胸腔鏡画像における教師なし画像変換を用いた肺領域抽出

Abstract

内視鏡手術において安全性向上の観点から関心領域の三次元構造の把握が必要である．術中の変形が小さい臓器については事前に撮影したCT画像を三次元構造の把握のために用いることができるが，術中変形の大きい臓器については変形推定が必要である．術中臓器の変形推定に関する研究は広く行われているが，変形推定を行うためにはカメラ画像内の二次元臓器領域の抽出が必要である．本研究では，術中変形が大きい臓器である肺を対象に，U-netを用いて胸腔鏡画像に対する肺領域抽出を試みた．胸腔鏡画像のうち喫煙者の画像については抽出精度が低下する問題に対し，正則化項を追加した損失関数に基づいたCycleGANを用いて教師なし画像変換を行い，抽出精度の向上を試みたので報告する．In endoscopic surgery, it is necessary to understand the three-dimensional structure of the target region to improve safety. For organs that do not deform much during surgery, preoperative CT images can be used to understand their three-dimensional structure, however, deformation estimation is necessary for organs that deform substantially. Even though the intraoperative deformation estimation of organs has been widely studied, two-dimensional organ region segmentations from camera images are necessary to perform this estimation. In this paper, we performed lung region segmentation method using U-net for the thoracoscopic image, which is an organ that deforms substantially during surgery. To solve the problem of low segmentation accuracy of smoker thoracoscopic images, we performed unsupervised image translation using a CycleGAN, which we added a regularization term to the loss function. This presentation reports the image translation results and its effect of lung region segmentation.