Lung Cancer segmentation and 3D visualization

Principal Investigator

Name
Hao Zhou

Degrees
Ph.D.

Institution
Dyad Medical, INC.

Position Title
Machine Learning Scientist

Email
hao@dyadmed.com

About this CDAS Project

Study
NLST (Learn more about this study)

Project ID
NLST-530

Initial CDAS Request Approval
Jun 21, 2019

Title
Lung Cancer segmentation and 3D visualization

Summary
We are trying to apply up-to-date deep learning techniques such as Mask R-CNN, 3-D Unet to our Lung Cancer CT images segmentation research. Our goal not only focuses on 2-D semantic segmentation, but also on 3-D voxel-wise segmentation and visualization. Lung Cancer classification and detection is a hard topic in machine learning research area, lack of voxel-wise nodule annotation data is one of the main constrains. We are requesting this exciting data set to help us train and improve our model. The high accuracy product will be potentially used by hospital as a supplementary tool of doctor's diagnosis.

Aims

1. Getting a good 3-D deep learning model for lung cancer semantic segmentation of CT images.
2. 3-D visualization and VR approach to help doctors evaluate the images more intuitively.
3. Using the Deep Learning model for more medical image classification/segmentation topics.

Collaborators

Ronny Shalev, Dyad Medical, INC.