Population groups
Two groups of participants were included: six healthy volunteers and ten patients with a thyrotropinoma (subsequently confirmed histologically following pituitary surgery). The healthy volunteers were control subjects in a study which included an assessment of the reproducibility of Met-PET findings on scans performed at a one-month interval (Clinical Trial ID: Research Registry 2070). Data from healthy volunteers allowed the most appropriate reference region for normalization of the pituitary uptake to be determined. The patient group was used to explore which subtraction technique provided the clearest localization of the pituitary tumor (matched with findings at surgery). The cohort of patients with TSH-secreting pituitary adenomas included 2 men and 8 women, with a mean age of 58 years (range 27–75). In each case, the patient had confirmed hyperthyroxinemia (raised free thyroid hormone levels) with non-suppressed TSH and met clinical and laboratory criteria for a diagnosis of autonomous tumoral TSH secretion [22]. All patients had microadenomas as reported by consultant neuroradiologists with expertise in pituitary disease and were subsequently reviewed at the regional pituitary multi-disciplinary team meeting. In four patients, these were visualized on standard clinical MRI [Spin Echo T1-weighted (T1SE) pre- and post-gadolinium and Fast T2-weighted SE (T2) FSE] and ranged from 6 to 10 mm in maximum diameter. A further three microadenomas were visualized on gradient recalled echo MRI (GRE, i.e., volumetric sequences used for PET to MRI registration) and ranged from 3 to 4 mm in maximum diameter. In the other three patients, the location of the microadenoma (4–5 mm in maximum diameter) could only be appreciated on GRE MRI when subsequently identified on Met-PET. Met-PET identified a clear adenoma in five patients, with suspected adenomas in a further three patients, but no definite abnormality in the remaining two cases. Each patient was treated with depot first generation somatostatin receptor ligand (SRL) (Lanreotide Autogel®, 90 mg 4-weekly for 3 doses) as per standard clinical practice in preparation for surgery (to alleviate symptoms and mitigate the perioperative risks associated with uncontrolled thyrotoxicosis) [22]. In eight patients, the suspected site of a TSH-secreting microadenoma as visualized on MRI and PET was confirmed on histology following transsphenoidal surgery. Two patients elected to continue with primary somatostatin receptor ligand therapy and both achieved sustained normalization of free thyroid hormone and TSH levels consistent with a tumoral origin of autonomous TSH secretion [23, 24]. These subjects were scanned according to clinical protocols prior to and following SRL therapy, which was anticipated to suppress Met-PET uptake in the tumor. This suppression of tumor function was confirmed clinically (resolution of symptoms) and biochemically (with normalization of thyroid hormone levels). Measurement of other pituitary hormones confirmed no other significant change in pituitary gland status.
Imaging procedure
Imaging was performed using a Discovery 690 PET/CT scanner (GE Healthcare, Chicago, Illinois, USA) 20 min (range 19–21 min) after administration of 382 MBq (range 293–411 MBq) of 11C-Methionine. One 15 cm bed position centered on the subject’s pituitary gland was acquired for 20 min. The images were reconstructed with OSEM iterative reconstruction using 3 iterations and 24 subsets, 128 × 128 matrix size, 2 mm Gaussian filter, scatter correction, CT-measured attenuation correction, time of flight and point spread function correction (SharpIR, GE). An unenhanced CT scan was acquired with 140 kV, fixed mA of 220, a rotation speed of 0.5 s, a pitch of 0984:1, 30 cm field of view, a slice thickness of 1.25 mm and a 1.25 mm spacing interval and reconstructed using filtered back projection.
MR imaging
MRI was performed with either a GE Optima™ MR450w 1.5 Tesla scanner (GE Healthcare, Chicago, Illinois, USA) or a GE Signa™ 3.0 Tesla scanner (GE Healthcare, Chicago, Illinois, USA) with a head coil using a fast spoiled gradient (recalled) echo (FSPGR) sequence. The sequence parameters were repetition time (TR) 11.5 ms, echo time (TE) 4.2 ms, slice thickness 1 mm and 256 × 256 matrix with 1 mm × 1 mm pixels. The images were acquired after a contrast injection of 0.1 mmol/kg gadobutrol (Gad).
