Skip to Main Content

An official website of the United States government

Deep learning of lung cancer images for segmentation and outcome prediction

Principal Investigator
Name
Olivier Gevaert
Degrees
Ph.D.
Institution
Stanford University
Position Title
Assistant Professor
Email
About this CDAS Project
Study
NLST (Learn more about this study)
Project ID
NLST-276
Initial CDAS Request Approval
Jan 31, 2017
Title
Deep learning of lung cancer images for segmentation and outcome prediction
Summary
Imaging genomics, also known as radiogenomics, is a burgeoning area of research that aims to link medical imaging with multi-omics molecular profiles of the same patients. Imaging genomics has shown its potential through its ability to predict clinical outcomes e.g. prognosis, and through predicting actionable molecular properties of tumors. In the past few years, deep learning in the form of Convolutional Neural Networks (CNNs) have emerged as the method of choice to analyze images outside of the medical domain by successfully solving complex tasks such as recognizing objects in images or labeling images. More recently, the first applications of deep learning have appeared in the medical domain focusing on diagnosis, segmentation and outcome prediction including applications in cancer medical imaging. Training CNNs on cancer images can be used to segment tumors, predict clinical outcome and predict molecular properties of cancer lesions.
Aims

To develop a deep learning framework for tumor segmentation from lung medical images in heterogeneous contexts
a. Develop algorithms based on convolution neural networks (CNN) to segment tumors
b. Evaluate the robustness of segmentations when applying algorithms in different contexts and centers
c. Evaluate segmentations by using them in a radiomics framework for supervised prediction of clinical characteristics of tumors

Collaborators

No additional collaborators beyond Stanford University.

Related Publications