Image Reconstruction (including machine learning)
.jpg "Felicia Seemann, PhD photo")
Felicia Seemann, PhD
Postdoctoral Fellow
National Heart, Lung, and Blood Institute, National Institutes of Health
Bethesda, Maryland, United States
Felicia Seemann, PhD
Postdoctoral Fellow
National Heart, Lung, and Blood Institute, National Institutes of Health
Bethesda, Maryland, United States
Ahsan Javed, PhD
Staff Scientist
National Institutes of Health, United States
Jaffar M. Khan, MD, PhD
Staff Clinician
National Heart, Lung, and Blood Institute, National Institutes of Health
Washington, District of Columbia, United States
Christopher G. Bruce, MD
Staff Clinician
National Heart, Lung, and Blood Institute, National Institutes of Health, District of Columbia, United States
Rachel Chae
Summer Intern
National Heart, Lung, and Blood Institute, National Institutes of Health, United States
Korel Yildirim, PhD
Catheter Prototyping Engineer
National Heart, Lung, and Blood Institute, National Institutes of Health
Chevy Chase, Maryland, United States
Amanda Potersnak, RT (R) (MR)
MRI Technologist
National Heart, Lung, and Blood Institute, National Institutes of Health, United States
Haiyan Wang
MRI Technologist
National Heart, Lung, and Blood Institute, National Institutes of Health, United States
Rajiv Ramasawmy, PhD
Staff Scientist
National Heart, Lung, and Blood Institute, National Institutes of Health
Bethesda, Maryland, United States
Robert J J. Lederman, MD
Principal Investigator
National Heart, Lung, and Blood Institute, National Institutes of Health
Bethesda, Maryland, United States
Adrienne E E. Campbell-Washburn, PhD
Principal Investigator
National Heart, Lung, and Blood Institute, National Institutes of Health
Bethesda, Maryland, United States
Shortness of breath is an early symptom of heart failure which is often exercise-induced through an increased pulmonary blood pressure and accumulation of lung water1. Lung water quantification is therefore of interest to unmask latent heart failure. Magnetic resonance imaging (MRI) measurements of lung water has been performed at rest and shortly after, but not during, exercise2–5. We develop a method for imaging dynamic changes in lung water. We validate it with a porcine model of mitral regurgitation and apply it to healthy subjects during supine exercise stress.
Methods:
Lung water was measured using self-gated 3D stack-of-spiral proton density weighted gradient echo sequence (TE/TR/θ=0.56ms/9ms/1°, 831 spiral interleaves, 5.0ms readout, 3.5x3.5x3.5mm resolution, 450x450x252mm FOV)6 at 0.55T MRI (prototype Siemens Aera)7. We used a sliding window image reconstruction with respiratory motion correction8–10 to derive time-resolved images (Fig 1). Short-axis cine11 and aortic flow were acquired to quantify cardiac output and mitral regurgitant fraction.
Automated image processing was used to derive 3D pixel-wise lung water density (LWD) maps with a neural network for lung segmentation4. LWD maps were computed from coil shading corrected images as the signal intensity ratio in the lungs to the surrounding body, assuming a 70% musculoskeletal water density5.
We used a porcine model of lung water through induction of mitral regurgitation12 (n=4, 40±2kg) by placing a suture across the anterior mitral leaflet12,13, externalized through a femoral vascular sheath. A regurgitant jet was dynamically induced inside the MRI by applying tension on the suture while imaging, triggering accumulation of lung water. Lung water accumulation was imaged over 1h and corroborated by simultaneous recordings of systemic and pulmonary arterial wedge pressures (PAWP).
Healthy subjects (n=9, 25±4yrs, 5 men) were imaged in transitions between rest and exercise using a supine pedal ergometer with 60-90W resistance.
Results:
Dynamic LWD imaging during induction of mitral regurgitation and during exercise was feasible (Fig 2-3).
A mitral regurgitant fraction of 53±21% and ΔLWD of 4.6±1.0% was achieved in the porcine model, with a ΔPAWP increase of 75±44% (6.5±3mmHg). Cardiac output decreased from 3.9±0.8 to 2.7±0.6L/min, p=0.007. Systemic pressure decreased by -23±15/-20±15%.
Healthy subject exercise imaging measured a ΔLWD of 14±12% after 10min of moderate exercise and 22±9% during vigorous exercise. We hypothesize this rise is explained by an increased intravascular pulmonary fluid. Cardiac output increased from 5.3±1.3 to 9.6±3.1L/min, p=0.003. Lung water did not change over 10min at rest (ΔLWD 0.2±2.5%, p=0.9).
Conclusion:
Dynamic changes in lung water density can be quantified during exercise stress using a sliding-window and motion corrected image reconstruction. Patient studies are warranted to determine if this method can unmask latent heart failure, where higher extravascular lung water is expected.