Synthetic human models for image-guided interventions
A significant number of medical interventions are currently driven by images. Examples include image-guided Radiation Therapy, in which the intervention can occur over very short (tracking, gating), relatively short (positioning), or longer (adaptation) time intervals, as well as image-guided interventional radiological as well as surgical procedures. These interventions share a common theme in that images are used to enhance a model of the patient that is used to drive or adjust strategic decisions. It is important to remember that imaging is not a true representation of the patient, and most imaging systems and derived actions cannot be considered instantaneous on the time scale of many precise interventions. As such, there is both instantaneous uncertainty in any action as well as a decay of the patient model derived from images that propagates with time from the imaging acquired. This project would seek to develop models of the patient that use imaging signals to enhance a prior existing model, and which would develop and contain uncertainty estimates related to the decision to be made. These uncertainty and best-guess estimates of the current state of the patient would propagate through time, and provide better guidance to the interventional process to determine the safety and likely effectiveness of the guidance-driven decision. In the event of uncertainty that could lead to error that exceeds likely benefit, such model-aided decisions could call for additional imaging before intervention, or could advise that the intervention in its current state is unwarranted. The current research plan involves gaining an understanding of specific interventional processes, the timing and sources of imaging signals that drive them, and the sources of uncertainty therein. As one or more target interventions are described, the hope would be to prototype a synthetic model of the local region/state of the patient that would be knowledgeable about the guidance decisions and inherent risks, and communicate such information in a concise fashion to the interventionalist.