The “question-mark” MR anatomy of the cervico-thoracic ganglia complex: can it help to avoid mistaking it for a malignant lesion on ⁶⁸Ga-PSMA-11 PET/MR?

Ganglia are consolidated parts of the sympathetic network located symmetrically along the spine. At the border of the chest and neck, the sympathetic inferior cervical ganglion (ICG) and the first thoracic ganglion (T1), may fuse to form a relatively large, variedly shaped, cervico-thoracic ganglion (CTG), called also the stellate ganglion (SG) (Figure 1A). Reported percentages of such fusions were scattered from 28% by Raveendran and Kamalamma1 through about 82–83%2,3; up to 100% according to Hoffman.4

Figure 1

The exact location of CTG/SG had been studied in detail on cadavers1, 2, 3,5, 6, 7 to find reliable anatomic landmarks for the needs of their therapeutic block or ablation to relieve pain, in different diseases including cardio-vascular and post-traumatic stress disorder8, 9, 10 or to assist avoiding their iatrogenic damage during surgery.

For a long time CTG/SG were not possible to be detected on any imaging because of their relatively small size. The first imaging study claims to be the magnetic resonance (MR) one, by Hogan and Ericson.11 However, even when CTG/SG started to be visible on imaging, their proper classification was not essentially crucial from oncological point of view, as erroneous taking them for normal or benign lymph nodes did not generally cause any harm to the patient.

The circumstances changed after the introduction of a new radiotracer to multimodal positron emission tomography/computer tomography (PET/CT) imaging. The prostate specific membrane antigen (PSMA)-targeted radiotracers labelled with radioactive gallium (Ga-68) used for the primary staging and follow-up of prostate cancer patients, turned out to gather also in sympathetic ganglia, including CTG/SG.12, 13, 14, 15 In such circumstances, the relatively small size of CTG/SG does not exclude them from suspicion of malignancy, when they show the radiotracer uptake on PET, because ⁶⁸Ga-PSMA ligand (PET/CT) has the functional potential do detect micrometastases even in non-enlarged, and therefore missed on only morphological CT assessment, lymph nodes.16,17

Detection of uptake in CTG/SG on multimodal PET/CT imaging, causes the possibility of mistaking them for malignant lesions, primarily metastatic lymph nodes12,13,15, which may lead to erroneous patient management and unnecessary treatment. The proposed hints for avoiding harmful misdiagnosis included the awareness of a possible pitfall, the anatomic location of CTG/SG13, CTG/SG special shape, i.e. often met band- and tear-drop configuration, and significantly higher intensity of uptake than in prostate cancer metastatic lymph nodes.15 However, the mentioned papers dealt with PET imaging combined with CT. In our study we have taken into consideration the MR-based multimodal PET imaging in order to check the shape of cervico-thoracic ganglia complex (CTG-C), in the view of PET factors hindering its proper identification (Figure 1B-D), including fused CTG/SG, not-fused CTG/SG and their connections with adjacent sympathetic ganglia if visible.

The retrospective analysis of 212 sites of a CTG complex was undertaken in 106 patients undergoing whole body ⁶⁸Ga-PSMA-11 PET/MR examination. The retrospective study was performed in accordance with the principles of the 1964 Declaration of Helsinki and all subsequent revisions and with national regulations. All patients had provided routine written informed consent before each examination.

The patients were males (age range 40–78 years, mean 63.70 ± 6.86 years; weight range 59–115 kg, mean 86 ± 29 kg) referred for routine primary staging or follow-up of prostate cancer between November 2015 and February 2018. All exams were performed using a multimodal PET/MR system (Biograph mMR scanner, Siemens, Germany, based on the 3T MR platform). All patients underwent the whole body MR and the whole body PET imaging about 83.04±19.96 minutes (range, 50–143 minutes) after injection of 168.43 ± 17.69 MBq (range, 115–210 MBq) of ⁶⁸Ga-PSMA-11.

