ECP 2023 Abstracts

S188 Virchows Archiv (2023) 483 (Suppl 1):S1–S391 13 There was no significant difference in the sensitivity and specificity of the AI model compared to the pathologists in detection of NM or INN. Conclusion: To conclude, the deep learning AI model showed excel- lent accuracy in detection of NM. Furthermore, the algorithm showed pathologist-level performance in detection of INN. Importantly, the AI model showed potential in differentiating between these two entities on routine stained WSIs. Use of such a tool could possibly reduce the need of immunohistochemistry and assist the pathologist in assess- ment of SNBs. Funding: The study was financed by grants from the Swedish state under the agreement between the Swedish government and the county councils, the ALF-agreement (Grant ALFGBG-973455) and by the Finnish Dermatopathologist Foundation. E-PS-05-034 Cutaneous sarcoma, still a challenge in the pathological diagnosis R. Niculescu*, I.G. Cocuz, A.H. Sabau, M.C. Popelea, O.S. Cotoi *Mures County Clinical Hospital, Romania Background & objectives: Cutaneous tumours with uncertain differ- entiation represent a group of neoplasms that due to their rarity lead to considerable diagnostic challenges. Another obstacle in the diagnosis of these tumours is the overlapping of morphological aspects with some common neoplasms (epithelial and melanocytic tumours). Methods: This case series report documents four types of cutaneous sarcoma arising in the dermis and subcutis: epitheloid, histiocityc, pleomorphic and with clear cell, diagnosed based on the histopatho- logical and immunohistochemical findings in our Pathology department between 2020 and 2022, all cases were with the clinical diagnosis of epidermoid carcinoma. All the samples were processed and interpreted strictly following the protocols. Results: At the microscopic analysis on H&E stain of epithelioid, clear cell and pleomorphic sarcomas, epithelioid and spindle cells were observed, but each with characteristic aspects such as: areas of necro- sis for the epithelioid sarcoma, multinucleated cells for pleomorphic sarcoma. The histiocytic sarcoma presented large and medium-sized cells, with large, irregular nuclei, with variable cytoplasm. Character- istic for all the lesions were marked cyto-nuclear atypia, a large num- ber of mitoses with a Ki67 between 30-50%. Immunohistochemistry: S100, HMB45,SOX10,Vimentin was positive in clear cell sarcoma, CD4,S100,CD68 positive in histiocytic sarcoma, S100,EMA,CKAE1/ AE3,Vimentin positive in epithelioid sarcoma, Vimentin,CD10,SMA positive in pleomorphic sarcoma. In all cases local recurrences were recorded after less than two years from the initial diagnosis. Conclusion: Recognition of morphologic clues on HE and immuno- histochemistry play a key role in the diagnosis of this types of tumours. Vimentin was the one with strong positivity in all four cases. Also for an accurate diagnosis it is required a close correlation between the clinical picture and the morphologic, immunohistochemical and molecular aspect of the tumour. One of the aspects, also observed in our case, which gives these tumours a high degree of severity is the risk of recurrence. E-PS-05-035 Automatic detection of common malignant lesions in whole slides dermatology histopathological images A. Nivaggioli*, N. Pozin, M. Clavel, M. Sockeel, S. Mazellier, E. Lanteri, M. Figuccio, C. Lefebvre, S. Kammerer-Jacquet, M. Bat- tistella, V. Rouleau, S. Sockeel *Primaa, France Background & objectives: The vast majority of malignant lesions in dermatopathology are either Melanoma, Basal-Cell Carcinoma (BCC), or Squamous-Cell Carcinoma (SCC). Automatically detecting such lesions, which further allows for automated depths and margins measurement, could aid pathologists make faster and more precise diagnosis. Methods: We gathered and labelled 1795 WSI containing both malig- nant and benign lesions and healthy tissue. We use a deep feature learning-based method to train a classifier on patches at zoom x20, to identify Melanoma, BCC, or SCC. Computer vision based methods then allow automatic measuring of lesion depth and margins. Results: Our validation dataset contains 392 slides with one of the common malignant lesions, and 500 slides of healthy tissue or benign lesion. Our algorithm is able to detect malignant lesions with an F1 score of 0.892. If a malignant lesion is detected, we can determine the lesion type with a balanced accuracy of 0.965. At the patch level, we are able to detect a lesion with a Precision-Recall AUC of 0.946. Thank to our deep feature learning based approach, models are trained in less than an hour on a single GeForce RTX 2080 Ti GPU. Conclusion: To our knowledge, we propose the first algorithm able to locate mentioned malignant lesions in whole slide histopathological images, while measuring depth and margins automatically. On-going developments should improve results further. Detection of benign lesions and classification of subtypes of malignant lesions are also being explored. E-PS-05-036 Melanoma arising in a giant congenital naevus in a 16-month-old infant: a case report W. Ouahioune*, N. Moulai, M. Guermi, Y. Taleb *Faculty Of Medicine Of Blida, Algeria Background & objectives: Melanoma arising in a giant congenital naevus is highly aggressive and potentially fatal tumour. Children and adults can be affected, but most melanomas appear during the first 5 years of life. The scalp and back are most commonly affected sites. Methods: We report the case of a 16-month-old infant who con- sults for a 03 cm nodule developed on a scalp congenital naevus. It was an unresectable naevus of 18 cm in size. This nodule was associated with a synchronous homolateral cervical adenopathy of 02 cm in size. Results: We received a skin resection piece measuring 12x10x1,4 cm, centred by a non-ulcerated blackish nodule of 3x2,5x1,6 cm. Histological examination showed a well limited dermo-hypodermic nodule of a melanocytic origin, which stood out from pre-existing con- genital naevus. Foci of tumour necrosis were found. The mitoses were numerous and atypia were marked. The patient died 06 months after diagnosis. Conclusion: The diagnosis of melanoma arising in giant congenital naevus must be made cautiously. It is imperative to eliminate a pro- liferative nodule in congenital naevus before confirming malignancy. E-PS-05-037 Blastic plasmacytoid dendritic cell neoplasm on the skin: case report of a rare disease D. Sá*, A. Coelho, J.R. Vizcaíno, F.E. Costa, F.S. Vieira, G. Carrola *Centro Hospitalar Universitário de Santo António, Portugal Background & objectives: Blastic plasmacytoid dendritic cell neo- plasm (BPDCN) is a rare neoplasm in which the proliferating cells derive from precursors of plasmacytoid dendritic cells. BPDCN char- acteristically has a predilection for cutaneous involvement and typi- cally occurs in elderly patients. Methods: We herein report a case of an 81-year-old man with hyper- tension and dyslipidaemia that resorted to a general surgery consulta- tion with a 4-month history of erythematous skin papules on the chest, back and head. He was submitted to a skin biopsy to characterize these lesions and we received a skin ellipse measuring 1,6x1x0,8 cm present- ing a violaceous epidermic surface.