Three-Dimensional Phase Resolved Functional Lung Magnetic Resonance Imaging

3D PREFUL MRI enables visualization of lung ventilation. What was our initial motivation to develop PREFUL MRI? It was the desire to capture the whole lung volume with higher isotropic image resolution compared to conventional 2D PREFUL MRI.

3D PREFUL MRI enables dynamic ventilation, depiction detection of lung disease, treatment monitoring, and hopefully in the future prognostication of patients with lung disease. In children PREFUL MRI is especially helpful because it's very patient friendly without radiation. The most significant advantage of 3D PREFUL MRI is its usage of a product sequence, which is available on state-of-the-art MR scanners, thereby facilitating its widespread adoption to other centers.

Additionally, in comparison to other 3D non enhanced Free-breathing MRI techniques, 3D PREFUL MRI has been recently validated to direct measurement of ventilation using 19 FCAs imaging and hyperized gas imaging, demonstrating strong correspondence between these methods. We have shown that 3D PREFUL MRI has been able to show ventilation heterogeneity and ventilation defects in patients with asthma and CBD. Furthermore, we were able to show treatment response effects in patients with cystic fibrosis after therapy.

So we think that in the future, 3D PREFUL MRI is an additive valuable tool in the clinic for patients with lung disease. The most recent 3D PREFUL developments include new sequence designs such as the use of spiral trajectories. These are relevant due to their lower hardware requirements and increased case-based sampling efficiency.

Additionally, simultaneous motor slice approach is gaining interest for significantly reducing scan time during 2D acquisition while enabling ventilation and perfusion weighted imaging. The next steps in 3D PREFUL MRI will involve further reducing acquisition times and employing deep learning based techniques for image reconstruction and image registration, as these steps are currently the most time consuming in the pipeline. Thereby we aim to significantly accelerate these processes, improving overall acquisition efficiency and enabling online reconstruction.

To begin, instruct the subject to lie on the magnetic resonance imaging or MRI table in a supine position. Using laser light navigation, position the patient so that the imaged area occupies the ISIS center of the bore. Perform a localizer sequence to obtain low resolution anatomical images in three planes to plan the 3D PREFUL exam.

After setting the acquisition parameters, set the number of radial projections so that the acquisition takes around eight minutes. Check if the whole thorax is included in the field of view and start the acquisition. Save the raw 3D PREFUL case-based data using the vendor-specific steps on the MR console.

Set the path to the raw data to evaluate the acquired data using the image reconstruction script. Estimate the MR system dependent gradient delays using the calibration projections with the S delay function in the BART. Apply a shift to each projection according to the calculated gradient delay, followed by handing filtering to the gradient delay corrected data using a periodic window in the XY plane and the Z direction.

For low resolution image reconstruction, select one eighth of the case-based data, centered around the middle index of the readout dimension, as well as one fourth of the case-based data centered around the middle index of the partition dimension. Reshape the matrices of case-based data and corresponding trajectories to include additional time dimensions with each time point comprising 14 radial projections. Utilizing the parallel imaging and compressed sensing pics command and BART perform image reconstruction.

To extract the respiratory gating signal, segment the lung parenchyma in each reconstructed low resolution 3D image using Otsu's thresholding and calculate the lung parenchyma volume for each time point. Apply a low pass filter with a cutoff frequency of 0.7 hertz to the extracted volume time series and calculate the breathing frequency. Then exclude extreme outliers from the time series.

Next, perform fast Fourier transform on the handing filter data along the slice direction to enable slice by slice image reconstruction. Reshape the case-based data matrix based on the extracted time series so that each respiratory bin contains a minimum of 100 spokes. Reshaped the case-based trajectories matrix to match the size of the case-based data and utilize 20%of projections from adjacent bins to enhance the signal to noise ratio of each respiratory bin.

Execute the picks command for image reconstruction for each slice. Save the variables, including the reconstructed images of one complete respiratory cycle, breathing frequency, and acquisition times in a DOTMAP file. Then invert the signal intensities of the reconstructed images and register them to a selected respiratory state using non rigid registration.

Perform filtering with Nadaraya-Watson kernel regression to the registered and reinverted images to interpolate 16 phases onto a uniformly spaced time grid. Eliminate signal variations unrelated to respiration using a low pass filter at 0.7 hertz to the interpolated data in the time domain. Set the parameters of window size to 5 5 5 and lambda to 0.001.

De-noise the lowpass filtered images using image guided filtering with the temporally averaged registered image as the guiding image. Calculate the regional ventilation or our vent maps in milliliter per milliliter for each respiratory Phase N.Next, segment the lung parenchyma of the end inspiratory 3D image using a convolutional neuronal network with NNU net architecture. Exclude lung vessels using a vessel recognition algorithm, and then calculate the ventilation defect or VD map of the R vent map of eighth phase, which represents the maximal ventilation.

Calculate the flow as the first derivative of R vent. Establish a reference flow volume loop or FVL for each slice computed from a healthy region defined by our vent values between 75th and 95th percentile. Determine the similarity between each lung parenchyma voxels FVL and the reference FVL using cross correlation with zero lag.

Then compute ventilation defect FEL correlation metric map. Compute ventilation time to peak or VTTP maps in percentage of respiratory cycle, followed by deviation of VTTP value from the peak inspiration at 50%Describe statistically the parameters are vent, FELCM, VTTP, and deviation of VTTP for all slices, providing median, mean, standard deviation values, and interquartile ranges. Finally, calculate the ventilation defect percentage values derived from the generated VD maps and export all calculated parameters to a spreadsheet file.

A 3D PREFUL MRA of a healthy volunteer indicated homogenous ventilation values for all the ventilation parameter maps. In the exemplary VD maps, ventilation defect percentage value for our vent was 3.6%and that for FELCM was 3%falling within the healthy normal range. Comparison of ventilation defect maps from 3D PREFUL MRI and 19 FMRI of a 54-year-old female chronic obstructive pulmonary disease patient, indicated good visual correlation of healthy and hypoventilated regions.

Comparison of 3D PREFUL MRI to measure regional responses after therapy of a lung of a cystic fibrosis subject revealed increased values of the FEL correlation metric as well as the reduction in VD after therapy.