Joel R. Wilkie

Comparison of radiographic texture analysis from computed radiography and bone densitometry systems.

Osteoporosis is a disease that results in an increased risk of bone fracture due to a loss of bone mass and deterioration of bone structure. Bone mineral density (BMD) provides a measure of bone mass and is frequently measured by bone densitometry systems to diagnose osteoporosis. In addition, computerized radiographic texture analysis (RTA) is currently being investigated as a measure of bone structure and as an additional diagnostic predictor of osteoporosis. In this study, we assessed the ability of a peripheral bone densitometry (PD) system to yield images useful for RTA. The benefit of such a system is that it measures BMD by dual-energy x-ray absorptiometry and therefore provides high- and low-energy digital radiographic images. The bone densitometry system investigated was the GE/Lunar PIXI, which provides 512 x 512 digital images of the heel or forearm (0.2 mm pixels). We compared texture features of heel images obtained with this PD system to those obtained on a Fuji computed radiography (CR) system (0.1 mm pixels). Fourier and fractal-based texture features of images from 24 subjects who had both CR and BMD exams were calculated, and correlation between the two systems was analyzed. Fourier-based texture features characterize the magnitude, frequency content, and orientation of the trabecular bone pattern. Good correlation was found between the two modalities for the first moment (FMP) with r=0.71 (p value<0.0001) and for minimum FMP with r=0.52 (p value=0.008). Root-mean-square (RMS) did not correlate with r=0.31 (p value>0.05), while the standard deviation of the RMS did correlate with r=0.79 (p value<0.0001). Good correlation was also found between the two modalities for the fractal-based texture features with r=0.79 (p value<0.0001) for the global Minkowski dimension and r=0.63 (p value=0.0007) for the fractal dimension from a box counting method. The PD system therefore may have the potential for yielding heel images suitable for RTA.

Temporal radiographic texture analysis in the detection of periprosthetic osteolysis

Periprosthetic osteolysis is one of the most serious long-term problems in total hip arthroplasty. It has been primarily attributed to the bodys inflammatory response to submicron polyethylene particles worn from the hip implant, and it leads to bone loss and structural deterioration in the surrounding bone. It was previously demonstrated that radiographic texture analysis (RTA) has the ability to distinguish between osteolysis and normal cases at the time of clinical detection of the disease; however, that analysis did not take into account the changes in texture over time. The goal of this preliminary analysis, however, is to assess the ability of temporal radiographic texture analysis (tRTA) to distinguish between patients who develop osteolysis and normal cases. Two tRTA methods were used in the study: the RTA feature change from baseline at various follow-up intervals and the slope of the best-fit line to the RTA data series. These tRTA methods included Fourier-based and fractal-based features calculated from digitized images of 202 total hip replacement cases, including 70 that developed osteolysis. Results show that separation between the osteolysis and normal groups increased over time for the feature difference method, as the disease progressed, with area under the curve (AUC) values from receiver operating characteristic analysis of 0.65 to 0.72 at 15 years postsurgery. Separation for the slope method was also evident, with AUC values ranging from 0.65 to 0.76 for the task of distinguishing between osteolysis and normal cases. The results suggest that tRTA methods have the ability to measure changes in trabecular structure, and may be useful in the early detection of periprosthetic osteolysis.

Investigation of physical image quality indices of a bone densitometry system

Osteoporosis is a disease characterized by a loss of bone mass and a deterioration of bone structure. Bone mineral density (BMD) measures bone mass and is currently the method used to diagnose osteoporosis, while computerized radiographic texture analysis (RTA) is being investigated as a measure of bone structure. The GE/Lunar PIXI peripheral bone densitometer (PD) system, which uses dual-energy subtraction to measure BMD, also provides a digital image of the heel or forearm. The goal of our current research was to evaluate the physical imaging properties of the PIXI system (pixel size of 0.2 mm) compared to a Fuji computed radiography (CR) system (pixel size of 0.1 mm) to determine its suitability for texture analysis from image data. Contrast was measured using a series of uniform images covering the useful clinical exposure range. Spatial resolution was characterized by the presampling modulation transfer function (MTF) determined by an edge method. Noise power spectra (NPS) for different exposures were calculated using a two-dimensional Fourier analysis method. The expectation modulation transfer function was measured and combined with the NPS data to calculate the noise-equivalent number of quanta. The slope of the characteristic curve of the peripheral densitometer (PD) system was found to be position dependent across the image, although this dependence was substantially reduced by use of the systems clinical-settings corrections. An MTF value of 0.5 was found at 0.5 cycles/mm for the densitometry system compared to the same value at 1.6 cycles/mm for the CR system. Unlike the CR system, the NPS of the densitometry system was found not to be directionally dependent and did not drop off at higher spatial frequencies.

