A5016052507- Medical Imaging Techniques and Applications
- Radiomics and Machine Learning in Medical Imaging
- AI in cancer detection
- COVID-19 diagnosis using AI
- Medical Image Segmentation Techniques
- Generative Adversarial Networks and Image Synthesis
- Brain Tumor Detection and Classification
- Advanced X-ray and CT Imaging
- Lung Cancer Diagnosis and Treatment
- Radiation Dose and Imaging
- Privacy-Preserving Technologies in Data
- Advanced Image Processing Techniques
- Hepatocellular Carcinoma Treatment and Prognosis
- Anomaly Detection Techniques and Applications
- Colorectal Cancer Screening and Detection
- MRI in cancer diagnosis
- X-ray Spectroscopy and Fluorescence Analysis
- Advanced MRI Techniques and Applications
- Nuclear Physics and Applications
- Medical Coding and Health Information
- Biomedical and Engineering Education
- Body Composition Measurement Techniques
- High-pressure geophysics and materials
- Medical Imaging and Analysis
- Electronic Health Records Systems
University of Tokyo Hospital
2017-2025
Waseda University
1997-2020
Kumamoto Health Science University
2013-2015
Kumamoto University
2013-2015
Kumamoto University Hospital
2014
Abstract The purposes of this study are to propose an unsupervised anomaly detection method based on a deep neural network (DNN) model, which requires only normal images for training, and evaluate its performance with large chest radiograph dataset. We used the auto-encoding generative adversarial (α-GAN) framework, is combination GAN variational autoencoder, as DNN model. A total 29,684 frontal radiographs from Radiological Society North America Pneumonia Detection Challenge dataset were...
Purpose: The target disorders of emergency head CT are wide-ranging. Therefore, people working in an department desire a computer-aided detection system for general disorders. In this study, we proposed unsupervised anomaly method using autoencoder and evaluated the performance our CT. Methods: We used 3D convolutional (3D-CAE), which contains 11 layers convolution block 6 deconvolution block. training phase, trained 3D-CAE 10,000 patches extracted from 50 normal cases. test calculated...
PurposeThe prevalence of type 2 diabetes mellitus (T2DM) has been steadily increasing over the years. We aim to predict occurrence T2DM using mammography images within 5 years two different methods and compare their performance.ApproachWe examined 312 samples, including 110 positive cases (developed after years) 202 negative (did not develop T2DM) methods. In first method, a radiomics-based approach, we utilized radiomics features machine learning (ML) algorithms. The entire breast region...
To generate synthetic medical data incorporating image-tabular hybrid by merging an image encoding/decoding model with a table-compatible generative and assess their utility. We used 1342 cases from the Stony Brook University Covid-19-positive cases, comprising chest X-ray radiographs (CXRs) tabular clinical as private dataset (pDS). generated (sDS) through following steps: (I) dimensionally reducing CXRs in pDS using pretrained encoder of auto-encoding adversarial networks (αGAN)...
The incidence rate for Type 2 Diabetes Mellitus (T2DM) has been increasing over the years. T2DM is a common lifestyle-related disease and predicting its occurrence before five years could help patients to alter their lifestyle ahead hence prevent T2DM. We intend investigate feasibility of radiomics features in using screening mammography images which benefit us terms preventability disease. This study examined prevalence 110 positive samples (developed after 5 years) 202 negative (did not...
Diagnostic radiologists need artificial intelligence (AI) for medical imaging, but access to images required training in AI has become increasingly restrictive. To release and use images, we an algorithm that can simultaneously protect privacy preserve pathologies images. address this, introduce DP-GLOW, a hybrid combines the local differential (LDP) with GLOW, one of flow-based deep generative models. By applying GLOW model, disentangle pixelwise correlation which makes it difficult...
Ultra-sparse-view computed tomography (CT) algorithms can reduce radiation exposure for patients, but these lack an explicit cycle consistency loss minimization and log-likelihood maximization in testing. Here, we propose X2CT-FLOW the maximum a posteriori (MAP) reconstruction of three-dimensional (3D) chest CT image from single or few two-dimensional (2D) projection images using progressive flow-based deep generative model, especially ultra-low-dose protocols. The MAP simultaneously...
The fact that accurate detection of metastatic brain tumors is important for making decisions on the treatment course patients prompted us to develop a computer-aided diagnostic scheme detecting tumors. In this paper, we first describe how extracted cerebral parenchyma region using standard deviation filter. Second, initial candidates were decided by sphericity and cross-correlation value with simulated ring template. Third, made true positive false templates obtained from actual clinical...
Standardized uptake values (SUVs) derived from
Abstract Purpose Many large radiographic datasets of lung nodules are available, but the small and hard-to-detect rarely validated by computed tomography. Such difficult crucial for training nodule detection methods. This lack can be addressed artificial synthesis algorithms, which create artificially embedded nodules. study aimed to develop evaluate a novel cost function networks detect such lesions. Embedding lesions in healthy medical images is effective when positive cases insufficient...
Abstract In this study, we investigated the application of distributed learning, including federated learning and cyclical weight transfer, in development computer-aided detection (CADe) software for (1) cerebral aneurysm magnetic resonance (MR) angiography images (2) brain metastasis contrast-enhanced MR images. We used datasets collected from various institutions, scanner vendors, field strengths each target CADe software. compared performance multiple strategies, a centralized strategy,...
A general-purpose method of emphasizing abnormal lesions in chest radiographs, named EGGPALE (Extrapolative, Generative and General-Purpose Abnormal Lesion Emphasizer), is presented. The proposed composed a flow-based generative model L-infinity-distance-based extrapolation latent space. trained using only normal an invertible mapping function from the image space to determined. In space, given unseen extrapolated so that point moves away X-ray hyperplane. Finally, moved mapped back...
The generation of synthetic medical records using Generative Adversarial Networks (GANs) is becoming crucial for addressing privacy concerns and facilitating data sharing in the domain. In this paper, we introduce a novel method to create hybrid that combine both image non-image data, utilizing an auto-encoding GAN (alphaGAN) conditional tabular (CTGAN). Our methodology encompasses three primary steps: I) Dimensional reduction images private dataset (pDS) pretrained encoder {\alpha}GAN,...