NUMERO: #20240102

Título: MACHINE LEARNING-BASED APPROACH TO INCIPIENT ORAL SQUAMOUS CELL CARCINOMA CLASSIFICATION

Nome do Apresentador: Cristina SALDIVIA-SIRACUSA

Categoria do Trabalho: Painel de pesquisa científica (PPC)

Área Temática: Estomatologia

Resumo: Objective: To assess and select the best-performing machine learning (ML) model from five candidates using clinical descriptors to achieve classification between in-situ and microinvasive oral squamous cell carcinoma (OSCC).
Study Design: This retrospective cross-sectional study used a dataset of 107 clinical records from patients diagnosed with incipient OSCC in 6 Latin American oral medicine services, including 70 in-situ and 37 microinvasive OSCC cases. A 5-fold cross-validation was used to train 5 classic ML algorithms with 30 data inputs. Evaluation metrics such as accuracy, precision, sensitivity and F1 were used to quantify the models' performance.
Results: Overall, the Gradient Boosting Classifier (GBC) model showed the best performance in classification when using a learning rate of 0.1, 100 estimators, and a 3-maximum depth, obtaining an accuracy of 0.55, precision of 0.65, sensitivity of 0.69, and F1 score of 0.66.
Conclusion: These preliminary ML approaches have shown promising performance as tools for automatic classification. In this study, GBC attained the best results distinguishing incipient OSCC, a diagnosis that currently poses a significant challenge. Further studies will be conducted to refine these methods and enhance their effectiveness in clinical practice.
This project was supported by FAPESP (grants number 2022/13069-8 and 2021/14585-7) and CNPq (307604/2023-3).

Autor 1:  Cristina SALDIVIA-SIRACUSA

E-mail 1:  [email protected]

Autor 2:   Diego Armando CARDONA CARDENAS

E-mail 2:  [email protected]

Autor 3 :  Caique Mariano PEDROSO

E-mail 3:  [email protected]

Autor 4:  Anna Luiza DAMACENO ARAÚJO

E-mail 4:  [email protected]

Autor 5:  Marcio Ajudarte LOPES

E-mail 5:  [email protected]

Autor 6:  Matheus CARDOSO MORAES

E-mail 6:  [email protected]

Autor 7:  Alan Roger SANTOS-SILVA

E-mail 7:  [email protected]