Catheter intervention for carotid and cerebrovascular diseases is rapidly developing treatment modality. A catheterization procedure involves the insertion of a flexible thin plastic catheter, into a blood vessel to provide an effective method of drawing blood or delivering medications and nutrients into a patient's blood stream. The intervention is directed to a method and apparatus for catheter navigation in 3D vascular tree exposures. The position of the catheter is detected and mixed into the 3D image of the pre-operatively exposed vascular tree reconstructed in a navigation computer.

In the catheter intervention, fluoroscopy may be used to confirm the position of the catheter and to manoeuvre it to the desired artery. Injection of contrast agent is used to visualize arteries. But fluoroscopy has potentially harmful ionization radiation. It also cannot provide depth information. Although the image shows the geometry of artery, it still cannot interpret position and pose in space. In addition, the injected contrast agent will disappear instantly, which causes repetition of injection. In whole procedure, patient and surgeon suffer from large dose radiation. The large dose contrast agent also tends to cause the patient allergy.

To solve the above problems, we considered using other safe methods to guide the catheter intervention. The intra-operative interventional imaging modalities include CT, fluoroscopy, MRI and US. CT provide high quality 3D image, however, it causes ionizing radiation. Non-real-time imaging also limits its application. Fluoroscopy and contrast agent are applied in clinic, their problems come from ionizing radiations, no depth information and allergy, although they can provide high quality image in real time. MRI is another modality providing 3D imaging. The main drawbacks are high cost of maintains and MR-compatible tools are need in operative region. By contrast, US is particularly well suited for intra-operative 3D anatomic visualization because of its safety, low cost and short acquisition time. An ultrasound exam is a safe diagnostic procedure that uses very-high-frequency sound waves to produce an image of many of the internal structures of the body. The sound waves prevent patients and surgeons from ionizing radiations. The short acquisition time makes US intra-operative imaging possible. Although US have drawbacks like poor image quality and narrow imaging region, the poor image quality can be improved by image processing algorithm. The narrow imaging region can be treated by reconstructing 2D images into 3D volume.

In order to realize US image-guided catheter intervention, a new image-guided catheter navigation system based on a three-dimensional freehand ultrasound and an electromagnetic tracking system is proposed. The method employs the freehand ultrasound imaging system to reconstruct two-dimension Doppler ultrasound images of the vessel. The reconstructed three dimension volume is augmented by segmentation of vessel cross section and registration of the catheter path obtained from a tracked catheter passing the vasculature.

In order to reconstruct two-dimension ultrasound image of vessel, a simple, automatic and robust calibration was developed for the freehand ultrasound imaging system without a phantom. A needle equipped with an electromagnetic tracking sensor is employed. The needle is moved in the vicinity of an ultrasound imaging plane taken by a fixed ultrasound probe also equipped with an electromagnetic EM tracking sensor. The ultrasound images, tracker's and needle tip's physical coordinates are recorded simultaneously. For each ultrasound image, the needle tip is recognized and the image coordinates are identified automatically. A point registration between the needle tip image and physical coordinates is performed to estimate the calibration matrix. A RANSAC algorithm is applied to minimize the registration error. A serial of experiments were performed to evaluate the performance of the proposed method. In all cases of our calibration process, the needle tip can be accurately identified in US images. There is 85% of the located needle tip drop a 1mm field close to the imaging plane. An accurate transformation matrix can be obtained using a tip data set of 20 frames. A calibration process with 20 data frames takes less than 10 minutes. This time is shorter than most previous method and a bit longer than some method using wall phantom. The FREs in different imaging depths are measured. The average FRE is 1.2mm, which is similar to other method using EM tracking system. The results of each imaging depth show that there is no significant difference between different imaging depths. Using the same data, the performance of RANSAC is compared with the conventional SVD. SVD is sensitive to these significant noisy points, but RANSAC can identify significant noisy point and produce an accurate result. The TRE of the US imaging system is evaluated by the tracked needle tip. The results show that the proposed method achieves similar accuracy to the previous manual method. A phantom with balls is used to evaluate the navigation accuracy of the proposed method. These results prove that, in contrast to previous methods, this is a simple, automatic and robust method which can achieve similar accuracy and efficient. The accuracy of the proposed method satisfies the need of the catheter navigation system.

