BJR
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
First published online December 10, 2007
British Journal of Radiology (2008) 81, 129-136
© 2008 British Institute of Radiology
doi: 10.1259/bjr/61371891
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by BONIATIS, I
Right arrow Articles by PANAYIOTAKIS, G
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by BONIATIS, I
Right arrow Articles by PANAYIOTAKIS, G

Full paper

A morphological index for assessing hip osteoarthritis severity from radiographic images

I BONIATIS, MSc 1 L COSTARIDOU, PhD 1 D CAVOURAS, PhD 2 I KALATZIS, PhD 2 E PANAGIOTOPOULOS, MD, PhD 3 and G PANAYIOTAKIS, PhD 1

1 University of Patras, School of Medicine, Department of Medical Physics, Patras, 2 Technological Educational Institute of Athens, Department of Medical Instrumentation Technology, Athens, 3 University of Patras, School of Medicine, Department of Orthopaedics, Patras, Greece

Correspondence: George S Panayiotakis, University of Patras, School of Medicine, Department of Medical Physics, 265 00 Patras, Greece. E-mail: panayiot{at}upatras.gr


   Abstract
Top
 Abstract
Introduction
 Methods and materials
 Results and discussion
 Conclusions
 References
 
A new method is proposed for assessing the severity of hip osteoarthritis (OA) based on radiographic hip joint space (HJS) morphology. 64 hips of patients with verified unilateral OA or bilateral OA were studied by digitizing the corresponding pelvic radiographs. Radiographic OA severity was assessed employing the Kellgren and Lawrence (KL) scale. Using custom-developed software, radiographs were enhanced, the margins of both HJSs were outlined, and 64 regions of interest (ROIs), corresponding to the delineated HJSs, were obtained. Employing custom-developed algorithms, an index ("joint space morphological index" – JSMI) evaluating alterations in the shape and size of HJS was introduced, calculated and normalized with respect to each patient's individual anatomy. The JSMI values were used to introduce classification rules concerning the characterization of a hip in accordance with the KL scale. For each patient in the unilateral OA group, the OA severity was expressed as the percentage of the HJS area difference between the patient's osteoarthritic and contralateral normal hip. The per cent HJS area difference and the JSMI values were used in the design of a regression model for providing a quantitative estimation of OA severity. The per cent HJS area difference correlated highly with the pathological JSMI values (r = –0.83, p<0.001). The implementation of the JSMI-based classification rules resulted in high classification accuracies for characterizing hips as normal or osteoarthritic, 90.6% (95% exact confidence interval (CI): 80.7–96.5%), as well as for discriminating among OA severity categories, 91.7% (95% CI: 77.5–98.2%). Additionally, a simplified approach of JSMI calculation is suggested for daily clinical use. These JSMI values (JSMI simplified) were found not to differ significantly from (p>0.05), and to be strongly correlated with (r = 0.96, p<0.001), the corresponding ones obtained by the computerized approach. Additionally, the implementation of classification rules based on JSMI simplified resulted in classification accuracies identical to the corresponding ones obtained for the JSMI-based rules. The proposed method may be utilized for evaluating OA and monitoring OA progression.


   Introduction
Top
 Abstract
 Introduction
Methods and materials
 Results and discussion
 Conclusions
 References
 
Osteoarthritis (OA) is a slowly developing disorder, characterized by non-uniform degeneration of articular cartilage and reparative formation of new bone, which results in stiffness and pain of the affected joint [1]. Despite the fact that MRI is increasingly being used in the investigation of the disease [2], plain film radiography is considered as the gold-standard imaging modality for assessing joint destruction related to OA. The characteristic radiographic alterations of the hip joint related to OA include hip joint space (HJS) narrowing (reflecting the progressive loss of articular cartilage), development of cysts in the subchondral bone, osteophyte formation, and increased density of subchondral bone in the form of subchrondal sclerosis [3]. Regarding the radiographic assessment of the severity of the disease, several grading scales have been proposed [4–6]. These scales are a subjective assignment of a severity grade to the studied hip joint, whereas the grades are defined on the basis of hip joint structural alterations visualized on plain radiographs. The Kellgren and Lawrence (KL) grading scale [4] is considered as the gold-standard despite its deficiencies [7].

