One of the major causes of death worldwide is coronary artery disease, for which stent implantation is a common treatment. Current diagnosis of coronary artery disease relies on visual examination of angiograms by operators to identify significant stenoses in arteries.
Visual examination of angiograms might lead to false diagnosis as it relies on estimation, which is subject to intra- and inter-operator variability. Yet, the accuracy of stenosis detection and its extent is very crucial in patient diagnosis. Unnecessary usage of stents places a huge financial cost on the patient and may complicate the heart condition further. On the other hand, missing stenoses that may be severe in coronary vessels could lead to health risks, such as myocardial infarction or death. These limitations of existing approaches are well known with significant errors demonstrated in diagnosis, with both under-‐ and over-‐calling of stenoses. Although techniques such as quantitative coronary angiography (QCA) are available, these still require significant human input and resource costs.
There is a great need for an automated system to quantitatively analyze angiograms to determine the extent and location of stenosis in arteries and recommend a stent (or other appropriate therapies) only if needed. Advent of new technologies provides better resolution and quality of medical images. However, many challenges in analysis of these images remain to be overcome. An automated system that analyzes these images reduces not only the cost of examination, but also unnecessary treatments and avoids under-‐treatment. Along with new image acquisition techniques that provide better resolution and quality, new image processing techniques help the physician perform accurate diagnosis.
Automated analysis of vasculature in coronary angiograms
The proposed technology is a combination of advanced image processing techniques and machine learning approaches used in pre-processing and vasculature segmentation. In the pre-processing part, artifacts are automatically removed from the video sequences to facilitate the segmentation of vessels. There are different kinds of artifact in angiogram videos, which make it very hard to be able to segment the vessels accurately. Stitches, pacemakers, and motion artifacts due to heart beat and movement of camera and patient are among artifacts, which exist in angiogram videos. After the pre-processing step, the system applies advanced digital image processing and machine learning methods to segment the vessels and calculate the diameter of each branch in the vasculatures. A decision support system aids cardiologists in diagnosing diseases using calculated quantitative parameters of vasculatures. The invention aims to reduce the human interaction as well as the computation time. The system is fully automated and is capable of performing the entire analysis without human intervention.
- Quantitatively analyze angiograms to determine the extent and location of stenosis in arteries and recommend a stent (or other appropriate therapies) only if needed.
- Fully automated vasculature segmentation software that creates results within a reasonable amount of time without human intervention.