Image registration
All registrations were performed in 3D Slicer [25] (version 4.10.2, 05–2019) using multi-modality rigid registration with six degrees of freedom, a maximum number of iterations of 1000 and a sampling ratio of 0.1%. To generate the required co-registered baseline and on-suppression Met-PET/MR images, the following steps were used (illustrated in Fig. 1).
-
1.
The on-suppression MR (MR2) was registered to the baseline MR (MR1) resulting in MR2 [Co-Reg].
-
2.
The baseline Met-PET (PET1) was registered to the baseline MR (MR1) to create PET1 [Co-Reg].
-
3.
The on-suppression PET (PET2) was registered to the on-suppression MR (MR2 [Co-Reg]) resulting in PET2 [Co-Reg].
-
4.
The PET2 [Co-Reg] was resampled into a matrix of the same size and shape as PET1 [Co-Reg] to create PET2 [Resampled].
Step 4 was required to enable voxel-wise manipulations to produce the subtraction images without the requirement for position and interpolation calculations while using the world coordinate system.
PET normalization
Normalized PET images (SUVr) were created by dividing the voxel values of the PET image by the mean voxel value in a reference region (Fig. 2). This process was implemented in a scripted 3D Slicer module (Fig. 2). The regions were drawn using a local thresholding tool on one or two representative slices of the PET images. The location of this reference region was optimized by finding which of the potential reference regions used for normalization (cerebellum, pons and gray matter) gave the most consistent pituitary tracer uptake in scans from healthy volunteers imaged one month apart. Once the images had been normalized to the reference regions, consistency was measured by finding the absolute percentage difference between the maximum signal in the pituitary gland at both time points. The region with the lowest mean absolute difference was considered the most appropriate region for normalization.
Subtraction images
Subtraction images were created by subtracting the on-suppression Met-PET image (PET2[Resampled]) from the baseline Met-PET image (PET1[Co-Reg]) (Fig. 3). This process was performed using SUVbw Met-PET images and the normalized SUVr (using the optimal reference region) Met-PET images to create two subtraction images for comparison (known as SUVbwSub and SUVrSub, respectively). This process was scripted using 3D Slicer’s native Python environment. The entire process, including registration, was completed in less than 5 min.
Assessment methodology
To quantitatively compare subtraction images using SUVbw images and SUVr images, ten patients who had confirmed lateralization of a microadenoma (< 10 mm) were assessed using a semi-automated technique. This was undertaken by assessing lateralization using a modified contrast-to-noise ratio equation which selected the maximum signal from each side of the pituitary gland. The midline of the gland was defined on the FSPGR MR sequences as the point of insertion of the infundibulum (pituitary stalk) viewed in the coronal plane (Fig. 4a). Thereafter, using the sagittal plane (so that the midline position could not be altered), the center of the gland was located (Fig. 4b). A rectangular volume of interest was centered at this point and was used to bisect the gland (Fig. 4c). The maximum value in each side of the gland was found, and the contrast between these maximum values was taken as a measure of lateralization. This contrast was then divided by the noise (standard deviation in the normalization ROI) to give a contrast-to-noise ratio (CNR) (see Eq. 1).
$${\text{CNR}} = ({\text{Max}}\;{\text{Signal}} - {\text{Max }}\;{\text{Contralateral}}\;{\text{ Signal}})/{\text{Standard}}\;{\text{ Deviation}}\;{\text{ in}}\;{\text{ reference }}\;{\text{ROI}}$$
(1)
where Max Signal can be either in the left or the right side of the pituitary gland (relative to the pituitary stalk) and the Max Contralateral Signal is from the opposite side of the gland to the Max Signal.
The mean CNR for both techniques was found, and the technique with the highest value was considered the most appropriate for highlighting the differences.