The presence of HJS narrowing on radiographic images of the hip joint has been considered as a defining criterion for epidemiological studies of hip OA [8], and the specific radiographic feature has been accepted as the most reliable index for monitoring disease progression [5]. Quantification of HJS narrowing requires measurements of radiographic HJS parameters, such as HJS width and HJS area [9]. HJS width can be measured with computerized [10, 11] or with manual [12–14] methods, whereas HJS area can be measured only with computerized methods [10, 11].

HJS narrowing causes alterations in the morphology of the HJS on radiographic images. In particular, the progressive and non-uniform loss of articular cartilage is expected to alter the shape and the size of the specific radiographic region in osteoarthritic hips. Thus, the introduction of an index, capable of evaluating morphological alterations of the HJS due to OA, could be of value for discrimination among OA severity categories as well as quantification of the severity of disease. Previous studies [8, 13] have introduced HJS width thresholds for characterizing a hip as normal or osteoarthritic. In previous studies performed by our group, the radiographic texture of HJS was used to evaluate hip joint alterations related to OA [15–17], as well as for quantitative assessment of the severity of disease [15]. In the present study, however, joint alterations of osteoarthritic hips are assessed through differences concerning not only the size but also the shape of the HJS due to OA. To the best of our knowledge, a similar approach has not been previously reported.

In particular, the objective of the present study was to develop a computerized method for evaluating the severity of hip OA by implementing image analysis algorithms on digitized pelvic radiographs. For this (i) a new quantitative index, capable of evaluating alterations in the morphology of radiographic HJS, was introduced and was normalized to a patient's individual anatomy; (ii) classification rules involving this morphological index were established for discriminating among various grades of hip OA severity; and (iii) a regression model was proposed for estimating OA quantitatively on each individual hip. In addition, an equivalent approach for the calculation of morphological index values, which is easily applied in daily clinical practice and does not rely on the use of a computerized system, but provides practically the same results as the computerized one, is proposed.


   Methods and materials
Top
 Abstract
 Introduction
 Methods and materials
Results and discussion
 Conclusions
 References
 
Patients and radiographs
64 hip images (18 normal and 46 osteoarthritic) were selected from standing anteroposterior (AP) weight-bearing pelvic radiographs of patients with verified unilateral or bilateral hip OA. Diagnosis of hip OA was performed according to the clinical and radiographic American College of Rheumatology criteria [18]. Radiographs were retrieved from the archives of our university hospital. All patients were candidates for total hip arthroplasty at the department of orthopaedics. 18 patients were verified for unilateral OA, 14 for bilateral OA. Patients' ages ranged between 49 years and 83 years, with a mean of 66.7 years. Radiographic assessment of OA severity was based on the KL grading-scale criteria and was performed by three experienced orthopaedists. According to the KL scale [4], hips with KL = 0, 1, 2, 3, 4 are characterized as normal, doubtful, mild, moderate and severe OA, respectively (Table 1Go). Based on the KL scale, our data were grouped into the following major OA severity categories: normal/doubtful (KL = 0, 1), mild/moderate (KL = 2, 3) and severe (KL = 4). Accordingly, hip distribution was 18/9/9 for the unilateral and 0/7/21 for the bilateral group to normal–doubtful/mild–moderate/severe OA categories, respectively.


View this table:
[in this window]
[in a new window]

  		Table 1. The Kellgren and Lawrence grading scale

 
Radiographs were performed on an X-ray unit (Siemens, Polydoros 50, Erlangen, Germany) with the following settings: 70–80 kVp, 100 cm focus to film distance, alignment of the X-ray beam 2 cm above the pubic symphysis, use of a fast screen and film cassette (30 cmx 40 cm). Radiographs were digitized at 12 bits (4096 grey levels) with a spatial resolution of 146 ppi (5.8 pixels mm–1), using a laser digitizer for medical applications [19]. Digitizer performance was evaluated employing a quality control protocol [20].

