Microvascular imaging in diffuse thyroid disease: diagnostic performance of subtraction ultra-micro angiography

Background Quantitative assessment of thyroid microvascularity may improve the ultrasound evaluation of diffuse autoimmune thyroid disease, particularly when conventional Doppler findings overlap. This prospective single-center study investigated whether subtraction Ultra-Micro Angiography (sUMA), quantified using the color pixel percentage (CPP) index, differentiates chronic autoimmune thyroiditis (CAT) from Graves’ disease (GD) and healthy thyroid parenchyma. Methods A total of 179 participants were included: 84 controls, 74 with newly diagnosed, untreated CAT, and 21 with newly diagnosed, untreated GD. All participants underwent standardized B-mode ultrasonography followed by sUMA using the same ultrasound platform and linear transducer. Three CPP sUMA measurements were obtained in each thyroid lobe, and the lobe medians were averaged to obtain the whole-gland CPP sUMA value. Interlobar and group differences, correlations, multivariable associations, apparent ROC performance, bootstrap cut-off distributions, and repeated stratified five-fold internal validation were analyzed. Results CPP sUMA values did not differ significantly between the right and left lobes. Mean CPP sUMA increased progressively from controls (23.25 ± 6.02%) to CAT (50.49 ± 13.58%) and GD (79.90 ± 8.86%). In CAT, CPP sUMA correlated positively with thyroid volume and TSH and inversely with FT4 and was significantly higher in overt than in subclinical disease. In GD, CPP sUMA correlated positively with FT3, FT4, and TSI. Differences between the diagnostic groups remained significant after adjustment for age, sex, BMI, and thyroid volume. Apparent ROC analysis showed high discrimination between CAT and controls (AUC 0.980; cut-off >31.55%; sensitivity 93.2%; specificity 90.5%) and between GD and CAT (AUC 0.966; cut-off >62.00%; sensitivity 100%; specificity 78.4%). GD and controls were completely separated in the analyzed cohort. During internal validation, mean accuracy was 88.5% for CAT versus controls, 82.0% for GD versus CAT, and 98.9% for GD versus controls. Conclusion CPP sUMA provided a quantitative tissue-level estimate of thyroid microvascularity and maintained high discriminatory performance during internal validation. However, the proposed cut-offs are hypothesis-generating and require confirmation in larger multicenter cohorts using different ultrasound platforms. Once externally validated, CPP sUMA may complement thyroid function testing, antibody assays, and conventional ultrasound in the assessment of autoimmune diffuse thyroid disease.