The goal of this research is to translate new and innovative MRI techniques to sick and challenging patient populations, like those that are born prematurely. Our studies include both proton MRI and hyperpolarized gas MRI in infants. This has begun to revolutionize our understanding of both parenchymal and airway components of disease.
HyPOINT (Hyperpolarized Imaging for New Therapies (Hypoint) in Pediatric CF)
Primary Objectives:
We have demonstrated that an AI-driven semantic quantification of lung structural alterations is feasible in CF by building an automated scoring system. Clinical validation expects that a biomarker reflects the clinical severity, correlates to a known outcome and may improve with an effective therapy. Our objective was to develop an algorithm enabling recognition of five structural alteration hallmarks on CT slices. We then aimed to assess the clinical validity of the quantitative scoring method by correlating to the patient's disease severity, as assessed by the CT Brody score. Additional objectives were to support the clinical validity to correlate to PFTs, assess variations in patients with and without lumacaftor/ivacaftor, and evaluate the reproducibility.
Our studies have demonstrated that AI-driven quantitative measurement of lung structural abnormalities on CT scanning in CF is, indeed, feasible, and can provide clinically important information in a broad range of patients using a wide range of CT scanners and CT techniques. The system showed good similarity and very good agreement with ground-truth identification of expert observers’ abnormalities, but dramatically quicker, with high reproducibility. Volumetric measurements showed a strong correlation to PFTs and a well-validated visual CT score at several time-points. The automated quantifications were found to sensitively detect longitudinal changes, either a reduction in CF patients with lumacaftor/ivacaftor treatment or an increase during the natural course of the disease. As a fully automated outcome measurement, the reproducibility was almost perfect.