MRI-based synthetic CT for assessment of the bony elements of the sacroiliac joints in children

Imaging of the sacroiliac (SI) joints in children is increasingly used for diagnosis and classification of juvenile spondyloarthritis (JSpA) [1]. Early diagnosis of sacroiliitis in children with JspA is important for therapy, especially with new biologic treatment options now available to delay progression and treat sacroiliitis in JspA [2, 3]. Historically the diagnosis of sacroiliitis was based on radiography. Radiography is not sensitive in early-stage disease [4] and has been replaced by MRI as the first line of investigation. MRI is excellent to detect active lesions of sacroiliitis and can also assess structural damage [2]. T1-weighted MRI images are the standard in evaluating structural lesions of sacroiliitis [5]. However, structural lesions including erosions, can be difficult to reliably assess on routine T1-weighted MRI scans in children [6]. The SI joint in adults is delineated as a sharply defined low signal black line on T1-weighted MRI images, which is rarely the case in children. Normal growth-related variations such as the absence of this black line representing the subchondral bone plate, blurring, and irregularity can mimic erosions, making this diagnosis more difficult in children [6, 7]. New ways to visualize and assess the bony structures of the SI joints on MRI would be helpful.

In most parts of the body, computed tomography (CT) is the method of choice to demonstrate bony anatomy [8]. In adults, CT has proven to be superior to MRI for assessing bone sclerosis and ankylosis in sacroiliitis [8]. Unfortunately, single-energy CT is not able to detect active lesions of sacroiliitis, which are key to diagnosing sacroiliitis according to the Assessment of SpondyloArthritis International Society (ASAS) criteria [9, 10]. CT can be used as a standard in adults in the evaluation of chronic sacroiliitis [1, 5, 11‐13], but it cannot routinely be used in children due to radiation exposure [2, 8].

Synthetic CT (sCT), a new MRI technique, is a deep learning-based technology performing a 3-dimensional MRI-to-CT mapping, which is learned from paired MRI and CT data, generating ‘CT-like’ images without ionizing radiation. The sCT is generated from an axial 3D T1-weighted radiofrequency spoiled multiple gradient-echo sequence (T1MGE) [14]. This technology has previously been clinically validated for the SI joints in adults but has not yet been validated for use in children [9, 14, 15].

The purpose of this study is to assess the equivalence of MRI-based sCT to conventional CT in normal SI joints in children in a semi-quantitative and quantitative assessment of bony morphology.

This prospective study was approved by the local ethics committee and written informed consent was obtained from all patients and their parents. Authors without conflicts of interest had full control of the inclusion of any data and information submitted for publication. None of the data from study participants has been reported previously.

In all patients, MRI was obtained within a maximum period of just over one month (range 5–33 days) after performing clinical CT or PET-CT scan. Six boys and 4 girls were included aged between 9 and 16 years (mean age 14 years), resulting in a total of 20 SI joints. There was no reported SI joint pathology in these patients.

In addition to the features that were scored semi-quantitatively and quantitatively, some obvious other incidental findings were depicted on CT as well as on sCT, such as spina bifida occulta (2 cases), ossification centers of the apophysis of the iliac crest (4 cases) and an enostoma (1 case) (Fig. 1).

One patient had a joint facet defect meeting the criterion of ≥ 3 mm, seen on both sCT and CT (Fig. 2). There were multiple irregularities seen on sCT and CT on the iliac and sacral side of the SI joint that did not meet the criterion of ≥ 3 mm (Fig. 3).

This prospective study was performed to test the equivalency of MRI-based sCT images with conventional CT images of (normal) SI joints in children. The sCT images were generated from the 3DT1MGE sequence using a deep learning-based method aiming at specific visualization of the osseous morphology by HU estimation [14].

We found that overall image quality was good for all bony structures on sCT and CT. Osseous morphology was correctly visualized on sCT for detection and visualization of lumbosacral transitional anomalies, fusion of sacral growth plates, cortical delineation of the SI joint, presence of ossified nuclei of the SI joint, and joint facet defects of the SI joint. Cortical delineation of the joint space was equally well seen on sCT and CT.

Subjectively, in some cases, the sCT images even demonstrated sharper cortical delineation than the CT and PET-CT images (Fig. S2). This made ossified nuclei and growth plates better visualized and more sharply delineated on sCT than on CT (Figs. 3, 4, and 5), explaining partly the variability in measurement of S1 and S2 vertebral bodies between sCT and CT. Suboptimal CT scan images may be due to the use of low-dose protocols conforming to the ALARA principle (as low as reasonably achievable), in which the lowest possible radiation dose to achieve the clinical diagnosis is used, which does not necessarily provide excellent images for evaluation of bony structures [19, 20]. This may explain why measurements of S1 and S2 vertebral bodies were not within the 1 mm equivalency margin, especially the endplate of the S2 vertebral body which was sometimes difficult to see on a midsagittal plane.

