NUMERO: #20220278

Título: Deep Learning for Automatic Segmentation of Clinical Photographs of Oral Premalignant and Malignant Lesions

Nome do Apresentador: Anna Luiza Damaceno ARAUJO

Categoria do Trabalho: Apresentação Oral de Pôster de Pesquisa Científica (PPC)

Área Temática: Estomatologia

Resumo: Objective: To explore the usage of Deep Learning (DL) induced models for automatic segmentation of premalignant and incipient malignant lesions on photographic images.
Study Design: A dataset of 308 clinical images from three institutions was used to design, train and evaluate DL-based models. For each image, ground-truth annotation was performed by 3 experts and combined via union of the labelled areas, thus minimizing false negatives. The dataset was split into two subsets, with 246 training and 62 test images, and 10-fold cross-validation was applied to the first subset. The experimental results were evaluated using mean pixel-wise Intersection Over Union (IoU).
Preliminary Results: The best performing model was a U-Net architecture with a 224x224 input. The downstack section of the U-Net was a VGG16 CNN pre-trained with the ImageNet dataset, fine-tuned with the training subset. The training used random horizontal and vertical flips as data augmentation. A mean IoU of 0.675 (±0.030 std) and mean accuracy of 0.865 (±0.020 std) were obtained.
Conclusion: These preliminary results demonstrate the feasibility of DL-powered models for automatic segmentation of premalignant and incipient malignant lesions. The induced model can be a reliable, fast and non-invasive screening tool for cancer detection.

Autor 1:  Anna Luiza Damaceno ARAUJO

E-mail 1:  [email protected]

Autor 2:   Eduardo Santos Carlos DE SOUZA

E-mail 2:  [email protected]

Autor 3 :  Isabel Schausltz Pereira FAUSTINO

E-mail 3:  [email protected]

Autor 4:  Cristina Saldivia SIRACUSA

E-mail 4:  [email protected]

Autor 5:  Marcio Ajudarte LOPES

E-mail 5:  [email protected]

Autor 6:  André Carlos Ponce de Leon Ferreira de CARVALHO

E-mail 6:  [email protected]

Autor 7:  Alan Roger Santos-Silva

E-mail 7:  [email protected]