Anthony Holmes MSC, PhD, Presenter: Employee, Optasia Medical Ltd

Joes Staal PhD, Abstract Co-Author: Employee, Optasia Medical Ltd

Brian Purvis, Abstract Co-Author: Employee, Optasia Medical Ltd

Tony Lacey BEng, PhD, Abstract Co-Author: Employee, Optasia Medical Ltd

Alan Brett PhD, Abstract Co-Author: Officer, Optasia Medical Ltd

The thoughtful implementation of image analysis techniques, combined with a clinician-centred user workflow can greatly improve precision and speed when scoring radiographs for RA progression, making feasible the use of such scoring at point-of-care.

Scoring of joint space narrowing (JSN) and erosions in radiographs by trained physicians, as an estimate of disease severity is a useful method of evaluating structural damage in RA. A number of studies, however, have demonstrated that scoring is highly evaluator-dependent, has limited precision, is sensitive to image acquisition technique and can be slow, taking up to 25 minutes to score a single baseline/follow-up pair of radiographs. Clearly, it would be advantageous to automate some of this process to achieve a more reproducible and rapid result, and improve data integrity through the reduction of transcription errors. The introduction of image-analysis enabled workflow improvements may also make feasible the use of radiographic scoring in a healthcare setting.

Clinicians were asked to evaluate a software prototype that contained the following key workflow features, on retrospective, IRB-approved, patient images:- VIRTUAL LIGHTBOX: multiple image display optimised to user’s monitor configuration. RESOLUTION INDEPENDENCE: images are displayed at twice the anatomical size, independent of monitor resolution. REGION/JOINT FIND: images are automatically segmented into scoring regions (DIPs, PIPs, MCPs and thumb, carpals) and joints are highlighted. User can display all timepoints for a given region. REGION/JOINT REGISTRATION: When scoring a joint in one image, the same joint at other timepoints is automatically shown at same scale, for reference. LINKED SCORING: scores are recorded at each joint to avoid transcription errors. ATLAS: integrated database of scoring examples to provide decision support.  

The features described were intended to improve precision and speed of scoring, and also to reduce transcription errors. Experienced radiologists and rheumatologists would probably not utilise the atlas. Further analysis is needed to quantify the workflow improvements and determine efficient utility.

Holmes, A, Staal, J, Purvis, B, Lacey, T, Brett, A, A Software-based Workflow Tool for the Scoring of Rheumatoid Arthritis on X-Rays.  Radiological Society of North America 2008 Scientific Assembly and Annual Meeting, February 18 - February 20, 2008 ,Chicago IL. http://archive.rsna.org/2008/6022089.html