Treatment Planning for Transskull Ultrasound Surgery and Therapy G.T. Clement, C. Connor, and K. Hynynen Department of Radiology, Harvard Medical School, Brigham and Women’s Hospital, 75 Francis St. Boston, MA 02115 Noninvasive treatment of brain disorders using focused ultrasound requires an accurate method for correcting ultrasound distortion. Previous studies indicate that control of ultrasound phase alone is sufficient for producing a focus in a known position through the skull. The present study concentrates on identifying practical methods that could be applied in a clinical setting. Ten ex vivo human calvaria are examined. Each sample is imaged in water using CT. The information is used to determine the inner and outer skull surfaces, the normal vectors along the surfaces, thickness as a function of position, and internal structure. Phase measurement over a series of points is obtained by placing a skull between a transducer and a receiver with the skull normal to the transducer. The data follows that predicted by a homogeneous skull model using a sound speed of 2650 m/s. However, large variance (S.D. = 1.05 rad) indicates the additional role of internal bone speed and density fluctuations. An algorithm is presented that corrects for phase aberrations and successfully produces a focus at a predetermined location through each of the skulls studied (N=10). Implementation of the method in a clinical setting discussed. INTRODUCTION Introduction A series of studies have demonstrated that focusing of ultrasound through the human skull in predetermined locations is possible. 1-3 Further, it has been shown that temperatures high enough to coagulate tissues can be achieved using this approach without producing excessive heating near the skull. 4 The final step necessary for practical implementation of a transskull procedure is a minimally invasive or, ideally, noninvasive treatment plan. Other work has suggested a minimally invasive procedure using a catheter- inserted hydrophone is possible. The present study concentrates on recent work toward achieving a completely noninvasive method. The central problem in developing a noninvasive approach lies in predicting the behavior of the ultrasound field after passing through the skull bone, which causes significant reflection, diffraction and absorption of the field. Successful prediction depends on developing correct and practical theory and methods. Practical considerations include obtaining accurate knowledge of the thickness and internal structure of the skull bone, precise registration between all points of the skull and the ultrasound array, and a computationally feasible model. Theoretical considerations involve finding a model as uncomplicated as possible without oversimplification of the problem. Our approach is guided by initial experiments, which demonstrate that driving transducer elements in an array so that the acoustic pressure produced by each element arrives at the intended focal location in phase is sufficient for produces a sharp focus through the skull. Expectedly, it was found that skull thickness and density variations are major variables in determining this phase. We have developed an algorithm to focus through the skull using thickness, density, and orientation obtained from CT images. The algorithm operates by projecting the field in the wavevector-frequency domain. Below, we outline the model and experimental procedure applied to ex vivo human skulls and discuss how this method could be applied in practice. THEORY AND METHODS The phase prediction algorithm for focusing through the skull operates in the wavevector frequency domain. The field from a small section of the transducer is propagated to the outer skull surface. At the outer surface the transmission coefficient is calculated as a function of incident angle before the transmitted (and refracted) field is propagated within the skull from the outer skull surface to the inner skull surface, and lastly from the inner skull surface into the brain tissue. The density within the skull is used to determine an effective skull sound speed for the relevant area. The present algorithm approximates a linear relation between density and effective sound speed. Data for the simulation is obtained from a digitized human head profile obtained using CT images. Both the coordinates of the skull surfaces as well as the internal density variation are obtained from these images. The calculation is performed only in bone lying within the beamwidth of the section being considered. Ten ex-vivo human calvaria are used in the study. Information about the shape and structure of an Proceedings of the 17th International Congress on Acoustics 2001 Rome, September 2-7, 2001