Early lung cancer diagnosis using AI
Research type
Research Study
Full title
Can a deep learning–based automatic detection (DLAD) algorithm decrease time to histological diagnosis of lung cancer – a retrospective cohort study
IRAS ID
311137
Contact name
Jenna Tugwell-Allsup
Contact email
Sponsor organisation
BCUHB
Duration of Study in the UK
0 years, 9 months, 1 days
Research summary
The use of artificial intelligence algorithms within radiology is evolving with new deep learning based automatic detection algorithms having the ability to detect numerous pathologies with high specificity and sensitivity. Chest X-rays can be difficult to interpret with their sensitivity in depicting lung cancer varying widely from 44% to 87%, depending on the study population and other factors; approximately 1-in-5 cancers can be missed. Chest X-rays are however used as first line imaging for numerous symptoms including those typically associated with lung cancer (cough, chest pain, shortness of breath, haemoptysis). Artificial intelligence software could aid the reporting clinician in the early detection of lung cancer, however, implementing such software into routine clinical practice is slow and complex. This is mainly due to their lack of local validation but also due to concerns of detection of false-positive and/or benign nodules which trigger further follow-up (e.g. CT). The aim of this study is to extract a retrospective sample of positive (based on histology) lung cancer patients and use their diagnostic imaging and historic chest X-rays to identify whether the AI algorithm can improve earlier detection of lung cancer. Also, a sample of normal chest X-rays (based on radiology report and a normal follow up CT within 8 weeks) will be evaluated using the same AI software to identify the possible rate of false positive findings.
REC name
N/A
REC reference
N/A