Mapping endometrial vascular functional gradients using depth-derived ultrasound localization microscopy (D-ULM) for early stratification of postinjury fibrotic remodeling and therapeutic guidance.

RATIONALE: Endometrial injury exhibits significant clinical heterogeneity in fibrotic outcomes. Recovery of microvascular perfusion is central to prognosis, while the depth of injury is an important factor influencing regenerative capacity. However, current clinical imaging techniques are limited by diffraction, resulting in insufficient resolution to reliably depict depth-dependent microvascular architecture of the endometrium, with consequent difficulty in evaluating injury depth or in making early prognostic distinctions. METHODS: A depth-derived ultrasound localization microscopy (D-ULM) approach is introduced to capture depth-dependent microvascular features in the endometrium. First, we established a normal depth reference band from healthy rats to provide a physiological baseline for identifying pathological vascular changes. Next, in an endometrial injury rat model, we applied unsupervised clustering of early D-ULM features to identify microcirculation phenotypes. Subsequently, we evaluated the correlation between D-ULM features, D-ULM clusters, and late-stage fibrosis area as well as immunofluorescence markers of hypoxia and inflammation, with a view to tying early microvascular changes to eventual recovery or to fibrotic remodeling. RESULTS: In healthy rats, depth-profile curves revealed a distinct transition zone between the deep and superficial layers, enabling the delineation of a quantifiable functional boundary of the endometrial microvasculature. Based on this depth-dependent layering, normal reference intervals were established for the seven D-ULM features. Among these, the reference interval for the vascular distribution center (com_depth_vessel) was 0.371-0.481. In injured cohort, unsupervised hierarchical clustering of day-3 D-ULM features identified three phenotypes: regenerative (Reg), inflammatory hyperperfusion (IH), and irreversible destruction (ID). The day-14 fibrosis area differed significantly among the phenotypes (P < 0.001), with the Reg phenotype showing the lowest fibrosis, while the IH and ID phenotypes exhibited markedly higher levels. Several D-ULM features were positively correlated with day-14 fibrosis, including com_depth_vessel (ρ = 0.707, q < 0.001), vessel_slope (ρ = 0.539, q = 0.017), vessel_auc (ρ = 0.811, q < 0.001), and mean_vessel_shallow (ρ = 0.846, q < 0.001); conversely, large_ratio_diff was inversely correlated (ρ = -0.568, q = 0.011). The IH phenotype exhibited elevated M1/M2 ratio, HIF-1α, and CD31 expression, whereas the ID phenotype showed increased M1/M2 ratio and HIF-1α but reduced CD31 expression by immunofluorescence. CONCLUSIONS: D-ULM enables depth-resolved imaging of endometrial microvascular functional gradients in a noninvasive manner. This approach provides an imaging framework for early stratification of injury phenotypes, and may guide personalized preventive strategies for patients at risk of postinjury fibrosis.