coronary-artery-stenosis-detection

Using Faster RCNN with 'inception resNetv2' as backbone architecture to detect stenosis regions within coronary artery from coronary CT angiography.

I followed this publication by Danilov et al. namely Real-time coronary artery stenosis detection based on modern neural networks¹ to detect stenosis region of coronary artery. Here the authors showed a brief comaparison between eight object dection architectures in coronary stenosis detection. Within them faster RCNN scored the highest.

Methodology

I utilized the object-detection-API by tensorflow to train our model and detect stenosis region of coronary artery.

Result

Dataset

Dataset is accessible by the following link: Angiographic dataset for stenosis detection²

Reference

Footnotes

1. Danilov, V.V., Klyshnikov, K.Y., Gerget, O.M. et al. Real-time coronary artery stenosis detection based on modern neural networks. Sci Rep 11, 7582 (2021), https://doi.org/10.1038/s41598-021-87174-2. ↩

2. Danilov, Viacheslav; Frangi, Alejandro; Klyshnikov, Kirill; Kutikhin, Anton; Ovcharenko, Evgeny; Gerget, Olga (2021), “Angiographic dataset for stenosis detection”, Mendeley Data, V1, https://doi.org/10.17632/ydrm75xywg.1 ↩