Histol Histopathol (2000) 15: 1051-1057 001: 10.14670/HH-15.1051 http:/www.hh.um.es Histology and Histopathology Ce/u/ar and Molecular Biology Potential use of spectral image analysis for the quantitative evaluation of estrogen receptors in breast cancer C. Rothmann 1 , l. Barshack 2 , A. Gil3, l. Goldberg 2 , J. Kopolovic 2 and Z. Malik 1 1 Life Sciences Deparment, Bar-llan University, Ramat-Gan, lsral, 2Pathology Deparment, Sheba Medica/ Center, Tel-Hashomer, Israel and 3 Applied Spectral lmaging, Migdal Haemek. Israel Summary. Evaluation of estrogen receptor (ER) content is an important factor in the choice of therapy and prognosis of breast cancer patients. In this study, we demonstrate a new spectral image analysis technique for objective and quantitative evaluations of stained specimens. The SpectraCube ™ system was used to analyze nuclear antigens in thirteen cases of breast cancer stained by the immunoperoxidase method with hematoxylin counterstain. Spectral imaging segregated the spectrum of diaminobenzidine (DAB) from the background color of hematoxylin and a spectral index was calculated. The spectral index essentially agreed with the pathologist's index (on a scale of O to 3) in seven out of the thirteen cases. A substantial number of ER positive pixels was detected in the two cases scored as O by the pathologist's index. In a test case scored as 1 by the pathologist's index we detected a significant number of pixels, representing 47% of the nuclei, with DAB-intensity values higher than the cut-of value of 1.2. These data suggest that spectral image analysis is a sensitive method providing intensive information with high reproducibility. Our spectral imaging method is highly flexible, enabling the user to define the spatial resolution of the analyzed specimen by choosing the number of pixels per one nucleus. Key words: Spectral image analysis, Quantification, Breast cancer, Prognostic factors, Estrogen receptors lntroduction Evaluation of hormone receptor expression in tumor cell nuclei is an integral part of routine breast cancer diagnosis (Lehr et al., 1998). Diagnosis is determined as a function of a number of factors such as TNM (tumor node metastasis) stage; tumor type and diferentiation; patient age; hormonal, cytotoxic and immune therapeutic Ofprint requests to: Prof. Zvi Malik, Microscopy Unit, Lite Sciences Faculty, Bar-llan University, Ramat-Gan 52900, Israel. Fax: 972 3 5345878. e-mail: malikz@mail.biu.ac.il modalities, and the duration of tamoxifen treatment. In addit ion, the estrogen (ER) and progesterone (PR) receptor status of the carcinoma provides important information for prognosis and choice of therapeutic approach (Bejar et al., 1998). Patients classified as ER positive/PR-positive are treated with tamoxifen, which has proven to produce a favorable response and survival advantage (Tonetti and Jordan, 1997). Additionally, tamoxifen is the only drug known to reduce the incidence of collateral disease and produces relatively few harmful side efects (Tonetti and Jordan, 1997). However, in routine breast cancer diagnosis, receptor expression is often quantified in arbitrary seores with high inter- and intraobserver variability. Therefore, a question is raised as to whether pathologists' subjective impressions of negative versus positive ER states - and their readings of grades of positivity - are reliable in determining appropriate oncological management. Immunohistochemical assays are most commonly used today to s tain ER. These me thods are quick, inexpensive and allow hormone receptor determination even if only small amounts of tumor tissue are available. Nevertheless there is no consensus regarding the method for scoring the results. Although the immunohisto chemical technique is easy to s tandardize, its interpretation relies solely on subjective visual estimates, yielding only qual itative or at best semiquantitative results (Baddoura et al., 1991; Lehr et al., 1998). Most pathologists distinguish between positive and negative results based on the percentage of positive tumor cells, the cutof being arbitrarily defined between 5 and 45% (Battifora et al., 1990; Raymond and Leong, 1990; Molino et al., 1995; Veronese et al., 1995; Pertschuk et al., 1996). The most commonly used method for semiquantifcation o f the hormone-receptor status is the so-called H-score. The H-scores take into consideration both the color intensity of the immunoreactive neoplastic cel l nuc lei and the percentage of posi tive cells in histological sections of parafin-embedded tissue stained with anti-hormone-receptor antibody by one of the commercially available immunohistochemical kits. Still, most of these scoring systems are cumbersome to