Skip to Main Content

An official website of the United States government

Detecting additional lung cancers on chest computed tomography (CT) using artificial intelligence: A retrospective study based on NLST dataset

Principal Investigator
Name
Dennis Robert
Degrees
M.B.B.S, M.M.S.T.
Institution
Qure.ai Technologies Private Limited
Position Title
Director of Clinical Research
Email
About this CDAS Project
Study
NLST (Learn more about this study)
Project ID
NLST-1433
Initial CDAS Request Approval
Jun 2, 2025
Title
Detecting additional lung cancers on chest computed tomography (CT) using artificial intelligence: A retrospective study based on NLST dataset
Summary
Lung cancer is one of the leading causes of cancer related death in US. Most people with lung cancer are currently diagnosed at an advanced stage. Early detection of malignant lung nodules improves survival rates. Screening with low dose computed tomography (LDCT) is recommended in high-risk individuals and few countries have started implementing screening programs. The shortage of radiologists and increasing workload present a challenge for effective implementation of screening and nodule management programs. Previous studies have reported the risk of missed lung cancers during screening and the lesion visible on retrospective assessment on chest CT scans. Artificial intelligence-based computer aided diagnosis (CAD) systems (like qCT, developed by Qure.ai) have shown promise in detection, characterisation lung nodules on chest CT and some of them are commercially available. While the diagnostic performance qCT has been evaluated in previous studies, the present study aims to evaluate the added value of AI based software qCT in detecting missed cancers.

The following questions are relevant to the present study:

1) Can CAD detect screen negative lung cancers in NLST screening dataset?

2) Can CAD detect lung cancers early in NLST screening dataset?
Aims

The study aims to evaluate performance of qCT in identifying missed LCs at screening in the NLST dataset

Primary objective:

- Proportion of screen negative cancers identified by qCT in LDCT arm of NLST trial

Secondary objectives:

- Detection performance of qCT amongst screen positive cancers and nodule datasets

- False positive rate of qCT (Screen positive + No nodule dataset)

- Proportion of cancers detected early by qCT among screen detected cancers and post screen cancers

Study Endpoints

Primary Endpoints:

· Proportion of screen negative cancers detected: Number of images correctly flagged by AI CAD based on lobe location/ Total number of screen negative lung cancers (Missed cancers)

Secondary Endpoints:

· Detection rate in screen positive cancers: Number of images correctly flagged by AI CAD based on lobe location/ Total number of screen positive cancers

· False positive rate in the screen positive + no nodules

· Detection rate of qCT in previous scans of screen detected lung cancers

· Performance of qCT in the nodule dataset of NLST

· Proportion of cancers detected based on Lung RADS for baseline scan: Number of images with LungRADs score of 3 and above based on AI results/Total number of screen negative lung cancers (Missed cancers)

Collaborators

Dr. Dennis Robert, Qure.ai
Dr. Santhosh S, Qure.ai
Manoj Tadepalli, Qure.ai
Ranjana Devi. Qure.ai
Saniya Pawar, Qure.ai
Deepak Adarsh, Qure.ai