Abstract
1636
Objectives: Today control of resection borders in tumor surgery is done using frozen sections, which only provides vague information about the location of possible remaining tumor tissue. Recently we introduced navigated nuclear probe imaging to provide an intra-operative surface reconstruction and visualization of the location of remaining tumor cells. The generated surfaces however are user-dependent and rely on statistically valid readings. To improve this, we compute and evaluate the confidence of the reconstruction in order to guide the scan for a complete and reliable radioactivity surface image.
Methods: Our existing tracked beta-probe surface reconstruction system was extended by computation and visualization of the confidence of the reconstruction. The confidence for each pixel of the reconstruction grid (5×5mm2 pixel size) is calculated using the simultaneously acquired probe readings and positions and an algorithm based on Pearson's goodness of fit test on the estimated activity within that pixel. The confidence is visualized as an image overlay onto the phantom. In an experiment we performed 10 scans with and without the visualization of the confidence with 3x3cm2 images of a tumor bed phantom with three 5×5mm2 residuals (tumor to background 50:10 kBq/ml Tc-99m).
Results: The reconstructed surfaces did not differ considerably between the modes (mean absolute distance between surfaces 3.6±0.5 cps). However the mean confidence was significantly higher for the guided scans (76±8% versus 66±10% for non-guided scans).
Conclusions: Visualizing the confidence during navigated nuclear imaging contributes to a less user-dependent procedure, through the guidance generating low confidence and thus potential incorrect surfaces is avoided.
- Society of Nuclear Medicine, Inc.