Quantitative evaluation of the tumour-stroma border by exponential regression analysis.

The architecture of the tumour margin is an essential feature for the histological diagnosis of certain neoplasms. In a previous study we have shown by computer simulations, that the degree of tumour cell motility and proliferation influences qualitative morphological criteria of the tumour border. Here we propose a method for an objective, quantitative description of the tumour cell distribution at the tumour-stroma border. 100 morphological patterns generated by computer simulation with different preset degrees of motility and proliferation were evaluated. On a measuring path starting at the most peripheral cell and going towards the tumour centre, the density of tumour cells for each distance from the tumour margin was calculated from the number of cells within a 5 x 5 matrix around each measuring point. An exponential regression function was adapted to the measured density distribution. The parameters of the regression curve and 'the goodness of fit' were tested for correlation with the preset simulation parameters. It was found that the degree of motility and proliferation can be deduced from the quantitative descriptors of the tumour margin (linear regression of the preset values and of the values estimated from the morphological analysis: r = 0.903; t = 20.7; P less than 0.001). Besides the evaluation of computer-simulated morphological patterns, the same measuring procedure can also be applied to histological slides of melanocytic skin tumours using automated image analysis.