Optimized neural network for vulnerable plaque detection in OCT images with noise tolerance and adaptive coefficient zeroing

Background: Optical Coherence Tomography (OCT) was a non-invasive imaging method that provides higher- resolution images of biological tissues. OCT is used in cardiovascular medicine to identify Astheroclerosis plaques, which are a kind of plaque in blood arteries that represent a significant chance of rupture along with subsequent cardiovascular problems. By analyzing OCT images, healthcare professionals can identify specific features of atherosclerosis plaques, including thin fibrous caps, fatty cores, and micro-calcifications. This allows early identification of life-threatening risks and targeted treatments. Still, the analysis method is subject to miscalculations and high workload. Objective: The objective of this research is to propose a novel approach called CVP-OCT-NTACZNN-PCBESA for detecting and classifying cardiovascular vulnerable plaques utilizing OCT imaging. The aim of CVP-OCTNTACZNN-PCBESA method is to provide healthcare practitioners with a more effective tool for early detection and better treatment of heart disease-causing plaques, by this reducing cardiovascular problems and advancing patient's life. Methods: After removing noise and improving image quality with Data-adaptive Gaussian Average Filtering (DAGAF), the recovered features are fed into NTACZNN. This neural network recognizes some Cardiovascular Vulnerable Plaques, including fibrocalcific plaque, fibroatheroma, thrombus, fibrous plaque, and microvessels. Result: The results of this research show that the CVP-OCT-NTACZNN-PCBESA method outperforms existing approaches in terms of accuracy, precision, detection quality score, and recall rate for identifying cardiovascular susceptible plaques. Especially, the CVP-OCT-NTACZNN-PCBESA outperforms DL-based models and diagnostic systems that use intravascular OCT imaging by 27.36% to 38.27% in accuracy and 19.40% to 33.42% in precision. Conclusion: In conclusion, the proposed CVP-OCT-NTACZNN-PCBESA method outperforms other existing approaches for detecting and characterizing cardiovascular susceptible plaques. Its improvements in accuracy and precision hold promise for earlier intervention and better patient outcomes in cardiovascular medicine.