Semi-Automated Software Detects Erosion Volume in Patients With RA

New software developed by researchers at the Brigham and Women’s Hospital Arthritis Center can successfully measure erosion volume on hand and wrist computed tomographic (CT) imaging, according to a report published in Arthritis & Rheumatology.

As articular bone erosions in rheumatoid arthritis (RA) correlate closely with functional disability, determination of accurate erosion volume is crucial. Authors of this study have developed a semiautomated software method to measure erosion volume seen on CT scans of the hand and wrist. This specially designed software facilitates the outlining of bony erosions in 3-dimensions, and then estimates erosion volume in that region.

Led by Jeffrey Duryea, PhD, of Brigham and Women’s Hospital and Harvard Medical School in Boston, Massachusetts, a team of researchers recruited 5 patients to undergo 2 sequential CT scans of each hand, 10-15 minutes apart. Patients fulfilled the American College of Rheumatology/European League Against Rheumatism classification criteria for RA, and all 5 patients were currently on disease-modifying antirheumatic drug (DMARD) therapy.

CT scans were evaluated using a semi-automated software tool to measure erosion volume in 6 subregions of both the hand and wrist. A board-certified radiologist initially identified the location of erosions, after which a research assistant marked the area outlined by the erosion margins using the new software. Finally, the radiologist performed a quality assurance evaluation.

For the vast majority of erosions, the software was able to determine the bone–erosion margin automatically. Researchers quantified reproducibility values using the intraclass correlation coefficient (ICC), the root mean square standard deviation (RMSSD), and the coefficient of variation (CV). The RMSSD and/or CV provide a direct assessment of the estimated differences in erosion volume measurement between the 2 readings.

The ICCs between the 2 separate image acquisitions were excellent, ranging from 0.97 to 1.00. At the hand level, the RMSSD was 15.6 mm³ with a CV of 7.3%, and the CVs at the 6 regions ranged from 7.6% to 21.0%. At the subject level, the RMSSD was 31.2 mm³ with a CV of 3.7%, and the CVs at the 6 regions ranged from 0.5% to 15.8%.

The results of the data showed that measurement error increases as a function of the total erosion, suggesting that the semiautomated software tool may be used to detect change, if the change in erosion volume is proportional to the total volume of the erosion.

“Articular bone erosions in RA correlate strongly with disability… no currently available treatment has been proven to reproducibly heal erosions,” wrote Dr Duryea. “To determine whether a drug therapy might heal erosions, erosion volume must be assessed accurately.”

Summary and Clinical Applicability

A novel semi-automated software method to measure erosion volume on hand and wrist CT scans was developed and was shown to detect structural changes in erosion volume.  This initial validation study was limited by the small number of subjects evaluated and the fact that advanced RA may present with osteopenia, which can potentially make the proper erosion margins difficult to define. Evaluation of the CT scans is also subject to factors specifically inherent to each particular radiologist and time spent obtaining measurements of erosion.

“Our data indicate that measurement error increases as a function of the total erosion volume in a particular anatomic region,” Dr Duryea and colleagues concluded. “This suggests that our semi-automated software tool may be used to detect change, if the change in erosion volume is proportional to the total volume of the erosion. Further longitudinal studies of therapies targeting erosion volume should help to address this issue.”

Reference

Duryea J, Russell R, Gravallese EM, et al. Development and validation of a semiautomated method to measure erosion volume in inflammatory arthritis by computed tomography scanning. Arthritis Rheumatol. 2016;68(2):332-336. doi: 10.1002/art_39459