Nuclear texture measurements in image cytometry.

DNA image cytometry is widely used in cytopathology as a means to obtain objective information concerning the diagnosis and prognosis of human cancer. Using specially designed devices, the high resolution spatial and photometric information is available in the images of a microscopic field. If quantitative DNA specific stains are used the chromatin distribution in the cell nuclei can be measured, which is one of the critical features for cytopathological analysis. In normal cells, changes in the chromatin appearance reflect changes in the activation patterns of genes. In tumors, dramatic changes in the nuclear chromatin appearance are common and have been associated with the progression of the disease. Features describing the chromatin distribution pattern are referred to as texture features. Nuclear texture features are sensitive to the differences between the various descriptive classes of chromatin patterns. In this paper we discuss the main categories of nuclear texture measurements. Texture features can be roughly divided into the following categories: 1) descriptive statistics of chromatin distribution; 2) discrete texture features; 3) range extreme; 4) markovian; 5) run length and 6) fractal texture features. Representative features of each of the above categories are discussed together with mathematical formulas, simple figures for explanation as well as images of typical cells which differ significantly in some texture features. Key references are also provided.