Estimating lung volumetric parameters via rapid, limited-slice, free-breathing thoracic dynamic MRI

PurposeWe present an observational study involving free-breathing short-scan-time dynamic MRI (dMRI) method that can be routinely used for computing dynamic lung volumes accurately. Materials and Methods(i) Full-resolution free-breathing sagittally-acquired 2D dMRI scans are gathered from 45 normal children via True-FISP sequence. Sparse dMRI (s-dMRI) scans are simulated from these datasets by subsampling in the spatio-temporal domains via a limited number NSS of selected sagittal locations and TSS of time instances (respectively, NFS and TFS for full scan). (ii) A 4D image is constructed from both full and sparse scans. Lungs are segmented from 4D image, and their volumes from full (VF) and sparse dMRI (VS) scans are computed. (iii) A regression model is fit for VF as a function of VS on a training set, and the full-resolution volume VP predicted by the model is estimated from VS. (iv) The deviation of VP from VF is analyzed on both synthesized sparse dMRI scans from a separate full-resolution test set and actual s-dMRI scans prospectively acquired from 10 normal children. ResultsWith NSS=5 (per lung) and TSS=40, the deviation of VP from VF was [~]2% with a total scan-time of [~]9 min (45-60 min for the full scan with NFS=15-22 (per lung) and TFS=80). These metrics become 0.4%, and <20 min for s-dMRI with NSS=15-22 (per lung) and TSS=40. Conclusions-dMRI is a practical approach for computing dynamic lung volumes that can be used routinely with no radiation concern, especially on patients who cannot tolerate long scan times.