Phantom imaging

The tubes were glued inside a plastic box for protection but no other fill was placed in this outer box. The box contained 8 tubes (two for each mixture A, B, C, D) in a 4x2 grid with 1.5 cm airgaps between the tube outer walls. For imaging, the long axis of each phantom tube was aligned with the z axis of the scanner. Two MRI cylindrical test bottles (diameter 12 cm, height 20 cm, volume 2 L) were placed on each side to support the MRI system (MAGNETOM Avanto, Siemens Healthcare, Erlangen, Germany) calibrations of reference frequency and B1 field, and a transverse slice midway along the 8 tubes within it was imaged following localisers. The adjustment volume, 20 cm x, 12 cm y and 2 cm z, was used to set only the scanner reference frequency because the preset first- and second-order shim settings (known as “Tune-Up” based on a 30 cm-diameter spherical phantom at isocentre) were used without specific adjustment (“shimming”) of these for the T1 mapping phantom.

The phantoms were kept in the same MRI room with a temperature logger, and imaged weekly for 1 year using consistent coil and phantom support arrangements, with the identical sequence version and parameters each week. The image parameters were the same subject to automatic calibrations of flip angle and reference frequency using an identical adjustment volume each week:

For T1, MOLLI FOV 360 ×306 mm, slice thickness 8 mm, flip angle 35°, parallel imaging with acceleration factor 2 was used at high-resolution (256 independent pixels over 360 mm and 144 over 306 mm (acquiring 126 by 7/8^th partial ky), TR/TE 2.6/1.1 ms, single-shot image acquisition duration 194 ms) at an electronically programmed heart rate of 75 bpm; and low-resolution (192 independent pixels over 360 mm and 128 over 306 mm (acquiring 112 by 7/8^th partial ky), TR/TE 2.4/1.0 ms, single-shot image acquisition duration 159 ms) at an electronically programmed heart rate of 100 bpm) versions, with pre-contrast 5(3)3 and post-contrast 4(1)3(1)2 variants. The use of different variations of the MOLLI sequence for the native and post-Gd values has been adopted in line with Schelbert et al. [21] as the two different variations allow better optimisation of images with long and short T1 respectively. This was unlikely to have influenced the longitudinal results as the 5(3)3 sequence was always used for imaging and comparison of the native values and the 4(1)3(1)2 for the post-Gd values.

For T2, spin-echo imaging FOV 230 × 108 mm, slice thickness 10 mm, TR/TE 4500/22-264 ms, flip angle 180°.

Mean T1 and T2 values were taken in ROIs on pixel-wise maps, examined for drift and analysed on a dedicated software package (CMR Tools, Cardiovascular imaging solutions, London, UK; Fig. 2). The coefficient of variation (CoV = 100× standard deviation/mean) % of 52 weeks was compared against 10 re-positioned repeats acquired within 2 h.

Title: Post processing T1 and T2 images. Each phantom mixture was included twice giving a total of four different mixtures. Legend: Panel a is the original inversion recovery image, panel b is the spin echo magnitude image. Panels c and d represent the final maps; panel C is the T1 map and panel D is the T2 map. The ROI mean values were taken well within the phantoms away from edge-ringing artefacts such as Gibbs artefact [26] using CMR Tools

Using the individual components for native and post-Gd blood and myocardium and assuming a “phantom haematocrit” of 0.425 we could calculate a test “extracellular volume fraction” (ECV) of the phantom for assessment of drift impact using the equation:

The temperature was measured continuously during the first 6 months using an MR-safe digital thermometer placed on top of the phantom box (during storage, not scanning) to the nearest 0.5 °C. For the remaining 6 months manual read-outs were made from a liquid crystal thermometer strip glued on top of the phantom box at each weekly time scan to the closest 0.5 °C and recorded. The T1 results are reported first as measured and second with application of a temperature-correction to take into consideration the small temperature variation each week. The temperature correction parameters are based on the results of the temperature experiments described below.

There is no standardised method of reporting changes in phantom values, so we calculated the overall change in phantom T1 in two ways, ensuring that we did not underestimate the effect: Firstly, the phantoms were scanned weekly for a year obtaining 52 values per phantom. Four successive weekly values per phantom were averaged resulting in 13 independent values per phantom. The difference between the highest and the lowest value divided by the highest value was calculated as the percentage change. Secondly, the first 4 values and last 4 values were averaged and their difference was divided by the highest value.