In order to safeguard against variations in a patient's individual anatomy, the length of a fixed line was used as a normalization factor. For each patient the line was drawn perpendicular to the femur shaft (see Figure 1Go, Line-N). To restrict variations in hip rotation, the width of the projected lesser trochanter was measured for both hips of each patient. According to the experienced orthopaedists, a threshold width difference less than or equal to 8 mm was considered acceptable. The specific threshold value was adopted as the median value of the width differences in a large sample of pelvic radiographs. Finally, only those radiographs that fulfilled this specific criterion were retained. Measurements were performed by means of custom-developed software [21, 22].


Figure 1
View larger version (151K):
[in this window]
[in a new window]

  		Figure 1. Determination of the hip joint space region of interest(HJS-ROI). tOs, acute angle defined by patient's standard anatomical landmarks encompassing the examined ROI. Line-N, fixed line used for normalization purposes.

 
Determination of hip joint space
Using Matlab (The MathWorks, Natic, USA), the contrast-limited adaptive histogram equalization (CLAHE) method [23] was applied on the digitized radiographs in order to enhance the image contrast emphasizing the articular margins of the hip joint. On each enhanced pelvic radiograph, two regions of interest (ROIs), corresponding to both HJSs, were manually segmented employing custom-developed software [21, 22].

The HJS-ROI was enclosed within an acute angle (see Figure 1Go, tOs), which was defined by the patient's standard anatomical landmarks, i.e. the centre of the femoral head (O), the lateral rim of the acetabulum (s) and the highest point of the homolateral sacral wing (t) [10]. Within this specified region, the articular margins of the hip joint were manually delineated by the orthopaedists in order to provide a more accurate outline of HJS. Manually segmented ROIs (see Figure 2Go) were further analysed as grayscale images.


Figure 2
View larger version (17K):
[in this window]
[in a new window]

  		Figure 2. Manually segmented hip joint space region of interest(HJS-ROI), corresponding to Figure 1Go.

 
Computer-based calculation of the joint space morphological index
In this study, a novel index capable of quantifying alterations in the shape and the size of radiographic HJS was introduced for the assessment of hip OA severity. This index, labelled as the "joint space morphological index (JSMI)", was generated from each HJS-ROI image according to the following automated steps:

  1. determination of the centre of "mass" ("centroid") of the manually segmented HJS-ROI (see Figure 3Go)
  2. tracing the exterior boundary of the HJS-ROI
  3. calculating the distances between the centroid and each point of the exterior boundary
  4. determination of the following distances (see Figure 3Go): D1, the minimum distance between the centroid and the upper boundary (roof of the acetabulum) of the HJS-ROI; D2, the minimum distance between the centroid and the lower boundary (upper margin of the femoral head) of the HJS-ROI; D3, the maximum distance between the centroid and the medial boundary of the HJS-ROI; and D4, the maximum distance between the centroid and the lateral boundary of the HJS-ROI
  5. the JSMI was calculated according to Equation (1):


    Formula 001



Figure 3
View larger version (23K):
[in this window]
[in a new window]

  		Figure 3. Determination of the D1, D2, D3 and D4 distances employed in the computerized calculation of the joint space morphological index. The centre of"mass" ("centroid") of the hip joint space region of interest (HJS-ROI), corresponding to Figure 2Go, is indicated with an asterisk.

 
In order to safeguard against variations related to each patient's individual anatomy, the JSMI was divided by the squared value of the length of Line-N (see Figure 1Go).

Simplified approach for calculating joint space morphological index
In an effort to provide a simplified approach for calculating the JSMI that could be easily performed in daily clinical routine by orthopaedists, without having to use a computer-based system, the following steps were followed (see Figure 4Go):


Figure 4
View larger version (19K):
[in this window]
[in a new window]

  		Figure 4. Simplified approach for computing the joint space morphological index. AB, line joining the extreme points of the lateral and the medial boundaries of the hip joint space region of interest(HJS-ROI). MN, vertical line, passing through the middle of line AB, connecting the upper and the lower boundary of the ROI.

