This study is the first to evaluate the specific properties of colour-coded DECT lesions in gout patients. Our study demonstrates that colour-coded DECT lesions in gout patients are heterogeneous in properties. Some colour-coded DECT lesions fulfil all characteristics of pure MSU depositions and therefore can be classified as “definite”, whereas other lesions have a higher DECT ratio and therefore must contain high-Zeff materials such as calcium and do not necessarily contain MSU crystals.
Most previous studies have not focused on the properties of colour-coded DECT lesions, but rather on the diagnostic accuracy of DECT scans in predicting gout at a patient level. Several studies have evaluated the specificity of colour-coded DECT lesions using either MSU microscopy [10,11,12,13,14] or other criteria as reference standards (e.g. physician evaluated diagnosis, ACR 1977 gout classification criteria) [14,15,16,17,18]. A recent systematic literature review calculated the pooled sensitivity and specificity of DECT examinations to be 0.81 and 0.91, respectively [19]. However, specificities for colour-coded DECT lesions varied markedly in these studies (0.48–1.00) [10,11,12,13,14,15,16,17,18,19] reflecting a varying proportion of false-positive findings, which could indicate that not all colour-coded DECT lesions truly represent MSU depositions. Bongartz et al. [11] showed that DECT demonstrated colour-coded lesions in 7 out of 41 non-gout patients, and all of these lesions were found within the cartilage/menisci in patients with knee osteoarthritis. The authors therefore concluded that DECT may have limited specificity in knee osteoarthritis [11]. In our patient cohort, only one non-gout patient showed colour-coded DECT lesions, and therefore, no conclusions can be drawn regarding diagnostic sensitivity or specificity of DECT scans at the patient level. However, the property analysis showed that the colour-coded DECT lesions in this non-gout patient had markedly different properties when compared to lesions in gout patients, where especially the mean DECT ratio was higher (1.26 vs. 1.06). This non-gout patient had a joint puncture negative for MSU crystals but positive for calcium pyrophosphate (CPP) crystals, and the patient did not fulfil the ACR/EULAR 2015 gout classification criteria neither at time of inclusion nor after 1 year. It is therefore possible that the colour-coded DECT lesions in this patient in fact represented CPP crystals rather than MSU crystals, but future studies are needed to investigate the discriminatory ability of DECT to distinguish these two crystal types.
Another explanation for colour-coded DECT lesions in non-gout patients might be small MSU depositions causing only “subclinical disease”, and some authors suggest this as an explanation for the false-positive findings [7]. However, this statement is not based on the assessment of lesion properties. When applying a DECT ratio cut-off well above 1.0, which is used in most studies [11, 12, 14, 18], one would expect some of the colour-coded DECT lesions to be other materials than MSU depositions (e.g. calcium-containing), since pure MSU crystals show equal HU values when scanned at high- and low-kV series (DECT ratio ≈ 1) [6]. The approximated but heavy right tailed normal distribution of DECT ratios in our study underlines that some colour-coded DECT lesions do contain high-Zeff materials and not necessarily represent MSU depositions.
In our study, we did not assess the effect of altering neither the DECT ratio cut-off nor the HU threshold, as the aim of our study was to evaluate colour-coded DECT lesions using factory default gout settings. However, two recent studies have investigated the effect of changing postprocessing protocols when performing DECT examinations [20, 21]. Both studies investigated the effect of lowering the threshold of attenuation from 150 HU to 120 HU [20] or 130 HU [21], respectively. Both studies found this to result in an improved visualization of MSU depositions, but lowering of HU threshold also resulted in an increased amount of artefacts, especially artefacts in well-known locations such as tendons, nail-beds and skin [20]. One of the studies found that the artefacts lead to misclassifications of patients [21], whereas the other study found specificity of DECT examination to remain unchanged [20].
Colour-coded DECT lesions were in our study most commonly found in the knee, MTP1 and midtarsal joints along with the quadriceps and patella tendons. These lesion locations are partly in line with the findings from another study evaluating the distribution of colour-coded DECT lesions [22]. This study also found the MTP1 joint to be the most common site for lesions involved in 57% of patients, while the knee and tarsal joints were among the 10 most prevalent locations [22]. Tendon lesions were seen more than twice as often in the Achilles tendon (36%) than in the quadriceps (16%) or patella tendons (12%) [22], whereas our study showed the regions to be involved equally frequent (52%, 52% and 48%, respectively).