Radiographic Texture Analysis in the Characterization of Trabecular Patterns in Periprosthetic Osteolysis

RATIONALE AND OBJECTIVES Periprosthetic osteolysis is a disease attributed to the bodys reaction to fine polyethylene wear debris shed from total hip replacements. The purpose of this preliminary study was to investigate the ability of radiographic texture analysis (RTA) to characterize the trabecular texture patterns on pelvic images for osteolysis and normal total hip arthroplasty (THA) cases. MATERIALS AND METHODS Fourier-based and fractal-based texture features were calculated for a database of digitized radiographs from 202 THA cases, 70 of which developed osteolysis. The features were calculated from regions of interest selected at two time points: less than 1 month after surgery, and at the first clinical indication of osteolysis (or randomly selected follow-up time for normal cases). Receiver operating characteristic (ROC) analysis was used to compare feature performance at baseline and follow-up for osteolysis and normal cases. RESULTS Separation between the RTA features for osteolysis and normal cases was negligible at baseline and increased substantially for the follow-up images. The directional Fourier-based feature provided the best separation with an A(z) value from ROC analysis of 0.75 for the follow-up images, in the task of distinguishing between normal and osteolytic cases. CONCLUSIONS The results from this preliminary analysis indicate that qualitative changes in trabecular patterns from immediately after surgery to the eventual detection of osteolysis correspond to quantitative changes in RTA features. It therefore appears that RTA provides information that could potentially be useful to aid in the detection of this disease.

Imputation methods for temporal radiographic texture analysis in the detection of periprosthetic osteolysis

Periprosthetic osteolysis is a disease triggered by the bodys response to tiny wear fragments from total hip replacements (THR), which leads to localized bone loss and disappearance of the trabecular bone texture. We have been investigating methods of temporal radiographic texture analysis (tRTA) to help detect periprosthetic osteolysis. One method involves merging feature measurements at multiple time points using an LDA or BANN. The major drawback of this method is that several cases do not meet the inclusion criteria because of missing data, i.e., missing image data at the necessary time intervals. In this research, we investigated imputation methods to fill in missing data points using feature averaging, linear interpolation, and first and second order polynomial fitting. The database consisted of 101 THR cases with full data available from four follow-up intervals. For 200 iterations, missing data were randomly created to simulate a typical THR database, and the missing points were then filled in using the imputation methods. ROC analysis was used to assess the performance of tRTA in distinguishing between osteolysis and normal cases for the full database and each simulated database. The calculated values from the 200 iterations showed that the imputation methods produced negligible bias, and substantially decreased the variance of the AUC estimator, relative to excluding incomplete cases. The best performing imputation methods were those that heavily weighted the data points closest to the missing data. The results suggest that these imputation methods appear to be acceptable means to include cases with missing data for tRTA.