For the augmentation of the vessel volume obtained from the free-hand system, a novel segmentation and reconstruction method based on catheter path was proposed. Two-dimension segmentation is performed on the vessel cross sections using a modified region growing algorithm combining grayscale distribution, shape feature and area regularity criterions. The segmented cross sections are interpolated into the new center line registered by the paths of catheters, which are typically placed in vessels. Based on the interpolated cross sections, a new vessel volume is reconstructed and rendered for navigation. The proposed method was evaluated by the data from a vessel phantom and in vivo. The segmentation results of in vitro trail show that the proposed 2D segmentation method can extract an accurate contour of the bifurcation than conventional region growing only depending on intensity distribution and shape features. During the catheterization, the bifurcation presents the node of the vessel network, which determines the position and orientation of the catheter tip moving. It is important to produce an accurate contour on the cross sections of bifurcation, because the segmented contour affects the reconstruction directly. With an accurate reconstructed bifurcation, surgeons can identify whether the catheter tip enters the target vessel or not. In the cross section of bifurcation, the conventional region growing extracted the contour of two jointed cross sections. Based on the incorrect contour, a larger ellipse was estimated, which would produce incorrect contour information and corrupt the newly reconstructed vessel volume. By contrast, the proposed segmentation method used the area threshold to constrain the growing process and produce a correct contour in each cross section. The diameter evaluation of the entire vessel proved the proposed segmentation could produce an accurate estimate of dimension. The segmentation results of in vivo trail show that the proposed segmentation based on intensity distribution of Doppler images could extract an accurate contour of the vessel from the artifacts caused by heart and tissue motion. The clear contours of vessel are helpful to realize high quality reconstruction. The segmented vessel contours are reconstructed based on the center line registered by catheter path. From the in vitro and in vivo results, cooperated with the segmentation of cross section, a high contrast vessel volume is obtained, which provides a good visualization of the scanned vessel. The catheter path is acquired by EM tracking system, which usually is affected by the close metallic objects. However, the catheter path presents the dimension and geometry of the vessel during catheterization. Based on the catheter path, the reconstructed vessel is closer to the vessel that the catheter passes. The tracked catheter tip and the reconstructed vessel augmented by catheter path provide more true spatial position relation, which helps the surgeon understand the position of the catheter tip in the vessel during catheter navigation.

The accuracy of the proposed catheter navigation system was verified by the catheter path obtained from the entire vessel. The distances between the registered center line and path line of the entire vessel are computed. The registration accuracy yields an average value of 1.7mm and the errors of the trunk and bifurcation where the catheter tip moves drop below 2mm. These results suggest the technique satisfies the accuracy of 3mm required by the US image-guided catheter intervention. US-based navigation does not require patient registration. Both the US images and catheter path can be obtained intra-operatively. Thus, this system is less susceptible to user- and procedure-dependent error sources associated with patient registration. The overall clinical accuracy of an US-based system is therefore expected to lie closer in value to the overall accuracy (laboratory test) than is the case with conventional pre-operative image-based systems. Acquisition, reconstruction, segmentation and augmentation of the US volume finish in 20 minutes. The time also can be shortened by introducing some accelerative algorithms. The utilization of the catheter navigation will not prolong the surgery time. These results prove that the developed US image-guided catheter navigation system is applicable to catheter intervention.

本論文は口腔がんの化学療法治療のひとつである超選択的動注法を支援するための画像誘導カテーテル治療支援に関する研究を取り扱っている。超選択的動注法は口腔がん組織の栄養動脈に対し,カテーテルを浅側頭動脈・後頭動脈から挿入し留置することで,抗がん剤の選択的投与を一定期間にわたり連続して行うことを可能とする治療法であり,高い臨床的有効性を示す。しかしカテーテル操作には熟練を要し,放射線造影下での長時間の操作が求められることから,術前・術中画像を用いたカテーテルナビゲーション手法の開発が期待されている。本論文は術中に3次元超音波画像計測可能な血管部位の形状計測結果と,カテーテルに装着した電磁気式位置トラッキングシステムにより得られるカテーテル先端軌跡により,術前に撮影した精細なX線CT画像から作成された血管モデルを変形することで,カテーテル治療中の血管位置を同定し,術中超音波では観察できない顎動脈・顔面動脈・舌動脈等の分岐部位置を推定し,カテーテル操作を誘導するカテーテルナビゲーションシステム実現に必要となる基盤技術開発を行っている。

