A5009032240- Advanced Image and Video Retrieval Techniques
- Video Analysis and Summarization
- Image Retrieval and Classification Techniques
- Multimodal Machine Learning Applications
- Music and Audio Processing
- Advanced Image Processing Techniques
- Infrastructure Maintenance and Monitoring
- Anomaly Detection Techniques and Applications
- Generative Adversarial Networks and Image Synthesis
- Domain Adaptation and Few-Shot Learning
- Image and Signal Denoising Methods
- Advanced Vision and Imaging
- Recommender Systems and Techniques
- Human Pose and Action Recognition
- Image Processing Techniques and Applications
- Visual Attention and Saliency Detection
- Radiomics and Machine Learning in Medical Imaging
- Gaze Tracking and Assistive Technology
- Industrial Vision Systems and Defect Detection
- AI in cancer detection
- Text and Document Classification Technologies
- Face and Expression Recognition
- Image Enhancement Techniques
- Neural Networks and Applications
- Speech and Audio Processing
Hokkaido University
2016-2025
Tokushima University
2005-2025
Tokushima University Hospital
2025
Hikone Central Hospital
2024
Kyoto Prefectural University of Medicine
2017-2024
Kyoto Prefectural University
2023-2024
Gifu University
2024
Japanese Red Cross Society Kyoto Daini Hospital
2023
Hokkaido University of Science
2008-2023
Kyoto first Red Cross hospital
2009-2022
PURPOSE: To evaluate the usefulness of fluid-attenuated inversion recovery (FLAIR) magnetic resonance (MR) imaging sequences in detection acute subarachnoid hemorrhage (SAH). MATERIALS AND METHODS: MR with FLAIR was performed a 0.5-T superconducting unit 20 patients (aged 30-72 years) SAH due to ruptured aneurysm and 27 control subjects 32-72 years). images were obtained 2 hours days after ictus. Findings evaluated compared computed tomographic (CT) findings. RESULTS: In all patients,...
To evaluate fluid-attenuated inversion-recovery (FLAIR) magnetic resonance (MR) imaging in the detection of subacute and chronic subarachnoid hemorrhage.The authors performed 19 FLAIR MR examinations at 0.5 T 14 adult patients with hemorrhage 3-45 days after ictus 22 control subjects. The on images was compared conventional spin-echo computed tomographic (CT) images.In hemorrhage, (100% detection) significantly superior to T1-weighted (36% detection, P < .01), T2-weighted (0% .02), CT (45%...
Abstract Background To improve the diagnostic accuracy of axillary lymph node (LN) metastasis in breast cancer patients using 2-[ 18 F]FDG-PET/CT, we constructed an artificial intelligence (AI)-assisted diagnosis system that uses deep-learning technologies. Materials and methods Two clinicians new AI retrospectively analyzed diagnosed 414 axillae 407 with biopsy-proven who had undergone F]FDG-PET/CT before a mastectomy or breast-conserving surgery sentinel biopsy and/or LN dissection. We...
Dataset distillation is an effective technique for reducing the cost and complexity of model training while maintaining performance by compressing large datasets into smaller, more efficient versions. In this paper, we present a novel generative dataset method that can improve accuracy aligning prediction logits. Our approach integrates self-knowledge to achieve precise distribution matching between synthetic original data, thereby capturing overall structure relationships within data. To...
We propose a novel continual self-supervised learning method (CSSL) considering medical domain knowledge in chest CT images. Our approach addresses the challenge of sequential by effectively capturing relationship between previously learned and new information at different stages. By incorporating an enhanced DER into CSSL maintaining both diversity representativeness within rehearsal buffer DER, risk data interference during pretraining is reduced, enabling model to learn more richer robust...
In sports training, personalized skill assessment and feedback are crucial for athletes to master complex movements improve performance. However, existing research on transfer predominantly focuses evaluation through video analysis, addressing only a single facet of the multifaceted process required acquisition. Furthermore, in limited studies that generate expert comments, learner's level is predetermined, spatial-temporal information human movement often overlooked. To address this issue,...
Composed Image Retrieval (CIR) provides an effective way to manage and access large-scale visual data. Construction of the CIR model utilizes triplets that consist a reference image, modification text describing desired changes, target image reflects these changes. For effectively training models, extensive manual annotation construct high-quality datasets, which can be time-consuming labor-intensive, is required. To deal with this problem, paper proposes novel triplet synthesis method by...
With the rapid development of recording and storage hardware, efficient methods to retrieve desired videos are required. Among video retrieval methods, cross-modal that aims at retrieving a target from natural language queries has attracted attention. Cross-modal is realized by learning common representation texts so their similarity can be calculated directly only based on contents. However, traditional approaches focus global features ignore fine-grained information such as single action...
Contrastive Language-Image Pre-training (CLIP) is vulnerable to adversarial attacks which cause misclassification by subtle modifications undetectable the human eye. Although training strengthens CLIP models against such attacks, it often degrades their accuracy on clean images. To tackle this challenge, we propose a novel defense strategy that leverages brain activity data. The proposed method combines features of with those examples, enhances robustness while maintaining high Experimental...
Neighbor embedding is widely employed to visualize high-dimensional data; however, it frequently overlooks the global structure, e.g., intercluster similarities, thereby impeding accurate visualization. To address this problem, paper presents Star-attracted Manifold Approximation and Projection (StarMAP), which incorporates advantage of principal component analysis (PCA) in neighbor embedding. Inspired by property PCA embedding, can be viewed as largest shadow data, StarMAP introduces...