 
  1. the length of the line AB connecting the extreme points A and B, corresponding to the lateral and medial boundaries, respectively, of the delineated HJS-ROI, was measured
  2. the line MN connecting the upper and the lower boundary of the HJS-ROI was drawn vertically from the middle of line AB, and its length was determined
  3. the JSMI values were calculated according to Equation (2):


    Formula 002


where (AB) and (MN) represent the lengths of the abovementioned lines. Finally, for normalization purposes, the JSMI-simplified values were divided by the squared value of the length of Line-N (see Figure 1Go).

Quantitative assessment of hip osteoarthritis severity
In an effort to provide a quantitative estimation of hip OA severity, a regression model was introduced. For each patient in the unilateral OA group:

  1. the number of pixels constituting the manually segmented HJS-ROI was determined, providing a computerized measurement for the HJS area
  2. a HJS narrowing index was derived according to Equation (3):


    Formula 003


    where HJSAnormal and HJSApath refer to the HJS area of the patient's contralateral normal and osteoarthritic hips, respectively [15]

  3. using non-linear regression analysis [24], the HJS narrowing and the JSMI values were employed in the design of a regression model, described by Equation (4):


    Formula 004


Hence, the HJS-narrowing index, described by Equation (3), was used as the dependent variable of the model, while the pathological JSMI, described by Equation (1), was utilized as the independent (predictor) variable.

Statistical analysis
The Student's paired t-test was used in order to investigate the existence of statistically significant differences between normal and osteoarthritic hips of the unilateral OA group for JSMI values. The correlation between the HJS-narrowing index and the pathological values of JSMI was evaluated by means of Pearson's correlation coefficient. Intraobserver as well as interobserver reproducibility concerning (i) the determination of the HJS-ROI, (ii) the measurement of the length of the normalization line (see Figure 1Go, Line-N) and (iii) the JSMI values were assessed by means of the coefficient of variation (CV) [24]. Accordingly, each one of the experienced orthopaedists evaluated separately all radiographs twice, with about a month's interval between evaluations. The scores obtained were utilized to calculate the CV, which provides an assessment of inter- and intraobserver reproducibility [24]; low coefficient values correspond to a high degree of reproducibility. The normality of distributions for the measured quantities was assessed by means of the Lilliefors test [25]. Non-linear regression analysis [24] was utilized for the determination of the model concerning the quantification of hip OA severity. In addition, Pearson's correlation coefficient was used to evaluate the correlation between the JSMI values obtained by the computerized and the simplified approach, whereas the Student's paired t-test was used in order to investigate whether these values differed significantly.

All the statistical processing was performed employing the "Matlab Statistics Toolbox", while the regression analysis task was carried out using the TableCurve software (AISN Software, Jandel Scientific, San Rafael, CA).


   Results and discussion
Top
 Abstract
 Introduction
 Methods and materials
 Results and discussion
Conclusions
 References
 
HJS narrowing has been considered as a reliable radiographic indicator of OA severity and it has been used to evaluate the progression of the disease [5]. The narrowing of radiographic HJS is associated with differentiations in both the shape and the size of the specific anatomical region due to the non-uniform loss of articular cartilage. In this study, these differentiations were quantified by the newly introduced JSMI. The rationale for considering JSMI as being capable of evaluating shape and size alterations of the HJS is based on its definition. In particular, and according to Equation (1), JSMI values for osteoarthritic hips are expected to be influenced by differentiations in the lengths of the distances D1, D2, D3, D4 due to OA. Utilizing digital image analysis terms, and considering the segmented HJS-ROI as an object within a digital image [26], the shape of the HJS-ROI (object) is expected to determine the position of its centre of "mass" ("centroid"). Thus, alterations in the shape of HJS, due to narrowing, are expected to differentiate the centroid position. Considering that the above-mentioned distances were defined taking the centroid of the segmented HJS-ROI as the reference point, differentiation in HJS shape is finally expected to influence the lengths of D1, D2, D3 and D4. On the other hand, alterations in the size of HJS are expected to influence, mostly, the distances D1 and D2. This can be justified considering that D1 and D2 express the minimum distances of the centroid from the superior and the inferior margins of the HJS, respectively. Thus, these distances, and subsequently their sum (D1+D2), are expected to be decreased because of the reduction of articular cartilage thickness.

