Key result
RFCN ResNet-101 V2 neural network achieves 0.94 mean average precision for real-time coronary stenosis detection.
Why the study?
Deep learning techniques may facilitate coronary artery stenosis detection on invasive angiography, but previous studies have failed to achieve superior accuracy and performance for real-time labeling.
Do deep learning methods provide accurate and fast real-time coronary artery stenosis detection on invasive angiography?
Population
Clinical angiography data of 100 patients
Comparison
Eight detectors based on different neural network architectures
Design
Model training and validation study
Authors
Loading...
Supports real-time AI assistance feasibility in angiography; hypothesis-generating pending prospective outcome validation.
Do deep learning methods provide accurate and fast real-time coronary artery stenosis detection on invasive angiography?
Deep learning models, particularly RFCN ResNet-101 V2, demonstrate high accuracy and sufficient speed to enable real-time coronary artery stenosis detection during invasive angiography.
Данилов et al. (2021) studied Coronary artery disease (n=100). Deep learning neural networks (e.g., RFCN ResNet-101 V2) vs. Reference labeling by interventional cardiologist was evaluated on Mean Average Precision (mAP) for stenosis detection. The RFCN ResNet-101 V2 neural network achieved an optimal accuracy-to-speed ratio for real-time coronary artery stenosis detection with a mean Average Precision of 0.94 and speed of 10 fps.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: