Lymphocytic infiltration is associated with a favourable prognosis and predicts response to chemotherapy in many cancer types, including the aggressive triple-negative breast cancer (TNBC). However, it is not well understood owing to the high levels of spatial heterogeneity within tumours, which is difficult to analyse by traditional pathological assessment. This paper describes an unbiased methodology to statistically model the spatial distribution of lymphocytes among tumour cells based on automated analysis of haematoxylin-and-eosin-stained whole-tumour section images, which is applied to two independent TNBC cohorts of 181 patients with matched microarray gene expression data. The novelty of the proposed methodology is the fusion of image analysis and statistical modelling for an integrative understanding of intratumour heterogeneity of lymphocytic infiltration. Using this methodology, a quantitative measure of intratumour lymphocyte ratio is developed and found to be significantly associated with disease-specific survival in both TNBC cohorts independent to standard clinical parameters. The proposed image-based measure compares favourably to a number of gene expression signatures of immune infiltration. In addition, heterogeneous immune infiltration at the morphological level is reflected at the molecular scale and correlated with increased expression of
CTLA4
, the target of ipilimumab. Taken together, these results support the fusion of high-throughput image analysis and statistical modelling to offer reproducible and robust biomarkers for the objective identification of patients with poor prognosis and treatment options.
Image analysis and statistical modelling for measurement and quality assessment of ornamental horticulture crops in glasshouses
