Quantifying early lung fibrosis using deep learning

Principal Investigator

Name
Joseph Jacob

Degrees
F.R.C.R., M.D.(Res).,

Institution
University College London

Position Title
Wellcome Trust Fellow

Email
j.jacob@ucl.ac.uk

About this CDAS Project

Study
NLST (Learn more about this study)

Project ID
NLST-468

Initial CDAS Request Approval
Jan 15, 2019

Title
Quantifying early lung fibrosis using deep learning

Summary
Idiopathic pulmonary fibrosis (IPF), the most common fibrosining lung disease has a median patient survival of only 3-5 years. The prevalence of FLD is increasing in the Western world partly related to ageing within the general population and partly because of increased awareness of the disease. Unfortunately FLD is typically diagnosed at an advanced stage once patients are already physically impaired and patient decline is often rapid.

Diagnosis in turn is often delayed as CT appearances and lung function test profiles are only well characterized for established, relatively late-stage disease. Recognising FLD at an early stage is therefore essential to create a treatment window where disease that has not yet become extensive could potentially be controlled. We aim to use deep learning mechanisms to quantify lung damage and disease progression.

Aims

Quantify CT features suggestive of early lung fibrosis
Quantify CT features indicating rapidly progressive CT phenotypes of lung fibrosis

Collaborators

Prof Daniel Alexander
Prof Geoff Parker
Mr Moucheng Xu
Dr Cheung Wing Keung
Mr Ashkan Pakzad
Dr Arjun Nair