El proceso de transformación Fases y mediaciones

8.4.1.1 Subject selection

Twenty-four adult patients were recruited (14 females, 10 males; mean age 43.9 ± 18.9

years, range 18-78). Recruitment occurred in two waves; the first was an opportunistic

sample (n=13) of whom only two exhibited prominent quantities of 18F-FDG BAT

(‘BAT06’ and ‘BAT13’). In addition two patients did not undergo concomitant PET/CT scans (‘BAT03’ and ‘BAT04’), and therefore their BAT status was not known. Consequently a second more defined group of subjects were recruited from the

population studied in chapters 3-4 (BAT13-18) on the basis of having high 18F-FDG

BAT uptake on PET/CT.

In addition six BAT negative controls were identified and recruited on the basis of

consistently absent 18_{F-FDG BAT uptake on serial PET/CT scans (BAT19-24). These}

patients were also selected to be age- and sex-matched to patients BAT13-18.

8.4.1.2 Scanning protocol

Scanning was performed on a GE Signa HDxt 3T MR scanner (General Electric, Milwaukee, USA) with a quad-channel cardiac receiver coil placed anteriorly across the upper thorax and cervical region.

A 3-point IDEAL FSE pulse sequence was performed with the following parameters:

TR/TE = 440/10.7 to 11.2 ms, flip angle = 90°, number of signal averages (NEX) = 3,

field of view = 300 x 300 to 380 x 380mm, voxel size (mm) 0.59 x 0.59 to 0.7 x 0.7.

Various slice thicknesses were obtained to explore to impact of volume averaging, i.e. 5

mm with 5.3 cm spacing (BAT01-13, 20) and 2.3mm with 2.8mm spacing (BAT03, 13- 24). Between 30 and 33 contiguous axial slices were acquired in an interleaved manner

from the upper cervical to mid-thoracic region during free breathing. In addition T2* mapping sequences were obtained with the same coverage as the IDEAL sequences for

12 patients (BAT13-24) with the following parameters: TR/TE1/ΔTE (ms) =

300/3.2/5.4, flip angle = 20°, number of signal averages = 0.75.

To assess MR changes within BAT over time, serial scans were obtained for some patients (BAT01, 06-08, 10, 11 and 13).

8.4.1.3 Image analysis

Following data reconstruction, fat fraction maps were produced offline from the separated fat and water images as described above using ImageJ. T2* maps were also produced offline using the MRIAnalysisPak plugin for ImageJ .

Fat fraction images maps were then thresholded with a lower limit set at 50% to exclude non-fat tissues.

Two methods of BAT identification were used: manual segmentation of fat in the

supraclavicular fossae, and transposition of 18F-FDG BAT ROIs from PET scans onto

the MR. WAT regions of interest were defined by freehand segmentation within the dorsal subcutaneous adipose tissue.

8.4.1.3.1 Method 1 - manual segmentation of supraclavicular fat

In chapter 5 we showed that 18F-FDG BAT, when present, occurs in the supraclavicular

fossae in 92.6% cases. ROIs were defined within the supraclavicular fossae using the

‘eroded range limited’ technique validated by Lundström et al. [268, 280], based on the

assumption that BAT, if present, will most likely occur in these anatomical areas. Crude ‘BAT’ ROIs were defined manually within the supraclavicular fossae on contiguous slices, after which a lower range limit on fat fraction of 50% was set to exclude non- fatty tissues, followed by morphological edge erosion to reduce edge effects from

Figure 8-11: Crude supraclavicular ROIs superimposed over thresholded fat fraction images.

(a) Superimposition of crude ROI over the 50% thresholded fat fraction image

(b) Following application of a morphological edge erode algorithm to exclude tissue interfaces

8.4.1.3.2 Method 2 - transposition of 18F-FDG BAT ROIs from PET/CT to MR

PET/CT scans containing 18F-FDG BAT ROIs were registered to the MR scans using

Mirada XD 3.4 image fusion software (Mirada Medical Ltd, Oxford, UK) employing a combination of automatic rigid and non-rigid registration, with manual placement of fiducial markers as necessary. Where multiple positive PET/CT scans had been

performed for a patient, the index scan with the greatest 18F-FDG BAT volume was

used. The quality of image registration was verified visually using the in-built ‘chequerboard’ visualisation tool (Figure 8-12 A). The extent of deformable registration applied was also assessed using a displacement grid (Figure 8-12 B), in which the length and direction of the arrows represent the magnitude and direction of deformation.

The magnitude of deformation was compared across different anatomical areas (viz.

neck, upper and mid mediastinum, supraclavicular fossae and axillae) to determine the

impact of arm position (i.e. whether the PET/CT was acquired with the arms elevated or

alongside the torso) on the degree of deformation required to register images. Areas where image registration proved unsuccessful (typically the axillae), or where large degrees of deformation were required, were not included in further BAT analysis.

Figure 8-12: Verification of PET/CT and MR co-registration.

(a) ’Chequerboard’ validation of co-registered PET/CT and MR images

(b) Displacement grid showing the degree of deformation required to register the images

18_{F-FDG BAT ROIs defined in chapter 4 were then transposed onto the MR scans}

(Figure 8-13 a). WAT ROIs were manually defined within the dorsal subcutaneous adipose tissue, and non-adipose tissue was excluded from analysis by thresholding the fat fraction images, with a lower limit set at 50% (Figure 8-13 b).

Figure 8-13: Position of 18F-FDG BAT and subcutaneous WAT ROIs.

(a) Transposed 18F-FDG BAT ROIs (red)

transposed upon the fat fraction map (red), and manually defined WAT ROIs (yellow)

(b) 18F-FDG BAT ROIs overlaid upon the thresholded fat fraction image to exclude non- adipose tissue (black)

8.4.1.4 Statistical analysis

The impact of arm position (i.e. whether PET/CT scans were performed with the arms

up or down) on the degree of deformation required for PET/CT and MR image co- registration was analysed using Kruskal Wallis analysis with Bonferroni post tests.

Calculated fat fraction and T2* values for BAT and WAT were compared using the unpaired t-test, and fat fraction across serial scans were compared using ANOVA. Analysis was performed using GraphPad Prism version 5.00 for Windows (GraphPad

Software, San Diego, California, USA). Significance was defined as p < 0.05.

8.4.1.5 Ethics

Written informed consent was obtained from human participants and appropriate ethical approval was obtained from the National Research Ethics Committee at Birmingham East, North and Solihull Research Ethics Committee (NHS REC reference 11/H1206/3). NHS Trust approval was obtained from the Research and Development office of University Hospitals Coventry and Warwickshire NHS Trust.