ABSTRACT
Since 1989 we performed stereotactic radiotherapy treatments of cerebral arterovenous malformations (AVM), estimating three-dimensional (3-D) localization and shape of target volumes by the Leksell stereotactic helmet on two orthogonal radiographic projections. Due to the limitations of this method, we developed a new technique for the localization of the target volume using digital subtraction angiography (DSA) and digital image processing. To achieve this result we first developed a method to correct nonlinear distortion of DSA images using spatial relocation of image pixels based on a calibration grid. We then developed an algorithm for localization of the target volume using two independent DSA projections. Target volume coordinates in the helmet system are calculated using two DSA acquisitions taken with a free angle (approximately 90 degrees), one in the AP and the other in the LL direction. The helmet can be freely positioned between the x-ray source and the image plane. The projections of eight reference points inserted in the helmet at a known location, are used to calculate the transformation matrix between the two coordinate systems. We performed numerical and experimental validation of the system. A hypothetical random error (up to 2 mm) on image coordinates of the reference points allowed to determine that the error in target localization was less than 0.2 mm. Using DSA images of target points with a known location within a phantom, the error between calculated and actual location was, on average, 0.30+/-0.13 mm (mean+/-SD), with a maximum error of 0.49 mm. The results of numerical and experimental validations show that the system we have developed allows fast and accurate localization of the center of the target volume and it is suitable for efficient guiding during stereotactic radiosurgery of AVM.
