Development of DCNN-based solution for lung cancer detection based on CT images.
Principal Investigator
Name
Noé Samaille
Degrees
M.D 2020, Applied Mathematics and Computer Science
Institution
IBM
Position Title
Data Scientist
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-648
Initial CDAS Request Approval
Mar 16, 2020
Title
Development of DCNN-based solution for lung cancer detection based on CT images.
Summary
Lung cancer is the one of the leading cause of death workdwide, accounting for 1.76 million deaths in 2018 out of 2.09 cases according to World Health Organization. As for other cancers, the best solution for lung cancer is early diagnosis and timely treatment.
Computed Tomography is one of the most common imaging tools for lung cancer diagnosis, used by radiologists to identify pulmonary nodules and evaluate their malignancy. It takes in average 10 minutes for radiologists to process a whole CT-scan and they drown under the amount of data to process, it is therefore very interesting for them to be augmented by a CAD system providing them with nodule predictions and classification to get a much quicker and much more precise diagnostic for their patients.
Deep Convolutional Neural Networks (DCNNs) are one of today's leading research field for object recognition and existing solution combining Computer Vision methods and DCNNs are among the best solutions for pulmonary nodule detection and classification.
The goal of this project is to build such CAD system inspired on these leading methods and base our research on improving them to provide a state of the art solution.
Computed Tomography is one of the most common imaging tools for lung cancer diagnosis, used by radiologists to identify pulmonary nodules and evaluate their malignancy. It takes in average 10 minutes for radiologists to process a whole CT-scan and they drown under the amount of data to process, it is therefore very interesting for them to be augmented by a CAD system providing them with nodule predictions and classification to get a much quicker and much more precise diagnostic for their patients.
Deep Convolutional Neural Networks (DCNNs) are one of today's leading research field for object recognition and existing solution combining Computer Vision methods and DCNNs are among the best solutions for pulmonary nodule detection and classification.
The goal of this project is to build such CAD system inspired on these leading methods and base our research on improving them to provide a state of the art solution.
Aims
- Develop a DCNN based solution for pulmonary nodule detection and lung cancer classification.
- Evaluate our solution by comparing it with top methods on public leaderboards such as the LUNA16 leaderboard.
Collaborators
Jean-Armand Broyelle, Cognitive Systems Lab PowerAI Technical Leader (IBM)
Emrick Sinitambirivoutin, Data Scientist (IBM)
Maxime Deloche, Data Scientist (IBM)