Article Text

Download PDFPDF
  1. M. Boesen1,
  2. M. Hinton2,
  3. J. Gonzalez-Zabaleta3,
  4. S. Beattie4,
  5. D. Schlichting4,
  6. T. Rooney4,
  7. O. Kubassova3
  1. 1University Hospital Bispebjerg-Frederiksberg, Copenhagen, Denmark
  2. 2Image Analysis Group, Philadelphia, United States of America
  3. 3Image Analysis Group, London, United Kingdom
  4. 4Eli Lilly and Company, Indianapolis, United States of America


Background: Magnetic resonance imaging (MRI) was used to confirm dose selection in a phase IIb clinical trial of baricitinib in patients with active rheumatoid arthritis (RA) on background methotrexate therapy (NCT01185353).[1] MRI data were retrospectively assessed for consistency, timing of post-contrast sequences following intravenous (IV) Gadolinium (Gd), and readability. Data were re-analyzed using a novel quantitative computer-aided methodology to extract the continuous volume of inflammatory changes.[2]

Objectives: The objective was to examine how image quality and timing of the post-contrast MRI sequence can impact MRI-based exploratory endpoints in RA clinical trials when using novel computer-aided analysis tools.

Methods: A total of 154 patients with definitive radiographic erosion had an MRI of the hand and wrist at baseline and at weeks 12 and 24. Three-dimensional T1-w fat-suppressed MRI sequences before and after IV Gd contrast were performed with dedicated coils. Due to the limited field of view, the coils were re-positioned during the image acquisition between the metacarpophalangeal (MCP) and finger joints and the wrist, following IV Gd injection, which introduced a time delay of the post-contrast sequences in the two anatomies in all patients.

Digital Imaging and Communications in Medicine (DICOM) headers of the MRIs were automatically assessed; the distribution of the time delay in minutes from Gd injection to post-contrast scan acquisition was calculated and the image quality and suitability for reading were evaluated (Figure 1). The time delays across MRI acquisitions at baseline and weeks 12 and 24 were also compared. Quality scores were assigned for each image using visual image quality assessment by an experienced reader blinded to treatment regimen, patient visits, and time after Gd. The images were categorized by quality based on total score. The reader used a proprietary software, to pre-define regions of interest (ROI) around the wrist and MCP joints (MCP-2 to MCP-5) in all three timepoints as a batch, avoiding adjacent blood vessels and possible artifacts. From these ROIs, the normalized volume of inflammation (NormI) was calculated in each joint relative to a standardized ROI in the thenar muscle. Quantitative Total Volume of Inflammation (QVI) was extracted automatically from all ROIs by counting the pixels that were enhanced two standard deviations above the intensity level of the normal muscle, allowing differentiation of areas with low-to-high inflammation.

Results: The timing of post-contrast images from Gd injection was closely linked to image quality. In up to 10% of MRI data, the delay from Gd injection to scan acquisition caused significant variation in signal intensities. This led to a perceived increase in enhanced synovial volume due to the known effusion effects of the contrast media over time, which did not correspond to real size of the underlying synovial volume and pathology (Figure 2).

Conclusion: The acquisition of MRIs in RA trials should be done in a methodical and systematic manner, where the quality of MRI scans and the correct timing of post-contrast sequences are optimized. The incorporation of unacceptable quality data will impact the interpretation of RA clinical trial data, especially when novel computer-assisted quantitative analysis methods for post-processing are used. Incorrect timing and inconsistency in image quality can be prevented by using coils covering the whole hand and/or a dynamic contrast-enhance (DCE)-MRI sequence immediately following IV Gd injection to ensure correct timing of the post-contrast MRI sequence.

References: [1]Peterfy C, et al. J Rheumatol. 2019 46: 887–895.

[2]Tripathi D, et al. IJRCI. 2014 2(S1):SR2. DOI: 10.15305/ijrci/v2iS1/89.

Disclosure of Interests: Mikael Boesen Consultant of: AbbVie, AstraZeneca, Eli Lilly, Esaote, Glenmark, Novartis, Pfizer, UCB, Paid instructor for: IAG, Image Analysis Group, AbbVie, Eli Lilly, AstraZeneca, esaote, Glenmark, Novartis, Pfizer, UCB (scientific advisor)., Speakers bureau: Eli Lilly, Esaote, Novartis, Pfizer, UCB, Mark Hinton: None declared, Javier Gonzalez-Zabaleta: None declared, Scott Beattie Shareholder of: Eli Lilly and Company, Employee of: Eli Lilly and Company, Douglas Schlichting Shareholder of: Eli Lilly and Company, Employee of: Eli Lilly and Company, Terence Rooney Shareholder of: Eli Lilly and Company, Employee of: Eli Lilly and Company, Olga Kubassova Shareholder of: IAG, Image Analysis Group, Consultant of: Novartis, Takeda, Lilly, Employee of: IAG, Image Analysis Group

Statistics from

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.