Coronary Analysis and Reporting 

Automated 3D reconstruction and evaluation of coronary arteries for interpreting coronary CTA studies in a standardized manner.
This ongoing research project explores ways to use AI to improve structured reports of Coronary CT Angiograms (CTA). For RSNA 2019, we developed an interactive prototype to elicit expert feedback from clinicians at the world’s largest radiology conference. The prototype demonstrates fast and automated coronary segmentation and labeling, which facilitates the reconstruction of arterial structures and displays in a variety of image formats.
  • Fast and automated reconstruction of the 3D coronary vascular tree structure
  • Colored vessel segment labeling helps doctors describe the location of pathological findings accurately
  • Provides a spherical expanded view which gives doctors an overall impression of the scan
  • Multiple VR, MPR, and cMPR techniques provide different perspectives of the 3D vascular structure
  • A variety of analysis modules including plaque detection and classification, stenosis measurement, myocardial bridge/stent detection, dominance determination, origin/course anomalies warning  are integrated into this post-processing system toward more efficient and accurate pathological assessment
  • Generates standardized reports which follow Society of Cardiovascular Computed Tomography guidelines
  • Stores reports as well as reconstructured, processed, and labeled images displaying identified pathology in PACS for later retrieval
  • Automated lesion detection and segmentation allows further quantitative analysis results including luminal & vessel diameter, stenosis length & degree, plaque volume, composition analysis, and plaque burden
  • Quantitative analysis results and descriptions of plaques are provided and arranged according to the major coronary branch systems. The measurement changes throughout a branch are displayed in a graph to facilitate pathological findings
  • Users can click on each plaque sign to quickly locate it; image contents identifying its pathology and related analysis results are presented simultaneously to save doctors’ time