In this study, we set a very strict 1-mm equivalence margin; all measures would have been within an equivalence margin of 2 mm.

The ongoing ossification process in children can result in unsharp cortical margins, making cortical assessment difficult. Multiple irregularities of the SI joints were seen in our study that did not meet the prespecified size definition of ≥ 3 mm (Fig. 3). Only 1 cortical irregularity met the definition and was observed on sCT as well as on CT by both readers (Fig. 2).

This is the first study using MRI-based sCT in children. In adults, several studies were published comparing sCT and CT in the SI joints, spine, and pelvis [9, 15, 21‐26]. In a study of the lumbar spine and hips, geometric analysis of MRI-based sCT showed similar measurements in comparison to CT [15, 25]. Jans et al. showed that the diagnostic accuracy of sCT was higher than that of T1-weighted MRI sequences for detecting erosions, sclerosis, and ankylosis in adult patients suspected of having sacroiliitis [9]. Using sCT in children can be of great benefit in the assessment of sacroiliitis as assessing pediatric SI joints is very challenging on MRI, mainly due to normal variability. In children, blurring and irregularity of the SI joint line is frequently seen on classical T1-weighted sequences, making it difficult to detect erosions. sCT could definitely improve the assessment of joint delineation and bone erosions in pediatric SI joints [6]. Cortical delineation of the SI joint was excellent in all but one patient (who had motion artifacts) in our study on both sCT and CT.

sCT has a key advantage in that it can be scanned in one examination together with the classical MRI sequences, allowing assessment of active and structural lesions concomitantly and thus possibly better diagnostic interpretation of sacroiliitis in children with JSpA. Scan time is just 5 min 12 s longer when including the sCT sequence. Importantly, sCT comes with no ionizing radiation, which is essential in the younger population as the radiosensitive reproductive organs are in close proximity to the SI joints. Even though low-dose CT shows promising results in adults with SpA, in pediatric patients it is particularly preferable to perform an examination without ionizing radiation such as sCT [11‐13].

sCT is also a 3D modality, thus allowing for images to be reconstructed in all planes, which could be especially beneficial to evaluate the rather complex SI joints [21].

Ankylosis and bony bridges of the SI joint are known structural lesions in sacroiliitis [3, 4, 10], however they were not expected in this normal population. In our study, bony bridges were scored in only 1 patient on sCT on the left side, not on CT, resulting in a kappa value of 0 (Table 3) (Table S1). Bony bridge-like features are a known artifact on sCT and a potential pitfall that has already been described in adults when using sCT for the diagnosis of sacroiliitis. Morbée et al. have shown that the vacuum phenomenon can be falsely seen as the bony bridge on sCT in SI joints [21]. In our case too, this rather thin white “artifactual” line crossing the SI joint on sCT (Fig. 1b.) did not mimic true ankylosis as can be seen in sacroiliitis, also no other features of sacroiliitis were present. In our normal study group, we also detected an incidental enostoma (Fig. 1e–f.) near the SI joint, which was equally seen on CT and sCT. This is promising as sCT might be used as well for assessing sclerosis as a structural lesion of sacroiliitis [21].

Our study has some limitations. The main limitation is that we did not image children with sacroiliitis, hence our study is limited to assessing image quality for normal appearances and normal variation. CT scans are not commonly performed in children with sacroiliitis due to radiation dose. Also, CT and MRI did not take place at the same day. The additional MRI scan was obtained within approximately 1 month after the CT scan. Also, different kinds of CT, including PET-CT were used, depending on the clinical question.

Another limitation is the low number of examined patients in this study. CT is in most children replaced by ultrasound or MRI in abdominal and pelvic pathology. Moreover, if an abdominal or pelvic CT was nevertheless performed, critical illness prevented these children from undergoing an additional MRI examination within one month. In this study, the youngest patient was 9 years old. In even younger age groups sedation might even be necessary, which was not approved by the ethics committee.

Future studies in larger groups of children, also including pathological scans of the SI joint, are required to further examine the added value of sCT for assessment of structural lesions (erosions, ankylosis, and sclerosis) of sacroiliitis on sCT in pediatric SI joints in comparison with the classically used MRI sequences and CT as gold standard where available.

By all our measures in a group of normal children, sCT was visually equivalent to CT for assessment of the bony morphology of pediatric sacroiliac joints. In a clinical setting, adding sCT to the MRI protocol may have additional value as it allows for easily visually interpretable visualization of the bony structures of the sacroiliac joints in children without the use of ionizing radiation.