Real-time movement compensation for synchronous robotic (HIFU) surgery
Abhilash Rakkunedeth Hareendranathan
Date of Issue2011
School of Mechanical and Aerospace Engineering
Open surgical techniques are being replaced by minimally invasive or noninvasive techniques in areas such as oncology, mainly due to reduction in tissue trauma and recovery time. Radiation therapy (RT), which is the most commonly used technique for noninvasive treatment, is known to cause tissue ionization which manifests into side effects such as edema, damage to epithelium and fatigue. It can also be a potential cause for cancer in rare cases. High intensity focused ultrasound (HIFU) is devoid of ionizing effects making it more suitable for organs such as the kidney which have low radiation tolerance. Although the therapeutic potential of HIFU has been proved, its clinical usage is still limited due to inaccuracies in the target tracking and control. HIFU surgery shares a few issues common to noninvasive surgery. Most significant among these is the movement of the target organ due to physiological processes such as respiration. Prevalent HIFU systems either ignore this movement or rely upon breath suppression. In this research, a movement model that can be incorporated into therapeutic system is proposed, so as to achieve real-time movement compensation. Aspects such as the image registration algorithm, movement control strategy and HIFU applicator design which are part of the therapeutic system were also addressed. The target application for this study was the treatment of kidney tumors such as the Renal Cell Carcinoma (RCC). Initial work deals with measurement and statistical analysis of the kidney from a set of healthy volunteers. The movement patterns observed were complex and subject-specific. Hence generic movement models cannot be used. Therefore, a prediction-correlation based model was proposed and a new empirical model was developed for kidney movement; this was used as the basis for non-linear predictors such as Unscented Kalman Filter (UKF), Extended Kalman Filter (EKF) and Adaptive Neuro Fuzzy Inference System (ANIFS). The movement model also comprised of a correlation network which maps the position of skin markers to current position of the kidney. The correlation module comprised of a new mapping function tuned using ANFIS. The prediction-correlation based approach has two advantages – firstly, it intuitively accounts for the subject specific nature of the movement and secondly, it reduces the length of prediction thereby improving the accuracy. This approach gives an accuracy of more than 92% for movement prediction. The maximum absolute error observed was 2.4 mm.