Prediction of osteoporosis with dental radiographs and age.

OBJECTIVES: In this study age and the trabecular pattern present on dental radiographs were used to predict the presence of osteoporosis. The objective was to evaluate the contribution of the trabecular pattern to the prediction. METHODS: In this project, 671 women between 45 and 71 years of age were recruited. Medical history was obtained and dental radiographs were made. Bone mineral density (BMD) was measured at three sites to assess the presence of osteoporosis according to the World Health Organization criteria. The radiographs were subjected to image analysis methods yielding measurements of the trabecular pattern. Thereafter, discriminant analysis was used to predict the presence of osteoporosis by means of the trabecular pattern and age. Sensitivity and specificity of age and the trabecular pattern were compared. Also, it was checked whether the inclusion of the trabecular pattern improved the sensitivity and specificity that were obtained when only age was used as the predictor. RESULTS: The sensitivity and specificity of the trabecular pattern present on dental radiographs were almost equal to those of age. However, combining age with the trabecular pattern increased the sensitivity from 0.71 to 0.75 and the specificity from 0.72 to 0.78; the latter increase was statistically significant. CONCLUSIONS: The trabecular pattern predicts the presence of osteoporosis just as well as age does. When combining the trabecular pattern with age, the sensitivity and specificity increased. Only the latter increase was statistically significant.

Selecting regions of interest on intraoral radiographs for the prediction of bone mineral density.

OBJECTIVES: A previous study showed that the trabecular pattern on dental radiographs correlates with femoral and spinal bone mineral density (BMD). The objective of this study was to determine if the correlation is affected by the size and location of the region of interest (ROI). METHODS: In a European research project on osteoporosis, BMD was measured at the left hip and the lumbar spine of 525 women. From all subjects, intraoral radiographs were made of the premolar region in the upper and lower jaws. Two ROIs were indicated manually on each scanned image. The smallest region involved only trabecular bone and the largest also included parts of the neighbouring teeth. The ROIs were subjected to automatic image analysis, yielding 26 measurements per ROI. Stepwise linear regression was used to predict femoral and spinal BMD. RESULTS: Inner and outer regions predicted BMD equally well. The radiographs of lower and upper jaw also predicted BMD equally well. Combining inner and outer regions did not improve the prediction of femoral and spinal BMD, but combining lower and upper jaws did. CONCLUSIONS: This study shows that it is possible to include parts of neighbouring teeth in the ROI used to assess the trabecular pattern and predict BMD. This simplifies the process of selecting the ROIs because no efforts have to be made to exclude neighbouring teeth. Combining ROIs of lower and upper jaws significantly improves the prediction of BMD.

Spatial orientation in bone samples and Young's modulus.

Bone mass is the most important determinant of the mechanical strength of bones, and spatial structure is the second. In general, the spatial structure and mechanical properties of bones such as the breaking strength are direction dependent. The mean intercept length (MIL) and line frequency deviation (LFD) are two methods for quantifying directional aspects of the spatial structure of bone. Young's modulus is commonly used to describe the stiffness of bone, which is also a direction-dependent mechanical property. The aim of this article is to investigate the relation between MIL and LFD on one hand and Young's modulus on the other. From 11 human mandibular condyles, 44 samples were taken and scanned with high-resolution computer tomography equipment (micro-CT). For each sample the MIL and LFD were determined in 72602 directions distributed evenly in 3D space. In the same directions Young's modulus was determined by means of the stiffness tensor that had been determined for each sample by finite element analysis. To investigate the relation between the MIL and LFD on one hand and Young's modulus on the other, multiple regression was used. On average the MIL accounted for 69% of the variance in Young's modulus in the 44 samples and the LFD accounted for 72%. The average percentage of variance accounted for increased to 80% when the MIL was combined with the LFD to predict Young's modulus. Obviously MIL and LFD to some extent are complementary with respect to predicting Young's modulus. It is known that directional plots of the MIL tend to be ellipses or ellipsoids. It is speculated that ellipsoids are not always sufficient to describe Young's modulus of a bone sample and that the LFD partly compensates for this.

