03102-A: Can Deep Learning (CNN) Detect Minimal Residual Disease (MRD) in Dogs Treated for Lymphoma?

Abstract

Lymphoma is a common and deadly form of cancer that occurs in the lymph nodes of dogs. Determining response to chemotherapy is an important goal of cancer treatment and critical to a patient’s long-term survivability. An essential part of this monitoring process includes sampling the affected lymph nodes with a needle and examination of the cells microscopically with cytologic exam. Additionally, these samples are used with molecular techniques to determine when the patient’s cancer is no longer detectable (minimal residual disease, MRD) and when the disease reoccurs. The associated costs for serial monitoring with these gold-standard molecular diagnostics can be prohibitive, leading to less frequent monitoring for disease relapse and later detection of disease remission. The use of computer image analysis (artificial intelligence, AI) in medicine is expanding exponentially, often with the goal of improving the detection of cancer. Convoluted neural networks (CNN) are one type of AI well adapted for medical image analysis.  

This study aims to demonstrate that the MRD status of dogs post-treatment for lymphoma can be detected by applying a CNN to digital cytology lymph node images. The immediate expected impacts of this study include earlier detection of disease remission, reducing the amount of chemotherapy required for the patient, improved post-treatment monitoring, more affordable monitoring, and ultimately longer patient survivability. The longer-term impact of this work will lead to the development of online decision support tools using these methods for the initial cytologic detection and monitoring of lymphoma and other types of neoplasms. 

Publication(s)