81690 research outputs found
Sort by
対話主義的授業論の探求 [全文の要約]
No full textこの博士論文全文の閲覧方法については、以下のサイトをご参照ください。https://www.lib.hokudai.ac.jp/dissertations/copy-guides
現代日本における公教育と学習塾の関係変容に関する検討 : 公設型学習塾の両義性を踏まえた補完的な教育機会保障の探究 [全文の要約]
No full textこの博士論文全文の閲覧方法については、以下のサイトをご参照ください。https://www.lib.hokudai.ac.jp/dissertations/copy-guides
ケンフェロールのミトコンドリア機能向上作用の解明 [全文の要約]
No full textこの博士論文全文の閲覧方法については、以下のサイトをご参照ください。https://www.lib.hokudai.ac.jp/dissertations/copy-guides
大雪山国立公園における周氷河地形のマッピング [全文の要約]
No full textこの博士論文全文の閲覧方法については、以下のサイトをご参照ください。https://www.lib.hokudai.ac.jp/dissertations/copy-guides
深層学習手法を用いた冠動脈解析のための画像前処理から3D モデリングに関する研究
No full textCoronary artery disease (CAD) remains a leading cause of mortality worldwide, necessitating advancements in diagnostic and treatment methods. This thesis explores novel deep learning approaches to improve the analysis of coronary artery imaging, addressing challenges in stenosis detection, and 3D reconstruction. Three key contributions are presented: (1) a Hessian-based image preprocessing method integrated with image fusion, enhancing visualization of coronary angiography; (2) an attention-enhanced YOLO-based framework for stenosis localization, improving detection accuracy in complex angiographic scenarios; and (3) an automated deep learning pipeline leveraging GhostNet and transformers for the extraction and 3D reconstruction of the right coronary artery (RCA) from CT images. Comprehensive experiments validate the efficacy of the proposed methods across multiple tasks in coronary artery analysis. The Hessian-based image preprocessing combined with image fusion significantly enhanced vessel visualization, leading to a 5% increase in detection accuracy (AP50 = 87.1%) using the YOLOv10-X model compared to raw dataset. The proposed attention-enhanced YOLO-based network achieved outstanding performance in stenosis localization, with an AP50 of 90.5% at the zero-degree left coronary artery (LCA) imaging angle, the highest accuracy across all tested angles. Additionally, the automated deep learning pipeline for the extraction and 3D reconstruction of the right coronary artery achieved a segmentation F1 score of 0.866 and a mean IoU of 0.835. The 3D reconstruction process provided superior visual clarity, making it suitable for clinical applications. These contributions collectively advance the accuracy, efficiency, and practicality of coronary artery disease analysis, offering transformative potential for real-time medical diagnostics and treatment planning. This study highlights the transformative potential of advanced deep learning architectures in CAD diagnosis, offering pathways for real-time, precise, and robust medical imaging solutions. Future research will focus on extending these techniques to broader datasets and refining latency for real-time clinical applications
妊婦看護職における職業性ストレスおよび自己管理行動とその関連要因 [全文の要約]
No full textこの博士論文全文の閲覧方法については、以下のサイトをご参照ください。https://www.lib.hokudai.ac.jp/dissertations/copy-guides
