Session Information
Session Type: Abstract Submissions (ACR)
Background/Purpose
Efficient and accurate evaluation of erosive damage to the MCP joints in RA patients is desirable in both clinical trials and clinical practice. The novel software Early Erosions in Rheumatoid Arthritis (EERA) hybridizes region growing and level-set segmentation algorithms, and can be used by a novice reader to semi-automatically quantify bone erosion volumes of the MCP joints captured by MRI. The objectives of this study were firstly to compare EERA to the Outcome Measures in Rheumatology Clinical Trials RA MRI score (RAMRIS), which is an established tool for evaluating erosive damage, and secondly to assess the inter- and intra-rater reliability of EERA.
Methods
Magnetic resonance images were acquired of both hands of 71 RA patients from a single rheumatology clinic at baseline and 2 years follow-up using a 1T magnet, 100mm cylindrical transmit and receive coil, and a 3D spoiled gradient echo sequence. Images were randomly distributed to 4 blinded musculoskeletal radiologists trained in RAMRIS. RAMRIS erosion scores for the 2nd through 5th MCP joints of both hands were evaluated and summed. A separate reader trained in EERA, but otherwise inexperienced with conventional quantification techniques, used EERA to evaluate and sum the volume of bone loss (in mm3) of the 2nd through 5th MCP joints of these same images. To assess reliability, a random subset of 20 images was evaluated by 2 additional novice readers similarly trained in EERA. All 3 readers scored these 20 images a second time after 72 hours. Spearman’s correlations were calculated to compare EERA and RAMRIS cross-sectionally and longitudinally, and intra-class correlation coefficients, ICC (2,1) with 95% confidence intervals (CI) were used to assess EERA reliability.
Results
Of the 71 participants [female: 78%, Caucasian: 72%, age: mean (standard deviation) 56.5 (12.8)yr, DAS28-ESR: 4.39 (1.42), symptom duration: 5.5 (5.7)yr], 52 (73%) were imaged at 2 years follow-up. EERA detected erosions in 43 (61%) participants, whereas RAMRIS detected erosions in 65 (92%) participants. The mean baseline erosion volume per participant was 63.4mm3 (129.7mm3), and mean baseline RAMRIS erosion score per participant was 10.8 (10.9). A Spearman’s rho=0.450 (p<0.001) showed moderate correlation between EERA erosion volumes and RAMRIS erosion scores at baseline, with similar correlation at 2 years follow-up (rho=0.496, p<0.001). The 2 year mean change in EERA erosion volume was 1.2mm3, and volume changes did not correlate significantly with 2 year RAMRIS erosion score changes (p=0.505). EERA reliability was excellent between all 3 raters, with an inter-rater reliability ICC of 0.945 (95% CI 0.887 to 0.959). Intra-rater reliability ICCs of 0.993 (95% CI 0.982 to 0.997), 0.979 (95% CI 0.949 to 0.992), and 0.933 (95% CI 0.834 to 0.973) were achieved for each reader.
Conclusion
EERA erosion volumes evaluated by a novice reader exhibited moderate correlation with RAMRIS erosion scores evaluated by trained radiologists. The excellent reliability of EERA suggests that it may have practical utility in a clinical setting, and future assessment of responsiveness is warranted.
Disclosure:
M. Jessome,
None;
I. Rodrigues,
None;
M. Tomizza,
None;
J. Barbosa,
None;
M. X. Koh,
None;
K. Beattie,
None;
W. G. Bensen,
None;
R. Bobba,
None;
A. Cividino,
None;
P. D. Emond,
None;
K. Finlay,
None;
C. Gordon,
None;
L. Hart,
None;
G. Ioannidis,
None;
E. Jurriaans,
None;
M. Larche,
None;
A. Lau,
None;
N. Parasu,
None;
R. Tavares,
None;
S. Tytus,
None;
H. Wu,
None;
J. D. Adachi,
None.
« Back to 2014 ACR/ARHP Annual Meeting
ACR Meeting Abstracts - https://acrabstracts.org/abstract/assessing-the-validity-and-reliability-of-a-novel-mri-semi-automated-algorithm-for-quantifying-bone-loss-in-the-hand/