森 紀美江¹, 向井 信彦², 近藤 貴大², 武井 良子¹, 山下 夕香里¹, 長谷川 和子^{1, 3}, 高橋 浩二¹

Kimie MORI¹, Nobuhiko MUKAI², Takahiro KONDO², Yoshiko TAKEI¹, Yukari YAMASHITA¹, Kazuko HASEGAWA^{1, 3}, Koji TAKAHASHI¹

¹昭和大学歯学部スペシャルニーズ口腔医学講座口腔リハビリテーション医学部門, ²東京都市大学大学院工学研究科, ³誠愛リハビリテーション病院

¹Division of Oral Rehabilitation Medicine, Department of Special Needs Dentistry, Showa University School of Dentistry, ²Graduate School of Engineering, Tokyo City University, ³Seiai Rehabilitation Hospital

キーワード : ultrasonic diagnosis, tongue movement, Japanese speech, three-dimensional tongue-shaped standard model

目的：既存の超音波診断装置を用いて発音時の舌運動の明瞭な3次元動画像を得ることは困難である．そこで，舌超音波画像のボリュームデータを基にパーソナルコンピュータ上で3次元動画像を作成する，3次元舌形状標準モデル（以下，3次元モデル）を考案し，日本語発音時の舌運動を観察することを目的に研究を開始した．本報では3次元モデルの構築方法について報告する．対象と方法：対象は34歳の女性で，安静位と日本語5母音の発音で得られる舌超音波画像を基にした．超音波診断装置上で舌尖を基準に4 mm幅の断層画像を取得し，舌表面の抽出のために変曲点を選択して制御点とした．次に，制御点を基にスプライン曲線を生成し，スプライン曲線を舌の断層画像に対応して並べることで舌の表面（スプライン曲面）を生成し，舌の3次元モデルを構築した．結果と考察：本手法により舌表面の3次元モデルの構築が可能になった．なお，制御点の選択を自動化することで3次元モデル構築時間の短縮化が図れ，多くの3次元画像を短時間で作成することができる．また，構築された3次元モデルを時系列に補間することで舌の3次元動画像が構築でき，詳細な舌運動の観察が可能になる．結論：本手法は，構音障害症例の発音時における舌運動様式の解明と臨床現場での病態説明に役立つと考えられる．

Purpose: It is difficult to obtain clear three-dimensional (3D) moving images with existing ultrasound systems. In order to investigate tongue movements during Japanese speech, we have developed a 3D tongue-shaped standard model (3D model) on a personal computer based on volume data of ultrasound tongue images. In this paper, the method for 3D model construction is discussed. Subjects and Methods: The subject was a 34-year-old female. Ultrasound tongue images while at rest and while producing five Japanese vowels were taken, and the tomographic data of the images were obtained with 4-mm slice intervals. To acquire the tongue surface line, inflection points of the tongue surface were selected as control points. A spline curve was derived based on the control points, and a tongue surface (spline surface) was generated from a set of spline curves that corresponded to the ultrasound slice images. Thus, the 3D model was constructed. Results and Discussion: By using our method, 3D models of the tongue surface were constructed. It will be possible to produce the 3D images in a shorter time by automating the selection of control points. In addition, it may be possible to observe tongue movement in detail by constructing 3D moving images based on the interpolation of a set of sequential 3D model data. Conclusion: By using our method, it may be possible to analyze disordered tongue movements during speech and explain the tongue status to patients in a clinical setting.