ECP 2022 Abstract Book

Virchows Archiv (2022) 481 (Suppl 1):S1–S364 13 (kappa values 0.98-1.0) and ICC values of 0.96-0.98, 0.91-0.95, and 0.96-0.98 in measuring the HER2 signals, CEP17 signals, and HER2/CEP17 ratio, by D-DISH respectively. Interobserver vari- ability between the four pathologists showed perfect agreement for genomic heterogeneity and group categorization, and moderate agreement for polysomy by D-DISH. Conclusion: Our study successfully validated the D-DISH test using the updated version of HER2 Dual ISH DNA Probe Cocktail Assay (Ventana Medical Systems, Inc., Tucson, AZ) as a substitute for FISH assay in accurately predicting the HER2 gene status with significant interobserver reproducibility. We conclude that this D-DISH test may be introduced in rouine diagnostic services as a reflex test for detecting HER2 gene status. OFP-10-006 Image-based identification of HER2 status in H&E-stained breast cancer slides O. Greenberg, A. Zubkov, L. Trejo, D. Nataf, I. Hayun, A. Avi- noam, I. Gazy, N. Paz-Yaacov, D. Hershkovitz* *Pathology Department, Tel Aviv Sourasky Medical Center, Tel Aviv, Israel Background & objectives: Determination of HER2 status is criti- cal for prognosis and guiding treatment decisions in breast cancer patients. Here we present a method to detect HER2 expression directly from routinely prepared diagnostic H&E slide images, using an image-based deep learning approach. Methods: A HER2-classifier was generated using 251 H&E- stained slide images and their relevant IHC and FISH status from the pathology department at Sourasky Medical Center. Convolu- tional neural network (CNN) analysis was performed on on-the-fly augmented images. Multiple instance learning (MIL) algorithms and ranking training schemes were applied to create the categori- cal HER2-classifier (positive/negative), powered by Imagene-AI. Results: The HER2-classifier was evaluated on a validation set including H&E images only of 104 retrospective cases. The model performance values were 87.5% sensitivity, 85.9% specificity, 86.14% accuracy and 0.89 AUC. To further evaluate the AI-solution additional 245 cases were analysed. In this set, a high proportion (n=9/20; 45%) of false callings was observed in samples with HER2 IHC=3. The IHC slides of samples with score 3 in the set were re-evaluated by two pathologists. While three of the 9 false-negative cases (33%) status was changed to 2, by at least one pathologist, there were no changes in the true-positive group (n=9). Conclusion: Implementation of Image-based solution to routine pathology workflow can support fast, cost-effective and stand- ardized method for biomarker detection. We evaluated the use of an AI-model to analyse HER2 status compared to conventional IHC and FISH methods. Analysis of 349 cases resulted in 85.3% accuracy. IHC, manually analysed by pathologists, is a subjective method with both intra- and inter-observer discordance’s reported. An AI-solution can support the routine workflow flagging cases where re-evaluation can support the pathologist analysis of dif- ficult cases. OFP-10-008 Application of molecular imaging as potential prognostic bio- marker for triple-negative breast cancer (TNBC) C. Villa*, P. Rainone, M. Cadamuro, S. Valtorta, S. Todde, R.M. Moresco, M. Lavitrano *Department of Medicine and Surgery, University of Milano- Bicocca, Monza, Italy Background & objectives: The administration of the antihyper- tensive syrosingopine in combination with metformin provided a synergistic effect against different tumours, including breast cancer. We aimed to investigate the response to metformin and/or syrosingopine, evaluating the [18F]FDG and [18F]FLT as potential early prognostic biomarkers. Methods: Balb/c female mice were inoculated subcutaneously with murine TNBC cells (4T1) and divided into six treatment groups: vehicle, cisplatinum, metformin, syrosingopine or cisplatinum plus metformin and metformin plus syrosingopine. The response to treatment was monitored by caliper measuring, for tumour volume and PET studies. Molecular biomarkers of glucose metabolism and tumour invasiveness were analysed by means of immunohistochemistry and q-RT-PCR. Results: A significant tumour growth inhibition (%TGI) has been observed only after metformin plus syrosingopine administration, confirming a synergistic effect after ten days of treatment. PET analyses revealed a significant reduction of [18F]FLT tumour uptake in cisplatinum plus metformin (*p<0.05) and metformin plus syrosingopine (*p<0.001) treated groups, whereas [18F]FDG uptake increased in all experimental conditions. Molecular analyses performed by both immunohistochemistry and q-RT-PCR demonstrated a significant decrease of the lactate transporter MCT4 levels in mice treated with cisplatinum and metformin plus syrosingopine (*p<0.05) as well as low levels of the EMT biomarker Snail only in the group treated with the combination of metformin plus syrosingopine (*p<0.05). Conclusion: Our data suggested that the combined administration of metformin plus syrosingopine is able to modulate the glucose metabolism and inhibit the tumour invasiveness and growth of cancer cells. Moreover, the use of [18F]FLT radiotracer may represent a potential biomarker for the early response to treatment of TNBC. OFP-10-009 Challenges of breast NEN: NEN G3 and genetic analysis Y. Liu*, M. Zhao *The Fourth Hospital, China Background & objectives: The Breast NEN(Br-NEN)was graded according to Nottingham standard, resulting in confusion between Br-NENG3 andNETG3. It was unclear whether therapeutic targets associated with gastroenteropancreatic NENare applicable in Br- NEN. This study aims to illustrate these issues. Methods: The 50 cases were re-diagnosed according to the 5thWHO classification criteria. The gastroenteropancreatic NEN characteristic immunohistological staining (SSTR1-5, DLL3, ISL- 1, INSM-1) were performed, and DNA was isolated from the pri- mary tumour for exome sequencing. In this study, NET G3 was set as Nottingham grade 3, excluding small cell carcinoma and large cell neuroendocrine carcinoma. Results: 1)Compared with G3 and NEC,the MYC pathway was enriched in G3, and the TP53 pathway was more common in the latter (P=0.036). 2)The G3 had more ZNF703 and FGFR1 genetic mutations than NEC, and was more likely to exhibit the genetic characteristics of Luminal B. 3)The SSTR2 and INSM1 positive expression was more common in the G1-2, P=0.006 vs 0.012, and showed decreasing trend in NEC compared with G3. No positive expression of DLL3 was found in G3except NEC. 4)The metastasis occurred in 3 cases (6% G3,11% NEC, 4% G1-2) and the breast cancer-related deaths occurred in 4 cases(12% S41