ECP 2023 Abstracts

S80 Virchows Archiv (2023) 483 (Suppl 1):S1–S391 13 AC: 96.8%), RM (SE: 100%, SP: 100%, AC: 100%), and NAM (SE: 28.6%, SP: 100%, AC: 81.5%) if PRAME was scored in the whole lesions (including N component for NAM and the dermal benign-looking component for NRLF/SNPMF). By contrast, if the PRAME was scored only in the M component for NAM and the atypical superficial component in NRLF/SNPMF, the diagnostic performance of PRAME remarkably improved (SE: 100%, SP: 100%, AC: 100%). Conclusion: Our results indicate that PRAME is a useful tool for the appropriate diagnosis of NRLF/SNPMF and its M mimickers (MERF, RM, and NAM). Furthermore, we found that in NAM, the accurate PRAME assessment restricted only to the histologically suspicious/ atypical components (M component for NAM and atypical superficial component in NRLF/SNPMF) is crucial to obtain satisfactory diagnos- tic performance, especially for SE. PS-08-007 A decision support system for the detection of cutaneous fungal infections using artificial intelligence N. Rappoport Levin, G. Goldinger*, A. Debby, J. Gildenblat, C. Sagiv, A. Barzilai *Sheba Medical Center, Israel Background & objectives: Fungal infections of the skin are very com- mon. We use the DeePathology STUDIO to create a proof of concept for an AI based decision support system for the diagnosis of fungal infection of the skin. Methods: We used skin biopsies of patients that were diagnosed for cuta- neous fungal infection at the Sheba Medical Center between 2014-2023. Samples were stained with PAS stain and digitized using the Philips IntelliSite scanner. With the DeePathology STUDIO fungal elements were annotated and used for the creation of the AI solution. The study was approved by the local IRB committee. Results: Two WSI (whole slide images) were used to train the AI solu- tion that detects tiles that contain fungal structures. We used 162 annota- tions of tiles that contain fungal elements and 74 tiles which were nega- tive for fungi. All tiles were annotated in association with the stratum corneum. For validation, the AI solution was deployed on 146 regions of interest in five slides. Accuracy metrics were calculated using the Validation Tool of the DeePathology STUDIO: Precision=0.8391, Recall=0.8441 and F1 score of 0.841. The AI solution was trained to be more tolerant to false positives than to false negatives as the main aim is to provide a screening tool. Conclusion: In this study, we provide the first evidence of an AI decision support algorithm for the detection of cutaneous fungal infection. Further work is planned for validating the algorithm on more cases and extend this work by automating the epidermal layer segmentation. PS-08-008 Comparison of the clinicopathological characteristics of NRAS- and BRAF-mutated primary cutaneous melanoma: a 7-year review from a tertiary referral centre P. Reddy*, D. Martin, A. Fabre, S.V.H.G. Strategic Skin Cancer Network *Histopathology Department, St Vincent’s University Hospital, Ireland Background & objectives: NRAS and BRAF mutations in primary cutaneous melanoma (PCM) show distinct characteristics such as thicker tumours and older age in NRAS-mutated cases. Our 7-year review compares the clinicopathological features of patients with PCM who underwent genetic testing at our institution. Methods: Our study included 123 patients who underwent genetic testing for PCM between January 2016 and December 2022. We ana- lysed clinical and pathological parameters including age, Clark level, Breslow thickness, ulceration, mitotic rate, and origin from a naevus. Categorical variables were analysed with a chi-square test and continu- ous variables with a Mann Whitney U test. Results: In our NRAS PCMs, NRASQ61X was the most common mutation (57/61) and V600E was the most common in the BRAF PCMs (52/62). NRAS PCMs were frequently observed on the leg (25/61) and had a nodular subtype (34/61). BRAF PCMs were more evenly distrib- uted between the leg (19/62) and the back (17/62) and both nodular (30/62) and superficial spreading (27/62) subtypes were commonly observed. Age, Breslow thickness, and mitotic rate were significantly associated with NRAS mutations (U=1048, p<0.05; U=150, p<0.05; U=183, p<0.05). Differences in Clark level and ulceration were not found to be significant. (p=0.054; p=0.23). BRAF mutated PCMs were more likely to originate from a naevus (17/62 vs 8/61, p=0.048). Conclusion: Our retrospective review is consistent with previous studies showing that NRAS PCMs are more commonly diagnosed in older patients, with greater Breslow thickness and higher mitotic rates. Additionally, we observed BRAF PCMs more commonly originated from naevi. These distinct clinicopathological characteristics may be associated with more aggressive tumour behaviour and as such can assist management decisions taken at multidisciplinary team meetings. PS-08-009 Lower frequency of cutaneous melanoma BRAF mutations in an Irish population S. Staunton*, R. Cummins, C. Ryan, M. Redmond *Beaumont Hospital, Ireland Background & objectives: Activating BRAF oncogene mutations in melanoma carry a worse prognosis. Management of advanced BRAF mutation-positive melanoma is changing rapidly. While BRAF muta- tions occur in 50% of advanced melanomas worldwide, there is emerg- ing evidence of lower frequency in an Irish population. Methods: A retrospective audit using the Laboratory Information Sys- tem of melanoma cases from 2018-2022 on which BRAF testing via next generation sequencing was performed using data from a single tertiary referral centre for melanoma mutation testing in Ireland. Both in-house and referred material was analysed. Results: Data from 508 cases was analysed. Mean age was 66.9 years. The most common specimen types were skin (44.3%, n=225), lymph nodes (17.7%, n=90), and brain (10.6%, n=54). BRAF mutations were present in 27.1% (n=138) of which 85.5% (n=118) were V600 mutations. NRAS and KIT mutations were present in 41.7% (n=212) and 23.6% (n=120) respectively, though not all cases were tested. Younger age was a significant predictor of BRAF mutation (p<0.00001) while gender was not (p=0.49). Superficial spreading melanoma was more likely than other subtypes to harbour a BRAF mutation (44.7%, p=0.001). BRAF muta- tion-positive melanomas were less likely to display ulceration (p=0.033) but more likely to display lymphovascular invasion (p=0.029). Conclusion: NICE states that standard treatment for advanced/meta- static melanoma is either immunotherapies or targeted therapy (alone or in combination). Patients with BRAF V600 mutation-positive mela- noma are likely to be offered a targeted therapy at some point in the treatment pathway, particularly if resistance occurs. As the prevalence of BRAF mutations in our centre is lower than that seen internationally, less treatment options may be available to our patient cohort, poten- tially leading to worse outcomes. PS-08-010 Brightfield multiplex immunohistochemistry assay for PD-L1 evaluation in challenging melanoma samples F. Ugolini*, S. Simi, L. Tinunin, G. Baroni, D. Massi *Section of Anatomic Pathology, Department of Health Sciences, University of Florence, Italy