Computer assisted detection of abnormalities in chest CT scans
Principal Investigator
Name
Joel Pinto
Degrees
Ph.D
Institution
Nuance Communications
Position Title
Research Manager
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-348
Initial CDAS Request Approval
Sep 18, 2017
Title
Computer assisted detection of abnormalities in chest CT scans
Summary
Lung cancer is the leading cause of cancer death in the world. According to the American cancer society about 155,000 people will die from lung cancer in 2017 in the United states alone and about 220,000 new cases will be diagnosed in the same time period [1] . Early detection of lung cancer (or any other form) is the key to survival and radiology imaging has been a boon in this direction. Advancements to radiology imaging techniques have led to an increase in both image utilization and volume in terms of number of images per exam. The large volume of images could potentially overburden radiologists and potentially lead to increase in oversight errors.
To address the above problem a radiologist could be assisted by a computer system that can automatically analyze the large number of scans/images and flag cases for immediate (and careful) attention from the radiologist. The radiologist can be further assisted by highlighting abnormalities e.g., a pulmonary nodule that might develop into cancer. Recent advancements in the field of computer vision, machine/deep learning, GPU computing has made this possible [2][3]. The CT dataset from NLST will be an important database in developing such a computer system.
[1] https://www.cancer.org/cancer/non-small-cell-lung-cancer/about/key-statistics.html
[2] https://www.kaggle.com/c/data-science-bowl-2017
[3] https://luna16.grand-challenge.org
To address the above problem a radiologist could be assisted by a computer system that can automatically analyze the large number of scans/images and flag cases for immediate (and careful) attention from the radiologist. The radiologist can be further assisted by highlighting abnormalities e.g., a pulmonary nodule that might develop into cancer. Recent advancements in the field of computer vision, machine/deep learning, GPU computing has made this possible [2][3]. The CT dataset from NLST will be an important database in developing such a computer system.
[1] https://www.cancer.org/cancer/non-small-cell-lung-cancer/about/key-statistics.html
[2] https://www.kaggle.com/c/data-science-bowl-2017
[3] https://luna16.grand-challenge.org
Aims
1) Develop machine learning algorithms to automatically detect abnormal lung nodules
2) Automatically flag cases that need urgent attention by the radiologist
Collaborators
Jin Chen, Nuance Communications
Dante De Nigris Moreno, Nuance Communications
Stefan Christof Hahn, Nuance Communications