Early detection of lung cancer with deep neural nets
Principal Investigator
Name
weiyi xie
Degrees
M.S.C
Institution
JianPei technology, Ltd
Position Title
Head of Dev department
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-260
Initial CDAS Request Approval
Dec 13, 2016
Title
Early detection of lung cancer with deep neural nets
Summary
In China, the number of lung cancer sufferers is significantly growing due to the air pollution and a large number of regular smokers. Early stage detection of lung cancer holds the opportunity to improve long-term survival rates. The goal of this project is to develop a computer aided diagnosis system(CADe and CADx) that helps radiologists finding malignancy earlier and more accurate in lung CT.
We have built several models based on deep neural networks to detect visual clues for lung cancers from CT(low dose, normal, contrast). One of our model on pulmonary nodule detection task have been published on the result page of LUNA 2016 result page(https://luna16.grand-challenge.org/results/)including a short description regarding the approaches we tried.
The project isn't limited to work on CT scans only from Chinese patients. the project outcomes will be evaluated on a large-scale test on database from different sources, e.g. different devices, geometrical regions, race, genders ,age group etc.
We have built several models based on deep neural networks to detect visual clues for lung cancers from CT(low dose, normal, contrast). One of our model on pulmonary nodule detection task have been published on the result page of LUNA 2016 result page(https://luna16.grand-challenge.org/results/)including a short description regarding the approaches we tried.
The project isn't limited to work on CT scans only from Chinese patients. the project outcomes will be evaluated on a large-scale test on database from different sources, e.g. different devices, geometrical regions, race, genders ,age group etc.
Aims
1) early detection of lung diseases that prone to be cancer based on its visual clues in Chest CT scans (CADe), specially for ground glass opacity type.
2) the classification of detected lesions into predefined malignancy levels(CADx).
3) fully automated segmentation of lung tissues, e.g. bronchi , trachea, vessels,
Collaborators
None