Abstract
Objective:
To explore the value of the quantitative parameter of intravoxel incoherent motion diffusion (IVIM-DWI) at 3.0 T MRI of the sacroiliac joint in differentiating different disease activity statuses of ankylosing spondylitis (AS) and to compare it with traditional diffusion-weighted imaging (DWI) and Spondyloarthritis Research Consortium of Canada (SPARCC) score.
Methods:
56 AS patients (active group, inactive group) and 24 healthy controls were included. Clinical data, quantitative parameters of IVIM-DWI MR images and the SPARCC scores were collected. The Kruskal–Wallis test was used to compare the differences between the groups. Receiver operating characteristic (ROC) curve analysis of histogram data and the SPARCC scores identified the efficacy of the three groups. The Spearman correlation coefficients were used to analyse the correlation between the quantitative IVIIM-DWI parameters and the SPARCC score.
Results:
The f (10th percentile) and SPARCC score of the active group were significantly higher than those of the inactive group. The f (10th, 25th, 50th percentiles), Dslow (average, entropy, 10th ~ 90 th percentiles), Dfast (kurtosis, skewness), ADC (average, 10th ~ 90 th percentiles) and the SPARCC score of the active group were significantly higher than the control group (p < 0.05). The AUC of the SPARCC score was the highest (0.799) in the identification between the active and inactive groups, and the sensitivity and specificity were 69.23 and 82.35%, respectively, at the cut-off value of 12. The SPARCC score was positively correlated with each percentile and the average value.
Conclusions:
Quantitative IVIIM-DWI parameters are helpful for the identification of different AS disease activity levels and are superior to traditional DWI. IVIM-DWI quantitative parameters had a good correlation with the SPARCC score.
Advances in knowledge:
A new MR technology-quantitative parameters of IVIM-DWI contribute to the identification of AS disease activity. IVIM-DWI quantitative parameters were well correlated with the SPARCC score.
Introduction
Ankylosing spondylitis (AS), a common chronic inflammatory rheumatic disease, has a prevalence of 0.1–1% worldwide, and the rate remains on the rise. 1 Disease activity was confirmed to be related to anxiety, depression, poor quality of life and work productivity. 2 As AS occurs alternately in both the active and chronic phases and the active phase will accelerate the development of the disease, it is particularly important to accurately judge the active phase and conduct timely intervention treatment.
At present, the active diagnosis of AS mainly depends on the activity score, sacroiliac joint MRI and laboratory examination. The AS activity scores commonly used include the Bath Ankylosing Spondylitis Disease Activity Index (BASDAI) and the Ankylosing Spondylitis Disease Activity Score (ASDAS). They are convenient and easy to operate but lack objectivity. 3 Laboratory tests such as erythrocyte sedimentation rate (ESR) and C-reactive protein (CRP) have certain sensitivity but low specificity. 4 The Spondyloarthritis Research Consortium of Canada (SPARCC) score is highly sensitive in detecting changes in AS activity and has high reliability among different scorers, 5 which is positively correlated with the BASDAI and ASDAS. 6,7
MRI is now widely used to detect sacroiliitis with AS because of its non-invasiveness and sensitivity. The short tau inversion recovery (STIR) sequence and fat suppression (FS) sequence are routinely used to detect the activity of AS, but they are not accurate enough. Published studies 8,9 believe that dynamic contrast-enhanced MRI (DCE-MRI) can detect the disease activity of AS. However, it is not appropriate to routinely apply in the imaging of AS because exogenous contrast agents will increase the risk of renal fibrosis. 10
Studies 11 have shown that diffusion-weighted imaging (DWI) is effective in diagnosing the activity of AS. Gaspersic et al 12 showed that the apparent diffusion coefficient (ADC) could be used to effectively monitor the therapeutic effect of AS. However, DWI cannot accurately reflect the diffusion information of pure water molecules in biological tissues. Intravoxel incoherent motion diffusion-weighted imaging (IVIM-DWI), which was proposed by Lebihan, 13 was originally used in the nervous system. It could distinguish the true-diffusion coefficient (Dslow), false-diffusion coefficient (Dfast), and tissue perfusion fraction (f).
This study attempted to outline the full-layer region of interest (ROI) and measure the histogram data of the IVIM-DWI parameters and to explore the identification ability of IVIM-DWI quantitative parameters between the groups with different AS activities. This is different from Zhao’s study 14 (conducted by outlining the IVIM-DWI quantitative parameters of single-layer ROI measurement). Moreover, we used the ASDAS to group the AS group, which is different from the BASDAI used by Zhao, and included CRP or ESR indicators on a patient’s subjective basis, which may make the results more accurate.
The purpose of our study was to evaluate the activity of AS using a new MR technique for clinical diagnosis and prognosis.
Methods and materials
Subjects
AS group
56 patients diagnosed with AS in the rheumatology and immunology department of the Second Affiliated Hospital of Soochow University from December 2019 to January 2021 were collected continuously, including 35 males and 21 females, aged 22–57 years old, with an average age of 33.8 ± 9.1 years old. The inclusion criteria were as follows: (1) a diagnosis of AS according to the Assessment in Spondyloarthritis International Society (ASAS) 2009 axial spondyloarthritis (axSpA) classification criteria; 15 (2) consistent scanning methods and parameters across all patients, yielding images that met the requirements for analysis. The exclusion criteria were as follows: (1) poor MR image quality and difficulty outlining the ROI; (2) recent trauma or infection; (3) history of rheumatism or tumour; and (4) contraindications to MRI scanning.
Control group
24 healthy controls were recruited during the corresponding period, including 14 males and 10 females, aged 23–43 years old, with an average age of 29.71 ± 6.05 years old. The inclusion criteria were as follows: (1) no history of rheumatism or tumour; (2) no history of lumbosacral pain; and (3) no recent trauma or infection. The exclusion criteria were as follows: (1) poor MR image quality and difficulty outlining the ROI; and (2) contraindications to MRI scanning.
Clinical data and groups
The name, gender, age, height and weight of each patient in the AS group and control group were recorded. CRP, ESR and HLA-B27 were recorded in the AS group within 1 week after MR examination, and ASDAS was performed in the AS group. A total of three groups: active group (39 cases): ASDAS-CRP ≥1.3; inactive group (17 cases): ASDAS-CRP <1.3; healthy control group (24 cases).
MRI protocols
All subjects underwent MR imaging on a 3T scanner (MAGNETOM Prisma, Siemens Healthcare, Germany) with a 32-channel body coil and 18-channel abdominal coil. The B value is (0, 10, 20, 30, 50, 80, 100, 200, 300, 600, 800, 1000 and 1500 s/mm2). The scanning parameters of each sequence are shown in Table 1.
Table 1.
Parameters of MRI protocols
| T 2W-TSE | T 1W-TSE | T 2W-FS | T 2W-FS | IVIM-DWI | |
|---|---|---|---|---|---|
| Plane | Oblique coronary | Oblique coronary | Oblique coronary | Axial | Axial |
| Repetition time (ms) | 3200 | 500 | 3200 | 3200 | 5100 |
| Echo time (ms) | 68 | 9.9 | 68 | 68 | 51 |
| Matrix | 320 × 256 | 320 × 256 | 320 × 256 | 320 × 256 | 134 × 134 |
| Field of view (mm²) | 210 × 210 | 210 × 210 | 210 × 210 | 210 × 210 | 380 × 306 |
| Thickness (mm) | 3 | 3 | 3 | 4 | 4 |
| slice gap (mm) | 0.3 | 0.3 | 0.3 | 0.3 | 0.3 |
| Time (min) | 3:54 | 2:27 | 3:54 | 3:54 | 6:06 |
FS, fat suppression; IVIM-DWI, intravoxel incoherent motion diffusion-weighted imaging; TSE, turbo spin echo.
