The Fred Hutch Biostatistics Program hosts seminars featuring presentations by Hutch and outside scientists to share their latest developments and recent research. Each seminar includes an hour-long presentation and discussion during which speakers showcase their work and findings.
This seminar will be held on Zoom due to the COVID-19 pandemic.
Biostatistics Seminar Series:
“Neural Style Transfer For Melanoma Classification”
Skin cancer is primarily diagnosed visually via dermoscopic analysis, biopsy, and histopathological examination. However, automated image classification of skin lesions is deemed challenging due to the irregularity and variability of the lesions’ appearances. In this work, we propose an adaptation of the Neural Style Transfer as a novel image pre-processing step for skin lesion classification problems. We represent each dermoscopic image as the style image and transfer the lesion's texture onto a homogeneous content image. This captures and transfers each lesion's main variability onto a unified, localized region, effectively allowing us to extract latent, low-rank features from the image. Using the International Skin Imaging Collaboration (ISIC) database, we show that the classification performance based on the extracted tensor features using the style-transferred images significantly outperforms methods on the raw images by more than 10%, and is also competitive with well-
studied, pre-trained CNN models through transfer learning.