Cardiologists perform millions of angiograms annually in the U.S. They currently “eyeball” the percent diameter stenosis to determine the presence of a serious blockage. The interpretation drives critical decisions regarding the need for further testing or surgical intervention (i.e. bypass surgery).

The prevalence of incorrect assessments has driven MCIRCC members to construct a fully automated, computer-based platform –AngioInsight–that quickly and accurately aids with interpretation of coronary angiogram videos obtained in routine practice, eliminating variability in human interpretation. When constructed, a public dataset of coronary angiogram videos will be available to encourage additional innovations and development of new algorithms.


Cardiologists’ visual estimation may be incorrectly assessing coronary stenosis severity in up to 20% of cases. As a result, many patients are exposed to unneccesary invasive stenting procedures or are not receiving the important treatments they need. Older computer-based techniques have had significant limitations that restrict the ability to adapt them into routine practice.


Pilot work conducted includes exploration of a proof-of-concept version of AngioInsight, exploring high quality, minimal disease, and limited artifact angiogram videos to train the algorithm. A promising number of correct identifications occurred. More work will be conducted to determine the accurate identifcation of more significant and complex blockages. The University of Michigan has received a patent (with another pending) and is commercializing AngioInsight.


The need for an accurate computer-based technique used to identify coronary stenosis led to the development of AngioInsight — a newly created Delaware corporation that will drive the coronary angiogram video interpretation technology forward.