Lung Cancer, Lung Neoplasms, Pulmonary Nodule, Multiple, Pulmonary Nodule, Solitary
Conditions
Keywords
Incidental lung nodules, Radiomics, Artificial Intelligence, Machine Learning
Brief summary
This study will collect retrospective CT scan images and clinical data from participants with incidental lung nodules seen in hospitals across London. The investigators will research whether machine learning can be used to predict which participants will develop lung cancer, to improve early diagnosis.
Interventions
DIAGNOSTIC_TESTMachine Learning Classification
Patient's scans will be used as input into in-house software to extract multiple radiomics features. These features will be used to develop a risk-signature which can predict malignancy risk. Patient scans will also be used as input into deep learning/convolutional neural network models to perform automated imaging classification.
Sponsors
RM Partners West London Cancer Alliance
Royal Brompton & Harefield NHS Foundation Trust
University College London Hospitals
Imperial College Healthcare NHS Trust
Lewisham and Greenwich NHS Trust
King's College Hospital NHS Trust
Epsom and St Helier University Hospitals NHS Trust
Institute of Cancer Research, United Kingdom
Guy's and St Thomas' NHS Foundation Trust
UCLH Biomedical Research Centre
Royal Marsden NHS Foundation Trust
Study design
Observational model
COHORT
Time perspective
RETROSPECTIVE
Eligibility
Sex/Gender
ALL
Age
18 Years to No maximum
Healthy volunteers
No
Inclusion criteria
* Age \> 18 * Baseline CT thorax imaging reported as having pulmonary nodule(s) between 5 and 30mm in the last 10 years. * Ground truth known (either scan data showing stability for 2 years (based on diameter) or one year (based on volumetry), complete resolution, or biopsy-proven malignancy. * Slice thickness \< 2.5mm.
Exclusion criteria
* • Absence of at least one technically adequate CT thorax imaging series (defined by visual inspection of presence of imaging data of the thorax in the DICOM record). * Slice thickness \> 2.5mm. * Imaging \> 10 years old. * Ground truth unknown.
Design outcomes
Primary
| Measure | Time frame | Description |
|---|---|---|
| Development of an imaging biobank | 1 year | The primary endpoint will be met if we are able to store baseline CT scans and the minimum clinical data set for 1000 patients. |
Secondary
| Measure | Time frame | Description |
|---|---|---|
| Discovery of a CT-thorax based radiomics profile to predict cancer risk. | 1 year | We aim to identify distinct clusters of radiomics variables to generate a radiomics predictive vector (RPV), which can be used to stratify patients according to malignancy risk. This vector will be used in multivariate analysis and compared to existing risk models. |
Countries
United Kingdom
Contacts
Primary ContactRichard Lee, MBBS PhD
Outcome results
None listed