This demonstrates better inter-rater agreeability with fusion imaging (0.851, 95% CI 0.784 -0.900) when compared to the standard post contrast imaging (0.64, 95% CI 0.376-0.780). Of particular note, there was a very large spread in scores using the post-contrast imaging as reflected by the wide confidence interval. This was based on a two-way random model where the same radiologists read all the same cases and were treated as a representative sample of radiologists as a whole.
We then compared the scores of patients who received treatment with biologic agents or DMARDs and those who did not. This was done for each reader and for the fusion and post contrast scores separately. We first employed a Levene’s test to assess for equality of variance and then performed a T-test between the two scores after. If there was no significant difference between the two groups on the Levene’s test (Sig. >0.05), then we interpreted the first line of the t-test. If there was a significant difference between the two groups (Sig <0.05), then we assessed the second row of the t-test.
Insert table
Reader 1: Significant difference in the OMERACT scores obtained between the two patient groups with fusion imaging (t = -2.22, p<0.04). No significant difference in the OMERACT scores obtained between the two groups with the post-contrast imaging.
Reader 2 found no significant difference between the OMERACT scores in the two patient groups on both the fusion and post-contrast images.
Reader 3 found no significant difference between the OMERACT scores in the two patient groups on both the fusion and post-contrast images.
Two of the three readers felt that fused DCE images provided additional benefit and our results show that they minimize inter-rater variability in image interpretation in this study. However, when comparing this against clinical outcome, only one reader demonstrated a significant difference in OMERACT scores with the fusion images between patient groups who ended up on treatment versus those who did not. Of note, there was no significant difference between these two patient groups on the current standard imaging.
Conclusion
Limitations:
1) Patients were not strictly anonymised. A degree of blinding was achieved by temporal separation between original image interpretation and this retrospective analysis. The most recent imaging was performed at least 8 months prior to image review.
2) The sample size of patients was small.
3) The field of view was altered for some studies for clinical reasons. These were kept in the study as reviewers felt it reflected a real world environment
4) The DIPs and PIPs and flexor tendon sheaths were not assessed on this occasion and can be assessed with further studies
5) Assessment of the fusion images was performed only in the coronal plane due to technical factors, but the standard images were reviewed in two orthogonal planes. This can be addressed by further studies
6) There was no assessment for motion artefact.
7) This study focused primarily on features of synovitis and did not assess features such as enthesitis and tenosynovitis.
8) Clinical review was performed by a single clinician and the criteria was not predefined
9) While a descriptive parameters of dynamic contrast enhancement characteristics such as the peak enhancement is easily understood, these are specific to the scanner used and field strength, which limits comparison \cite{Maijer2016}.
Benefits
We have shown that the fusion of peak enhancement DCE is possible and reduced levels of inter-observer variability in the assessment for synovitis. This also correlated with the confident clinical diagnosis of an inflammatory arthritis for one reader.
Time resolved MRA is useful sequence in the assessment of synovitis. This does not significantly add to scanning time for post contrast imaging and is technically feasible. Automating the processing has generated reproducible images that are achievable within a clinical setting and time frame.
Replacing standard pre and post contrast imaging with a TRICKS MRA sequence and post processed peak enhancement fusion images could significantly reduce scanning time for patients (by up to a quarter at our institution), potentially increasing throughput and reducing waiting times. Further assessment with a larger sample size with clinical correlation is needed to assess this further.