ECP 2023 Abstracts

S236 Virchows Archiv (2023) 483 (Suppl 1):S1–S391 13 *Hospital del Mar, Spain Background & objectives: Stromal tumour infiltrating lymphocytes (sTIL) are linked to better outcomes in triple-negative and HER2-posi- tive breast cancer (TNBC and HER2+). Specific data on the prognostic impact of plasma cells or lymphocytes within sTIL, and different cell types is still under study. Methods: We validated a deep learning sTIL scoring model (smsTIL) based on the segmentation of various cell types in whole slide images (WSI). Focusing on HER2+ and TNBC, we assessed the concordance between sTIL visual scoring and smsTIL in 130 WSI. Furthermore, we analysed 175 WSI to correlate smsTIL and different cell types with clinical data and patient outcomes. Results: The segmentation model classified 596510 cells. We found a high correlation between sTIL and smsTIL (R=0.76,P<2.2e-16), and a statistically significant association between smsTIL and overall survival (OS) in TNBC and HER2+ (P=0.0021, P=0.041). In TNBC, this association was also observed for lymphocytes and plasma cells separately (P=0.00069, P=0.0048). Low tumour cells and necrosis, and high ‘other’ cells resulted in reduced OS (P=0.039, P=0.029, P=0.04). Univariate analysis showed that smsTIL value (P=0.016), age (P=0.017), TNM status (T2: P=0.052; T3: P=0.02; T4: P=0.00024), nodal status (P=0.0023) and metastases (P=0.00019) were associated with OS. Multivariate analyses showed that smsTIL value (P=0.009), nodal status (P=0.025) and metastasis (P=0.015) were independently associated with OS. Conclusion: We developed a deep learning algorithm to assess sTIL and confirmed that the values quantified by the pathologist and the algorithm are highly correlated, providing evidence that smsTIL can be implemented in daily routine. We demonstrated the prognostic value of smsTIL in HER2+ and TNBC, being an independent predictor of OS in TNBC. Moreover, lymphocytes and plasma cells individual quanti- fication play an individual prognostic role in TNBC, as well as tumour cells, necrosis and ‘other’ cells included in the stroma. E-PS-08-017 Unravelling the immune involvement of sentinel lymph nodes in melanoma metastasis by hyper-multiplexed imaging T. Haran*, I. Milo, S. Deis, E. Pikarsky, L. Keren *Department of Pathology and Department of Immunology and Can- cer Research, Institute for Medical Research Israel-Canada, Hadassah Medical School - Hebrew University, Jerusalem, Israel Background & objectives: We sought to refine our understand- ing of tumour-draining lymph node biology and architecture to identify characteristics predictive of future disease progression and unravel tumour-immune interactions in early phases of melanoma dissemination. Methods: Multiplexed Ion Beam Imaging-Time of Flight (MIBI-TOF) is a novel technology that utilizes metal-labelled antibodies coupled to mass spectrometry, allowing simultaneous localization of up to 40 proteins at subcellular resolution in situ. We used this technology to examine sentinel lymph nodes from a cohort of 80 melanoma patients and identify factors other than microscopic tumour foci associated with distant metastasis. Results: A 40-plex antibody panel was assembled to enable concurrent labelling and identification of melanoma cells and various immune cell subtypes and states. We first compared metastasis negative nodes from patients with subsequent disease progression (n = 20) with those with sustained remission (n = 28). Intriguingly, lymph nodes from patients with subsequent progression showed a distinct pro-inflammatory phe- notype, characterized by the presence of active secondary lymphoid follicles, an increased proportion of Ki67-positive lymphocytes, and decreased numbers of anti-inflammatory M2-polarized CD163-positive macrophages. Conversely, evidence of T-cell exhaustion was seen in lymph nodes bearing overt metastatic disease (n = 32), with clustering of PD-1-positive CD8-positive T-cells at the periphery of tumour foci. Conclusion: Tumour-negative sentinel lymph nodes from melanoma patients with subsequent disease progression can be distinguished from those from patients with prolonged remission. In patients who go on to develop distant metastasis, an apparently vigorous immune reaction is seen even when no tumour cells can be detected in the node. That this immune response is later overwhelmed suggests an overactive reaction followed by immune exhaustion. Identification and targeting of these patients for early immunotherapy may prevent this exhaustion and thereby disease progression. E-PS-08-018 A novel pathological image evaluation method for prostatectomy specimens using the homology profile method M. Ishigaki*, Y. Nomura, K. Nakane, M. Nishio, M. Watanabe, K. Uchida *Mie University Graduate School of Medicine, Department of Medical Science, Department of Tumour Pathology, Japan Background & objectives: The homology profile method (HPM) is a mathematical concept that analyses the connections between components. Its usefulness in detecting various cancers on pathology images has been reported. We investigated the usefulness of HPM in cancer detection using pros- tatectomy specimens. Methods: HE-stained specimens of prostatectomy performed at Mie University Hospital were scanned using Leica APERIO CS2 and then converted from svs format to jpg format. Images of 3000x3000pixel including cancerous and non-cancerous areas were arbitrarily extracted and divided into 100 images of 300x300pixel. The diagnostic ability of cancer was examined based on the Betti num- ber calculated from HPM for these images. Results: The Betti number in one dimension was defined as b1, and the maximum value was defined as b1max. The Mann-Whitney U test was performed for b1max in cancerous and non-cancerous areas, and there was a significant difference between the two groups. (p<0.05) The AUC from the ROC curve was 0.873 for the highest value and 0.402 for the lowest value. The b1max contained noise in certain regions that did not depend on path- ological morphology, and these were removed. After removal, the b1max was similarly significantly different between the two groups. (p<0.05) The AUC was 0.944 even for the lowest value, with a specificity of 0.980 and a sensitivity of 0.800. Conclusion: HPM was able to detect prostate cancer using currently used criteria. Unlike AI, the Betti number calculated by HPM is invari- ant and does not need to be relearned even if the diagnostic criteria are changed in the future. The HP method is also easy to use as a stand- alone system, thus avoiding the possibility of medical information leak- age through the network. In the future, HPM may be used together with AI as a simple, low-risk diagnostic aid. E-PS-08-019 Volumetric scanning achieved by collecting multiple images in Z plane for WSI image acquisition enables more accurate tumour microenvironment analysis in breast cancer cases J. Jain*, P. Perugupalli, R. Gupta, D. Anand, V. Singh, H. Mulcahndani *PRAMANA INC, India Background & objectives: Tumour microenvironment plays an impor- tant role in cancer progression and treatment response prediction in breast cancer. Here we present a study of 30 breast cancer cases where tumour microenvironment analysis was done in WSI with and without capturing Z stacks. Methods: Volumetric scanning of 30 Breast cancer cases was done using Pramana HT scanner with Z stacks (collecting multiple images in