Development of a Prospectively Motion Corrected Free-breathing FLASH Sequence.

Thesis / Dissertation


Permanent link to this Item
Journal Title
Link to Journal
Journal ISSN
Volume Title
Respiratory motion of the heart is a fundamental challenge to cardiac MR imaging (CMR). This motion is frequently compensated for with breath-holding and acceptance-window methods. In situations where breath-holding is not viable, navigated free-breathing with an acceptance window can be used. This method results in inefficient acquisitions, creating longer scan times. This dissertation outlines the implementation of an adaptive predictor-observer control system in a FLASH sequence. The control system predicts the position of the diaphragm throughout the imaging segments based on multiple diaphragm position measurements acquired during the nonimaging segments. The position of the imaging slice is then prospectively adjusted using a linear scaling factor to perform slice following of the heart. Typically, a generalized scaling factor of 0.6 is used but this does not compensate for the variation amongst subjects nor the 3D nature of the heart. The performance of the control system was tested on phantoms and in 8 healthy volunteers. All imaging was performed on a 3T Skyra (Siemens AG, Erlangen). Initial phantom testing was performed utilizing a motion rig that simulates tidal breathing motion. Five sets of ECG-triggered FLASH acquisitions were performed in each healthy volunteer: (i) breath-holds (BH), (ii) free-breathing with no motion correction, (iii) freebreathing navigated-FLASH with a 4mm acceptance window (gated), (iv) free-breathing navigated-FLASH adapted to utilize the control system, and (v) a set of low-resolution cine FLASH images (TR=86ms, 50 images). The log data from the acquisitions with the control system adapted sequence were then analysed to measure the accuracy of the control system's predictions. Images acquired with the standard BH sequence were compared to those from the control system adapted sequence, the acceptance window sequence, and the uncorrected free-breathing sequence. Finally, the set of cine images were segmented at the lung-liver interface and around the heart. The edge of the lungliver interface and an edge of the heart were tracked to calculate the proportional change of the diaphragm's position to the heart's position, for each subject. The error between the control system's predicted position of the diaphragm and estimated actual position was within the 4mm acceptance window used by the gated sequence. The root mean squared error (RMSE) was below 3mm for many of the acquisitions and below 4mm for all except three acquisitions. The resultant images show improved quality using the control system compared with no correction and similar quality when compared to the gated acquisition, although artifacts due to the expansion and contraction of the chest wall remained. Tracking the edge of the lung-liver interface and the heart yielded variable tracking factors across subjects (0.58 to 1.02). Although the slice following of the control system is accurate, tracking during nonlinear sections of the breathing cycle remains challenging. There remains a risk of large tracking inaccuracy if errors in median calculation occur. The linear tracking factors relating diaphragm positions to the heart positions are constant for each subject but vary greatly between each subject, indicating the need for further research into subject specific tracking factors for each individual acquisition. The control system adapted acquisition can provide similar image quality to the gated acquisition, for certain tracking factors, whilst maintaining 100% imaging efficiency through the respiratory cycle.