68 CLINICAL UTILITY OF RADIOGRAPHIC TEXTURE ANALYSIS PERFORMED ON DENSITOMETRIC CALCANEAL IMAGES.

Background Currently, the major challenge in osteoporosis research and clinical practice is to improve the assessment of fracture risk. Bone fragility is determined by bone quantity measured as bone mineral density (BMD) and by bone quality or structure that cannot be easily assessed using currently available methods. Radiographic texture analysis (RTA) is an image-based noninvasive method of evaluating bone structure through computerized analysis of trabecular pattern of bone radiographs. We have applied RTA to calcaneal images obtained using a portable densitometer. RTA yields the following features: RMS (root mean square variation, a measure of the variability in the radiographic texture pattern, the relative difference in the contrast between light and dark areas with higher values corresponding to the strong rich trabecular pattern), and its directional measure, sdRMS (standard deviation of RMS), which is a measure of anisotropy of the trabecular pattern, with higher values corresponding to the more directional texture pattern, and Minkowski fractal (MINK), a measure of roughness-smoothness of the trabecular pattern. In this study of patients referred for BMD testing at the University of Chicago, we examined (1) which biologic factors influence RTA and (2) how well RTA differentiates subjects with high and low bone fragility. Methods In 837 patients (age 19-95 years, 759 women), BMD was measured at the lumbar spine, proximal femur, and calcaneus; clinical risk factors were obtained using a questionnaire; and RTA was performed on densitometric heel images. Fragility was defined as the presence of vertebral fractures detected on VFA (vertebral fracture assessment - densitometric spine image). The association of RTA with biologic factors or with prevalent vertebral fractures was examined using regression analysis. Results (1) Relationship between RTA and biologic factors: In a multivariate regression analysis, significant predictors of sdRMS were BMI (p < .001), gender (p < .001), age (p = .001), race (p = .004), and heel BMD (p = .009). For MINK, the only significant predictors were BMI (p < .001) and heel BMD (p < .001). (2) Utility of RTA in assessing fragility was examined in a subset of 333 postmenopausal women who had no secondary causes of and were not receiving therapy for osteoporosis. In a univariate logistic regression analysis with the presence of prevalent vertebral fractures as a binary outcome, the separations between 49 women with and 284 without vertebral fractures using RTA and using BMD measurement were similar (p = .002 and p < .001 for sdRMS and MINK, respectively; p = .01, p = .003, and p < .001 for spine, heel, and hip BMD, respectively). In a multivariate logistic regression, there was a significant association of vertebral fractures with RTA, BMD, and age with odds ratio (OR) of having a vertebral fracture of 2.0 for each decade increase in age (p < .001), OR = 1.8 for a 1-unit decrease in hip BMD T-score (p = .003), and OR = 1.7 for a 1 SD decrease in sdRMS (p = .003) or MINK (p = .008). Conclusion These results suggest that RTA of heel images obtained using a portable densitometer characterizes bone properties not measured by BMD. Specifically, RTA provides assessment of fragility that is not captured by the currently used measurements such as BMD and clinical risk factors (age). Addition of RTA to BMD and clinical risk factors can improve the stratification of fracture risk. Further studies are needed to determine whether RTA would be useful in selecting patients for different forms of osteoporosis therapy.

WE-C-I-609-05: Temporal Radiographic Texture Analysis for the Detection of Periprosthetic Osteolysis

Purpose: We have been investigating temporal radiographic texture analysis (RTA) as a method to help detect the disease periprosthetic osteolysis associated with total hip arthroplasties. This disease is a common but difficult to detect long-term complication for total hip replacement patients. It typically goes unnoticed on radiographs until at least seven years after the operation. The goal of our research is to assess the ability of temporal RTA to detect osteolysis before it is visible radiographically. Method and Materials: We obtained digitized pelvis radiographs from 84 total hip replacement cases from the Anderson Orthopaedic Research Institute. Each case included a baseline image taken shortly after surgery and follow-up images taken at various time intervals. The cases were assessed for osteolysis by an orthopaedic surgeon and regions of interest (ROIs) were selected within the osteolytic region (or a comparable region for normal cases) on the final image of each case. These ROIs were then visually matched on all previous images. Fourier-based, fractal-based and correlation-based features were calculated for each ROI. To measure temporal trends in feature values, we calculated the slope of the least squares fitted line for each case using data through five year and nine year time ranges, respectively. Temporal feature performance was examined using Receiver Operating Characteristic (ROC) analysis. Results: Forty-four cases were determined to have osteolysis while forty were normal. Az values from ROC curves ranged from 0.6 to 0.75 for the task of distinguishing between osteolysis and normal cases for both time ranges. Conclusion: Temporal RTA appears to have the potential to help detect periprosthetic osteolysis before visual radiographic appearance of the disease. More development of temporal RTA and analysis with a larger patient database is therefore warranted. Conflict of Interest: M.L.G. is a shareholder in R2 Technology, Inc. (Sunnyvale, CA).