Researchers: Michael Roberts, Tom McLellan, Muhunthan Thillai, Alessandro Ruggiero, Ke Chen, Carola-Bibiane Schönlieb

Idiopathic Pulmonary Fibrosis (IPF) is a progressive fibrotic lung condition of unknown cause. In current clinical practice the assessment of the progression of fibrosis or response to treatment is based on a visual assessment of the high resolution Computed Tomography (CT) scan images performed by a specialist thoracic radiologist.

This project developed in collaboration with the Interstitial lung disease unit at Royal Papworth Hospital has the aim to develop a bespoke image analysis and machine learning pipeline. In a multi-resolution and multi-tasking approach lungs (on downsampled low-resolution images), airways, fibrosis, vessels etc (on the high-resolution images) are segmented and analysed, and from there predictive biomarkers are extracted that can be used to quantify and subsequently predict progression of IPF.

Photo caption: Cross sectional CT scan of the lungs from a patient with IPF. Note the honeycombed areas at the periphery of the lung lobes which are characteristic of this disease