Digital image analysis protocol for determining the radiocarpal joint space in the rheumatoid arthritic wrist

This paper describes a simple protocol for measuring the joint space of the rheumatoid arthritic (RA) wrist from projection radiographs. The protocol is implemented using a computer algorithm based upon the Interactive Data Language platform. The computerized algorithm features a user-friendly graphical interface to aid the operator to measure joint space parameters, namely distance and area, of the wrist vertebral morphometry at the radiocarpal region. Dual-energy X-ray absorptiometry (DXA) radiograph of a standard hand and wrist phantom was evaluated using the measurement protocol to determine the accuracy and precision of the protocol. The accuracy, parameterised by the systematic error, returned a mean of 5.20% for distance and is equal to 3.49% for area measurement. The precision of the measurement protocol, parameterised by the coefficient of variation (CV), for distance returned a mean of 1.96%; the CV for area measurement equals 2.1%. Three observers participated to investigate the repeatability (intra-observer) and reproducibility (inter-observer) of the measurement protocol, parameterised by the CV, using DXA radiographs from a healthy volunteer and a RA patient. The inter-observer repeatability for distance measurement for the respective observers returned mean values of 10.9%, 7.7% and 11.4% for the healthy wrist. However, the results revealed improved repeatability for the RA wrist; the CV for the respective observers returned mean values of 7.7% 7.1% and 10.0%. The inter-observer repeatability for area measurement for the respective observers returned mean values of 10.2%, 7.1% and 10.1% for the healthy wrist. However, the results revealed improved repeatability (in two out of the three observers) for the RA wrist; the CV for the respective observers returned mean values of 6.8% 6.5% and 10.8%. Student's t-test analysis of the intra-observer repeatability revealed that the measurements of distance and area were generally not intra-observer sensitive. On the other hand, student's t-test analysis of the inter-observer reproducibility revealed that half of the distance measurements were inter-observer sensitive; whereas the remaining were not. Similar findings were obtained for area measurements. Overall the results reveal that the variabilities in accuracy and precision tests and the repeatability and reproducibility tests were typically 10% or less. These findings, in addition to the versatility and simplicity of the digital image analysis protocol, lend to the potential of using the protocol to complement the acquisition of bone mineral density data derived from DXA for diagnosing the progression of RA in patients.