A5066648889- Radiomics and Machine Learning in Medical Imaging
- Medical Imaging Techniques and Applications
- Atomic and Subatomic Physics Research
- Lung Cancer Diagnosis and Treatment
- AI in cancer detection
- Advanced MRI Techniques and Applications
- MRI in cancer diagnosis
- Chronic Obstructive Pulmonary Disease (COPD) Research
- COVID-19 diagnosis using AI
- Endometrial and Cervical Cancer Treatments
- Radiation Dose and Imaging
- Advanced X-ray and CT Imaging
- Medical Imaging and Pathology Studies
- Lymphoma Diagnosis and Treatment
- Respiratory Support and Mechanisms
- Medical Imaging and Analysis
- Bladder and Urothelial Cancer Treatments
- Cardiac Imaging and Diagnostics
- Reliability and Agreement in Measurement
- Radiology practices and education
- Medical Image Segmentation Techniques
- Myasthenia Gravis and Thymoma
- Salivary Gland Tumors Diagnosis and Treatment
- Insurance, Mortality, Demography, Risk Management
- Artificial Intelligence in Healthcare and Education
Kobe University
2012-2024
Kyoto University
2018-2024
Kobe University Hospital
2020-2021
Kyoto University Hospital
2019
Foundation for Biomedical Research and Innovation
2015-2017
This study aimed to develop and validate computer-aided diagnosis (CXDx) system for classification between COVID-19 pneumonia, non-COVID-19 the healthy on chest X-ray (CXR) images. From two public datasets, 1248 CXR images were obtained, which included 215, 533, 500 of pneumonia patients, samples, respectively. The proposed CADx utilized VGG16 as a pre-trained model combination conventional method mixup data augmentation methods. Other types models compared with VGG16-based model. Single...
Purpose To prospectively compare the diagnostic capability of short inversion time inversion-recovery (STIR) turbo spin-echo (SE) imaging, diffusion-weighted (DW) magnetic resonance (MR) and fluorodeoxyglucose (FDG) combined positron emission tomography (PET) computed (CT) in N stage assessment patients with non–small cell lung cancer (NSCLC). Materials Methods This prospective study was approved by institutional review board, written informed consent obtained from all patients. A total 250...
To prospectively compare the capabilities of dynamic perfusion area-detector computed tomography (CT), magnetic resonance (MR) imaging, and positron emission (PET) combined with CT (PET/CT) use fluorine 18 fluorodeoxyglucose (FDG) for diagnosis solitary pulmonary nodules.The institutional review board approved this study, written informed consent was obtained from each subject. A total 198 consecutive patients 218 nodules underwent CT, MR FDG PET/CT, microbacterial and/or pathologic...
Natural language processing using models has yielded promising results in various fields. Language can help improve the workflow of radiologists. This retrospective study aimed to construct and evaluate for automatic summarization radiology reports. Two report datasets from MIMIC Chest X-ray (MIMIC-CXR) database Japan Medical Image Database (JMID) were included this study. The MIMIC-CXR is an open comprising chest radiograph JMID a large computed tomography magnetic resonance imaging reports...
Abstract We hypothesized that, in discrimination between benign and malignant parotid gland tumors, high diagnostic accuracy could be obtained with a small amount of imbalanced data when anomaly detection (AD) was combined deep leaning (DL) model the L 2 -constrained softmax loss. The purpose this study to evaluate whether proposed method more accurate than other commonly used DL or AD methods. Magnetic resonance (MR) images 245 tumors (22.5% malignant) were retrospectively collected....
Abstract This retrospective study aimed to develop and validate a deep learning model for the classification of coronavirus disease-2019 (COVID-19) pneumonia, non-COVID-19 healthy using chest X-ray (CXR) images. One private two public datasets CXR images were included. The dataset included from six hospitals. A total 14,258 11,253 in 2 455 dataset. based on EfficientNet with noisy student was constructed three datasets. test set 150 evaluated by radiologists. Three-category accuracy...
Abstract This study aimed to develop a versatile automatic segmentation model of bladder cancer (BC) on MRI using convolutional neural network and investigate the robustness radiomics features automatically extracted from apparent diffusion coefficient (ADC) maps. two-center retrospective used multi-vendor MR units included 170 patients with BC, whom 140 were assigned training datasets for modified U-net five-fold cross-validation 30 test assessment performance reproducibility features. For...
We aimed to develop and evaluate an automatic prediction system for grading histopathological images of prostate cancer. A total 10,616 whole slide (WSIs) tissue were used in this study. The WSIs from one institution (5160 WSIs) as the development set, while those other (5456 unseen test set. Label distribution learning (LDL) was address a difference label characteristics between sets. combination EfficientNet (a deep model) LDL utilized system. Quadratic weighted kappa (QWK) accuracy set...
The purpose of this study is to investigate the effect a novel reconstruction algorithm, adaptive iterative dose reduction using 3D processing, on emphysema quantification by low-dose CT.Twenty-six patients who had undergone standard-dose (150 mAs) and (25 CT scans were included in retrospective study. Emphysema was quantified several quantitative measures, including index given percentage lung region with low attenuation (lower than -950 HU), 15th percentile density, size distribution...
Combinations of data augmentation methods and deep learning architectures for automatic pancreas segmentation on CT images are proposed evaluated. Images from a public dataset were used to evaluate the models. Baseline U-net chosen models segmentation. Methods included conventional methods, mixup, random image cropping patching (RICAP). Ten combinations Four-fold cross validation was performed train these with methods. The dice similarity coefficient (DSC) calculated between results manually...
Abstract Endometrial cancer (EC) is the most common gynecological tumor in developed countries, and preoperative risk stratification essential for personalized medicine. There have been several radiomics studies noninvasive of EC using MRI. Although segmentation usually necessary these studies, manual not only labor-intensive but may also be subjective. Therefore, our study aimed to perform automatic on MRI with a convolutional neural network. The effect input image sequence batch size...
Purpose: The purpose of this study was to develop and evaluate lung cancer segmentation with a pretrained model transfer learning. constructed from an artificial dataset generated using generative adversarial network (GAN). Materials Methods: Three public datasets containing images nodules/lung cancers were used: LUNA16 dataset, Decathlon NSCLC radiogenomics. used generate for the help GAN 3D graph cut. Pretrained models then dataset. Subsequently, main Finally, radiogenomics model. Dice...
The objective of our study was to prospectively compare the capability dynamic area-detector CT analyzed with different mathematic methods and PET/CT in management pulmonary nodules.Fifty-two consecutive patients 96 nodules underwent CT, PET/CT, microbacterial or pathologic examinations. All were classified into following groups: malignant (n = 57), benign low biologic activity 15), high 24). On total, arterial, systemic arterial perfusions calculated using dual-input maximum slope method;...