動体追尾放射線治療のための複数周辺臓器の多次元特徴量に基づく膵癌変位推定法

Abstract

本研究はX線画像上で目視不可能な臓器における動体追尾放射線治療への応用のために, 複数臓器の多次元特徴量に基づく腫瘍変位推定法を提案する. 25 症例の 4D-CT から抽出した 10 時相の臓器形状データを対象に臓器形状テンプレートとの形状間位置合わせを行うことによって頂点単位の統計的変位モデルを得た. 胃, 肝臓, 左右腎臓, 十二指腸の多次元特徴量からLASSO回帰を用いて膵癌の重心変位を求めた結果, 胃, 十二指腸, 左腎の重心変位が推定に有効な特徴量として選定され, $2.75pm1.70$ mmの誤差で変位推定が可能であった.

This paper proposes a method to estimate displacements of pancreatic cancer based on multidimensional features of multiple surrounding organs for real-time radiation therapy. Multidimensional features which consist of displacements of vertices between maximal inspiratory state and maximal expiratory state are obtained by shape registration to target 4D-CT data. The result of the experiments showed that the proposed method could estimate the centroid of the displacements of pancreatic cancer with an error of $2.75 pm 1.70$ mm using LASSO regression from the features of stomach, liver, left and right kidney and duodenum. And the proposed method selected the center of displacements of liver, duodenum and left kidney as effective features.