Textural analysis of optical scattering for identification of cancer in breast surgical specimens

Textural analysis of tissue scattering images is proposed for healthy versus tumor discrimination. Scattering center density varies from normal to tumor tissues and this variation is translated into different textures in the scattering power map. Adipose tissue shows low autocorrelation values while tumor tissues present higher entropies than normal tissue. Consequently, a combination of autocorrelation and entropy values allows ready tissue discrimination by a supervised linear classifier. The proposed approach has been validated over a set of 29 breast tissue samples achieving a sensitivity of 73.59% and specificity of 82.40%.