This R21 research aims to contribute to enhanced spatiotemporal image reconstruction and analysis to achieve safer neurovascular procedures. Current intra-procedural cerebral angiographic images are two-dimensional projections, which require demanding real-time interpretation by the operator. In a time-sensitive environment, the operator must perform the mental mapping between 2D X-Ray angiographic projection time lapse images and the complex 3D structure of the cerebrovasculature in order, among other, to detect and locate proximal and distal thromboembolic events that may have arisen during an intervention. Failure to detect partially or completely obstructive clots during an intervention such as carotid stenting or aneurysm coiling, could result in blood flow interruption to downstream brain tissue leading to ischemic stroke. Currently available intra-procedural tomographic 3D vascular maps do not contain local blood flow information. Desired dynamic 3D (3D+T) images are currently unavailable because: 1) blood flow across the vasculature is too rapid for available X-ray based tomography units, and 2) such technology, were it available, would subject the patient to considerable additional radiation exposure and contrast agent injection. Detecting intra-procedural thromboembolic events would enable their treatment to avoid a stroke. The likelihood of successful detection in turn would significantly benefit from the ability to 1) reconstruct 3D+T images to estimate and monitor blood flow in real time from multiple perspectives, 2) obtain quantitative information of segmental blood flow and regional perfusion, and 3) detect and locate procedure-induced spatio-temporal changes and blood flow anomalies (e.g. formed or embolized thrombus or plaque fragment, intimal dissection, or iatrogenic vasospasm), provided that this capability occurs in a timely fashion without additional distracting human intervention or radiation exposure to the patient. To address these issues, we developed (Aim 1) a variational energy formulation for simultaneous smoothing and segmentation that fuses in a unified fashion prior information and measurement data. This approach simultaneously evolves flow over the entire vasculature and avoids propagation and accumulation of errors downstream. Next, analytical methods will be developed for extracting information from blood flow patterns to enable the detection and localization of abnormalities situated in areas not directly imaged. Both the reconstruction and anomaly detection and localization algorithms will be tested using phantom data to demonstrate the ability to deal with ambiguity, followed by benchtop data from the flow model incorporating simulated dynamic branch occlusions. The algorithms will also be validated by retrospective analysis of 3D and dynamic 2D (2D+T) images collected from previous procedures with known adverse events such as thromboembolism. The results will form the basis for expansion of this approach to algorithm optimization and prospective clinical evaluation.
It is difficult to interpret from two-dimensional X-ray images what is inside our body. They do not reveal always so well the nature of the blood flow inside the body, nor the possible occurrence of blood clots. This research overcomes this limitation by providing radiologists and surgeons with three-dimensional images of blood flow in vessels. These images are easier to interpret, more revealing and will make it possible for doctors to detect, locate and take actions to correct abnormalities such as clots, as well as perform a safer and faster surgery thus saving lives and money.
Showing the most recent 10 out of 24 publications
