Clustering and Phenotyping Interstitial Lung Disease
Moving beyond a label: Understanding disease behaviour in interstitial lung disease
Royal Brompton Hospital
Duration of Study in the UK
3 years, 0 months, 1 days
Pulmonary fibrosis or Interstitial Lung Disease (ILD) encompasses a wide spectrum of progressive respiratory disorders where normal lung is replaced by scar tissue. Clinically this causes cough, shortness of breath and in the majority of cases ultimately death from respiratory failure. The commonest of these conditions, idiopathic pulmonary fibrosis (IPF), affects over 32 000 individuals in the UK and accounts for 1 in every 100 deaths in this country each year. Despite the recent approval of two antifibrotic drugs for IPF the 5-year survival rate remains 25%, far worse than many common cancers.
The diagnosis of ILD is frequently challenging. The currently accepted gold standard for diagnosis is multi-disciplinary team (MDT) assessment of clinical, radiological and histopathological data. It is well recognized, however, that many aspects of this process have major flaws, with each component plagued by problems of inter- and intra-observer variability. This process is lengthy and our current diagnostic tests, including lung biopsy, also carry a significant risk to patients. Despite all of this up to 15% of cases of ILD remain unclassifiable, leaving patients in limbo with no diagnosis and no clear therapeutic strategy.
Given the importance of an accurate diagnosis and the risks associated with lung biopsy, a computer based system for diagnosis in ILD would be of great value. We aim to dissect the complexity of ILD using non biased Machine learning approaches on data to sidestep conventional classifications, identify patients at risk of rapid progression, speed up the diagnostic pathway and ultimately avoid the morbidity and mortality associated with the current diagnostic pathway.
London - Bloomsbury Research Ethics Committee
Date of REC Opinion
29 Nov 2019