Our goal is to explore the frontiers for developing reconstruction strategies which overcome to the fullest extent possible the factors limiting detection and localization of lesions in emission imaging. The principle guiding our research is the use of task-based assessment of image quality. During the past funding period we have observed unexpected and intriguing results, the exploration of which we propose to further investigate.
Our specific aims are to: 1) Probe what dictates when the use of attenuation correction (AC) in reconstruction will and will not improve the accuracy of lesion detection and localization in lung, soft tissue, and bone for chest and abdomen imaging; 2) Explore an innovative approach for formulating reconstruction strategies based on the enhancement of detection-task performance as opposed to exact restoration of the degradation; 3) Examine the hypothesis that priors based on anatomical information from dual-modality imaging will improve the accuracy of lesion detection in maximum a posteriori (MAP) reconstruction; 4) Investigate the hypothesis that respiratory motion correction methods can be developed which will promote AC, improve lesion detection, and facilitate the use of anatomical priors with neuroendocrine tumor imaging; 5) Explore extensions of multiclass (scanning) channelized human-model observers that will enable agreement with human observers for our proposed study conduction methodologies. The clinically significant procedures of single photon emission computed tomographic (SPECT) imaging for neuroendocrine tumors and for bone imaging of back-pain have been selected as the test-beds for our investigations. We will base our investigations on the hypothesis that the formulation of a reconstruction strategy is best performed when the slices and detection task closely approximate their clinical application. To closely emulate the clinic we emphasize free-response observer studies which employ representative populations of hybrid images and volumetric display. Hybrid images are actual clinical acquisitions to which Monte Carlo simulated lesions are added. The formulation and comparison of strategies for the quantification task will serve as a counter-point to our studies of detection. The result of these investigations will be the definition of ways to turn acquired pictures of molecular imaging tracers within a patient into clearer pictures of the three-dimensional location of these tracers from which physicians can make an accurate diagnosis and evaluate therapy options. ? ? ?
Showing the most recent 10 out of 14 publications
