Magnetic Resonance Imaging for Lung Radiotherapy a Prospective Study

Firstly, 3 stage III NSCLC patients receiving radiotherapy will be imaged, each for a single MRI session using TWIST and HASTE sequences. Initial sequence parameters will be those determined during the preceding technical development, but these will be fine-tuned to maximize tumour visualisation as this part of the study progresses, achieving the most practically useful trade-off between image resolution and noise, qualitatively and quantitatively assessed by a radiologist to determine and fine-tune image quality.

Then a further 12 patients will be imaged, each for two MRI sessions taking place during the radiotherapy schedule and separated by at least a week. Each MR session will consist of the following sequence: 15 seconds of TWIST, 15 seconds of HASTE, 90 seconds off, 15 seconds of TWIST and 15 seconds of HASTE. For each patient an on-treatment 4D cone-beam CT will also be collected (standard process), alongside the diagnostic quality planning 4DCT. Patient breathing coaching will be consistent between CT and MR, as will patient positioning; that is, patients will be imaged with their arms above their heads. The images will be analyzed to determine -

1. Do extents of tumour movement seen in TWIST 4D-MR images differ from those seen in planning 4D-CT scans, judging the movement extent according to differences in internal target and gross tumour volumes (ITVs and GTVs) defined from the two sets of images, and in the range of motion of the tumour centre of mass? This may well be the case, since the 4D-MR scans catalogue movement over several breathing cycles, whereas 4D-CTs describe a single composite cycle, synthesised from slices collected at various times over multiple cycles.
2. How reproducible is the movement seen at the two MR imaging sessions? Additionally, how reproducible is the movement seen within each MR imaging session?
3. How similar according to volume, Dice similarity index (percentage of overlap) and Haussdorf distance (maximum distance between the contours of two structures) are GTVs outlined on single phases of 4D-CT and TWIST 4D-MR images, after rigidly registering the centres-of-mass of the two GTVs to allow for movement?
4. How consonant are tumour contours defined on single slice HASTE MR images with those defined on a phase of the 4D TWIST images? Answering question 1 will allow us to determine the utility of gauging tumour movement over extended 4D-MR imaging sessions, rather than from 4D-CT sessions which have to be short to avoid excessively irradiating patients. Question 2 will cast further light on the same issue, allowing us to determine the stability over the course of RT schedules of motion assessed over the course of around 1 minute during an individual TWIST scan.

Answering question 3 will allow us to understand how fully 4D-MR images can be used within the treatment planning process. If outlined GTVs differ greatly between MRI and CT, then 4D-MRI might only provide more complete movement data; whereas if CT and MRI-based GTVs are similar the 4D-MRI may have more uses in treatment planning, particularly if some tumour regions are more clearly visible on MRI than on CT.

Question 4 will allow us to gauge the accuracy and precision of tumour definition on real-time single MRI slices, compared to definition on 4D-MR and 4D-CT. Answering this question is an essential precursor to the development of automatic algorithms