Reproducibility and repeatability of a semi-automated pipeline to quantify trapeziometacarpal joint angles using dynamic computed tomography.

BACKGROUND: The trapeziometacarpal (TMC) joint is a mechanically complex joint and is commonly affected by musculoskeletal diseases such as osteoarthritis. Quantifying in vivo TMC joint biomechanics, such as joint angles, with traditional reflective marker-based methods can be difficult due to the joint's location in the hand. Dynamic computed tomography (CT) can facilitate the quantification of TMC joint motion by continuously capturing three-dimensional volumes over time. However, post-processing of dynamic CT datasets can be time intensive and automated methods are needed to reduce processing times to allow for application to larger clinical studies. The purpose of this work is to introduce a fast, semi-automated pipeline to quantify joint angles from dynamic CT scans of the TMC joint and evaluate the associated error in joint angle and translation computation by means of a reproducibility and repeatability study. METHODS: Ten cadaveric hands were scanned with dynamic CT using a passive motion device to move thumbs in a radial abduction-adduction motion. Static CT scans and high-resolution peripheral quantitative CT scans were also acquired to generate high-resolution bone meshes. Abduction-adduction, flexion-extension, and axial rotation angles were computed using a joint coordinate system. Reproducibility and repeatability were assessed using intraclass correlation coefficients, Bland-Altman analysis, and root mean square errors. Target registration errors were computed to evaluate errors associated with image registration. RESULTS: We found good repeatability for flexion-extension, abduction-adduction, and axial rotation angles. Reproducibility was moderate for all three angles. Joint translations exhibited greater repeatability than reproducibility. Specimens with greater joint degeneration had lower repeatability and reproducibility. We found that the difference in resulting joint angles and translations were likely due to differences in segment coordinate system definition between multiple raters, rather than due to registration errors. CONCLUSIONS: The proposed semi-automatic processing pipeline was fast, repeatable, and moderately reproducible when quantifying TMC joint angles and translations. This work provides a range of errors for TMC joint angles from dynamic CT scans using manually selected anatomical landmarks.

Internal calibration for opportunistic computed tomography muscle density analysis.

INTRODUCTION: Muscle weakness can lead to reduced physical function and quality of life. Computed tomography (CT) can be used to assess muscle health through measures of muscle cross-sectional area and density loss associated with fat infiltration. However, there are limited opportunities to measure muscle density in clinically acquired CT scans because a density calibration phantom, allowing for the conversion of CT Hounsfield units into density, is typically not included within the field-of-view. For bone density analysis, internal density calibration methods use regions of interest within the scan field-of-view to derive the relationship between Hounsfield units and bone density, but these methods have yet to be adapted for muscle density analysis. The objective of this study was to design and validate a CT internal calibration method for muscle density analysis. METHODOLOGY: We CT scanned 10 bovine muscle samples using two scan protocols and five scan positions within the scanner bore. The scans were calibrated using internal calibration and a reference phantom. We tested combinations of internal calibration regions of interest (e.g., air, blood, bone, muscle, adipose). RESULTS: We found that the internal calibration method using two regions of interest, air and adipose or blood, yielded accurate muscle density values (< 1% error) when compared with the reference phantom. The muscle density values derived from the internal and reference phantom calibration methods were highly correlated (R2 > 0.99). The coefficient of variation for muscle density across two scan protocols and five scan positions was significantly lower for internal calibration (mean = 0.33%) than for Hounsfield units (mean = 6.52%). There was no difference between coefficient of variation for the internal calibration and reference phantom methods. CONCLUSIONS: We have developed an internal calibration method to produce accurate and reliable muscle density measures from opportunistic computed tomography images without the need for calibration phantoms.

Open-source image analysis tool for the identification and quantification of cortical interruptions and bone erosions in high-resolution peripheral quantitative computed tomography images of patients with rheumatoid arthritis.