The procedure of ⁶⁸Ga-PSMA-11 preparation and the imaging protocol was the same as already described in Bialek and Malkowski.^{18 68}Ga-PSMA-11 was synthesized as follows. ⁶⁸Ge/⁶⁸Ga generator (Eckert & Ziegler Rdiopharma GmbH, Berlin, Germany) was eluted with 5ml of sterile, ultra-pure 0.1M hydrochloric acid, in order to obtain sterile, endotoxin-free solution of Ga-68 chloride. For labelling a vial containing 20ug sterile and endotoxin-free lyophilisate of PSMA-11 (GMP) (ABX, Radeberg, Germany) and a vial containing 60 mg of sodium acetate was used. To the above set 2ml of Ga-68 chloride was added and mixed for 10-20s. to complete dissolution. Subsequently, the mixture was incubated for 10 min. at 95°C. The labelled tracer was purified on a column of Sep-Pak Light C18 (Waters) and filtered on a 0.22 μm pore size filter (MILLEX-GV, Merck). Radiochemical purity (≥95%) was confirmed by thin-layer chromatography, checked on iTLC-SG bands in ammonium acetate-methanol (1:1) solution.

To reduce patients discomfort during the long PET/MR examination time, the patient’s arms were placed alongside the body. PET and MR imaging were performed simultaneously. MR sequences included: axial T2-wighted TSE fat-saturated 5mm-slice, 400mm field-of-view (FOV) images, respiratory gated in the region of the chest and abdomen, axial T1-weighted Vibe Dixon 3mm-slice, 430mm FOV images, breath-holded in the region of the chest and abdomen. Attenuation correction was calculated according to the manufacturer’s protocol basing on a fast 3D FLASH based MR VIBE (volume interpolated breath-hold examination) sequence.

PET imaging was performed with an acquisition time of 5 min per bed position in caudocranial direction starting from the pelvis. Acquired PET sinograms were reconstructed with the HD-PET algorithm (point-spread function) using 3 iterations, 21 subsets, a Gaussian filter: the Full Width at Half Maximum (FWHM) 4.0 mm, an image matrix of 172. Performed separately, pelvis and lower limbs PET/MR imaging were not analyzed in the current study.

Image analysis

Retrospective analysis of ⁶⁸Ga-PSMA-11 PET/MR scans, fused and not-fused, included the estimation of the radiotracer uptake (SUV_max normalized by body weight) and morphologic features (the size, shape and position) of CTG complex, as well as the symmetry of their location and symmetry of visual uptake intensity. Analysis and quantification was performed on a Syngo.via Viewer workstation (Siemens, Germany).

We used the name “cervico-thoracic ganglia complex” (CTG-C) for the complex of ganglia located in the typical area (described in the discussion), because it is not possible to ascertain reliably on a whole-body MR part of a PET/MR scan if the detected ganglia structures are fused or not, as it is not possible to follow the whole sympathetic chain. The observed complex of ganglia connected by visible strands, possibly included sometimes a T2 sympathetic ganglion or a vertebral ganglion.

Asymmetry of PSMA-ligand uptake in CTG-C (in intensity, in the level of maximal intensity, or both), as well as increased uptake in visual assessment and independently, when SUV_max amounted at least 2, were regarded as risk factors of mistaking CTG-C for metastases or other malignant lesions. The background ⁶⁸Ga-PSMA-11 activity was measured in gluteal muscles (GM).

The analysis of the CTG-C shape consisted of two steps. First, the form of all CTG-C was characterized descriptively. Subsequently, the described configurations were reevaluated and categorized. When reviewing CTG complexes we noticed repeating configurations, not previously identified, resembling the shape of a question-mark (thinner or thicker, normal, reversed, mirror or upside-down, more or less straight including bludgeon-like shape), or a part of a question-mark similar to a kidney or comma, as well as exclamation-mark forms (Figure 2). These shapes were categorized as “typical”.

Figure 2

The tumor-like shapes: oval, binodular or longitudinal, were regarded as “mistakable”, potentially suspicious of malignancy, including lymph node metastasis.

The non-specific shapes (wavy with small nodules, bent or unreliable to be assessed), were named “other”, and were also treated as not suspicious of malignancy.

On whole-body MR scans the CTG-C was identifiable in 100% on the left (L) side and in 97% (103/106) on the right (R) side, and were located symmetrically or partially symmetrically in 90% (95/106) of patients.

