12 th European Congress on Digital Pathology previously European Congress on Telepathology and international Congress on Virtual Microscopy College des Bernardins, close to île Saint-Louis Paris, France, 18–21 June 2014 Automated image analysis is superior to manual reading of HER2 expression in breast cancer Rossing HH, Talman ML, Vainer B Department of Pathology, Rigshospitalet, University of Copenhagen, Denmark BACKGROUND Human epidermal growth factor receptor 2 (HER2) is a receptor for circulating growth factor that stimulates cell proliferation. Overexpression, or amplification, of HER2 oncogene plays an important role in the development and progression of approx % of breast cancers and is the target for the anti-cancer treatment drug trastuzumab (Herceptin). Analysis of HER2 expression in breast cancer is essential for selecting the correct patients for this treatment and relies on semi-quantitative assessment using immunohistochemical (IHC) methods. In approx. 20% of the patient cases, HER2-IHC stainings are inconclusive, and its necessary to supplement with a fluorescence in situ hybridization (FISH) analysis in order to determine whether the HER2 gene is truly amplified. The aim of the study was to validate the digital, automated image analysis algorithm HER2-CONNECT TM (Visiopharm), which is developed to discriminate between normal protein expression and protein overexpression, with the intention of minimizing the number of inconclusive IHC scores. MATERIALS AND METHODS Twelve tissue micro arrays (TMAs) were manufactured containing two cores from each of 156 breast cancer patients (number of cancer cores = 312). Each core was reviewed in order to ensure the presence of invasive carcinoma and to exclude analysis of noninvasive DCIS components. Immunostaining (IHC) was performed with Dako’s ready-to-use HER2 test, and TMA specimen slides were scanned by a Pannoramic Midi whole slide scanner (3DHISTECH, Budapest, Hungary). One batch-analysis of the HER2-CONNECT algorithm was run including all the samples. The HER2-CONNECT algorithm evaluates the immunoreaction of HER2 based on cell membrane connectivity by using a dynamic measure (between 0 and 1) of the size of HER2-stained membrane fragments. The automatically read connectivity translates into the classic diagnostic score for HER2 protein expression (0, 1+, 2+ or 3+) in agreement with the ASCO/CAP guidelines. The algorithm was compared to manual reading of the HER2 protein expression (IHC), and the inconclusive (2+) protein expression samples were compared to manual reading of the HER2 gene expression using fluorescence in situ hybridization (FISH). FISH was performed with ZytoLight ® SPEC HER2/CEN 17 Dual Color Probe (Zytovision, Bremerhaven, Germany). Manual reading demonstrated a sensitivity of 95.5% and a specificity of 81.0% with 17.0% inconclusive samples. Using the digital HER2-CONNECT test, both sensitivity and specificity increased (97.7% and 98.1%, respectively), and only 1.9% of the cases were deemed inconclusive. Total agreement when comparing HER2-CONNECT reading with manual IHC reading supplemented by FISH in borderline (2+) cases was 98.1%. Using the HER2-CONNECT digital image analysis algorithm based on IHC detection of HER2 protein expression on tumor cell membranes, less than 2% of the cases were inconclusive. Application of the automated image analysis HER2-CONNECT algorithm to assess the HER2 protein expression in breast cancer instead of manual reading decreases the need for supplementary FISH testing by almost 90%. This entails a significant impact on cost reduction and turn-around-time. HER2-CONNECT will make the assessment of HER2 fully automatic, fast and objective and implementing such automation will ensure that more cancer patients benefit from the correct treatment. In the end this will increase the survival rate of breast cancer patients. RESULTS CONCLUSION