Identification of bone erosions and quantification of erosion volume is important for rheumatoid arthritis diagnosis, and can add important information to evaluate disease progression and treatment effects. High-resolution peripheral quantitative computed tomography (HR-pQCT) is well suited for this purpose, however analysis methods are not widely available. The purpose of this study was to develop an open-source software tool for the identification and quantification of bone erosions using images acquired by HR-pQCT. The collection of modules, Bone Analysis Modules (BAM) - Erosion, implements previously published erosion analysis techniques as modules in 3D Slicer, an open-source image processing and visualization tool. BAM includes a module to automatically identify cortical interruptions, from which erosions are manually selected, and a hybrid module that combines morphological and level set operations to quantify the volume of bone erosions. HR-pQCT images of the second and third metacarpophalangeal (MCP) joints were acquired in patients with RA (XtremeCT, n = 14, XtremeCTII, n = 22). The number of cortical interruptions detected by BAM-Erosion agreed strongly with the previously published cortical interruption detection algorithm for both XtremeCT (r2 = 0.85) and XtremeCTII (r2 = 0.87). Erosion volume assessment by BAM-Erosion agreed strongly (r2 = 0.95) with the Medical Image Analysis Framework. BAM-Erosion provides an open-source erosion analysis tool that produces comparable results to previously published algorithms, with improved options for visualization. The strength of the tool is that it implements multiple image processing algorithms for erosion analysis on a single, widely available, open-source platform that can accommodate future updates.

Assessing the Sensitivity of Dual-Energy Computed Tomography 3-Material Decomposition for the Detection of Gout.

OBJECTIVES: The aim of this study was to assess the accuracy and precision of a novel application of 3-material decomposition (3MD) with virtual monochromatic images (VMIs) in the dual-energy computed tomography (DECT) assessment of monosodium urate (MSU) and hydroxyapatite (HA) phantoms compared with a commercial 2-material decomposition (2MD) and dual-thresholding (DT) material decomposition methods. MATERIALS AND METHODS: Monosodium urate (0.0, 3.4, 13.3, 28.3, and 65.2 mg/dL tubes) and HA (100, 400, and 800 mg/cm 3 tubes) phantoms were DECT scanned individually and together in the presence of the foot and ankle of 15 subjects. The raw data were decomposed with 3MD-VMI, 2MD, and DT to produce MSU-only and HA-only images. Mean values of 10 × 10 × 10-voxel volumes of interest (244 µm 3 ) placed in each MSU and HA phantom well were obtained and compared with their known concentrations and across measurements with subjects' extremities to obtain accuracy and precision measures. A statistical difference was considered significant if P < 0.05. RESULTS: Compared with known phantom standards, 3MD-VMI was accurate for the detection of MSU concentrations as low as 3.4 mg/dL ( P = 0.75). In comparison, 2MD was limited to 13.3 mg/dL ( P = 0.06) and DT was unable to detect MSU concentrations below 65.2 mg/L ( P = 0.16). For the HA phantom, 3MD-VMI and 2MD were accurate for all concentrations including the lowest at 100 mg/cm 3 ( P = 0.63 and P = 0.55, respectively). Dual-thresholding was not useful for the decomposition of HA phantom. Precision was high for both 3MD-VMI and 2MD measurements for both MSU and HA phantoms. Qualitatively, 3MD-VMI MSU-only images demonstrated reduced beam-hardening artifact and voxel misclassification, compared with 2MD and DT. CONCLUSIONS: Three-material decomposition-VMI DECT is accurate for quantification of MSU and HA concentrations in phantoms and accurately detects a lower concentration of MSU than either 2MD or DT. For concentration measurements of both MSU and HA phantoms, 3MD-VMI and 2MD have high precision, but DT had limitations. Clinical implementation of 3MD-VMI DECT promises to improve the performance of this imaging modality for diagnosis and treatment monitoring of gout.

Advancements in Osteoporosis Imaging, Screening, and Study of Disease Etiology.