Osteoporosis and the general dental practitioner: reliability of some digital dental radiological measures.

OBJECTIVES: Dental radiographs are relatively inexpensive and are regularly made of a large fraction of the adult population; therefore, they represent an enormous potential as a screening tool for osteoporosis. Monitoring the population by means of dual X-ray absorptiometry (DXA), which is currently the most accepted method for diagnosing osteoporosis, involves enormous costs and facilities. In previous studies, it was shown that the radiographic trabecular pattern shows correlations with the bone mineral density (BMD) as measured by DXA. The objective of this study was to assess the reproducibility of the quantitative analysis of the trabecular pattern on dental radiographs. METHODS: Six regions of interest were selected manually on three digital radiographic images of 20 women. This process was performed 10 times resulting in 1200 image samples. For each image sample 26 parameters were measured. The reliability of the parameters was evaluated by means of Cronbach's alpha. RESULTS: Of the values of Cronbach's alpha 83% is at least 0.9 and 99% is at least 0.8. CONCLUSIONS: The measurements of the parameters used in this study are very reproducible. Therefore, the manual selection of the regions of interest does not introduce large amounts of noise. The imaging parameters potentially offer an accurate tool for the prediction of BMD values.

Detection of proximal caries with high-resolution and standard resolution digital radiographic systems.

AIMS: The aim of this study was to: (1) compare the diagnostic accuracy of the high-resolution and standard resolution settings of four digital imaging systems for caries diagnosis and (2) compare the effect on the diagnostic accuracy of reducing the high-resolution image sizes to the standard resolution dimensions, and vice versa. MATERIALS AND METHODS: 90 extracted human premolars were mounted in groups of 5 in plaster blocks, containing 4 test teeth and 1 non-test tooth. Two blocks at a time were placed in a jig to simulate a bitewing radiograph. Radiographs were taken using four digital systems (Planmeca Dixi 2; Gendex Visualix HDI; Dürr Vistascan; Digora Optime), each at two resolution settings. Next, the teeth were sectioned and a total of 65 surfaces were incorporated in the study. Additionally, the bicubic interpolation method was applied to reduce the high-resolution original images and to enlarge the standard resolution images. The original, reduced and enlarged images were randomly shown to five observers in two random sessions. The observers were asked to assess caries depth on a 4-point scale. The observers' scores were compared with the results from a histological examination. Data were analysed using the statistical theory for multivariate discrete data. Cohen's kappa was used to determine the agreement with the gold standard. RESULTS: None of the comparisons between the spatial resolution settings, or the comparisons between increased or reduced image size and the original image sizes, showed significant differences in the probability of caries detection (chi2=26.59, df=26, P approximately 0.50). The four digital systems used in this study differ significantly in the probability of caries detection (chi2=41.55, df=24, P<0.02). Compared with the gold standard, the Gendex Visualix HDI CCD sensor has the highest measure of agreement (kappa=0.31). CONCLUSION: Caries diagnosis does not improve when using high-resolution settings compared with the standard settings. The use of bicubic convolution interpolation for zooming has no detectable effect on caries diagnosis and therefore is recommended to use when enlarging or reducing radiographs. The probability of caries detection, however, was different for the sensor systems of the different manufacturers.

Prediction of bone mineral density with dental radiographs.

