NUMERO: #20260585

Título: Artificial Intelligence in Histopathological Diagnosis of Head and Neck Squamous Cell Carcinoma: A Scoping Review

Nome do Apresentador: René Andrés Martínez-Flores

Categoria do Trabalho: Pôster - Investigação Original (IO)

Área Temática: Oncologia

Resumo: OBJECTIVE: To characterize tasks, AI paradigms, training strategies, and performance metrics in histopathological image analysis of head and neck squamous cell carcinoma (HNSCC).
STUDY DESIGN: Scoping review following Joanna Briggs Institute and PRISMA-ScR guidelines. PubMed, Scopus, and Web of Science were searched (May 2025). Primary studies using human H&E images, AI models, quantitative metrics, and histopathological validation were included.
RESULTS: Of 3,208 records, 40 studies were included. Most focused on oral cavity tumors (95%) and used deep learning (82.5%). Classification predominated (82.5%), mainly tumor versus non-tumor (81.8%). Segmentation (10%) and detection (7.5%) were less frequent. Performance ranged from 85.8%-100% accuracy, F1-score 0.84-1, AUC-ROC 0.90-1, and Dice 0.82-0.94. External validation was limited.
CONCLUSION: AI models show strong performance, mainly in classification. However, methodological heterogeneity, limited external validation, and scarce annotated data hinder clinical translation. Future studies should target prognostic histopathological features.

Autor 1:  René Andrés Martínez-Flores

E-mail 1:  [email protected]

Autor 2:   Valentina Aránguiz-Calderón

E-mail 2:  [email protected]

Autor 3 :  Meyby Vega-Pillampel

E-mail 3:  [email protected]

Autor 4:  Felipe Cáceres-Villanueva

E-mail 4:  [email protected]

Autor 5:  Constanza Jiménez-Lizama

E-mail 5:  [email protected]

Autor 6:  Juan Pablo Vásconez-Hurtado

E-mail 6:  [email protected]

Autor 7:  Sven Niklander-Ebensperger

E-mail 7:  [email protected]