PURPOSE OF REVIEW: The purpose of this review is to inform researchers and clinicians with the most recent imaging techniques that are employed (1) to opportunistically screen for osteoporosis and (2) to provide a better understanding into the disease etiology of osteoporosis. RECENT FINDINGS: Phantomless calibration techniques for computed tomography (CT) may pave the way for better opportunistic osteoporosis screening and the retroactive analysis of imaging data. Additionally, hardware advances are enabling new applications of dual-energy CT and cone-beam CT to the study of bone. Advances in MRI sequences are also improving imaging evaluation of bone properties. Finally, the application of image registration techniques is enabling new uses of imaging to investigate soft tissue-bone interactions as well as bone turnover. While DXA remains the most prominent imaging tool for osteoporosis diagnosis, new imaging techniques are becoming more widely available and providing additional information to inform clinical decision-making.

Bone and joint enhancement filtering: Application to proximal femur segmentation from uncalibrated computed tomography datasets.

Methods for reliable femur segmentation enable the execution of quality retrospective studies and building of robust screening tools for bone and joint disease. An enhance-and-segment pipeline is proposed for proximal femur segmentation from computed tomography datasets. The filter is based on a scale-space model of cortical bone with properties including edge localization, invariance to density calibration, rotation invariance, and stability to noise. The filter is integrated with a graph cut segmentation technique guided through user provided sparse labels for rapid segmentation. Analysis is performed on 20 independent femurs. Rater proximal femur segmentation agreement was 0.21 mm (average surface distance), 0.98 (Dice similarity coefficient), and 2.34 mm (Hausdorff distance). Manual segmentation added considerable variability to measured failure load and volume (CVRMS > 5%) but not density. The proposed algorithm considerably improved inter-rater reproducibility for all three outcomes (CVRMS < 0.5%). The algorithm localized the periosteal surface accurately compared to manual segmentation but with a slight bias towards a smaller volume. Hessian-based filtering and graph cut segmentation localizes the periosteal surface of the proximal femur with comparable accuracy and improved precision compared to manual segmentation.

The utility of multi-stack alignment and 3D longitudinal image registration to assess bone remodeling in rheumatoid arthritis patients from second generation HR-pQCT scans.

BACKGROUND: Medical imaging plays an important role in determining the progression of joint damage in rheumatoid arthritis (RA). High resolution peripheral quantitative computed tomography (HR-pQCT) is a sensitive tool capable of evaluating bone microarchitecture and erosions, and 3D rigid image registration can be used to visualize and quantify bone remodeling over time. However, patient motion during image acquisition can cause a "stack shift" artifact resulting in loss of information and reducing the number of erosions that can be analyzed using HR-pQCT. The purpose of this study was to use image registration to improve the number of useable HR-pQCT scans and to apply image-based bone remodeling assessment to the metacarpophalangeal (MCP) joints of RA patients. METHODS: Ten participants with RA completed HR-pQCT scans of the 2nd and 3rd MCP joints at enrolment to the study and at a 6-month follow-up interval. At 6-months, an additional repeat scan was acquired to evaluate reliability. HR-pQCT images were acquired in three individual 1 cm acquisitions (stacks) with a 25% overlap. We completed analysis first using standard evaluation methods, and second with multi-stack registration. We assessed whether additional erosions could be evaluated after multi-stack registration. Bone remodeling analysis was completed using registration and transformation of baseline and follow-up images. We calculated the bone formation and resorption volume fractions with 6-month follow-up, and same-day repositioning as a negative control. RESULTS: 13/57 (23%) of erosions could not be analyzed from raw images due to a stack shift artifact. All erosions could be volumetrically assessed after multi-stack registration. We observed that there was a median bone formation fraction of 2.1% and resorption fraction of 3.8% in RA patients over the course of 6 months. In contrast to the same-day rescan negative control, we observed median bone formation and resorption fractions of 0%. CONCLUSIONS: Multi-stack image registration is a useful tool to improve the number of useable scans when analyzing erosions using HR-pQCT. Further, image registration can be used to longitudinally assess bone remodeling. These methods could be implemented in future studies to provide important pathophysiological information on the progression of bone damage.