NUMERO: #20240871

Título: DEEP CONVOLUTIONAL NEURAL NETWORK FOR ACCURATE CLASSIFICATION OF MYOFIBROBLASTIC LESIONS ON WHOLE-SLIDE IMAGES

Nome do Apresentador: Anna Luiza Damaceno ARAUJO

Categoria do Trabalho: Painel de pesquisa científica (PPC)

Área Temática: Patologia Oral

Resumo: Objective: This study aimed to investigate the utilization of a Deep Convolutional Neural Network for classifying myofibroblastic lesions into benign and malignant categories based on whole-slide images (WSI).
Methods: A ResNet50 model, pre-trained with weights from ImageNet, was trained using a cohort of 20 patients (11 benign and 9 malignant cases). Following annotation of tumor regions, the WSIs were fragmented into smaller patches (224x224 pixels). The images were then non-randomly divided into training (309,810 patches), validation (42,319 patches), and test (42,061 patches) subsets, maintaining an 80:10:10 ratio. The CNN training was caried out for 75 epochs utilizing a batch size of 4, the Adam optimizer, and a learning rate of 0.00001.
Results: ResNet50 achieved an accuracy of 97.1%, precision of 99.4%, sensitivity of 94.7%, specificity of 99.5%, F1 score of 97%, and AUC of 99%, indicating nearly flawless performance in distinguishing between benign and malignant tumors, despite the small sample size.
Conclusion: These results suggest that ResNet successfully learned and accurately classified myofibroblastic lesions, exhibiting high accuracy during training and robust generalization capabilities on unseen data.

Autor 1:  Anna Luiza Damaceno ARAUJO

E-mail 1:  [email protected]

Autor 2:   Daniela GIRALDO-ROLDÁN

E-mail 2: 

Autor 3 :  Giovanna Calabrese dos SANTOS

E-mail 3: 

Autor 4:  Matheus Cardoso MORAES

E-mail 4: 

Autor 5:  Luiz Paulo KOWALSKI

E-mail 5: 

Autor 6:  Alan Roger SANTOS-SILVA

E-mail 6: 

Autor 7:  Pablo Agustin VARGAS

E-mail 7: