Author + information
- Vlad Zimin1,
- Yazan Gharaibeha2,
- Juhwan Lee2,
- Gabriel Tensol Rodrigues Pereira1,
- Mohamad Amer Alaiti1,
- Dean Jia1,
- Armando Vergara-Martel1,
- Luis Augusto Palma Dallan1,
- Elder Zago1,
- Patricia Pizzato1,
- Hiram G. Bezerra1 and
- David L. Wilson2
Calcified plaques can negatively impact stent expansion, requiring meticulous procedure planning from interventional cardiologists. Optical Coherence Tomography (OCT) is the only intravascular imaging technique with the ability to image the extent of calcification, but it is cumbersome and prone to variability.
We investigated whether our software can accurately and semi-automatically detect calcified plaques with the future goal to create a model to predict lumen change after applying pre-stent pressure.
We used 45 previously segmented OCT pullbacks for calcification, performed by the Cardiovascular Imaging CoreLab in Cleveland, totaling 2,640 frames to feed our deep-learning algorithm. The images were analyzed at a pixel level between lumen, “other”, and calcium. Analysis was compared to the human gold standard result for calibration purposes. After that step proved to be successful, we applied our algorithm to a prospectively acquired ex vivo vessel and compared it to the human gold standard, blinded to the deep-learning results.
|Predicted “Other”||Predicted “Lumen”||Predicted “Calcified Plaque”|
|True “Other”||14,208,138(95.1 ± 2.8)||220,241(1.4 ± 1.2)||493,874 (3.3 ± 2.0)|
|True “Lumen”||17,574 (1.5 ± 2.4)||1,112,913 (98.0 ± 2.4)||4,768 (0.4 ± 0.5)|
|True “Calcified Plaque”||32,154 (13.7 ± 6.9)||7,817 (3.3 ± 2.4)||193,697 (82.8 ± 6.8)|
We conclude that our deep-learning algorithm can correctly label the lumen and other tissues with 95% accuracy and >80% accuracy for calcium. This opens the path for real-time, accurate calcification segmentation to be applied to a Finite Element Model (FEM). This work will provide a fundamental understanding of the interaction between the reconstructed heterogeneous lesion and pre-stenting to provide clinical guidance for choosing the interventional techniques in the presence of calcification.