The SPARCC score
The SPARCC score was calculated for all the patients in the AS group. The scoring rules were as follows: six sacroiliac joint lipid-suppression T 2WI serial oblique coronal scans, which showed a larger layer of synovium. The score was calculated according to the following three aspects: (1) the score of the range of involvement: each side of the sacroiliac joint at each level was divided into four quadrants. The score of high signal bone marrow oedema in each area was one point, and no score was 0 points. The total score of bilateral sacroiliac joints at 6 levels was 48 points. (2) Score of oedema intensity: one point was added to the signal intensity of each side of the sacroiliac joint lesion at each level, which was close to or higher than the signal intensity of the anterior iliac vein at the same level. The total score of bilateral sacroiliac joints at 6 levels was 12 points; (3) Score for the depth of oedema: one point was added to the depth of oedema lesions on each side of the sacroiliac joint at each level, and the total score of bilateral sacroiliac joints at 6 levels was 12 points. The total score was 72 points, including the range of involvement, intensity of oedema and depth of oedema.
Image analyses
ROI segmentation and image processing
IVIM-DWI images of subjects were uploaded to Firevoxel software. Combined with T 2WI-FS oblique coronal and T 2WI-FS axial images, all bone marrow oedema areas with high signal under bilateral sacroiliac joint surface were selected as ROI regions. On the IVIM-DWI image with a b value of 1500 s/mm2, the lesions were delineated layer by layer along the edge, while the bone cortex and necrotic cystic lesion areas were avoided Figure 1. In the control group and the patient group without obvious bone marrow oedema, four ROIs were placed symmetrically under the bilateral sacroiliac joint surface, with three layers on each side and attention was paid to avoiding the bone cortex 16 (Figures 2 and 3). Two radiologists with 3 and 8 years of experience delineated the ROI simultaneously, and the three-dimensional ROI was obtained by multilayer fusion.
Figure 1.
A~ F A male, 32 years old, with AS in the active group; A: Axial T 2WI-FS shows bilateral bone marrow oedema under the sacroiliac articular surface. B: Schematic diagram of ROI selection on IVIM-DWI image with b value of 1500 mm2/s; C ~ F: False colour images of Dslow, Dfast, f and ADC of the lesion area, with mean values of 0.543 × 10−3 mm2/s, 61.461 × 10−3 mm2/s, 0.185 and 0.804 × 10−3 mm2/s, respectively. ADC, apparent diffusion coefficient; AS, ankylosing spondylitis; IVIM-DWI, intravoxel incoherent motion diffusion-weighted imaging; FS, fat suppression; ROI, region of interest; T 2WI, T 2 weighted imaging.
Figure 2.
A ~ F An inactive group of AS patients, male, 24 years old; A: Axial T2WI-FS shows the bilateral sacroiliac articular surface. B: Schematic diagram of ROI selection on IVIM-DWI image with b value of 1500 mm2/s; C ~ F: False colour images of Dslow, Dfast, f and ADC of the lesion area, with average values of 0.314 × 10−3 mm2/s, 146.805 × 10−3 mm2/s, 0.123 and 0.392 × 10−3 mm2/s, respectively. ADC, apparent diffusion coefficient; AS, ankylosing spondylitis; IVIM-DWI, intravoxel incoherent motion diffusion-weighted imaging; FS, fat suppression; ROI, region of interest; T 2WI, T 2 weighted imaging.
Figure 3.
A~ F A healthy control group of AS patients, female, 24 years old; A: Axial T2WI-FS shows the bilateral sacroiliac articular surface. B: Schematic diagram of ROI selection on IVIM-DWI image with b value of 1500 mm2/s; C ~ F: False colour images of Dslow, Dfast, f and ADC of the lesion area, with average values of 0.261 × 10−3 mm2/s, 145.287 × 10−3 mm2/s, 0.110 and 0.323 × 10−3 mm2/s, respectively. ADC, apparent diffusion coefficient; AS, ankylosing spondylitis; IVIM-DWI, intravoxel incoherent motion diffusion-weighted imaging; FS, fat suppression; ROI, region of interest; T 2WI, T 2 weighted imaging.
Dslow, Dfast and f were calculated according to the diffusion-weighted double exponential model as Sb/S0=(1 f)exp(-b·Dslow)+f·exp[-b(Dfast)]. 13 Using the DWI single exponential model as Sb/S0 = exp(-b*ADC), 17 ADC values are obtained on IVIM-DWI images with b values of 0 and 1500 s/mm2 respectively. Dslow is the true-diffusion coefficient related to the diffusion of water molecules, Dfast is the pseudo-diffusion coefficient related to the microcirculation perfusion, f represents the percentage of the microcirculation perfusion effect in the overall diffusion effect, and ADC is the apparent diffusion coefficient.
Statistical analysis
Statistical analysis was performed using SPSS 23.0 and MedCalc 19.6.4. The Shapiro–Wilk test was used to test the normality of continuous measurement data, and Levene’s test was used to test the homogeneity of variance of the data. Data consistent with a normal distribution are represented by ‾x ± s, while those inconsistent with a normal distribution are represented as the median±interquartile interval. The χ2 test was used to compare the counting data. The Mann–Whitney U test was used to compare the difference in measurement data between the inactive group and the active group. The consistency test of quantitative parameter histogram data obtained from ROIs was carried out by two imaging physicians. An intraclass correlation coefficient (ICC) ≥0.75 indicated good consistency. The Kruskal–Wallis test was used to compare the differences in measurement data between the active group, the inactive group, and the control group, and the Bonferroni method was used for correction. Receiver operating characteristic (ROC) analysis was applied to quantitative parameter histogram data and the SPARCC scores to identify the efficacy of the active, inactive, and control groups and to calculate the area under the curve (AUC). The maximum Youden index was used to determine the critical value, and the sensitivity and specificity of each parameter for differential diagnosis were calculated. The Spearman correlation coefficients were used to analyse the correlation between IVIM-DWI quantitative parameters and the SPARCC score. p < 0.05 was considered statistically significant. The average values of each parameter, such as the 10th percentile, 25th percentile, 50th percentile, 75th percentile, 90th percentile, entropy, kurtosis and skewness, were recorded, and pseudo-colour maps were generated.
Results
Clinical data
In this study, a total of 56 patients were included, including 39 patients in the active group (27 males, 12 females, aged 22–57 years), 17 patients in the inactive group (8 males, 9 females, aged 24–45 years), and 24 patients in the healthy control group (14 males, 10 females, aged 23–43 years). There were no statistically significant differences in gender, age, height, weight, or HLA-B27 expression among the three groups (p > 0.05). The Mann–Whitney U test showed that the CRP, ESR, ASDAS-CRP, ASDAS-ESR and the SPARCC scores of the active group were greater than those of the inactive group (p < 0.05) (Table 2).
Table 2.
