AIMS: To determine whether the measurement of vascularity can be used to differentiate follicular adenomas from follicular carcinomas or to reflect the prognosis of follicular carcinomas and papillary carcinomas of the thyroid gland, and to compare four methods of assessing vascularity. METHODS: Tissue sections from 26 papillary carcinomas, 15 follicular adenomas, and 15 follicular carcinomas were stained with an antibody to CD34. A computerised image analysis system was used to calculate, for each tumour, mean endothelial areas and the mean endothelium to tumour epithelial nucleus area ratio from 10 systematically selected fields across one dimension of the tumour ("systematic field" analysis) or from the three most vascularised fields of the tumour ("hot spot" analysis). A European Organisation for Research on Treatment of Cancer (EORTC) prognostic index was calculated for each papillary carcinoma and follicular carcinoma. RESULTS: Significant differences in vascularity between the three tumour groups could only be shown by comparing mean endothelial area values measured from hot spots. While the hot spot median mean endothelial area of follicular carcinomas was significantly greater than that of follicular adenomas, there was a large overlap between the two groups. For follicular carcinomas, higher hot spot mean endothelial area values were related to worse prognosis as indicated by the EORTC prognostic indices. No association between vascularity and prognosis was found for the papillary carcinomas, regardless of the method of assessing vascularity. CONCLUSIONS: Measuring endothelial area from hot spots using a computerised image analysis system is a sensitive method of assessing the vascularity of thyroid tumours. While vascularity measurement cannot be recommended as a practical tool for differentiating between malignant and benign follicular tumours, the suggestion that vascularity may reflect prognosis for follicular carcinomas deserves further study.