第1章では口腔がんの超選択的動注法を用いた化学療法の概要と,3次元超音波再構成技術の概要を示し,第2章で本研究の目的を述べている。第3章では超選択的動注法に対するカテーテルナビゲゲーション実現のための課題と具体的なナビゲーション手法を提案している。第3章では血管の3次元再構築のために必要となる2次元超音波画像診断装置プローブの電磁気式位置トラッキングシステムによる3次元超音波画像計測システムの実装に必要となる,超音波断層像と3次元位置計測装置手法の相対関係を校正のための新しい手法を提案している。具体的には電磁気式位置トラッキングシステムのポインタのみを用い,特定の校正用ファントムを用いることなく超音波画像座標と,電磁気式位置トラッキング座標系の対応を示すデータ列を収集し,RANSACアルゴリズムを適用して,その変換行列を算出する手法を提案・実装した。従来報告されていた幾何学的形状が既知のファントムを用いる校正方法より有意に短い時間(約10分程度)で, 従来報告されていた制度と同程度のRMS誤差で1.2mmの校正ができることを実験により示した。またこの手法で構成したフリーハンド超音波画像計測システムを用い,精度2.1mmでカテーテルナビゲーションが可能であることを示している。第4章では第3章で開発したフリーハンド3次元超音波システムを用いで表在血管像を3次元再構築する方法として,血管の超音波ドップラ画像の特徴量と形状に関する拘束を考慮することで,ノイズやアーティファクトを多く含む血管超音波画像から,血管分岐部も含めて安定してセグメンテーションする手法,ならびに血管に挿入したカテーテルの先端軌跡から血管の中心線軌跡を推定する手法を提案している。ファントム実験ならびに動物実験により,安定して血管セグメンテーションが可能であり,血管がカテーテル挿入により変形したとしてもある程度その変形を補正して血管再構築が可能であることを示している。第6章では第3-5章で開発した技術を総合して,超選択的動注法支援のためのナビゲーションシステムを試作し,血管分岐部を持つ樹脂製血管ファントムモデルを用いて,超選択的動注法でのカテーテル操作を模擬した実験を行っている。具体的には術前CT画像計測に相当する設計データ,血管像のフリーハンド三次元再構築,カテーテル挿入操作とカテーテル先端軌跡の計測,フリーハンド三次元再構築血管3次元モデルとカテーテル先端軌跡の計測結果をもとに,術前CT画像に相当する設計データを変形し,カテーテル挿入時の血管分岐部位置の推定を行う実験を実施している。その結果カテーテル操作や,体位変換による周囲組織の変形を模擬した血管位置の変位があったとしても,血管分岐部を3mm程度の誤差で推定できる可能性を示唆するデータを得た。第7章では本研究全体に対する考察と,臨床応用可能なシステム開発のために残された課題を議論し,第8章で結論を述べている。

以上の研究成果は,臨床的効果が優れていることが示されている口腔がんの化学療法治療のひとつである超選択的動注法を,より簡便に実施可能とする画像誘導ナビゲーションの実現に向けて,その基礎となる複数の手法を提案している。ファントム実験により人体に対する影響が少ない術中超音波画像計測と,カテーテルに集積可能な電磁気式位置トラッキングシステムを用いて,術中放射線被曝量を抑制しつつ,患者の体位姿勢変化やカテーテル操作による血管変形が存在したとしても安定して目的の血管分岐部近傍までカテーテルを誘導できる可能性を示しており,臨床的に有用なシステム実現への基礎的な知見を与えている。

よって本論文は博士(工学)の学位請求論文として合格と認められる。