Article Text
Abstract
Background Currently, radiographs are used to semi-quantitatively assess bone destruction in patients with rheumatoid and psoriatic arthritis. However, an exact quantification of bone erosions may be important for evaluating disease activity and treatment efficacy. Based on high-resolution peripheral QCT (hr-pQCT) we developed a new segmentation technique, allowing the quantification of volume and shape of bone erosions.
Objectives To develop a new technique for semiautomatic 3D segmentation and quantification of volume, surface area and shape of erosions in in-vivo images of the metacarpophalangeal (MCP) joints of patients with RA.
Methods High resolution peripheral quantitative CT images from 25 RA patients (mean age ± SD 54.2±13.7 years) were acquired on an XtremeCT scanner (Scanco Switzerland) with an isotropic voxel size of 82μm. The scans covered 3cm. Erosions were quantified in second to fourth MCP joint. The 3D segmentation consisted of the following steps: 1. automatic 3D segmentation of the periosteal bone surface 2. placement of a seed point in each erosion, 3. automatic 3D segmentation of each erosion including the separation with the marrow space and the protrusion of the cortex (to differentiate erosions from cysts). 4. interactive corrections when necessary. Here we report initial intra-operator analysis precision results determined for erosion volume, surface area and sphericity of 56 erosions detected in the 25 patients.
Results The figure shows two multi-planar reformations of a segmented erosion. Mean (min/max) values for volume, surface area and sphericity were 7.4mm3 (0.37mm3/54.7mm3), 31.1mm2 (3.48mm2/159.0mm2) and 0.58 (0.39/0.72). Manual correction was necessary for 27 erosions, in particular if the boundary between erosion and marrow space was ambiguous. The analysis was repeated three times by the same operator. Root mean squares of coefficients of variation (rmsCV) were 4.1%, 4.7% and 2.5% for volume, surface area and sphericity. Root mean squares of standard deviations (rmsSD) were 0.35mm3, 2.15mm2 and 0.01. The rmsCV was dominated by 4 manually edited erosions, which were rather small (<4.3mm3). Exclusion of these 4 improved the rmsCV to 2.3%, 3.3% and 2.2%.
Conclusions Compared to standard CT and MRI imaging techniques, HR-pQCT provides improved spatial resolution and may become a preferred modality to quantify erosions in RA. For these scans we developed a semiautomatic image analysis technique, with which in addition to a precise measurement of the volume of the erosions an assessment of their shape is provided. Potential limitations are an accurate quantification of large erosions that are difficult to separate from bone marrow as well as vast deteriorations of the periosteal surface present in advanced RA. While HR-pQCT currently offers the best spatial resolution for 3D in-vivo imaging of the hands, the scan range is limited and scan times are rather long (approx. 5 min) with the consequence of frequent movement artifacts.
Disclosure of Interest None Declared