Clinical characteristics of ankylosing spondylitis patients and volunteers
| Characteristic | Active group (n = 39) | Inactive group (n = 17) | Control group (n = 24) | p value |
|---|---|---|---|---|
| Gender (male/female) | 27/12 | 8/9 | 14/10 | 0.45 |
| Age (years) | 32 ± 11 | 32.00 ± 5.97 | 29.71 ± 6.05 | 0.12 |
| Height (cm) | 168.03 ± 7.15 | 166.94 ± 6.70 | 166.25 ± 6.77 | 0.60 |
| Weight (kg) | 67.13 ± 12.04 | 64.29 ± 8.28 | 64.67 ± 11.49 | 0.58 |
| HLA-B27 (+/–) | 32/7 | 14/3 | NA | 0.98 |
| CRP (mg/L) | 7.50 ± 18.10 | 2.00 ± 3.10 | NA | <0.001 |
| ESR (mm/L) | 17.00 ± 21.25 | 4.00 ± 5.25 | NA | 0.001 |
| ASDAS-CRP | 2.70 ± 1.45 | 1.17 ± 0.53 | NA | <0.001 |
| ASDAS-ESR | 2.38 ± 1.50 | 1.12 ± 0.52 | NA | <0.001 |
| SPARCC | 21.00 ± 39.50 | 0.00 ± 11.75 | NA | <0.001 |
| Pairwise comparison of clinical characteristics between active group, inactive group and control group | ||||
| Characteristic | G1/G2 | G1/G3 | G2/G3 | |
| Gender (male/female) | 0.14 | 0.42 | 0.54 | |
| Age (years) | 0.75 | 0.09 | 0.36 | |
| Height (cm) | 0.49 | 0.25 | 0.67 | |
| Weight (kg) | 0.21 | 0.35 | 0.99 | |
ASDAS, Ankylosing Spondylitis Disease Activity Score; CRP, C-reactive protein; ESR, erythrocyte sedimentation rate; SPARCC, Spondyloarthritis Research Consortium of Canada.
G1: Active group; G2: Inactive group; G3: Control group.
Observer consistency detection results
The histogram analysis results of quantitative IVIM-DWI parameters Dslow, Dfast, f and ADC measured by two imaging physicians showed good consistency, with an ICC range of 0.774 ~ 0.998. The results measured by senior physicians were used for subsequent studies (Table 3).
Table 3.
Consistency detection results of histogram analysis results of Dslow, Dfast, F and ADC values
| Parameter | Dslow (95% CI) | Dfast (95% CI) | f (95% CI) | ADC (95% CI) |
|---|---|---|---|---|
| Mean | 0.986 (0.978 ~ 0.991) | 0.990 (0.984 ~ 0.994) | 0.955 (0.931 ~ 0.971) | 0.996 (0.994 ~ 0.998) |
| Skewness | 0.939 (0.906 ~ 0.960) | 0.973 (0.959 ~ 0.983) | 0.830 (0.735 ~ 0.891) | 0.796 (0.682 ~ 0.869) |
| Kurtosis | 0.961 (0.940 ~ 0.975) | 0.987 (0.980 ~ 0.992) | 0.917 (0.873 ~ 0.946) | 0.774 (0.648 ~ 0.855) |
| Entropy | 0.933 (0.898 ~ 0.957) | 0.806 (0.713 ~ 0.871) | 0.886 (0.827 ~ 0.925) | 0.797 (0.700 ~ 0.895) |
| 10th | 0.973 (0.958 ~ 0.983) | 0.950 (0.923 ~ 0.968) | 0.947 (0.918 ~ 0.965) | 0.986 (0.978 ~ 0.991) |
| 25th | 0.981 (0.971 ~ 0.988) | 0.982 (0.972 ~ 0.988) | 0.906 (0.857 ~ 0.939) | 0.990 (0.984 ~ 0.993) |
| 50th | 0.983 (0.973 ~ 0.989) | 0.965 (0.946 ~ 0.978) | 0.929 (0.892 ~ 0.954) | 0.995 (0.993 ~ 0.997) |
| 75th | 0.987 (0.979 ~ 0.991) | 0.944 (0.913 ~ 0.963) | 0.937 (0.904 ~ 0.959) | 0.996 (0.993 ~ 0.997) |
| 90th | 0.990 (0.984 ~ 0.993) | 0.866 (0.781 ~ 0.919) | 0.824 (0.738 ~ 0.883) | 0.996 (0.994 ~ 0.998) |
ADC, apparent diffusion coefficient.
IVIM-DWI quantitative parameter histogram analysis and the SPARCC score comparison among the active, inactive and control groups (Table 4)
Table 4.
Comparison of IVIM-DWI quantitative parameter histogram analysis results of active group, inactive group and control group
| Parameter | Active group | Inactive group | Control group | p-value |
|---|---|---|---|---|
| f mean | 0.134±0.043* | 0.117±0.026* | 0.116±0.022* | 0.246 |
| f skewness | 3.424±2.375* | 4.387±1.535* | 3.259±0.738* | 0.051 |
| f kurtosis | 8.125±3.486* | 26.560±23.106* | 10.416±6.107* | 0.069 |
| f entropy | 2.942±0.555* | 2.803±0.229* | 2.767±0.262* | 0.036 |
| f 10th | 0.020±0.050* | 0.000±0.005* | 0.000±0.000* | <0.001 |
| f 25th | 0.054±0.037* | 0.042±0.060* | 0.000±0.000* | <0.001 |
| f 50th | 0.101±0.038* | 0.078±0.039* | 0.055±0.032* | <0.001 |
| f 75th | 0.147±0.058* | 0.140±0.037* | 0.143±0.037* | 0.341 |
| f 90th | 0.225±0.205* | 0.185±0.088* | 0.248±0.109* | 0.12 |
| Dslow mean | 0.692±0.297 | 0.483±0.494 | 0.254±0.055 | <0.001 |
| Dslow skewness | 0.361±0.863* | 1.080±1.250* | 1.730±1.050* | <0.001 |
| Dslow kurtosis | -0.415±2.060* | 2.130±4.854* | 5.482±7.447* | <0.001 |
| Dslow entropy | 3.837±0.464* | 3.630±0.312* | 3.606±0.290* | 0.014 |
| Dslow 10th | 0.389±0.230 | 0.273±0.326 | 0.119±0.077 | <0.001 |
| Dslow 25th | 0.491±0.250 | 0.310±0.365 | 0.178±0.073 | <0.001 |
| Dslow 50th | 0.664±0.309 | 0.370±0.482 | 0.242±0.059 | <0.001 |
| Dslow 75th | 1.023±0.690 | 0.553±0.614 | 0.310±0.042 | <0.001 |
| Dslow 90th | 1.179±0.721 | 0.588±0.716 | 0.386±0.041 | <0.001 |
| Dfast mean | 76.536±53.816 | 75.585±99.255 | 144.779±34.447 | <0.001 |
| Dfast skewness | 4.380±2.370* | 3.764±1.265* | 2.588±0.507* | 0.001 |
| Dfast kurtosis | 16.718±26.008* | 15.074±20.073* | 7.647±3.691* | 0.005 |
| Dfast entropy | 1.474±0.380* | 1.372±0.216* | 1.605±0.134* | 0.007 |
| Dfast 10th | 3.181±2.182 | 2.831±1.570 | 1.882±0.558 | 0.042 |
| Dfast 25th | 6.736±1.650 | 5.648±2.808 | 6.103±1.058 | 0.365 |
| Dfast 50th | 10.036±6.600 | 9.918±14.103 | 20.733±235.028 | <0.001 |
| Dfast 75th | 22.392±236.145 | 19.073±235.420 | 250.000±0 | <0.001 |
| Dfast 90th | 250.000±228.241 | 250.000±24.994 | 250.000±0 | 0.705 |
| ADC mean | 0.853±0.346 | 0.572±0.575 | 0.347±0.102 | <0.001 |
| ADC skewness | 0.450±1.504* | 1.162±2.520* | 1.833±1.179* | 0.006 |
| ADC kurtosis | -0.856±4.774* | 3.475±9.663* | 6.305±5.402* | <0.001 |
| ADC entropy | 3.832±0.912* | 3.608±0.388* | 3.566±0.399* | 0.185 |
| ADC 10th | 0.482±0.271 | 0.322±0.382 | 0.158±0.095 | <0.001 |
| ADC 25th | 0.607±0.291 | 0.376±0.435 | 0.220±0.090 | <0.001 |
| ADC 50th | 0.802±0.356 | 0.453±0.588 | 0.303±0.067 | <0.001 |
| ADC 75th | 1.200±0.796 | 0.661±0.720 | 0.406±0.079 | <0.001 |
| ADC 90th | 1.268±0.475 | 0.784±0.585 | 0.509±0.099 | <0.001 |
ADC, apparent diffusion coefficient; IVIM-DWI, intravoxel incoherent motion diffusion-weighted imaging.
