Development of a neural network model to predict patients' clinical outcome in lung cancer
Principal Investigator
Name
Jean Yoo
Degrees
Pharm.D.
Institution
Lunit
Position Title
Researcher
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-432
Initial CDAS Request Approval
Jul 31, 2018
Title
Development of a neural network model to predict patients' clinical outcome in lung cancer
Summary
Although digital pathology has been actively adapted recently, computational support in cancer diagnosis still remains limited to measuring specifically stained compartments due to absence of robust computer algorithm to detect various features on H&E stained slides. As tumor micro-environment rises as a key to comprehend immune dynamics in various cancers, histopathology may bring rich information about what happens around the tumor before/during treatment then this will lead to precision treatment based on each individual's immune profile. However, for a human, understanding the whole slide with eyes is regarded impossible because of abundant information including mixed thousands of normal/tumor cells and lymphocytes while a convolutional neural network can do. In this project, using lung pathological images from NLST database and clinical data, we will develop accurate neural network model which will assess patients' prognosis and clinical outcome based upon features from H&E stained tumor samples.
Aims
1) Detection of compartments including tumor epithelium, lymphocytes, and tumor stroma using a convolutional neural network
2) Development of a prognostic machine learning model in lung cancer
3) Exploring the potential of imaging biomarkers
Collaborators
All members at the R&D center of Lunit