| 電子情報通信学会 研究会発表申込システム 講演論文 詳細 |
技報閲覧サービス [ログイン] 技報アーカイブ |
| トップに戻る | 前のページに戻る | [Japanese] / [English] |
| 講演抄録／キーワード | ||
| 講演名 | 2026-03-06 13:38 Depthmap-based 2D-3D Reconstruction of the Wrist Bones from a Single-View Radiograph for the Diagnosis of Deformity and Treatment Planning ○Yanis Tahrat・Yi Gu（NAIST）・Ryoya Shiode・Kunihiro Oka（Osaka Univ.）・Soufi Mazen・Yoshinobu Sato・Yoshito Otake（NAIST）  MI2025-104 |
|
| 抄録 | （和） | Single-view radiography is the standard for hand deformity assessment but lacks the three-dimensional spatial information essential for precise surgical planning. While depthmap-based 2D–3D reconstruction offers a potential solution, its application to the wrist region is challenged by complex bone geometry and significant anatomical overlap. This study proposes a framework for 2D–3D reconstruction of the radius, ulna, and carpal bones from a single radiograph. The core of the method is a dedicated CT-to-X-ray registration pipeline designed to extract quantitative depth information. The framework integrates 2D deep learning for bone segmentation and landmark detection to constrain the search space. Registration is formulated as a gradient-driven optimization problem, employing a gradient correlation similarity metric and the Covariance Matrix Adaptation Evolution Strategy (CMA-ES) to achieve robust pose estimation. The proposed framework was validated on a large dataset consisting of 510 CT volumes and 476 radiographs. Registration quality was assessed through visual evaluation of DRR–radiograph overlays, demonstrating stable and reproducible alignment across cases and Target Registration Error value. This optimized registration enabled the generation of high-fidelity, millimeter-scale quantitative depth maps, which can be used as the basis for training tasks, including 2D–3D bone reconstruction models. By integrating single-view radiography with quantitative depth estimation, this work bridges the gap toward clinically actionable 3D data and enhances conventional radiographic assessment with a geometry-aware tool for orthopedic deformity diagnosis and treatment planning. |
| （英） | Single-view radiography is the standard for hand deformity assessment but lacks the three-dimensional spatial information essential for precise surgical planning. While depthmap-based 2D–3D reconstruction offers a potential solution, its application to the wrist region is challenged by complex bone geometry and significant anatomical overlap. This study proposes a framework for 2D–3D reconstruction of the radius, ulna, and carpal bones from a single radiograph. The core of the method is a dedicated CT-to-X-ray registration pipeline designed to extract quantitative depth information. The framework integrates 2D deep learning for bone segmentation and landmark detection to constrain the search space. Registration is formulated as a gradient-driven optimization problem, employing a gradient correlation similarity metric and the Covariance Matrix Adaptation Evolution Strategy (CMA-ES) to achieve robust pose estimation. The proposed framework was validated on a large dataset consisting of 510 CT volumes and 476 radiographs. Registration quality was assessed through visual evaluation of DRR–radiograph overlays, demonstrating stable and reproducible