Beschreibung:
<jats:title>Abstract</jats:title>
<jats:p>In many cancers, tumor-infiltrating lymphocytes (TILs) indicate levels of tumor immunogenicity and are a strong predictor of survival. An understanding of the phenotype and spatial distribution of TILs within tumor regions would be advantageous for characterizing host response. However, visual TIL assessment is prone to error and multimarker quantitation is difficult with standard methods. Here we present a multi-marker, computer-aided event-counting method for the determining the phenotypes of lymphocytes in melanoma sections using a novel multispectral imaging (MSI) approach. A section of a tissue microarray containing 120 melanoma cores was stained for CD3, S100, Foxp3 and hematoxylin. This was imaged using MSI and the individual staining of each marker separated from each other using spectral unmixing. The images were analyzed using software which had been trained to recognize the tumor area based on the S100 staining pattern. Then the Foxp3 status of each CD3+ TIL was then determined. Results indicate that machine-learning software can be trained to accurately recognize tumor regions within each core. MSI enabled the accurate quantitation of three immunostains in the sample without crosstalk. Within the tumor region of each core it was possible to count the CD3+ TILs and then determine the Foxp3 status of each. This multimarker phenotyping and counting approach shows the potential for broad applicability in the assessment of TILs in many solid tumors.</jats:p>