Quantificação automática de tonalidades em imagens de imunohistoquímica e classificação de tumores

Abstract

Immunohistochemistry is considered as a crucial technique and is widely used in many medical research laboratories as well as for clinical diagnostics. It consists in antigens detection in tissue sections using labeled antibodies as specific reagents. Antigen-antibody interactions are then visualized by a marker such as fluorescent dye, enzyme, colloidal gold or radioactive element. Colloidal gold has been successfully used to identify immunohistochemical reactions by means of light microscopy, with fast detection, high precision, low cost, and furthermore requiring small amount of sample. The classification of a malignant or benign tumor in immunohistochemical images can be made by measuring color levels or the geometric appearance. However, the anatomopathologist is generally the only professional who can do such analyses. There exits few softwares that assist in such analyses, but they are very expensive, or process only images in gray scale (8 bits), which lead to a loss of information, or demands highly qualified staff to be operated. The aim of this work was to develop a software with friendly interface that make possible a classification of the tumors into malignant or benign, by means of a quantitative analysis of the diverse pixels tonalities of the image. The results obtained with the developed software show a good agreement with the method generally used by anatomopathologists to validate their own measurements. © Springer-Verlag Berlin Heidelberg 2007.