Since the computation of JSMI involves the D1, D2, D3 and D4 distances (see Figure 3Go), the osteoarthritic condition is expected to differentiate JSMI values between normal and osteoarthritic hips. Statistical analysis revealed the existence of statistically significant differences (p<0.001) between normal and osteoarthritic hips for the JSMI values, indicating the capacity of JSMI to signify morphological alterations of the radiographic HJS due to OA. It has to be noted, however, that a medial pattern of femoral head migration (FHM) occurs in about 20% of patients [27], which might have an effect on the sensitivity of the proposed method. The results revealed that only one in five medial FHM cases of the present study was misclassified (95% exact CI: 1–72%), suggesting there may not be much interference in the findings of the proposed method.

The measured quantities were found to follow normal distributions. In particular, according to the results of the Lilliefors test for normality, the measured values were found to have come from distributions that were fitted adequately by (theoretical) normal distributions, which had the same mean and variance as the measured ones [25]. All measurements were found to be reproducible. Intraobserver reproducibility was, on average, high for the HJS area measurements (CV = 3.4%), the length of Line-N measurements (CV = 2.7%), and the JSMI values (CV = 3.6%). Similarly, interobserver reproducibility was also high, CV = 4.2%, CV = 3.1%, and CV = 4.1% for HJS area measurements, length of Line-N measurements and JSMI values, respectively. Regarding the patients of the unilateral OA group, the mean values (±standard deviation (SD)) of normalized JSMI values for osteoarthritic and contralateral normal hips were 0.023 (±0.011) and 0.077 (±0.021), respectively. The corresponding numbers for bilateral OA patients were 0.037 (±0.025).

According to Conrozier et al [10], in patients with unilateral OA, the contralateral unaffected hip can be used for obtaining normal reference values. Within this context, a reference threshold value for normal JSMI greater than 0.056 (mean±SD = 0.077±0.021) was calculated from the JSMI of the contralateral normal hips (see Table 2Go). A verification of this finding was obtained by subjecting to this threshold the JSMI of the normal and osteoarthritic hips. Referring to the unilateral OA group, 17 out of 18 (94.4% (95% CI: 72.7–100%)) of normal and 18 out of 18 (100% (95% CI: 84.7–100%)) of osteoarthritic hips were characterized correctly. As a further test of the classification rule, the latter was applied to the hips of the bilateral OA group, and again high accuracy results were obtained by classifying correctly 23 out of 28 (82.1% (95% CI: 63.1–93.9%)) hips. For all hips in the study, the overall classification accuracy regarding the characterization of a hip as normal or osteoarthritic was 90.6% (95% CI: 80.7–96.5%) (see Figure 5Go). The specificity accuracy was 94.4% (correct classification for 17 out of 18 of normal hips, 95% CI: 72.7–100%), while the sensitivity accuracy was 89.1% (95% CI: 76.4–96.4%), since 41 out of 46 osteoarthritic hips were characterized properly. The normalization of JSMI measurements rendered threshold values to be independent of the patient's individual anatomy variations. This normalization was considered to be necessary, since the individual's anatomy was found to influence the quantitative estimation of HJS parameters, such as the HJS width [28].


View this table:
[in this window]
[in a new window]

  		Table 2. Classification rules for assessing hip osteoarthritis severity, defined on the basis of JSMI

 

Figure 5
View larger version (20K):
[in this window]
[in a new window]

  		Figure 5. Joint space morphological index values for normal(•) and osteoarthritic hips (x). Horizontal solid line corresponds to the threshold value regarding the characterization of a hip as normal or osteoarthritic.

 
Rules for assessing hip OA severity were defined on the basis of mean and SD values of JSMI, obtained for hips assigned by the experienced orthopaedists to the KL scale severity categories. Referring to Table 2Go, a hip was characterized as mild/moderate OA if its JSMI was in the range 0.025–0.056, as of severe OA if its JSMI was lower than 0.025, and as of normal/doubtful OA if its JSMI was greater than 0.056.

