Deep Learning-Driven Study of an Intelligent Analysis System for Atherosclerotic Ultrasound Images of Hypertensive Patients

A deep learning-based system was developed for automatic assessment of atherosclerotic lesions in carotid ultrasound images. Using improved Mask R-CNN, the system achieved plaque detection with 93.5% accuracy and IMT measurement with 0.112 mm error. Combined with clinical indicators, the risk prediction model reached an AUC of 0.912. Clinical applications in five hospitals showed reduced diagnosis time from 36.8 to 8.4 minutes and improved diagnostic consistency to <tex xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">$\text{91.0\%}$</tex>. The system significantly enhanced the efficiency and accuracy of atherosclerosis screening and risk assessment.