[home] [Personal Program] [Help]

---

Session: Surgery & Intervention
Session starts: Friday 27 January, 14:00
Presentation starts: 14:15
Room: Room 531

---

Ana Cordon Cordon (University of Twente)
Yoeko Mak (University of Twente)
Momen Abayazid (University of Twente)

Abstract:
Liver cancer is one of the most common causes of cancer death [1]. Percutaneous liver biopsy and tumor ablation are two procedures widely implemented in the diagnosis and treatment of hepatic lesions. Respiratory motion limits the accuracy of abdominal/thoracic percutaneous procedures [2]. The tumor’s position changes with time, and current medical imaging modalities fail to provide simultaneous high-quality real-time visualization [3]. Motion tracking techniques offer the possibility to estimate the position of a tumor in real-time using surrogate signals [4]. This work aims to evaluate the implementation of an RGB-D camera as a surrogate signal to create a motion model that estimates the respiratory-induced motion in the liver. The proposed approach uses the RGB-D camera to create point clouds that capture the respiratory-induced changes in the abdominal area and use it as an input for the motion model. A correspondence model is created implementing supervised learning approaches. The RGB-D camera was evaluated in two scenarios, a robotic liver phantom and a human subject experiment. An electromagnetic sensor and ultrasound images served as ground truth for the phantom and human subject experiments, respectively. A robotic manipulator was included in the phantom set-up to perform the needle steering and insertion tasks. The performance of the correspondence models and the robotic manipulator were evaluated using the ground truth. The created motion model for the robotic liver phantom showed a mean absolute error of 0.59 mm and 2.66 mm. The coefficients of determination for the anterior-posterior and superior-inferior directions were 0.96 and 0.70, respectively. The implementation of the robotic manipulator displayed a needle insertion error of 3 mm. The human subject experiment results displayed a coefficient of determination above 0.75 in all sessions. To conclude, abdominal point cloud reconstructions can serve as a surrogate to estimate in real-time the respiratory-induced motion in the liver. This work incorporates a real-time tumor tracking algorithm implemented by a robotic manipulator which performs the needle steering task. Further work should account for the tool-tissue interaction that takes place while performing the needle insertion, as well as other sources of motion that might affect the position of the tumor.