For verification purposes, the classification rules were applied to the data concerning the unilateral OA group and the results were compared with the results concerning OA assessment using the KL grading scale (see Table 3Go). 17 out of 18 normal hips were correctly classified (classification accuracy 94.4% (95% CI: 72.7–100%)), 8 out of 9 hips were successfully classified in the mild/moderate category (classification accuracy 88.9% (95% CI: 51.8–99.7%)), and 8 out of 9 hips were correctly assigned to the severe category (classification accuracy 88.9% (95% CI: 51.8–99.7%)). 33 out of 36 hips were assigned to the correct severity category, resulting in an overall classification accuracy of 91.7% (95% CI: 77.5–98.2%). This relatively high classification score, although it was accomplished by verifying the JSMI-based rules on the data from which the index values have been derived, may be indicative of the reliability of the suggested approach. At this point it must be stressed that the exclusion of the distances D3 and D4 from the index definition resulted in lower classification accuracies for the discrimination between normal and osteoarthritic hips as well as among grades of OA severity (see Table 4Go). The specific finding indicates that not only the D1 and D2 but also the D3 and D4 distances contribute to the determination of hip OA.


View this table:
[in this window]
[in a new window]

  		Table 3. Comparison of results obtained by the KL scale and the proposed JSMI-based classification rules

 

View this table:
[in this window]
[in a new window]

  		Table 4. Comparison of overall classification accuracies accomplished by the JSMI-based classification rules, when D3 and D4 were (i) excluded from and (ii) included in the index definition

 
Besides discriminating among various OA-severity categories, the JSMI was used as the independent (predictor) variable of a regression model for quantitative estimation of the severity of the disease. The latter was expressed by means of an index (see Equation (3)) that had been introduced in a previous study performed by our group [15] and had been found to bear good correlation with the KL grading scale. The index takes values on a scale from 0 to 100 and quantifies OA severity by expressing the HJS narrowing as a percentage of HJS area difference between the osteoarthritic and the contralateral normal HJS. Mean values (±SD) of HJS area for osteoarthritic and contralateral normal hips were 33.7 (±20.3) mm2 and 105.0 (±23.8) mm2, respectively. These values differed significantly (p<0.001) from the HJS areas of the osteoarthritic hips, being smaller than the contralateral normal ones. Statistical analysis revealed that the correlation between the HJS-narrowing index and the pathological JSMI values was strong and significant (r = –0.83, p<0.001). The negative sign indicates that JSMI is expected to obtain lower values for higher values of the HJS-narrowing index. Thus advanced stages of the disease (corresponding to higher index values) are expected to be characterized by lower values of JSMI. The relationship between the HJS-narrowing index and the pathological JSMI values is shown in Figure 6Go, where it can be observed that plotted data followed a well-behaved regression curve.


Figure 6
View larger version (16K):
[in this window]
[in a new window]

  		Figure 6. Hip joint space-narrowing index versus normalized pathological values of joint space morphological index. Solid curve is the best fit to the data points ({blacksquare}). Horizontal solid lines indicate the regions of OA severity according to HJS-narrowing index values.

 
By combining the JSMI threshold values (Table 2Go) and the regression Equation (4), percentages corresponding to OA severity categories were calculated. Thus, index values for severe OA were >68.0%, for mild/moderate OA within the interval [5.7–68.0%] and for normal or doubtful OA less than 5.7%.

The relatively high classification scores achieved in this study suggest that the proposed computerized approach is reliable and reproducible, and in good agreement with the KL grading scale. However, in order to evaluate the sensitivity of the JSMI to small changes in OA progression, and thus to verify if the proposed index has any advantage over KL or joint space width scores, longitudinal studies are required.

In an attempt to expand and facilitate the applicability of the proposed method under circumstances of daily clinical practice routine, where a computerized environment may not be available, an alternative simplified approach was proposed. As JSMI values are derived according to Equation (2), the distances involved in the calculation of the JSMI-simplified values (see Figure 4Go) can be measured manually on a pelvic radiograph. Since the length of Line-N (see Figure 1Go) can also be measured manually, the generation of normalized JSMI values can now be performed without the use of a computer system. Statistical analysis revealed that the computerized and the corresponding JSMI values obtained by the simplified approach did not differ significantly (p>0.05). In addition, as presented in Figure 7Go, the two sets of JSMI values demonstrated strong and significant correlation (r = 0.96, p<0.001). Practically this means that the JSMI values, obtained according to the simplified approach, could be utilized for discrimination among OA severity categories and the quantitative assessment of the severity of disease.