Our subgroup analyses revealed that some common locations for colour-coded lesions (knee joint, midtarsal joints and quadriceps tendon) showed both definite MSU depositions and uncertain lesions, whereas two common locations for colour-coded lesions (MTP1 joint and patella tendon) exclusively showed definite MSU depositions. In line with our results, Bongartz et al. [11] found the anatomic region used for diagnostic assessment of importance for the overall specificity of the DECT examination. Bongartz et al. focused their primary analysis on an index joint, where DECT scans had a sensitivity of 0.90 and a specificity of 0.83 when compared to MSU crystal microscopy. Secondary analyses expanded the examination to include all green pixels of the scanned area (primarily feet and knees) resulting in a higher sensitivity (0.95) but a significant drop in specificity (0.56) [11].
In order to avoid false-positive DECT examinations, a specificity-optimized evaluation set for reading DECT scans of gout patients would be beneficial. One approach could be to simply lower the DECT ratio in the factory default gout settings, thereby excluding more high-Zeff lesions. However, our study does not support this strategy, since this would not have excluded the non-gout MSU-imitating lesions. Instead, a sole focus on the MTP1 joint or the patella tendon would have increased the specificity without reducing sensitivity of DECT examination in gout patients. In order to avoid image noise artefacts, which are often seen when analysing small lesions [7], we furthermore—based on lesion properties—introduced a size criterion where we excluded lesions < 6 voxels. For measurements in a two-dimensional plane, this would equal ≥ 1 mm in lesion diameter. This proposed size criterion is identical to previously proposed size criteria. Mallinson et al. [7] proposed several potential artefacts, and these artefacts are also included in the ACR/EULAR 2015 gout classification criteria. Here it is stated that “nailbed, submillimeter, skin, motion, beam hardening, and vascular artifacts should not be interpreted as DECT evidence of urate deposition” [2], where submillimeter artefacts are described as “colouring of single pixels or areas smaller than 1 mm” [7]. The size criterion proposed in our study did not change our findings at the patient level, since all patients—which had colour-coded DECT lesions—also had lesions > 5 voxels/≥ 1 mm. A specificity-optimized evaluation set for DECT examinations based on our patient cohort would therefore include only the evaluation of large lesions (diameter ≥ 1 mm) located in either the MTP1 joints or the patella tendons, but this evaluation set requires further prospective validation.
The strengths of our study include that quantitative property analysis for all colour-coded DECT lesions was combined with location analysis in a large number of joint/tendon regions enabling us to determine the distribution pattern of lesions differing in DECT properties. Since quantitative lesion properties were automatically included in the analyses (except for obvious artefacts like nail-bed artefacts), no reader variance was introduced, thereby making interreader reliability assessment unnecessary. Diagnosis of all patients was based on MSU crystal microscopy thereby securing that all study-defined gout patients truly had gout. The DECT scanner used in our study (Siemens Somatom Force) was a third-generation dual-source DECT scanner, which has higher spectral separation and higher precision [23, 24] compared with DECT scanners used in other studies [10,11,12,13,14,15,16,17,18]. Image noise can partly be reduced through improved postprocessing software and through increased spectral separation in the newer generations of DECT scanners which by itself reduces image noise in dual-energy calculations [23]. Emerging techniques may allow for dramatically improved spectral imaging, as it is seen in photon-counting CT, where every pixel gives exact physical material and/or tissue information [25]. In gout, this results in images with a finely detailed punctate pattern of MSU crystal depositions in contrast to the clump-like appearance on DECT [26].
A major limitation of this study was the small number of patients included. Although the numbers of lesions were high (4033 lesions), they were derived from a small number of patients (n = 22), where especially the numbers of non-gout patients showing colour-coded DECT lesions were low (n = 1). However, the aim of this study was not to establish the diagnostic accuracy of DECT at the patient level but to evaluate the properties of colour-coded DECT lesions in gout patients. A thorough investigation of the properties of colour-coded DECT lesions across joint/tendon regions in non-gout patients compared to such lesions in gout patients should be evaluated in future studies.