Performance evaluation and generalisability of an AI algorithm for lung cancer detection
Principal Investigator
Name
Lizzie Barclay
Degrees
B.SC., MB.Ch.B
Institution
Annalise-AI Pty Ltd
Position Title
Associate Medical Director
About this CDAS Project
Study
PLCO
(Learn more about this study)
Project ID
PLCOI-1234
Initial CDAS Request Approval
May 30, 2023
Title
Performance evaluation and generalisability of an AI algorithm for lung cancer detection
Summary
Lung cancer is the 2nd most commonly diagnosed cancer and the leading cause of cancer related death worldwide, with more than 2.2 million cases of lung cancer recorded in 2020. The survival rate of lung cancer is low, largely attributed to the fact that diagnosis predominantly occurs in advanced stages. Screening programs have been implemented in some regions but rely on the use of low-dose CT (LDCT).
Although its utility for lung cancer screening remains highly debated, chest x-ray (CXR) remains one of the first investigations performed in the workup of suspected lung cancer. Although early detection of actionable nodules identified on CXR has generally been regarded as having a low sensitivity, improved detection rates have been reported when CXR interpretation is augmented by computer aided detection (CAD) devices, with increased accuracy measures.
The PLCO dataset provides a highly valuable CXR dataset of confirmed lung cancer cases and other abnormalities. We plan to utilise the PLCO dataset to evaluate the performance and generalisability of a comprehensive AI algorithm.
Aims
Evaluate the performance of an AI algorithm for the detection of triaging chest x-rays with findings suspicious for lung cancer and requiring further imaging.
Evaluate the performance of an AI algorithm for the detection of other abnormalities on chest x-ray.
Collaborators
Melissa Ryan
Leslie Cass