The mean thickness of the CTG-C, i.e. minimal transverse diameter, was 4.31 ± 1.20 mm (range, 1.5–8 mm), mean width, i.e. maximal diameter on a transverse MR plane, was 13.7 ± 5.18 mm (range, 4–29 mm) and mean length (in cranio-caudal orientation) was 13.1 ± 3.72 mm (range, 5–25 mm). Diameters with respect to the right and left side are presented in Table 1.

All identified CTG-C showed ⁶⁸Ga-PSMA-11 uptake, with mean SUV_max 2.65 ± 0.8 (range 1.06–6.23). Uptake values with respect to the right and left side are presented in Table 2.

With regard to the left CTG-C, there was a positive correlation of a medium size (r = .318; p = .001) between its thickness and SUV_max. Analysis regarding CTG-C of both sides taken together revealed a positive, statistically significant and small-size correlation with SUV_max (r = .179; p = .01). As for the right CTG-C, no correlation nor tendency was found between the thickness and SUV_max (r = .028; p = .782).

Noticeably increased, possible to be taken for malignant, ⁶⁸Ga-PSMA-11 uptake, SUV_max≥2, was detected in 87.7% (93/106) of the L-CTG-C and 74.5% (79/106) of the R-CTG-C (with respect to both sides see Figure 3); whereas the background in gluteal muscles presented mean SUV_max of 1.04 ± 0.32 (range, 0.46–1.81). In visual assessment, as routinely performed in clinical practice, suspicious of malignancy appearance of CTG-C was observed in 70.8% of the L-CTG-C (75/106) and 60.4% (64/106) of the R-CTG-C (with respect to both sides see Figure 3). The main criteria of possible malignancy on PET were the PSMA-ligand avidity and asymmetry of uptake (in the intensity, in the level of maximum intensity or both) in CTG-C (Figure 1B-D;Figure 3A,B;Figure 4A-D).

Figure 3

Figure 4

On MR part of the whole-body PET/MR, in the majority of patients (66.0% L-CTG-C, 70/106; 53.8% R-CTG-C, 57/106), the CTG complex presented the following shapes, reflecting their anatomic structure and connections: a question-mark (L-CTG-C: 23.6%, 25/106; R-CTG-C: 22.6%, 24/106), an exclamation-mark (L-CTG-C: 3.8%, 4/106; R-CTG-C: 2.8%, 3/106), or a part of a question-mark resembling kidney (L-CTG-C: 27.4%, 29/106; R-CTG-C: 12.3%, 13/106) or comma (L-CTG-C: 11.3%, 12/106; R-CTG-C: 16.0%, 17/106) (Figure 3E, F;Figure 5).

Figure 5

The tumor-like CTG-C shape, possible to be mistaken for malignancy, was detected in 28.3% (30/106) on the left side and in 40.6% (43/106) on the right side, probably due to MR artifacts obscuring finer linear elements of the CTG-C and not allowing for the tracing of the entire CTG-C tract (Figure 3C, D). The rest (L-CTG-C: 5.7%, 6/106; R-CTG-C: 5.7%, 6/106) of CTG-C presented nonspecific shape.

In cases, where visual molecular assessment of PET revealed suspicious, potentially mistakable with malignancy, increased PSMA-ligand uptake in CTG-C, the recognition of a “typical” or an “other” shape of underlying CTG-C on MR part of the examination generated a rise in the accuracy of their proper identification. For the right CTG-C from 39.6% to 76.4%, χ2(1) = 29.46; p < .001, for the left CTG-C from 29.2% to 77.4%, χ2(1) = 49.29; p < .001, for both from 34.4% to 76.8%, χ2(1) = 77.41; p < .001. The effect size was moderately large, with V = .37, V = .48 and V = .43 for the right, left and both CTG-C, respectively. Recognizing the shape of the CTG-C as a “typical” or an “other” in MR allowed us to classify as “not-suspicious” 60.9% of the right CTG-C, 68% of the left CTG-C and 64.7% of all CTG-C which were treated as “suspicious” after sole PET assessment.

