Deep learning using medical image for CAD research.
Principal Investigator
Name
LIANJUN Mei
Degrees
Master, Statistics
Institution
Terralogic Technologies
Position Title
R&D Director
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-224
Initial CDAS Request Approval
Jun 17, 2016
Title
Deep learning using medical image for CAD research.
Summary
In recent years, deep learning technology has been developed greatly and widely used in different industries, such speech recognition, image recognition, robotics, and so on. We want to use deep learning technologies or develop new algorithms to analyze medical images. Our ambition is to find effective tools to detect lung cancer by computer automatically.
During the past 30 years, the lung cancer patients increased 465% in China because of worse and worse air quality and other environmental problems. What’s worse, the health-care resources are distributed unevenly in China. More than 80% high-quality medical resources are centralized in tier-1 and tier-2 cities such as Shanghai, Beijing, Guangzhou, and Shenzhen. In the rural areas or West/Middle of China, there is a lack of well-trained doctors and right devices. Today, poor access and high fees are the two major challenges in China's healthcare system. For long-term, we want to use mobile technology embedded our CAD algorithms to touch every potential patient, and make better products and services much more accessible. Patients can use our app or something like that to diagnose themselves first. We want to democratize knowledge and information to the common people and help them to make better decisions.
Project steps:
Step 1: label data
Prepare lung cancer images, benign nodule patients, and other lung images.
Step 2: Data pre-processing
Rescale CT Attenuation images in order to better capture lung cancer patterns in CT images.
Data augmentation in order to reduce over-fitting on image recognition when training.
Step 3: Data training
Train labeled lung cancer medical image. 60% of the data will be used as training data set, 20% will be used as validation set and the last 20% will be used as test data set.
Details about CAD algorithms
Purpose:
The purpose is to detect lung cancer automatically using CAD from benign nodules or other lung diseases, such as ILD.
Methods:
We are planning to use deep convolutional neural networks (CNN) and deep contract learning method to train lung medical images.
During the past 30 years, the lung cancer patients increased 465% in China because of worse and worse air quality and other environmental problems. What’s worse, the health-care resources are distributed unevenly in China. More than 80% high-quality medical resources are centralized in tier-1 and tier-2 cities such as Shanghai, Beijing, Guangzhou, and Shenzhen. In the rural areas or West/Middle of China, there is a lack of well-trained doctors and right devices. Today, poor access and high fees are the two major challenges in China's healthcare system. For long-term, we want to use mobile technology embedded our CAD algorithms to touch every potential patient, and make better products and services much more accessible. Patients can use our app or something like that to diagnose themselves first. We want to democratize knowledge and information to the common people and help them to make better decisions.
Project steps:
Step 1: label data
Prepare lung cancer images, benign nodule patients, and other lung images.
Step 2: Data pre-processing
Rescale CT Attenuation images in order to better capture lung cancer patterns in CT images.
Data augmentation in order to reduce over-fitting on image recognition when training.
Step 3: Data training
Train labeled lung cancer medical image. 60% of the data will be used as training data set, 20% will be used as validation set and the last 20% will be used as test data set.
Details about CAD algorithms
Purpose:
The purpose is to detect lung cancer automatically using CAD from benign nodules or other lung diseases, such as ILD.
Methods:
We are planning to use deep convolutional neural networks (CNN) and deep contract learning method to train lung medical images.
Aims
Find effective CAD algorithms based on medical images to help people in developing countries.
Collaborators
Hangzhou Dianzi University