This overall goal of this SBIR project is to develop a fully automated bone removal method for Dual Energy Computed Tomography (DECT) angiography scans. Dual energy scans offer the opportunity to better understand the material decomposition of anatomy, thus allowing for new methods to visualize and understand a wide range of diseases and conditions. In Phase I of this proposal we will develop and evaluate the main algorithmic components of our automated bone segmentation method, evaluate the potential impact on CTA workflow, and design a prototype user interface. We will also design, conduct, and analyze a preliminary evaluation of the automatically produced bone suppressed images with respect to manual segmentations. Algorithm development and evaluation will be performed using an existing database of dual energy clinical CT images, provided by GE Healthcare. In Phase II we will further improve the robustness of the method to include more diverse data from different dual-energy scanners and different anatomy, perform a larger clinical evaluation, and develop a commercial product. The ultimate goal of this work is to develop and sell this technology as an automated bone segmentation and removal product. This proposal is a partnership between Stanford University, which has extensive clinical expertise in developing computational aids for medical image interpretation, and Kitware, a small business with experience in medical visualization and software development. Currently, a fully robust and automated bone removal system does not exist, and the proposed novel solution has the potential to significantly improve current head and neck CTA interpretation making this a highly innovative and important project.
The specific aims of the research are to: 1. Develop the key components of a fully automated dual-energy CTA bone segmentation and removal method consisting of: a. An algorithm component to perform the initial decomposition of anatomy (bone, vessels, air, soft tissue) based on dual-energy data. b. An algorithm component to recover vascular regions erroneously classified as bone by algorithm component (a). c. A final algorithm component to remove any non-vascular regions erroneously classified as vessels by the algorithm component (a) above, including the removal of partial volume bone fragments and high intensity fragments introduced by noise. 2. Develop and evaluate a prototype application incorporating these three algorithm components. The application will display the result of automated bone removal with a traditional 2D slice display and 3D MIP/volume renderings. 3. Perform a pilot study evaluating the accuracy of the automated bone removal relative to state of the art manual techniques while documenting the improvement in the workflow.
The goal of this project is to develop a fully automated bone removal method for Dual Energy Computed Tomography (DECT) angiography scans. The proposed DECT and algorithmic solution has the potential to significantly improve current head and neck CTA interpretation.
