A method is described for computing a statistically significant difference between dysplasia and Barrett's esophagus (both with and without inflammatory component) using a discriminate function with diffuse reflectance measurements performed at a minimum of four different wavelengths of 485, 513, 598, and 629 nm. The discriminate function found depends both on local blood fraction volume THB and oxygenation SO2. A pull-back approach of spectral data acquisition is disclosed which takes into account tissue motility in esophagus and measurement geometry peculiarities. The pull-back approach provides a significant improvement of measurement reproducibility and reduction of data deviation by 75-100%, resulting in a better discrimination between different histological groups.