Figure 7
View larger version (22K):
[in this window]
[in a new window]

  		Figure 7. Scatter diagram indicating the strong linear correlation between the joint space morphological index(JSMI) values obtained by the computerized and the simplified approach. Solid line represents the straight line fitting the data.

 
Classification rules based on the values of JSMI simplified are shown in Table 5Go. As can be observed, these thresholds were only slightly different from the corresponding ones defined on the basis of JSMI. When these revised thresholds were applied to the unilateral OA patients and the results were compared with assessment employing the KL grading scale, the overall classification accuracy achieved was 91.7% (95% CI: 77.5–98.2%), since 33 out of 36 hips were assigned to the correct categories (see Table 6Go). As can be observed from Table 6Go, the implementation of the JSMI-simplified rules resulted in classification scores identical to the corresponding ones obtained by the implementation of the JSMI-based rules (see Table 3Go). Finally, as in the case of JSMI for the D3 and D4 distances, the exclusion of (AB) from the JSMI-simplified index resulted in lower scores regarding the determination of hip OA (see Table 7Go).


View this table:
[in this window]
[in a new window]

  		Table 5. Classification rules for assessing hip osteoarthritis severity, defined on the basis of JSMI simplified

 

View this table:
[in this window]
[in a new window]

  		Table 6. Comparison of results obtained by the KL scale and the proposed JSMI simplified-based classification rules

 

View this table:
[in this window]
[in a new window]

  		Table 7. Comparison of overall classification accuracies accomplished by the JSMI simplified-based classification rules, when (AB) was (i) excluded from and (ii) included in the index definition

 

   Conclusions
Top
 Abstract
 Introduction
 Methods and materials
 Results and discussion
 Conclusions
References
 
A morphological index is proposed for assessing hip OA severity on pelvic radiographs employing digital image analysis algorithms. The index characterized with high accuracy radiographic hip joint spaces as normal, of mild–moderate or of severe OA, and additionally it evaluated OA severity quantitatively. Finally, an easy to use, simplified but effective version of the index was formed for use in everyday clinical practice.


   Acknowledgments
 
The first author was supported by a grant by the State Scholarship Foundation (SSF), Greece. The authors thank the staff of the Departments of Orthopaedics and Radiology for their contribution to this work.

Received for publication April 6, 2006. Revision received April 29, 2007. Accepted for publication May 15, 2007.


   References
Top
 Abstract
 Introduction
 Methods and materials
 Results and discussion
 Conclusions
 References
 