In the separated evaluation of the “typical” MR CTG-C shape the rise in accuracy was from 39.6% to 76.4%, χ2(1) = 29.46; p < .001 for the right CTG-C, from 29.2% to 73.6%, χ2(1) = 41.71; p < .001 for the left CTG-C, from 34.4% to 75%, χ2(1) = 70.4; p < .001 for both. The effect size was moderately large, with V = .37, V = .44 and V = .41 for the right, left and both CTG-C, respectively. Recognizing the shape of the CTG-C as “typical” in MR allowed us to classify as “not-suspicious” 60.9% of the right CTG-C, 62.67% of the left CTG-C and 61.9% of all CTG-C which were treated as “suspicious” after sole PET assessment.

The CTG/SG can be identified along the sympathetic chain on the border between the neck and thorax, anteriorly and slightly caudad to the head of the first rib11,12, inferior and anterior to the transverse process of C712,19, lateral and posterior to the lateral edge of the longus colli muscle (at the level of T1)11, inferior to the subclavian artery12, medial, posterior, medial and anterior or medial and posterior to the vertebral artery3,11. Which was confirmed also in the current study.

In our study we have tracked down a configuration of the cervico-thoracic ganglia complex not categorized in previous MR study exploring CTG/SG.11,19 The majority of CG-C in our work resembled a question-mark in various configurations: thinner or thicker, mirror, reversed or mirror and upside-down, more or less straight including bludgeon-like shape and an exclamation-mark, or a part of a question-mark: without a dot, resembling kidney or comma. Our “dot” in the question-mark or exclamation-mark sign might be T1 sympathetic ganglion (when CTG is not fused) or T2 ganglion (when CTG is fused) in case of a normal or a mirror shape, or a vertebral ganglion – in case of an upside-down appearance.

Even the oval, binodular or longitudinal shapes, classified as tumor-like, were usually partially surrounded by unsuccessful fat saturation artifacts, which probably obscured the finer sympathetic branches or chain and in more favorable conditions might let the recognition of a “typical” shape.

Many of presented in our work shapes or their parts correspond to the shapes described previously on CT or anatomical studies.15

On CT it is not possible to trace fine structures as nerves and their connections, therefore only thicker oval or longitudinal parts of a neuronal network are identifiable. That is the reason why MR-based multimodal PET/MR, which can depict strand of the sympathetic chain, may be more specific in identifying sympathetic ganglia. Nevertheless, the data available with PET/CT have been already proved to be entirely sufficient for correct interpretation of findings.15

The analysis of the CTG-C shape appears important in the context of potentially suggesting metastasis or other malignancy PSMA-ligand uptake on multimodal PET/MR, which was present in our study in 65.6% in visual assessment and in 81.1% when regarding abnormal SUV_max of at least 2. We have chosen the SUV_max 2, because it was used also in some previous papers as a cut-off value for prostate cancer metastases16,20,21 and because it constitutes the reported lower range in lymph nodes metastases in prostate cancer14,22, as well as it exceeded the upper range of the background SUV_max (1.81) in our study.

Elevated PSMA-ligand uptake in CTG/SG may be properly recognized as not alarming under a few conditions. The most important of them is awareness of the diagnostic reader that such a possibility exists, second is the typical location, subsequently – symmetry. The listed aspects were already underlined by authors of PET/CT studies.13,15 Further aspects, including the likelihood on prostate cancer metastases on the level of the cervico-thoracic junction and the intensity of prostate cancer metastases15 are discussed below.

We have observed, that morphological symmetry of CTG-C location (beginning and ending on the same or very similar level) did not necessarily overlap with the point of maximal PSMA-ligand uptake, therefore morphologically symmetric CTG-C seemed often to have asymmetric foci of maximal uptake when viewed on fused PET/MR images, facilitating a diagnostic mistake. Therefore, the additional observation of a specific underlying MR shape below the suspicious uptake on PET constitutes the additional useful tip aiding proper diagnosis. Reasons for asymmetry of CTG-C location on MR might include also, apart from true anatomic asymmetry, the asymmetric patient position resulting in the transverse MR plane not exactly perpendicular do the true long axis of the body.

