Technology Development of Deep Learning based Chest X-Ray Reading
Principal Investigator
Name
Jing Xiao
Degrees
Ph.D
Institution
PingAn Technology(Shenzhen) Company, Ltd.
Position Title
The General Manager of Big Data Platform Division
Email
About this CDAS Project
Study
PLCO
(Learn more about this study)
Project ID
PLCO-283
Initial CDAS Request Approval
Jul 5, 2017
Title
Technology Development of Deep Learning based Chest X-Ray Reading
Summary
Chest X-ray remains one of the major clinical diagnosis modalities for various lung diseases, such as lung nodule, interstitial lung diseases, hypoinflation, pneumothorax, cardiomegaly, etc. More importantly, X-ray imaging is often the first imaging diagnosis tool which applies to large population for preventative care. The collaborative research in more precision X-ray diagnosis could help diagnosis of diseases in early stages, or give reliable assessment on the risk of disease, based on which therapies or prevention measures could be targeted. In this way, the effect of medical measures could be maximized with minimum resources. It could provide a more accurate approach for disease prevention. Firstly, it could identify the individuals that have higher risk of disease and require prevention measures. Secondly, it could help clinicians to select appropriate prevention methods based on the person’s condition. In addition to preventative care, chest X-ray plays a more crucial role in precision treatment. It provides relatively accurate information and results for a variety of conditions with respect to treatment decision, treatment planning, and treatment follow-up. For example, X-ray imaging provides a road map for certain surgeries, and doctors regularly use CT scans to monitor the patient response to treatments and detect tumor recurrence or metastasis. The research thrust in comprehensive chest X-ray image understanding can potentially grow into a widely accepted lung imaging prescreening tool.
Aims
To improve and develop scalable deep learning empowered technical solutions to detect or tag major diseases from chest X-ray images in high precision, via mining a large collection of patient data in both imaging and text domains (~100K patient cases)
Collaborators
Le Lu, National Institutes of Health Clinical Center (NIHCC)