A number of laboratories are utilising both hypoxia and perfusion markers to spatially quantify tumour oxygenation and vascular distributions, and scientists are increasingly turning to automated image analysis methods to quantify such interrelationships. In these studies, the presence of regions of necrosis in the immunohistochemical sections remains a potentially significant source of error. In the present work, frozen MCa-4 mammary tumour sections were used to obtain a series of corresponding image montages. Total vessels were identified using CD31 staining, perfused vessels by DiOC 7 staining, hypoxia by EF5/Cy3 uptake, and necrosis by haematoxylin and eosin staining. Our goal was to utilise image analysis techniques to spatially quantitate hypoxic marker binding as a function of distance from the nearest blood vessel. Several refinements to previous imaging methods are described: (1) hypoxia marker images are quantified in terms of their intensity levels, thus providing an analysis of the gradients in hypoxia with increasing distances from blood vessels, (2) zonal imaging masks are derived, which permit spatial sampling of images at precisely defined distances from blood vessels, as well as the omission of necrotic artifacts, (3) thresholding techniques are applied to omit holes in the tissue sections, and (4) The relationship between tumour hypoxia and therapeutic response has been well documented in the literature for quite some time, and direct measures of tumour oxygenation have been shown to correlate with both long-term survival (Hockel et al, 1996) and the occurrence of distant metastases (Brizel et al, 1996). Two of the more prominent techniques for measuring tumour hypoxia are: (1) the Eppendorf electrode for determination of tumour pO 2 levels, and (2) nitroimidazole hypoxia markers, e.g., EF5, NITP, and pimonidazole, which covalently bind to hypoxic tumour cells and allow immunohistochemical or flow cytometric determination of hypoxia distributions (Koch et al, 1995;Hodgkiss and Wardman, 1992;Raleigh et al, 1987). The presence of tumour necrosis can have substantial effects on each of these types of measurements. With the Eppendorf technique, necrotic regions result in a reduction in measured pO 2 levels that is not reflective of a corresponding reduction in clonogenic survival of the tumour cells (Fenton et al, 1995). This leads to overestimates of the fraction of radiobiologically resistant tumour cells. In the case of hypoxia markers, regions of necrosis can also lead to inaccurate predictions in overall tumour hypoxia. Since these drugs are not metabolized in necrotic areas, such regions appear well oxygenated and lead instead to an underestimate of overall tumour hypoxia. Several recent studies have described methods for quantifying hypoxia marker distribution as a function of distance from either perfused or anatomical blood vessels. The first (Rijken et al, 2000) is an elegant, multiparameter analysis of vascularity, perfusion, and hypoxia that characterises uptake of two hypoxia markers,...