Long-term carotid plaque progression and the role of intraplaque hemorrhage: A deep learning-based analysis of longitudinal vessel wall imaging

Fig. 1. Flow chart of the study cohort

Fig. 2. Illustration of the calculations for long-term and short-term progression rates of plaque burden. For each subject, the %WV of plaques in the left and right carotid arteries was measured at all available time points. For each plaque, a linear model was fitted to determine the long-term progression rate, expressed as the annualized absolute change in %WV (%/year). Short-term progression rates (%/year) were calculated as the annualized difference in %WV between consecutive scans. %WV percent wall volume

Fig. 3. Example of bilateral plaque burden increases over five vessel wall imaging sessions spanning 5 years, demonstrating symmetrical plaque progression on both the left and right sides. The vessel wall contours were generated using deep learning-based segmentation, with red contours delineating the vessel lumen and blue contours delineating the vessel outer wall boundary

Fig. 4. Upper panels: correlation of long-term and short-term growth rate of plaque burden between bilateral carotid arteries, demonstrating systemic influences on bilateral plaque progression. Lower panels: correlation of bilateral short-term progression in the two subgroups of carotid arteries with (right) and without IPH (left). The partial correlation coefficient between left and right plaque progression rates was calculated after adjusting for age, sex, history of hypertension, and statin use. IPH intraplaque hemorrhage

Fig. 5. Box plots demonstrating the effects of presence and volume of IPH on plaque burden. In (a) and (c), %WV and annualized changes in %WV were compared based on presence or not of IPH. In (b) and (d), %HV was divided into three equal-sized bins, comparing %WV and annualized changes in %WV across different degrees of IPH volume. %WV percent wall volume, IPH intraplaque hemorrhage, %HV percent hemorrhage volume

Fig. 6. Example of plaque progression over four vessel wall imaging sessions spanning five years, demonstrating an increase in plaque burden, marked by the occurrence of IPH at Scan 3 and subsequent volume expansion at Scan 4. The vessel wall and IPH contours were generated using deep learning-based segmentation, with red and blue contours delineating the vessel lumen and vessel outer wall boundary on MERGE, and orange contours delineating IPH boundary on SNAP. IPH intraplaque hemorrhage, MERGE motion-sensitized driven equilibrium prepared rapid gradient echo sequence, SNAP simultaneous non-contrast angiography and intraplaque hemorrhage imaging