Search

Cardiovascular diseases are a leading cause of mortality worldwide. This study introduces an innovative end-to-end pipeline for automated measurement of Carotid Intima-Media Thickness (CIMT) from ultrasound images, a crucial step in assessing cardiovascular risk. The process begins with the localization of the Region of Interest (ROI) using the You Only Look Once (YOLO) v5, followed by the application of various image enhancement methods. For precise segmentation of the Intima-Media Complex (IMC), a Deep Learning model is employed, featuring an encoder-decoder architecture, Self-Organizing Neural Networks (Self-ONN), multi-stage feature fusion, and deep supervision. The Simultaneous Truth and Performance Level Estimation (STAPLE) algorithm is then utilized to combine these segmented masks, producing a final mask for precise CIMT measurement. This method achieved a Dice score of 82.03% and an Intersection over Union of 69.55% on the public Carotid Ultrasound Binary Segmentation (CUBS) dataset. Furthermore, the novel CIMT calculation technique demonstrated a mean squared error of 0.049 mm and a mean average error of 0.166 mm. The utilization of YOLOv5 for ROI selection significantly improved the accuracy of CIMT measurements, ensuring the most relevant regions are considered for analysis. The application of the STAPLE algorithm for prediction consensus demonstrates significant promise in producing optimal segmentation masks. In conclusion, this research, along with its continual enhancements, holds promise for facilitating the integration of human expertise and Deep Learning technologies, thereby refining diagnostic processes and contributing to the advancement of healthcare standards. • Complete CIMT solution: ROI detection, IMC segmentation, mask post-processing. • Innovative IMC segmentation: deep supervised fusion in encoder-decoder. • Unique CIMT calculation from segmentation mask. • In-depth interobserver variability analysis on public dataset subset. • Extensive exploration of image enhancement's impact on automation. [ABSTRACT FROM AUTHOR]