There is consensus to use the bone mineral density (BMD) for the operational definition of the degree of osteoporosis and the risk of osteoporotic fractures. Dual X-ray absorptiometry (DXA) is the common technique to determine BMD. Because of high costs and limited availability of DXA equipment it is worthwhile to look for alternative diagnostic techniques. As part of a larger study, the Osteodent project, we investigated if the trabecular pattern on dental radiographs can be used to predict BMD and to identify the subjects with osteoporosis and increased risk of osteoporotic fractures. In four clinical centers 671 women with an average age of 55 years were recruited. BMD values were measured by DXA equipment at the femoral neck, total hip, and spine. One panoramic and two intraoral radiographs were made. From 525 women a complete set of BMD values and radiographs was obtained. Four regions of interest on the radiographs were selected manually and then processed automatically. On all regions of interest mean and standard deviation of the gray values were measured and several features describing the shape of the binarized trabecular pattern. Multiple regression was used to predict BMD of total hip and spine by means of the radiographic measurements combined with age. It was found that age accounts for 10% of the variation in total hip BMD and 14% of the variation in spinal BMD. When all measurements on the dental radiographs are used the explained variation increases to 22% and 23%. The areas under the ROC curves are comparable to those of commonly used screening instruments for osteoporosis. It is concluded that prediction of DXA measurements of BMD by means of quantitative analysis of the trabecular pattern on dental radiographs is feasible.

A sensitive method for measuring spatial orientation in bone structures.

OBJECTIVES: This article introduces the newly developed line frequency deviation (LFD) method for measuring the orientation of the trabecular structure and shows that it is more sensitive than the mean intercept length (MIL) method that is commonly used. METHODS: The LFD method, which has been developed to measure the orientation of bone on two-dimensional X-ray images, was expanded to handle three-dimensional shapes. For the purpose of comparison, both the LFD and the MIL methods were applied to micro CT scans of 24 trabecular bone samples as well as to 24 simple synthetic samples. LFD and MIL values were calculated in various directions and collected in polar plots. Next, the anisotropy was quantified by calculating the coefficient of variation as well as by fitting ellipsoids through the plots. RESULTS: The MIL method yielded smooth rather spherical ellipsoidal polar plots with almost no sensitivity for changes in structure. The LFD method yielded more slender polar plots and more sensitivity for geometrical changes. The LFD method yielded significantly more anistropy and larger variation in anisotropy. CONCLUSIONS: The LFD method is a more sensitive descriptor of spatial orientation of bone structures than the MIL method.

End-user survey for digital sensor characteristics: a pilot questionnaire study.

OBJECTIVE: To survey end-user opinions on dental digital sensor characteristics for the design of a new X-ray imaging sensor. MATERIALS AND METHODS: 100 questionnaires were sent out to dentists and dental radiologists. The questionnaire consisted of six parts related to dental sensors. A: Details about the respondent; B: Prioritization of most important aspects of digital sensors; C: Rating advantages and D: disadvantages of digital sensors; F: Dental features that need to be enhanced by digital sensors; G: End-user comments. RESULTS: Fifty-six questionnaires were returned. Contrast resolution and imaging time were assessed as the most and the least important aspects, respectively. Aspects considered as advantages by approximately 80% of respondents were: optimal contrast resolution, increased specificity, increased discrimination between diseased and healthy tissue, increased sensitivity, lower dose to the patient by more than 20%; as a disadvantage, increased patient dose was given. Dental features considered important by over 80% of the respondents were for intraoral radiographs: the visibility of caries, periapical, periodontal, and bone lesions; for panoramic radiographs: reduction of ghost images, image sharpness and bone. Bone and soft tissues on cephalograms and bone on temporomandibular joint (TMJ) tomograms were mentioned by over 70%. CONCLUSION: The most desired characteristic for a new sensor is contrast resolution. Dose saving is not considered as important unless it is more than 20%. Examination time is not a major issue. Caries, periapical and periodontal, and bone lesions for intraoral radiographs, reduction of ghost images, image sharpness and bone for panoramic radiographs were the most frequently mentioned dental features that should be enhanced by digital sensors.

An efficient filling algorithm for counting regions.

Region filling has many applications in computer graphics and image analysis. Some region filling tasks can be performed by fast scan line filling algorithms. Other region filling tasks require seed filling algorithms which are more general but slower. This paper introduces a seed filling algorithm that is designed to count regions irrespective of their shape. The method is described and its performance is compared with three alternative algorithms by applying them to a collection of 34 test images. The four methods showed complete agreement with respect to the counted numbers of regions. The proposed method was found to be fastest and requiring least memory.