Histographic analysis of oedema and fat in inflamed bone marrow based on quantitative MRI

Objective

To demonstrate proof-of-concept for a quantitative MRI method using histographic analysis to assess bone marrow oedema and fat metaplasia in the sacroiliac joints.

Materials and methods

Fifty-three adolescents aged 12–23 with known or suspected sacroiliitis were prospectively recruited and underwent quantitative MRI (qMRI) scans, consisting of chemical shift-encoded (at 3 T) and diffusion-weighted imaging (at 1.5 T), plus conventional MRI (at 1.5 T) and clinical assessment. qMRI scans produced proton-density fat fraction (PDFF) and apparent diffusion coefficient (ADC) maps of the sacroiliac joints (SIJs), which were analysed using an in-house software tool enabling partially automated ROI definition and histographic analysis. Logistic regression and receiver operating characteristic (ROC) analyses assessed the predictive performance of ADC- and PDFF-based parameters in identifying active inflammation (oedema) and structural damage (fat metaplasia).

Results

ADC-based parameters were associated with increased odds of oedema (all p < 0.05); ROC-AUC was higher for histographic parameters representing the upper end of the ADC distribution than for simple averages. Similarly, PDFF-based parameters were associated with increased odds of fat metaplasia (all p < 0.05); ROC area-under-the-curve was higher for histographic parameters representing the upper end of the PDFF distribution than for simple averages. Both ADC- and PDFF-based histographic parameters demonstrated excellent inter- and intra-observer agreement (ICC > 0.9).

Conclusions

ADC-based parameters can differentiate patients with bone marrow oedema from those without, whilst PDFF-based parameters can differentiate patients with fat metaplasia from those without. Histographic analysis might improve performance compared with simple averages such as the mean and median and offers excellent agreement within and between observers.

Key Points

• Quantitative MRI with histographic analysis can identify bone marrow oedema (an active inflammatory lesion) and fat metaplasia (a ‘chronic’ inflammatory lesion) in patients with spondyloarthritis.

• The use of histographic analysis might improve the performance of quantitative MRI for detecting bone marrow oedema and fat metaplasia compared with simple averages such as the mean and median.

• Bone marrow oedema and fat metaplasia are known to be of diagnostic and prognostic significance, and the proposed method could support clinical decisions around biologic (and other) therapies in spondyloarthritis.

Electronic supplementary material

The online version of this article (10.1007/s00330-020-06785-x) contains supplementary material, which is available to authorized users.

Magnetic resonance imaging of sacroiliitis in children: frequency of findings and interobserver reliability

BACKGROUND

Clinicians increasingly rely on imaging in juvenile idiopathic arthritis (JIA) to identify sacroiliitis and guide treatment. However, there is limited evidence about magnetic resonance imaging (MRI) for sacroiliitis in children, and interobserver reliability is variable.

OBJECTIVE

Identify the frequency of MRI findings in children with suspected sacroiliitis, calculate inter-reporter reliability and assess the value of diffusion-weighted imaging and contrast-enhanced sequences.

MATERIALS AND METHODS

We retrospectively reviewed 3 years of sacroiliac joint MRI records for suspected sacroiliitis in patients <21 years at a United Kingdom tertiary referral paediatric hospital. Five radiologists (panel of three radiologists and two independent radiologists) reviewed all MRI examinations using a pictorial checklist to identify oedema, effusions, diffusion-weighted signal abnormality, enhancement, erosions and sclerosis. The frequency of panel findings was reported. Interobserver agreement was calculated using the Cohen kappa coefficient.

RESULTS

An MRI diagnosis of sacroiliitis was made in 12 of 99 examinations (12%). The findings in all scans included oedema (9%), erosions (8%), diffusion-weighted signal abnormality (6%), abnormal enhancement (6%) and effusion (4%). All scans with abnormal contrast enhancement had other MRI features of sacroiliitis. Interobserver agreement was slight to moderate.

CONCLUSION

Oedema and erosions were the most common findings. Inter-reporter reliability was variable with at best moderate agreement for the presence of sacroiliitis and erosions. The use of contrast enhancement for diagnosing sacroiliitis in children with JIA may be questionable.