The difference in f (entropy, 10th percentile, 25th percentile, and 50th percentile) between the active group, the inactive group, and the control group was statistically significant (p < 0.05). Dslow (mean value, skewness, kurtosis, entropy, 10th percentile, 25th percentile, 50th percentile, 75th percentile, and 90th percentile) was significantly different between the active group, the inactive group and the control group (p < 0.05). The difference in Dfast (mean value, skewness, kurtosis, entropy, 10th percentile, 50th percentile, and 75th percentile) among the active group, the inactive group and the control group was statistically significant (p < 0.05). ADC (mean value, skewness, kurtosis, 10th percentile, 25th percentile, 50th percentile, 75th percentile, and 90th percentile) was significantly different between the active group, the inactive group, and the control group (p < 0.05).
Pairwise comparison of IVIM-DWI quantitative parameters and the SPARCC scores among the active, inactive, and control groups (Table 5)
Table 5.
IVIM-DWI quantitative parameter analysis results and SPARCC evaluation groups were compared pairwise
| Parameter | Inactive–active | Active–control | Inactive–control | |||
|---|---|---|---|---|---|---|
| Standard error | p- value | Standard error | p-value | Standard error | p- value | |
| f 10th | 5.620 | 0.017 | 5.017 | <0.001 | 6.130 | 0.626 |
| f 25th | 6.465 | 0.080 | 5.771 | <0.001 | 7.051 | 0.040 |
| f 50th | 6.753 | 0.258 | 6.028 | <0.001 | 7.366 | 0.186 |
| Dslow mean | 6.753 | 0.620 | 6.029 | <0.001 | 7.366 | 0.001 |
| Dslow skewness | 6.753 | 0.308 | 6.029 | <0.001 | 7.366 | 0.130 |
| Dslow kurtosis | 6.753 | 0.037 | 6.029 | <0.001 | 7.366 | 0.089 |
| Dslow entropy | 6.753 | 0.101 | 6.029 | 0.027 | 7.366 | 1.000 |
| Dslow 10th | 6.753 | 0.594 | 6.029 | <0.001 | 7.366 | 0.013 |
| Dslow 25th | 6.753 | 0.755 | 6.029 | <0.001 | 7.366 | 0.003 |
| Dslow 50th | 6.753 | 0.550 | 6.029 | <0.001 | 7.366 | 0.002 |
| Dslow 75th | 6.753 | 0.481 | 6.029 | <0.001 | 7.366 | 0.001 |
| Dslow 90th | 6.753 | 0.629 | 6.029 | <0.001 | 7.366 | ＜0.001 |
| Dfast mean | 6.754 | 0.714 | 6.029 | <0.001 | 7.366 | 0.015 |
| Dfast skewness | 6.754 | 1.000 | 6.029 | 0.001 | 7.366 | 0.040 |
| Dfast kurtosis | 6.754 | 1.000 | 6.029 | 0.006 | 7.366 | 0.049 |
| Dfast entropy | 6.754 | 0.905 | 6.029 | 0.044 | 7.366 | 0.010 |
| Dfast 10th | 6.754 | 1.000 | 6.029 | 0.053 | 7.366 | 0.230 |
| Dfast 50th | 6.747 | 1.000 | 6.023 | <0.001 | 7.359 | 0.004 |
| Dfast 75th | 6.591 | 1.000 | 5.883 | <0.001 | 7.189 | 0.001 |
| ADC mean | 6.754 | 0.304 | 6.029 | <0.001 | 7.366 | 0.020 |
| ADC skewness | 6.754 | 0.749 | 6.029 | 0.004 | 7.366 | 0.351 |
| ADC kurtosis | 6.754 | 0.139 | 6.029 | <0.001 | 7.366 | 0.497 |
| ADC 10th | 6.754 | 0.521 | 6.029 | <0.001 | 7.366 | 0.013 |
| ADC 25th | 6.754 | 0.518 | 6.029 | <0.001 | 7.366 | 0.005 |
| ADC 50th | 6.754 | 0.542 | 6.029 | <0.001 | 7.366 | 0.003 |
| ADC 75th | 6.754 | 0.282 | 6.029 | <0.001 | 7.366 | 0.015 |
| ADC 90th | 6.754 | 0.346 | 6.029 | <0.001 | 7.366 | 0.010 |
| SPARCC | 6.381 | 0.002 | 5.696 | <0.001 | 6.960 | 0.094 |
ADC, apparent diffusion coefficient; IVIM-DWI, intravoxel incoherent motion diffusion-weighted imaging; SPARCC, Spondyloarthritis Research Consortium of Canada.
The statistically significant data in Table 4 were further compared in pairs, and the Bonferroni method was used to correct the results. The f (10th percentile) and the SPARCC score of the active group were greater than those of the inactive group, and the Dslow kurtosis of the active group was smaller than those of the inactive group. Dslow (mean, 10th percentile, 25th percentile, 50th percentile, 75th percentile, and 90th percentile), Dfast (skewness, kurtosis), f (25th percentile), ADC (mean, 10th percentile, 25th percentile, 50th percentile, 75th percentile, and 90th percentile) in the inactive group were all greater than those in the control group. Dfast (mean value, entropy, 50th percentile, and 75th percentile) were all lower than those of the control group. The f (10th percentile, 25th percentile, and 50th percentile), Dslow (mean, entropy, 10th percentile, 25th percentile, 50th percentile, 75th percentile, and 90th percentile), Dfast (skewness and kurtosis), ADC (mean, 10th percentile, 25th percentile, 50th percentile, 75th percentile, and 90th percentile), the SPARCC score of active group were higher than those of the control group, the Dslow (skewness and kurtosis), Dfast (mean value, entropy, 50th percentile, 75th percentile), ADC (skewness and kurtosis) of the active group were lower than those of the control group (p < 0.05).
IVIM-DWI quantitative parameter histogram analysis and SPARCC scoring results to identify efficacy in the active group and the inactive group (Table 6) (Figure 4)
Table 6.
IVIM-DWI quantitative parameter histogram analysis and SPARCC scoring results to identify the activity group, efficacy of inactive group and control group
| Parameter | Cut-off | AUC | AUC 95% confidence interval (%) | Sensitivity (%) | Specificity (%) | p-value | |
|---|---|---|---|---|---|---|---|
| Active vs inactive group | Dslow kurtosis | 1.512* | 0.735 | 60.0 ~ 84.4 | 84.62 | 58.82 | 0.001 |
| Dslow entropy | 3.782* | 0.680 | 54.1 ~ 79.9 | 66.67 | 70.59 | 0.012 | |
| f 10th | 0.016* | 0.708 | 57.1 ~ 82.2 | 51.28 | 94.12 | <.001 | |
| f 25th | 0.053* | 0.673 | 53.4 ~ 79.2 | 61.54 | 76.47 | 0.019 | |
| ADC kurtosis | 1.253* | 0.685 | 54.7 ~ 80.2 | 71.79 | 70.59 | 0.012 | |
| ADC 75th | 0.787 | 0.676 | 51.9 ~ 82.3 | 74.40 | 70.60 | 0.038 | |
| ADC 90th | 0.993 | 0.677 | 52.5 ~ 82.9 | 74.40 | 64.70 | 0.036 | |
| SPARCC | 12* | 0.799 | 67.1 ~ 89.5 | 69.23 | 82.35 | <.001 | |
| Active vs control group | Dslow mean | 0.338 | 0.910 | 81.1 ~ 96.8 | 79.49 | 100.00 | <.001 |
| Dslow skewness | 0.896* | 0.832 | 71.7 ~ 91.5 | 76.92 | 87.50 | <.001 | |
| Dslow kurtosis | 2.699* | 0.895 | 79.2 ~ 95.8 | 87.18 | 83.88 | <.001 | |
| Dslow entropy | 3.800* | 0.697 | 56.8 ~ 80.6 | 64.10 | 79.17 | 0.003 | |
| Dslow 10th | 0.244 | 0.867 | 75.8 ~ 93.9 | 74.36 | 100.00 | <.001 | |
| Dslow 25th | 0.288 | 0.893 | 79.0 ~ 95.7 | 79.49 | 100.00 | <.001 | |
| Dslow 50th | 0.332 | 0.910 | 81.1 ~ 96.8 | 79.49 | 100.00 | <.001 | |
| Dslow 75th | 0.374 | 0.926 | 83.2 ~ 97.7 | 82.05 | 100.00 | <.001 | |
| Dslow 90th | 0.476 | 0.925 | 83.0 ~ 97.6 | 82.50 | 100.00 | <.001 | |
| Dfast mean | 106.882 | 0.851 | 73.9 ~ 92.9 | 82.05 | 87.50 | <.001 | |
| Dfast skewness | 3.066 | 0.752 | 62.7 ~ 85.2 | 66.67 | 87.50 | <.001 | |
| Dfas kurtosis | 12.445 | 0.717 | 58.9 ~ 82.3 | 61.54 | 95.83 | 0.001 | |
| Dfast entropy | 1.359 | 0.670 | 54.0 ~ 78.3 | 43.59 | 100.00 | 0.013 | |
| Dfast 10th | 2.290 | 0.647 | 54.4 ~ 78.1 | 61.54 | 87.50 | 0.012 | |
| Dfast 50th | 11.810 | 0.803 | 68.4 ~ 89.3 | 66.67 | 91.67 | <.001 | |
| Dfast 75th | 250.000 | 0.807 | 68.8 ~ 89.6 | 71.79 | 91.67 | <.001 | |
| f Entropy | 3.032* | 0.670 | 54.0 ~ 78.3 | 38.46 | 91.67 | 0.013 | |
| f 10th | 0* | 0.782 | 66.0 ~ 87.6 | 56.41 | 100.00 | <.001 | |
| f 25th | 0.021* | 0.864 | 75.5 ~ 93.8 | 74.36 | 91.67 | <.001 | |
| f 50th | 0.090* | 0.822 | 70.5 ~ 90.7 | 64.10 | 91.67 | <.001 | |
| ADC mean | 0.427 | 0.874 | 76.6 ~ 94.4 | 79.49 | 95.83 | <.001 | |
| ADC skewness | 0.978* | 0.747 | 62.1 ~ 84.8 | 69.23 | 91.67 | <.001 | |
| ADC kurtosis | 1.253* | 0.784 | 66.2 ~ 87.8 | 71.79 | 91.67 | <.001 | |
| ADC 10th | 0.298 | 0.874 | 76.6 ~ 94.4 | 76.92 | 100.00 | <.001 | |
| ADC 25th | 0.347 | 0.896 | 79.4 ~ 95.9 | 76.92 | 100.00 | <.001 | |
| ADC 50th | 0.395 | 0.907 | 80.7 ~ 96.6 | 79.49 | 100.00 | <.001 | |
| ADC 75th | 0.462 | 0.874 | 76.7 ~ 94.5 | 82.05 | 95.83 | <.001 | |
| ADC 90th | 0.645 | 0.873 | 76.5 ~ 94.3 | 82.05 | 95.83 | <.001 | |
| SPARCC | 0* | 0.936 | 84.5 ~ 98.2 | 87.18 | 100.00 | <.001 | |
| Inactive vs control group | Dslow mean | 0.338 | 0.863 | 71.9 ~ 95.0 | 64.71 | 100.00 | <.001 |
| Dslow kurtosis | 2.434* | 0.738 | 57.7 ~ 86.2 | 64.71 | 83.33 | 0.006 | |
| Dslow 10th | 0.244 | 0.778 | 62.1 ~ 89.3 | 64.71 | 100.00 | 0.003 | |
| Dslow 25th | 0.288 | 0.819 | 66.7 ~ 92.1 | 64.71 | 100.00 | <.001 | |
| Dslow 50th | 0.332 | 0.858 | 71.3 ~ 94.7 | 64.71 | 100.00 | <.001 | |
| Dslow 75th | 0.368 | 0.885 | 74.6 ~ 96.3 | 76.47 | 95.83 | <.001 | |
| Dslow 90th | 0.476 | 0.912 | 78.1 ~ 97.8 | 82.35 | 100.00 | <.001 | |
| Dfast mean | 84.273 | 0.775 | 61.7 ~ 89.0 | 64.71 | 95.83 | 0.001 | |
| Dfast skewness | 3.466* | 0.757 | 59.8 ~ 87.7 | 58.82 | 95.83 | 0.002 | |
| Dfast kurtosis | 12.445* | 0.760 | 60.1 ~ 87.9 | 58.82 | 95.83 | 0.002 | |
| Dfast entropy | 1.363* | 0.804 | 65.0 ~ 91.1 | 58.82 | 100.00 | <.001 | |
| Dfast 50th | 12.489 | 0.770 | 61.2 ~ 88.6 | 70.59 | 91.67 | 0.001 | |
| Dfast 75th | 21.439 | 0.828 | 67.8 ~ 92.8 | 64.71 | 100.00 | <.001 | |
| f Skewness | 3.451* | 0.772 | 61.5 ~ 88.8 | 76.47 | 79.17 | 0.001 | |
| f Kurtosis | 12.888* | 0.762 | 60.4 ~ 88.1 | 76.47 | 79.17 | 0.002 | |
| f 25th | 0.021* | 0.743 | 58.2 ~ 86.6 | 58.82 | 91.67 | 0.001 | |
| f 90th | 0.195* | 0.733 | 57.2 ~ 85.9 | 70.59 | 79.17 | 0.008 | |
| ADC mean | 0.427 | 0.784 | 62.8 ~ 89.7 | 64.71 | 95.83 | <.001 | |
| ADC 10th | 0.239 | 0.773 | 61.6 ~ 88.9 | 76.47 | 83.33 | 0.003 | |
| ADC 25th | 0.347 | 0.806 | 65.3 ~ 91.3 | 64.71 | 100.00 | <.001 | |
| ADC 50th | 0.394 | 0.838 | 69.0 ~ 93.4 | 64.71 | 100.00 | <.001 | |
| ADC 75th | 0.462 | 0.814 | 66.1 ~ 91.8 | 76.47 | 95.83 | <.001 | |
| ADC 90th | 0.604 | 0.843 | 69.6 ~ 93.8 | 82.35 | 91.67 | <.001 | |
| SPARCC | 0* | 0.735 | 57.4 ~ 86.1 | 47.06 | 100.00 | <.001 |
ADC, apparent diffusion coefficient; AUC, area under the curve; IVIM-DWI, intravoxel incoherent motion diffusion-weighted imaging; SPARCC, Spondyloarthritis Research Consortium of Canada.
