Using machine learning to classify subtypes of rhabdomyosarcoma

Jingyan Zhang

Document Type

Article

Publication Date

8-9-2024

Keywords

JGM

JAX Location

In: Student Reports, Summer 2024, The Jackson Laboratory

Abstract

Rhabdomyosarcoma (RMS) is a soft-tissue pediatric cancer, accounting for 5% of all solid tumor cases. The prognosis and treatment outcomes vary significantly depending on the subtype of RMS and its mutational profile – however, disparities exist in accessing genetic screening and expert pathology services, often influenced by geographical and socioeconomic factors. Our objective is to leverage neural network models to categorize RMS samples solely based on histopathology images obtained from biopsies. 386 whole-slide images (WSI) of RMS biopsies from various institutions in the United States were consolidated to test binary classification models and explore model behavior on the images. The results show that a native model – prior developed in the Chuang Lab – exhibits high accuracy in classification by relying on characteristic histological morphologies of RMS subtypes.

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

COinS