Título: PAN-CANCER ANALYSIS OF FIBROBLASTS FOR PREDICTING THE PROGRESSION OF POTENTIALLY MALIGNANT ORAL DISORDERS
Nome do Apresentador: Luisa Ammirabile AUGUSTO
Categoria do Trabalho: Painel de pesquisa científica (PPC)
Área Temática: Patologia Oral
Resumo: Potentially malignant oral disorders (OPMDs) may progress to squamous cell carcinoma (SCC), a significant public health concern due to its incidence and prevalence. Evidence suggests that fibroblasts acquire distinct phenotypes, such as cancer-associated fibroblasts (CAFs), during malignant transformation, possibly preceding epithelial changes. Objective: This pilot study aims to evaluate whether an AI model trained to identify CAFs can predict which OPMD cases will progress to SCC. Study Design: We performed a pan-cancer analysis including 1,000 cases of colon cancer, lung adenocarcinoma, SCC, hepatocellular carcinoma, and triple-negative breast cancer. Fibroblasts were segmented using morphological masks generated by Cellpose and labeled with the VIM⁺, PAN-CK⁻, CD31⁻, CD45⁻ panel. An AI model was trained on 3 million cells to recognize CAFs based on morphological features extracted from geojson masks. The model was then applied to 12 OPMD cases (9 that progressed to SCC and 3 that did not) to assess whether fibroblast profiles alone could predict malignant transformation. Results: The model achieved over 90% accuracy in fibroblast identification and successfully distinguished the cases that progressed. Conclusion: Our pilot AI model effectively identified CAFs and their association with malignancy, supporting its potential use as a predictive tool for OPMD progression and early cancer detection.
Autor 1: Luisa Ammirabile AUGUSTO
E-mail 1: [email protected]
Autor 2: Khoa HUYNH
E-mail 2: [email protected]
Autor 3 : Luiz Fernando Ferraz da SILVA
E-mail 3: [email protected]
Autor 4: Jinze LIU
E-mail 4: [email protected]
Autor 5: Kevin M. BYRD
E-mail 5: [email protected]
Autor 6: Bruno Fernandes MATUCK
E-mail 6: [email protected]
Para baixar o aplicativo, escolha abaixo: