Use of Radiomics and AI To Predict Malignancy in Indeterminate Lung Nodules
Principal Investigator
Name
Richard Lee
Degrees
MBBS MRCP MA (Cantab) PhD
Institution
The Royal Marsden Hospital
Position Title
Consultant Respiratory Physician, Early Diagnosis Champion, RM Partners Lung Pathway Chair
Email
About this CDAS Project
Study
NLST
(Learn more about this study)
Project ID
NLST-593
Initial CDAS Request Approval
Nov 19, 2019
Title
Use of Radiomics and AI To Predict Malignancy in Indeterminate Lung Nodules
Summary
Indeterminate lung nodules require interval imaging to assess cancer risk, which generates anxiety for patients and places a significant resource-burden on the National Health Service.
A proportion of patients who score highly on existing risk models, such as those with large (>15mm) nodules, will have negative biopsies, and there is a need to improve patient stratification.
We have previously developed radiomics algorithms to predict the risk of malignancy in incidental lung nodules, but these have not been tested in cohorts of exclusively larger nodules.
A retrospective clinical database of lung nodule patients investigated at the Royal Brompton Hospital has been developed, which will be used to test the accuracy of our existing tools in predicting malignancy in 15-30mm nodules, as well as to develop novel radiomics and AI approaches in this setting. Existing lung-nodule datasets, such as the NLST data, will be used to validate our predictive models.
A proportion of patients who score highly on existing risk models, such as those with large (>15mm) nodules, will have negative biopsies, and there is a need to improve patient stratification.
We have previously developed radiomics algorithms to predict the risk of malignancy in incidental lung nodules, but these have not been tested in cohorts of exclusively larger nodules.
A retrospective clinical database of lung nodule patients investigated at the Royal Brompton Hospital has been developed, which will be used to test the accuracy of our existing tools in predicting malignancy in 15-30mm nodules, as well as to develop novel radiomics and AI approaches in this setting. Existing lung-nodule datasets, such as the NLST data, will be used to validate our predictive models.
Aims
1) To establish an anonymised retrospective database of clinical information and scan images from patients investigated for lung nodules nodules at the Royal Brompton Hospital and other sites across the London Cancer Alliance.
2) To use this database to create novel radiomics algorithms to predict the risk of malignancy in both large (>=15mm) and small (<15mm) nodules.
3) To validate these algorithms in independent cohorts, including the NLST dataset.
Collaborators
Professor Eric Aboagye, Professor of Cancer Pharmacology and Molecular Imaging, Imperial College London.
Professor Anand Devaraj, Consultant Radiologist, The Royal Brompton and Harefield NHS Foundation Trust.
Dr. Benjamin Hunter, Clinical Research Fellow, The Royal Marsden NHS Foundation Trust.