTM (Long short-term memory). The proposed method is applied to 26 respiratory sounds, and the 0.90 of accuracy, sensitivity of 0.97, and specificity of 0.90 is obtained.
Classification of Respiratory Sounds by Generated Image and Improved CRNN
Naoki Asatani,Tohru Kamiya,Shingo Mabu,Shoji Kido 제어로봇시스템학회 2021 제어로봇시스템학회 국제학술대회 논문집 Vol.2021 No.10
The death toll from respiratory illness reached nearly 8 million in 2019. Auscultation is used to diagnose for respiratory illness. Highly accurate diagnosis is required to reduce the number of deaths. However, unlike diagnostic imaging, auscultation of respiratory sounds could not visualize the diagnostic results. In addition, since there is a problem that the experience of a doctor affects the diagnosis results, it is required to develop a diagnostic system for quantitative analysis. In recent years, the development of a diagnostic system using the ICBHI 2017 Challenge Respiratory Sound Database has been carried out in the field of respiratory sound analysis. However, the proposed system still has accuracy problems. Therefore, in this study, we improve the proposed method by classifying the improved CRNN (Convolutional Recurrent Neural Network) by inputting multiple respiratory sound images. As a result, Sensitivity: 0.64, Specificity: 0.83, Average Score: 0.74, Harmonic Score: 0.72 were obtained, and excellent results were achieved compared with other methods.
Koki Minami,Huimin Lu,Hyoungseop Kim,Shingo Mabu,Yasushi Hirano,Shoji Kido 제어로봇시스템학회 2019 제어로봇시스템학회 국제학술대회 논문집 Vol.2019 No.10
Auscultation of respiratory sounds is very important for discovering the respiratory disease. However, there is no quantitative evaluation method for the diagnosis of respiratory sounds until now. It is necessary to develop a system to support the diagnosis of respiratory sounds. In addition, there are few studies using dataset suitable for generating realistic classification models that can be used in clinical sites in algorithm development for automatic analysis of respiratory sounds. We describe the development of an algorithm for the automatic classification of the large-scale respiratory sound dataset used in ICBHI 2017 Challenge as containing crackles, containing wheeze, containing both, and normal. Our approach consists of two major components. Firstly, transformation of one-dimensional signals into two-dimensional time-frequency representation images using short-time Fourier transform and continuous wavelet transform. Secondly, classification of transferred images using convolutional neural networks. In this paper, we apply our proposed method to 920 respiratory sound data, and achieve score of 28[%], harmonic score of 81[%], sensitivity of 54[%] and specificity of 42[%].
A Detection Method for Liver Cancer Region Based on Faster R-CNN
Muki FURUZUKI,Huimin LU,Hyoungseop KIM,Yasushi HIRANO,Shingo MABU,Masahiro TANABE,Shoji KIDO 제어로봇시스템학회 2019 제어로봇시스템학회 국제학술대회 논문집 Vol.2019 No.10
In recent years, liver cancer has become the fourth-largest number of deaths in the world. Surgery is a typical treatment for liver cancer. Therefore, advance information about the number and size of cancer is important for surgery. Multi-phase CT images are well known diagnostic method. By extracting the region of the liver and the region of cancer from the obtained CT image, the shape can be finally restored in 3D. In this paper, as a preliminary step to construct an image analysis method for efficiently extracting cancerous regions in multi-phase CT, we propose a method of obtaining a rectangular region as a rough cancerous region of interest. As a method, after preprocessing the input image, using Faster R-CNN, the region of interest including the cancer region is extracted as a rectangle. As a result of applying this method to 11 cases of arterial phase of multi-phase CT, the detection performance was different depending on the network model adopted for backbone part.
Detection of Lung Nodules from Temporal Subtraction Image Using Deep Learning
Kohei TAMAI,Noriaki MIYAKE,Humin LU,Hyoungseop KIM,Seiichi MURAKAMI,Takatoshi AOKI,Shoji KIDO 제어로봇시스템학회 2019 제어로봇시스템학회 국제학술대회 논문집 Vol.2019 No.10
In recent years, the number of death due to lung cancer is increasing year by year worldwide. Early detection and early treatment of lung cancer are important. Especially, early detection of the abnormalities on thoracic MDCT images detection of small nodules is required in visual screening. Although a CT apparatus is used for the examination, the burden on the image interpretation doctor is large due to the high performance of the CT, so the diagnostic accuracy may be reduced. In this paper, we propose an image analysis method to detect abnormal shadows from chest CT images automatically. The initial lesion candidate areas are extracted by using temporal subtraction technique that emphasizes temporal change by subtracting from a current image to previous one which is obtained same subject. The image of the area is given as input and classification is performed by CNN (Convolutional Neural Network). In the discrimination experiment based on our proposed method, 90.26 [%] of true positive rates and 13.58 [%] of false positive rates are obtained from the 49 clinical cases.