School of EE Seminar- Yosef Kashi- Visual-System Inspired Algorithm for Enhanced Visibility in Coronary Angiograms
https://us02web.zoom.us/j/86813756849?pwd=SEdJUE84VkhITTNJakdSZUJEKzNpdz09
Meeting ID: 868 1375 6849
Passcode: c2nkDw
Electrical Engineering Systems Seminar
Speaker: Yosef Kashi
M.Sc. student under the supervision of Dr. Hedva Spitzer
Wednesday, February 2nd, 2022, at 15:00
Visual-System Inspired Algorithm for Enhanced Visibility in Coronary Angiograms
Abstract
The current gold standard for detection of coronary artery disease and invasive treatment is conventional X-ray coronary angiography (CAG).
A lot of effort has been invested in previous studies to better expose and enhance visibility which is a key factor for improved diagnosis and treatment. Yet, visibility of the blood vessels and their luminal content, especially the small ones, is still limited in spite of its substantial clinical significance. We propose an algorithm for Cineangiograms, whose main part is inspired by a specific mechanism of the visual system. It enables performing multiscale line completion, integrated with ROI detection framework. We computed binary vascular ROI estimation, with additional functions that enable continuation, separately for each scale, followed by integrated scales to detect content within the vascular channels. In the last stage, the ROI is incorporated to the original image with further adjusting the appropriate dynamic range of the image.
The results have been evaluated by consensus of 2 cardiologists who reported on the better visibility of the coronary blood vessels, especially the small ones. The better identification of various catheterization tools within the blood vessels, as well as the lesions themselves (for the first time), may all together lead to better diagnostic and procedural capabilities. Nevertheless, also noted was the appearance of disturbing vertebral and bony artifacts, which can mimic small blood vessels and may lower diagnostic accuracy, especially while viewing still images, and less so in cine mode. Ongoing research addresses this challenge. Our current algorithm has a potential to be performed in real time, without the need of annotated data, and even in low contrast conditions (during dye clearance or lower radiation).
השתתפות בסמינר תיתן קרדיט שמיעה = עפ"י רישום שם מלא + מספר ת.ז. בצ'ט