Rischpler et al.15 in the ample PET/CT study reported cervical sympathetic ganglia as having often a characteristic band-shaped (57.5%) and teardrop (38.5%) configuration. The band-shape may reflect described in our current study elements of a question mark and a tear-drop may be close to our comma shape.

Shapes of SG described in MR-based study encompassed: fusiform, triangular, or globular but concerned only 9 volunteers.11

Shapes of SG described in anatomical cadaver studies included: spindle, dumbbell, and an inverted “L” shape (with the two latest demonstrating a definite “waist”)7, to which Marcer et al.6 had added a perforated (by the vertebral artery usually at the superior portion) and a truncated form (where the vertebral artery created a shortened and flattened form of the ganglion). Kwon et al.5 categorized SG shapes as fusiform-rounded, fusiform-elongated, and bilobed.

However, the shapes assessed basing on dissection studies may be deprived of their fine spacial arrangement visible on MR, but lost during preparation. The choice of the classification is subjective and may be free and depends on the imagination of the author. Comparing between pathology and imaging bears the burden of different condition of the organism at the time of examination, and therefore the results may be difficult to equate. The overall better tissue delineation on MR does not take the general diagnostic sufficiency away from CT.

The binodular shape in our study may reflect anatomically described dumbbell or bilobed form. The fusiform shape after dissection may lose its bent projection depicted on MR in the form of a question-mark. The truncated or “L” shape form probably might partially correspond to the kidney shape on MR. The association of CTG-C shape with the “question-mark” does not claim to be a novel discovery of that shape itself, but may better help an unexperienced reader to remember the problem.

Nevertheless, the results of previous papers prove, that data available with application of PET/CT are sufficient to correctly classify sympathetic ganglia on multimodal imaging with PSMA-ligands.15 Especially that in case of prostate cancer, metastases as high as at the level of CTG-C are not so often and that prostate cancer metastases adjacent to ganglia show significantly more intense PSMA-ligand uptake15, and we would like to stress the above conclusion.

The thickness range of CTG-C was in our study 1.5–8 mm, whereas Perlov and Vehe (1935) reported 3–10 mm for satellite ganglia. The mean thickness of CTG-C in our study (4.31 mm) was in close concordance with anatomical cadaver studies (4.5 mm by Marcer et al.6), but smaller than 8 mm reported by Jamieson et al.2 The other diameters also differed in comparison with cadaver studies, the length was shorter, which was most probably caused by the different character of the study method: section – and imaging, and possibly because of different measurement policies and the measurement plane. Similarly, Hogan and Erickson11 in MR study reported obtaining smaller cephalocaudal SG dimension than in previous dissection studies, of just over 1cm, which corresponded to our results of about 13 mm.

In our MR-based study mean SUV_max was similar, and only slightly higher (2.75 in the L-CTG-C and 2.54 in the R-CTG-C) than reported in CT-based multimodal PET studies: by Kanthan et al.13: 2.4 in the L-CTG-C and 2.2 in the R-CTG-C, and by Rischpler et al.15: 2.4.

The indisputable and undeniable, however unavoidable shortcoming of current study, identically to previous studies, is the impossibility of reliable confirmation of the true nature of studied ganglia, for obvious reasons of unattainable histopathological result. A few researches in previous works concerning celiac sympathetic ganglia tried to make a restitution for that by examining cadavers with CT or MRI and later performing the HP analysis.14,23 Their conclusions confirmed the sufficiency of using anatomical landmarks for locating the ganglia.

To sum up, special properties of a novel PET PSMA-based radiotracer, physiologically accumulating also in anatomically normal sympathetic ganglia, including cervico-thoracic ones, impelled researchers to study meticulously all possible properties of these ganglia to avoid mistaking them for malignancy. In our study we associated discovered detailed anatomical CTG-C shapes with a form of a question-mark, exclamation-mark or a part of a question-mark to facilitate an unexperienced reader proper interpretation of the PSMA-ligand PET/MR examination.

The characteristic shape of the cervico-thoracic ganglia complex (resembling a question-mark or its part), reflecting CTG complex anatomy, helps in proper recognition of CTG-C on multimodal whole-body ⁶⁸Ga-PSMA-ligand PET/MR imaging, when detectable uptake might lead to considering pathology.