3 Tesla magnetic resonance imaging and computerised image analysis in the evaluation of rheumatoid hand joints

Abstract

Objective: To explore development and testing of a new patient positioning device during wrist MRI. To evaluate rheumatoid arthritis (RA) wrist 3T MRI with computerised image analysis tools including bone segmentation techniques, semiquantitative scoring and dynamic contrast enhanced (DCE-MRI) for synovitis measures, with similar assessments on healthy subjects. To examine the relationship between X-ray/MRI disease activity measures and bone mineral density loss using computerised digital X-ray radiogammetry (DXR). Methods: Preliminary study was about development and evaluation of a new patient positioning device for wrist imaging and its evaluation against current standard positions. In the main longitudinal MRI study, 13 RA and 10 healthy subjects (HV) were recruited with 10 RA and 7 HV completing the study. Hand/wrist MRI and radiographs were performed over a year and assessed using semi-quantitative scoring and exploratory computerised image analysis tools, which included pilot work on bone segmentation techniques, including manual and semi-automated methods; bone mineral density loss assessment using DXR-online and its comparison with MRI disease activity measures; and dynamic contrast enhanced MRI (DCE-MRI) analysis using Dynamika software. Result: A developed bridge patient positioning device allowed for comfortable, good quality and reproducible imaging as compared to standard positions (Bridge vs hand by the side vs hand above head position: comfort score (out of 10) (mean±std) – 7.3±0.7 vs 7.1±0.8 vs 6.1±1.6, and image quality (Signal to noise/contrast to noise ratio) – 6.1±1.7/3.1±0.5 vs 5.3±1.5/2.6±0.5 vs 7.7±1.1/4±0.5), with subsequent use in a longitudinal study (Comfort score: RA/HV:9.1/8.1). The semi-automated bone segmentation method was much quicker than manual technique (10/8 vs 165/132 minutes for two readers), with good inter and intra-observer similarity for manual method and in between the two methods, though the semi-automated method failed in an advanced RA patient. The manual segmentation using SliceOmatic software was also time consuming - 3256 minutes for 2196 slices, but with image registration and transformation a visual analysis of future images was possible. In the main longitudinal study, there were stable moderate DAS 28 disease activity scores (Day 1 – 3.9 and week 52 – 4.0). MRI disease activity measures showed good correlation. Bone marrow oedema (BME) correlated with erosions, and automated early 3 month rate of metacarpal digital x-ray radiogrammetric bone mineral density loss (RC-BMD) correlated with 1 year wrist BME change (p=0.035). No significant change was seen for MRI, radiographic or any disease activity measure over the year, though there was small increase in MRI erosion: 2.4 (1.6%), BME: 0.4 (0.8%) and radiographic: 1.8 (0.4%) mean scores. In HV, no radiographic erosions were seen, but MRI showed erosion-like changes, low grade BME and low-moderate synovial enhancement. It was also demonstrated that dynamic contrast enhancement does occur in healthy volunteers, and the inherent variability of perfusion measures obtained with the quantitative DCE-MRI method was small both in HV and stable RA patients on routine treatment. Conclusion: A new MRI wrist patient positioning device was developed, tested and successfully used in a longitudinal study. New and exploratory computerised image analysis techniques in RA, including bone segmentation, DXR and DCE-MRI have a potential role in longitudinal RA clinical trials.