Morphological analysis of cutaneous T-cell lymphoma unaging using attention-based deep learning

Authors

Amaris Huang

Document Type

Article

Publication Date

Summer 2022

Keywords

JGM

JAX Location

In: Student Reports, Summer 2022, The Jackson Laboratory

Abstract

Cutaneous T-Cell Lymphoma (CTCL) is a rare cancer that develops in the lymph tissues, in which white blood cells, T-Lymphocytes (T-Cells), develop abnormalities. The most common form of CTCL is Mycosis Fungoides (MF), which is a generally slow progressing disease, and if monitored well, many patients have a normal life expectancy; however, patients with advanced stage MF may undergo large cell transformation (LCT), resulting in patients resisting multiple forms of therapy and having poorer prognosis. This study proposes using two attention-based deep-learning techniques to develop computational tools that would be able to analyze CTCL tissue biopsies and help predict LCT risk. Through adding spatial attention mechanisms to Convolutional Neural Networks (CNN) and utilizing Vision Transformers (ViT), the models were able to provide statistically significant improvements in distinguishing between indolent and precursor tissue tiles.

Please contact the Joan Staats Library for information regarding this document.

COinS