alignment across cases and Target Registration Error value. This optimized registration enabled the generation of high-fidelity, millimeter-scale quantitative depth maps, which can be used as the basis for training tasks, including 2D–3D bone reconstruction models. By integrating single-view radiography with quantitative depth estimation, this work bridges the gap toward clinically actionable 3D data and enhances conventional radiographic assessment with a geometry-aware tool for orthopedic deformity diagnosis and treatment planning. |
|
| キーワード | （和） | CT-to-X-ray registration / single-view radiography / depth map / wrist / radius and ulna / CMA-ES / U-Net / |
| （英） | CT-to-X-ray registration / single-view radiography / depth map / wrist / radius and ulna / CMA-ES / U-Net / | |
| 文献情報 | 信学技報, vol. 125, no. 395, MI2025-104, pp. 167-170, 2026年3月. | |
| 資料番号 | MI2025-104 | |
| 発行日 | 2026-02-26 (MI) | |
| ISSN | Online edition: ISSN 2432-6380 | |
| 著作権に ついて |
技術研究報告に掲載された論文の著作権は電子情報通信学会に帰属します．(許諾番号：10GA0019/12GB0052/13GB0056/17GB0034/18GB0034) | |
| PDFダウンロード | MI2025-104 |
|
| 研究会情報 | |
| 研究会 | MI |
| 開催期間 | 2026-03-05 - 2026-03-06 |
| 開催地（和） | 沖縄県青年会館 |
| 開催地（英） | OKINAWAKEN SEINENKAIKAN |
| テーマ（和） | 医用画像処理一般・メディカルイメージング連合フォーラム |
| テーマ（英） | Medical Imaging, etc. |
| 講演論文情報の詳細 | |
| 申込み研究会 | MI |
| 会議コード | 2026-03-MI |
| 本文の言語 | 英語 |
| タイトル（和） | |
| サブタイトル（和） | |
| タイトル（英） | Depthmap-based 2D-3D Reconstruction of the Wrist Bones from a Single-View Radiograph for the Diagnosis of Deformity and Treatment Planning |
| サブタイトル（英） | |
| キーワード(1)（和/英） | CT-to-X-ray registration / CT-to-X-ray registration |
| キーワード(2)（和/英） | single-view radiography / single-view radiography |
| キーワード(3)（和/英） | depth map / depth map |
| キーワード(4)（和/英） | wrist / wrist |
| キーワード(5)（和/英） | radius and ulna / radius and ulna |
| キーワード(6)（和/英） | CMA-ES / CMA-ES |
| キーワード(7)（和/英） | U-Net / U-Net |
| キーワード(8)（和/英） | / |
| 第1著者 氏名（和/英/ヨミ） | Yanis Tahrat / Yanis Tahrat / |
| 第1著者 所属（和/英） | Nara Institute of Science and Technology (略称： NAIST) Nara Institute of Science and Technology (略称： NAIST) |
| 第2著者 氏名（和/英/ヨミ） | Yi Gu / Yi Gu / |
| 第2著者 所属（和/英） | Nara Institute of Science and Technology (略称： NAIST) Nara Institute of Science and Technology (略称： NAIST) |
| 第3著者 氏名（和/英/ヨミ） | Ryoya Shiode / Ryoya Shiode / |
| 第3著者 所属（和/英） | Osaka University (略称： Osaka Univ.) Osaka University (略称： Osaka Univ.) |
| 第4著者 氏名（和/英/ヨミ） | Kunihiro Oka / Kunihiro Oka / |
| 第4著者 所属（和/英） | Osaka University (略称： Osaka Univ.) Osaka University (略称： Osaka Univ.) |
| 第5著者 氏名（和/英/ヨミ） | Soufi Mazen / Soufi Mazen / |
| 第5著者 所属（和/英） | Nara Institute of Science and Technology (略称： NAIST) Nara Institute of Science and Technology (略称： NAIST) |
| 第6著者 氏名（和/英/ヨミ） | Yoshinobu Sato / Yoshinobu Sato / |
| 第6著者 所属（和/英） | Nara Institute of Science and Technology (略称： NAIST) Nara Institute of Science and Technology (略称： NAIST) |
| 第7著者 氏名（和/英/ヨミ） | Yoshito Otake / Yoshito Otake / |
| 第7著者 所属（和/英） | Nara Institute of Science and Technology (略称： NAIST) Nara Institute of Science and Technology (略称： NAIST) |
| 講演者 | 第1著者 |
| 発表日時 | 2026-03-06 13:38:00 |
| 発表時間 | 13分 |
| 申込先研究会 | MI |
| 資料番号 | MI2025-104 |
| 巻番号（vol） | vol.125 |
| 号番号（no） | no.395 |
| ページ範囲 | pp.167-170 |
| ページ数 | 4 |
| 発行日 | 2026-02-26 (MI) |
[研究会発表申込システムのトップページに戻る]