In the last decades medical imaging has evolved from purely diagnostic purposes to other areas, like interventional imaging. While diagnostic imaging is often threedimensional (3D) with a high resolution, interventional imaging is more constrained in space and time of acquisition. Therefore, interventional imaging is frequently twodimensional (2D) and images are of lesser quality. By spatialy registering the 2D images with the 3D images, the interpretation of the imaged scene (anatomy) is enhanced, allowing to easier follow the interventional plan that was derived by using the pre-interventional 3D images. For this reason image registration is the key enabling factor for image-guided minimally invasive endovascular interventions. In this work all steps of the problem has been thoroughly addressed , from the development of novel methods, validation of methods and translation to clinical practice. All key elements of the augmented image-guided endovascular intervention are described: (a) reference image datasets with gold standard registrations, (b) development of a novel 3D-2D registration method, (c) objective and comparative validation of state-of-the-art methods, (d) registration initialization, and (e) automated detection of misregistrations.
F.35 Other
COBISS.SI-ID: 10644564It is known from literature that the sensitivity of visual assessment of smaller intracranial aneurysms is not satisfactory. Computer–aided diagnosis may shorten visual inspection and increase detection sensitivity by directing a diagnostician to suspicious locations in cerebral vasculature. For detection of blob-like structures, like aneurysms, multiscale enhancement filters, based on eigenvalues of the Hessian matrix, have already been proposed. Their main drawbacks are that they produce nonuniform responses within the structures they are supposed to enhance and that they are rather slow. In this paper, we propose a blob enhancement filter, which can be directly calculated from the Hessian matrix without eigenvalue decomposition. Using two metric functions, we compared the proposed filter with other blob enhancement filters on a synthetic image and on fifteen 3D digitally subtracted angiograms. The response of the proposed filter is more uniform and is computed six times faster than the other filters. Filtration of blood vessels, aneurysms, and consequent segmentation have been implemented into a computer-aided system that delivers 3D realistic models of the intracranial vasculature and can be displayed in real 3D.
F.22 Improvement to existing health/diagnostic methods/procedures
The project PI in Associate Editor of IEEE Transactions on Medical Imaging, which is according to the impact factor the 16th among 248 journals in the field of Electrical Engineering.
C.04 Editorial board of an international magazine