Figure 4.
a. Dslow kurtosis, f10% digit and SPARCC score were used to differentiate the ROC curve of the active group and the inactive group; b. Dslow average, Dslow 50% bits, Dslow 75% bits, Dslow 90% bits, ADC50% digit and SPARCC score were used to differentiate the ROC curve between the active group and the control group; c. Dslow 75% digit and Dslow 90% digit were used to differentiate the ROC curve between inactive group and control group. ADC, apparent diffusion coefficient; ROC, receiver operating characteristic; SPARCC, Spondyloarthritis Research Consortium of Canada.
In the discrimination task between the active group and the inactive group, the AUC for the SPARCC score was largest at 0.799. When the cut-off value was 12, the sensitivity and specificity were 69.23 and 82.35%, respectively. In IVIM-DWI parameters, the AUC of Dslow (kurtosis) was the largest (0.735). When the cut-off value was 1.512, the sensitivity and specificity were 84.62 and 58.82%, respectively. The AUC of f (10th percentile) was larger (0.708). When the cut-off value was 0.016, the sensitivity and specificity were 51.28 and 94.12%, respectively (Figure 4a). In the differentiation between the active group and the control group, the SPARCC score had the largest AUC (0.936), and the sensitivity and specificity were 87.18 and 100.00%, respectively, when the cut-off value was 0. For the IVIM-DWI parameters, the maximum AUCs of Dslow (75th percentile) and Dslow (90th percentile) were 0.926 and 0.925, respectively. When cut-off values of 0.374 × 10−3 mm2/s and 0.476 × 10−3 mm2/s are applied, the sensitivity and specificity were 82.05 and 100.00% and 82.50 and 100.00%, respectively. The AUCs of Dslow (average value) and Dslow (50th percentile) are larger, both of which are 0.910. When the cut-off values were 0.338 × 10−3 mm2/s and 0.332 × 10−3 mm2/s, the sensitivity and specificity, respectively, were 79.49 and 100.00% for the former cut-off and 79.49 and 100.00% for the latter (Figure 4b). In the differentiation between the inactive group and the control group, the maximum AUC of Dslow (90th percentile) was 0.912, and the sensitivity and specificity were 82.35 and 100.00%, respectively, when the cut-off value was 0.472 × 10−3 mm2/s.The AUC for Dslow (75th percentile) was larger at 0.885, and the sensitivity and specificity were 76.47 and 95.83%, respectively, when taking the cut-off of 0.368 × 10−3 mm2/s (Figure 4c).
Correlation between the SPARCC score and IVIM-DWI quantitative parameters
Spearman correlation results showed that the percentile and average value of IVIM-DWI parameters Dslow and ADC were positively correlated with the SPARCC score, while Dslow (kurtosis) and ADC (kurtosis, skewness) were negatively correlated with the SPARCC score. f (skewness, kurtosis, entropy, 10th–50th percentiles) was positively correlated with the SPARCC score; Dfast (mean, 10th percentile, 50th–90th percentiles) was negatively correlated with the SPARCC score. Dfast (skewness, kurtosis) was positively correlated with the SPARCC score. Among them, Dfast (mean) had the best correlation with the SPARCC score (r = −0.689, p < 0.001)(Figure 5)(Table 7).
Figure 5.
Scatter plot of correlation analysis between SPARCC score and Dfast mean value. SPARCC, Spondyloarthritis Research Consortium of Canada.
Table 7.
Correlation between IVIM-DWI quantitative parameters and SPARCC score
| Parameter | Dslow | Dfast | f | ADC | ||||
|---|---|---|---|---|---|---|---|---|
| Rvable | p-value | R value | p-value | R value | p-value | R value | p-value | |
| mean | 0.510 | <.001 | −0.689 | <.001 | −0.160 | 0.238 | 0.492 | <.001 |
| Skewness | −0.228 | 0.091 | 0.621 | <.001 | 0.349 | 0.008 | −0.348 | 0.009 |
| Kurtosis | −0.494 | <.001 | 0.612 | <.001 | 0.383 | 0.004 | −0.535 | <.001 |
| Entropy | 0.591 | <.001 | −0.242 | 0.072 | 0.380 | 0.004 | 0.601 | <.001 |
| 10th | 0.468 | <.001 | −0.559 | <.001 | 0.448 | 0.001 | 0.480 | <.001 |
| 25th | 0.469 | <.001 | 0.099 | 0.470 | 0.547 | <.001 | 0.469 | <.001 |
| 50th | 0.517 | <.001 | −0.446 | 0.001 | 0.426 | 0.001 | 0.494 | <.001 |
| 75th | 0.531 | <.001 | −0.576 | <.001 | −0.092 | 0.500 | 0.519 | <.001 |
| 90th | 0.510 | <.001 | −0.378 | 0.004 | −0.378 | 0.060 | 0.504 | <.001 |
IVIM-DWI, intravoxel incoherent motion diffusion-weighted imaging; SPARCC, Spondyloarthritis Research Consortium of Canada.
Discussion
Analysis of subjects’ clinical data
In this study, there was no significant difference in HLA-B27 test results between the AS active group and the inactive group, indicating that the expression of HLA-B27 may not be used as a target to identify the activity of AS. A study 18 pointed out that the sensitivity of ASDAS-CRP was higher than that of ASDAS-ESR, so in this study, ASDAS-CRP was used to group AS patients. Our study manifested that the CRP, ESR, ASDAS-CRP and ASDAS-ESR in the active group were significantly higher than those in the inactive group (p < 0.05).
The value of IVIM-DWI quantitative parameter histogram analysis in distinguishing among the active AS, inactive AS, and control groups
IVIM-DWI is a double exponential model based on the fitting of multiple b values. However, at present, there is no sufficient research on the optimal combination of b values. Theoretically, the more b values there are, the more accurate the results will be. 19 Studies have shown 20 that the higher the value of b is, the smaller the influence brought by perfusion; 600 s/mm2 is the best b value. 21,22 In this study, we chose 0, 10, 20, 30, 50, 80, 100, 200, 300, 600, 800, 1000 and 1500 s/mm2, which can show bone marrow oedema more sensitively but also has good stability and reproducibility.
The results of this study suggested that the average Dslow and all percentiles of the active group were significantly higher than those of the inactive group and the control group, which was consistent with the previous results of Zhao et al. 14 In general, in infectious sacroiliac arthropathy caused by inflammatory cells and vasogenic oedema, the diffusion of water molecules is usually limited due to the increase in protein-rich fluid and the decrease in Dslow. 23,24 However, the main pathophysiological manifestation of AS sacroiliac arthritis is bone marrow oedema. According to Bozgeyik et al, 25 bone marrow oedema will lead to an increase in water molecules in the extracellular space, which indicates that the increased activity of AS disease will lead to the increased diffusion of local water molecules in the bone marrow under the sacroiliac joint surface and an increase in the Dslow value. In this study, the mean value and percentiles of Dslow in the active group were higher than those in the inactive group, but the differences were not statistically significant. The reasons may be that the number of cases in this study was insufficient. Second, both the active group and the inactive group had cases of bone marrow oedema and the non-bone marrow oedema group, which would also affect the results of the study. The Dslow (kurtosis) of the active group was significantly lower than that of the inactive group, and the Dslow (skewness and kurtosis) of the active group was significantly lower than that of the control group. With the increase in the disease activity state, the Dslow value increased, and the skewness and kurtosis decreased. Our study showed that the results of the average value of ADC and kurtosis were similar to the ADC histogram study of Wang et al. 17 ; furthermore, there were more parameters in our histogram study.
The tissue perfusion correlation coefficient f is related to blood flow. 26 f (10th percentile) in the active group was significantly higher than that in the inactive group and the control group. f (10th percentile, 25th percentile, 50th percentile) in the active group was significantly higher than that in the control group, and f (25th percentile) in the inactive group was significantly higher than that in the control group, which is similar to the results of previous studies. This may be due to the formation of pannus under the sacroiliac joint surface during the active stage of AS, resulting in the increased amount of blood and serum transported to the bone marrow cavity through the capillaries.
In previous studies, there was no significant difference in the Dfast mean value between the active group, chronic group and control group, while in this study, Dfast (mean value, entropy, 50th percentile, 75th percentile) showed a significant difference between the inactive group and control group and between the active group and control group. This may be because the grouping used in this study was based on ASDAS, while previous studies all used BASDAI, and the differences in grouping led to differences in results, suggesting that higher accuracy can be achieved using the ASDAS score with CRP or ESR than by relying only on the BASDAI score, which is affected by patients' subjective interpretations.
In addition, this study measured the quantitative parameters of ADC and the histogram of the data at the same time. The results indicated that all the histograms of ADC data between the active group and the control group were significantly different. Moreover, the average ADC and all percentiles of the active group were significantly greater than those of the control group, and the average ADC and all percentiles of the inactive group were also significantly greater than those of the control group, consistent with previous research conclusions. 17 However, all the histogram data of ADC showed no significant difference between the active group and the inactive group, suggesting that quantitative parameters of IVIM-DWI may be superior to traditional DWI in the detection of AS disease activity.
The results of this study showed that only Dslow (kurtosis) and f (10th percentile) had statistical significance in discriminating between active and inactive groups, and the AUC (0.744) of Dslow (kurtosis) was higher than that of f (10th percentile). Moreover, in the discrimination between the active group and the control group, the AUC of Dslow (75th percentile, 90th percentile) was the largest (0.926, 0.925), which was significantly greater than that of all the ADC histogram parameters. In differentiating the inactive group from the control group, the AUC of Dslow (90th percentile) was the largest (0.912), which was greater than that of all statistically significant ADC histogram parameters. This indicates that the IVIM-DWI parameter Dslow can better reflect the real water molecule diffusion information in the tissue, is superior to the traditional DWI quantitative parameter ADC, and has the greatest significance in the identification of AS disease activity. Moreover, we can observe that in the histogram data of quantitative parameters in Dslow, higher percentiles (75th, 90th) have higher AUCs and sensitivities than lower percentiles (50th) and means, suggesting that higher percentiles may be more sensitive to the differential diagnosis of AS disease activity.
Comparison of IVIM-DWI quantitative parameter histogram analysis and the SPARCC score
In this study, IVIM-DWI quantitative parameters were compared with the SPARCC score in the diagnosis of AS disease activity. The results showed that in the discrimination between the inactive group and active group and between the active group and control group, the SPARCC score had the largest AUCs of 0.799 and 0.936, respectively, which were higher than the quantitative IVIM-DWI parameters. This may be because the SPARCC score mainly relies on the visual observation of the graders, and some patients in the inactive group and active group do not have visible bone marrow oedema signals under the MRI sacroiliac joint surface, while bone marrow oedema or other subtle pathological changes may have been present at this time, and IVIM-DWI quantitative parameters can distinguish these subtle changes.
Limitations
First, the sample size of this study was not sufficient, and the number of different groups was not balanced enough. In particular, there were fewer cases with obvious bone marrow oedema under the sacroiliac joint surface, which may lead to errors in the results. Second, the grouping used is influenced by the subjectivity of patients according to ASDAS, which may lead to bias of the results. Third, the AS patient group in this study was not pathologically confirmed.
Conclusion
Quantitative parameters of IVIM-DWI contribute to the identification of AS disease activity and are superior to traditional DWI, suggesting that quantitative parameters of IVIM-DWI have a certain application value in the assessment of AS disease activity.
Footnotes
The authors Li Liu and Zhimin Zhou contributed equally to the work.
Contributor Information
Li Liu, Email: a19852911837@163.com, Department of Imaging, Linyi Central Hospital, Linyi, Shandong, China .
Zhimin Zhou, Email: 103626078@qq.com, Department of Radiology, The Second Affiliated Hospital of Soochow University, Suzhou, Jiangsu, China .
Sunyu Hua, Email: huasunyu@163.com, Department of Radiology, The Second Affiliated Hospital of Soochow University, Suzhou, Jiangsu, China .
Leixi Xue, Email: xueleixi2002@163.com, Department of Rheumatology and Immunology, The Second Affiliated Hospital of Soochow University, Suzhou, Jiangsu, China .
Jiangtao Zhu, Email: jiangtaozhu@126.com, Department of Radiology, The Second Affiliated Hospital of Soochow University, Suzhou, Jiangsu, China .
Rong Liu, Email: liurongtc@126.com, Department of Radiology, The Second Affiliated Hospital of Soochow University, Suzhou, Jiangsu, China .
Yong Li, Email: szfeyly@163.com, Department of Radiology, The Second Affiliated Hospital of Soochow University, Suzhou, Jiangsu, China .
REFERENCES
- 1. Proft F, Poddubnyy D, et al. . Ankylosing spondylitis and axial spondyloarthritis: recent insights and impact of new classification criteria . Ther Adv Musculoskelet Dis 2018; 10: 129–39. 10.1177/1759720X18773726 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Sağ S, Nas K, Sağ MS, Tekeoğlu İ, Kamanlı A . Relationship of work disability between the disease activity, depression and quality of life in patients with ankylosing spondylitis . J Back Musculoskelet Rehabil 2018. ; 31: 499 – 505 . doi: 10.3233/BMR-169657 [DOI] [PubMed] [Google Scholar]
- 3. Solmaz D, Yildirim T, Avci O, Tomas N, Akar S, et al. . Performance characteristics of the simplified version of ankylosing spondylitis disease activity score (sasdas) . J Clin Rheumatol 2016. ; 35: 1753 – 58 . doi: 10.1007/s10067-015-3147-z [DOI] [PubMed] [Google Scholar]
- 4. Zochling J, Braun J, van der Heijde D, et al. . Assessments in ankylosing spondylitis . Best Pract Res Clin Rheumatol 2006; 20: 521–37. 10.1016/j.berh.2006.03.011 [DOI] [PubMed] [Google Scholar]
- 5. Landewé RBM, Hermann K-GA, van der Heijde DMFM, Baraliakos X, Jurik A-G, et al. . Scoring sacroiliac joints by magnetic resonance imaging. a multiple-reader reliability experiment . J Rheumatol 2005; 32: 2050–55. [PubMed] [Google Scholar]
- 6. Zhang P, Yu K, Guo R, Shah S, Morelli JN, et al. . Ankylosing spondylitis: correlations between clinical and mri indices of sacroiliitis activity . Clin Radiol 2015. ; 70: 62 – 66 . doi: 10.1016/j.crad.2014.09.015 [DOI] [PubMed] [Google Scholar]
- 7. Cui Y, Zheng J, Zhang X, Zeng H, Luo R, et al. . Evaluation of treatments for sacroiliitis in spondyloarthropathy using the spondyloarthritis research consortium canada scoring system . Arthritis Res Ther 2016; 18: 38. 10.1186/s13075-016-0916-2 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. Zhang M, Zhou L, Huang N, Zeng H, Liu S, et al. . Assessment of active and inactive sacroiliitis in patients with ankylosing spondylitis using quantitative dynamic contrast-enhanced mri . J Magn Reson Imaging 2017; 46: 71–78. 10.1002/jmri.25559 [DOI] [PubMed] [Google Scholar]
- 9. Chu GY, Zhang MC, Yang HS, et al. . Dynamic contrastenhanced mri in predicting the activity of sacroiliac arthritis in ankylosing spondylitis[j] . Chin J Radiol 2015. ; 49: 895 – 99 . [Google Scholar]
- 10. Braun J, Bollow M, Eggens U, König H, Distler A, et al. . Use of dynamic magnetic resonance imaging with fast imaging in the detection of early and advanced sacroiliitis in spondylarthropathy patients . Arthritis Rheum 1994; 37: 1039–45. 10.1002/art.1780370709 [DOI] [PubMed] [Google Scholar]
- 11. Dallaudière B, Dautry R, Preux P-M, Perozziello A, Lincot J, et al. . Comparison of apparent diffusion coefficient in spondylarthritis axial active inflammatory lesions and type 1 modic changes . Eur J Radiol 2014. ; 83: 366 – 70 . doi: 10.1016/j.ejrad.2013.10.009 [DOI] [PubMed] [Google Scholar]
- 12. Gaspersic N, Sersa I, Jevtic V, Tomsic M, Praprotnik S, et al. . Monitoring ankylosing spondylitis therapy by dynamic contrast-enhanced and diffusion-weighted magnetic resonance imaging . Skeletal Radiol 2008; 37: 123–31. 10.1007/s00256-007-0407-2 [DOI] [PubMed] [Google Scholar]
- 13. Le Bihan D, Breton E, Lallemand D, Grenier P, Cabanis E, et al. . MR imaging of intravoxel incoherent motions: application to diffusion and perfusion in neurologic disorders . Radiology 1986; 161: 401–7. 10.1148/radiology.161.2.3763909 [DOI] [PubMed] [Google Scholar]
- 14. Zhao Y, Li S, Liu Z, Chen X, Zhao X, et al. . Detection of active sacroiliitis with ankylosing spondylitis through intravoxel incoherent motion diffusion-weighted mr imaging . Eur Radiol 2015; 25: 2754–63. 10.1007/s00330-015-3634-2 [DOI] [PubMed] [Google Scholar]
- 15. Sieper J, Rudwaleit M, Baraliakos X, Brandt J, Braun J, et al. . The assessment of spondyloarthritis international society (asas) handbook: a guide to assess spondyloarthritis . Ann Rheum Dis 2009; 68 Suppl 2: ii1-44. 10.1136/ard.2008.104018 [DOI] [PubMed] [Google Scholar]
- 16. Le Bihan D, Breton E, Lallemand D, Aubin ML, Vignaud J, et al. . Separation of diffusion and perfusion in intravoxel incoherent motion mr imaging . Radiology 1988; 168: 497–505. 10.1148/radiology.168.2.3393671 [DOI] [PubMed] [Google Scholar]
- 17. Wang F, Chu C, Zhu L, Zhao C, Wei Y, et al. . Whole-lesion adc histogram analysis and the spondyloarthritis research consortium of canada (sparcc) mri index in evaluating the disease activity of ankylosing spondylitis . J Magn Reson Imaging 2019. ; 50: 114 – 26 . doi: 10.1002/jmri.26568 [DOI] [PubMed] [Google Scholar]
- 18. van der Heijde D, Lie E, Kvien TK, Sieper J, Van den Bosch F, et al. . ASDAS, a highly discriminatory asas-endorsed disease activity score in patients with ankylosing spondylitis . Ann Rheum Dis 2009; 68: 1811–18. 10.1136/ard.2008.100826 [DOI] [PubMed] [Google Scholar]
- 19. Lemke A, Stieltjes B, Schad LR, Laun FB, et al. . Toward an optimal distribution of b values for intravoxel incoherent motion imaging . Magn Reson Imaging 2011; 29: 766–76. 10.1016/j.mri.2011.03.004 [DOI] [PubMed] [Google Scholar]
- 20. Thoeny HC, De Keyzer F, Oyen RH, Peeters RR, et al. . Diffusion-weighted mr imaging of kidneys in healthy volunteers and patients with parenchymal diseases: initial experience . Radiology 2005; 235: 911–17. 10.1148/radiol.2353040554 [DOI] [PubMed] [Google Scholar]
- 21. Lories RJ, Baeten DLP, et al. . Differences in pathophysiology between rheumatoid arthritis and ankylosing spondylitis . Clin Exp Rheumatol 2009. ; 27: S10 - 4 . [PubMed] [Google Scholar]
- 22. Li BB, Ren CP, Li Y, et al. . Magnetic resonance diffusion weighted imaging sequence shows the best b value of sacroiliac joint imaging of healthy volunteers[j] . Chinese J Tissue Engineering Research 2013. ; 17: 3124 – 31 . [Google Scholar]
- 23. Khoo MMY, Tyler PA, Saifuddin A, Padhani AR . Diffusion-weighted imaging (dwi) in musculoskeletal mri: a critical review . Skeletal Radiol 2011; 40: 665–81. 10.1007/s00256-011-1106-6 [DOI] [PubMed] [Google Scholar]
- 24. ] Kawada S, Takahara T, Ichikawa T, et al. . The Apparent Diffusion Coefficient (ADC) as a Potential Marker of Inflammation Associated with Body Abscesses[J] . Tokai J Exp Clin Med,2017,42:58-63 . [PubMed] [Google Scholar]
- 25. Bozgeyik Z, Ozgocmen S, Kocakoc E, et al. . Role of diffusion-weighted mri in the detection of early active sacroiliitis . AJR Am J Roentgenol 2008; 191: 980–86. 10.2214/AJR.07.3865 [DOI] [PubMed] [Google Scholar]
- 26. Le Bihan D . Intravoxel incoherent motion perfusion mr imaging: a wake-up call . Radiology 2008; 249: 748–52. 10.1148/radiol.2493081301 [DOI] [PubMed] [Google Scholar]