小橋研究室 | Kobashi Laboratory

Publications

For detailed publication lists, please visit the links below. Total: - peer-reviewed papers / Presentations: -

Research

Recent Research Topics (selected outcomes from the last ~10 years)

Based on publications mainly from 2016–2025, we summarize our recent outcomes into five research themes. “Representative papers” link directly to DOI pages.

1. Neuroimaging for stroke care: hematoma expansion prediction & brain MRI classification

We develop machine learning models for early risk prediction (e.g., hematoma expansion) and differential diagnosis using CT/MRI.

Key achievements:
• Clinically oriented prediction models for hematoma expansion
• Improved discrimination using attention mechanisms in brain MRI

Representative papers (DOI):
Clinically Applicable Machine Learning Approach to Predict Intracerebral Hematoma Expansion
Differentiating Idiopathic Normal Pressure Hydrocephalus Using Multi-Head Attention CNN with Diversify-Loss in Brain MRI

2. Orthopedics & sports medicine: ultrasound CAD for elbow osteochondritis dissecans (OCD)

We study deep learning-based CAD to support early detection of sports-related disorders using ultrasound images.

Key achievements:
• Deep learning detection pipelines suitable for clinical workflow
• Lesion localization + detection for practical CAD implementation

Representative papers (DOI):
Deep Learning-Based Computer-Aided Diagnosis of Osteochondritis Dissecans of the Humeral Capitellum Using Ultrasound Images
Deep learning-based osteochondritis dissecans detection in ultrasound images with humeral capitellum localization

3. Dental panoramic X-rays: detection/classification of teeth, prostheses, and third molar impaction

We develop automated detection and recognition methods for panoramic radiographs to support clinical tasks and improve efficiency.

Key achievements:
• Robust teeth/prosthesis detection combining deep learning and prior knowledge
• Clinically relevant classification tasks such as third molar impaction

Representative papers (DOI):
Optimization technique combined with deep learning method for teeth recognition in dental panoramic radiographs
Teeth and prostheses detection in dental panoramic X-rays using a deep learning-based object detector and a priori knowledge-based algorithm
Automated Detection and Classification of Mandibular Third Molar Impactions in Panoramic X-Rays

4. Trauma imaging: anatomy-aware fusion and attention for reliable fracture diagnosis

We explore anatomy-guided feature fusion and attention mechanisms to improve robustness and reliability in fracture diagnosis.

Key achievements:
• Anatomy-aware fusion for robust evidence aggregation
• Designs that support explainability (where the model “looks”)

Representative paper (DOI):
Invisible Yet Detected: PelFANet with Attention-Guided Anatomical Fusion for Pelvic Fracture Diagnosis

5. Outcome prediction & explainable AI: decision support with clinical data

We build outcome prediction models (e.g., ESWL treatment outcome) and incorporate explainability (e.g., SHAP) to support clinical decision-making.

Key achievements:
• Explainability-enabled outcome prediction using SHAP
• Feature engineering and evaluation for practical deployment

Representative paper (DOI):
Prediction of Ureter ESWL Outcome by Machine Learning and Model Interpretation Approach Using SHAP

Projects

*Auto-generated from KAKEN ( NRID)

Research Achievements

Contact Information