  1. Creamer P. Osteoarthritis pain and its treatment. Curr Opin Rheumatol 2000;12:450–5.[CrossRef][Medline]
  2. Raynauld J-P. Quantitative magnetic resonance imaging of articular cartilage in knee osteoarthritis. Curr Opin Rheumatol 2003;15:647–50.[CrossRef][Medline]
  3. Ory PA. Radiography in the assessment of musculoskeletal conditions. Best Pract Res Clin Rheumatol 2003;17:495–512.[CrossRef][Medline]
  4. Kellgren JH, Lawrence JS. Radiological assessment of osteoarthrosis. Ann Rheum Dis 1957;16:494–501.[Free Full Text]
  5. Altman RD, Fries JF, Bloch DA, Carstens J, Cooke DT, Genant H, et al. Radiographic assessment of progression in osteoarthritis. Arthritis Rheum 1987;30:1214–25.[Medline]
  6. Lane NE, Nevitt MC, Genant HK, Hochberg MC. Reliability of new indices of radiographic osteoarthritis of the hand and hip and lumbar disc degeneration. J Rheumatol 1993;20:1911–18.[Medline]
  7. Spector TD, Cooper C. Radiographic assessment of osteoarthritis in population studies: whither Kellgren and Lawrence? Osteoarthritis Cartilage 1993;1:203–6.[CrossRef][Medline]
  8. Croft P, Cooper C, Wickham C, Coggon D. Defining osteoarthritis of the hip for epidemiologic studies. Am J Epidemiol 1990;132:514–22.[Abstract/Free Full Text]
  9. Hochberg MC. Quantitative radiography in osteoarthritis: Analysis. Baillieres Best Pract Res Clin Rheumatol 1996;10:421–8.[CrossRef]
  10. Conrozier T, Tron AM, Balblanc JC, Mathieu P, Piperno M, Fitoussi G, et al. Measurement of the hip joint space using computerized image analysis. Rev Rhum Engl Ed 1993;60:105–11.
  11. Gordon CL, Wu C, Peterfy CG, Duryea J, Klifa C, Genant HK. Automated measurement of radiographic hip joint space width. Med Phys 2001;28:267–77.[CrossRef][Medline]
  12. Lequesne M. Chondrometry: quantitative evaluation of joint space width and rate of joint space loss in osteoarthritis of the hip. Rev Rhum Engl Ed 1995;62:155–8.[Medline]
  13. Ingvarsson T, Hägglund G, Lindberg H, Lohmander S. Assessment of primary osteoarthritis: comparison of radiographic methods using colon radiographs. Ann Rheum Dis 2000;59:650–3.[Abstract/Free Full Text]
  14. Hilliquin P, Pessis E, Coste J, Mauget D, Azria A, Chevrot A, et al. Quantitative assessment of joint space width with an electronic calliper. Osteoarthritis Cartilage 2002;10:542–6.[CrossRef][Medline]
  15. Boniatis I, Costaridou L, Cavouras D, Panagiotopoulos E, Panayiotakis G. Quantitative assessment of hip osteoarthritis based on image texture analysis. Br J Radiol 2006;79:232–8.[Abstract/Free Full Text]
  16. Boniatis I, Costaridou L, Cavouras D, Panagiotopoulos E, Panayiotakis G. A computer-based image analysis method for assessing the severity of hip joint osteoarthritis. Nucl Instrum Meth A 2006;569:610–3.
  17. Boniatis I, Costaridou L, Cavouras D, Kalatzis I, Panagiotopoulos E, Panayiotakis G. Assessing hip osteoarthritis severity utilizing a Probabilistic Neural Network based classification scheme. Med Eng Phys 2007;29:227–37.
  18. Altman R, Alarcón G, Appelrouth D, Bloch D, Borenstein D, Brandt K, et al. The American College of Rheumatology criteria for the classification and reporting of osteoarthritis of the hip. Arthritis Rheum 1991;34:505–14.[Medline]
  19. Lumiscan 75, system specifications. Lumisys Inc. 1998; http://www.lumisys.com/support/manuals.html
  20. Efstathopoulos EP, Costaridou L, Kocsis O, Panayiotakis G. A protocol-based evaluation of medical image digitizers. Br J Radiol 2001;74:841–6.[Abstract/Free Full Text]
  21. Sakellaropoulos P, Costaridou L, Panayiotakis G. An image visualisation tool in mammography. Med Inform Internet Med 1999;24:53–73.[CrossRef][Medline]
  22. Sakellaropoulos P, Costaridou L, Panayiotakis G. Using component technologies for web-based wavelet enhanced mammographic image visualization. Med Inform Internet Med 2000;25:171–81.[CrossRef][Medline]
  23. Pizer SM, Amburn EOP, Austin JD, Cromartie R, Geselowitz A, Greer T. Adaptive histogram equalization and its variations. CVGIP (Computer Vision, Graphics and Image Processing) 1987;39:355–68.[CrossRef]
  24. van Belle G, Fisher LD, Heagerty PJ, Lumley T. Biostatistics. A methodology for the health sciences. Hoboken, New Jersey: Wiley-Interscience; 2004.
  25. Lilliefors HW. On the Kolmogorov-Smirnov test for normality with mean and variance unknown. J Am Stat Assoc 1967;62:399–402.[CrossRef]
  26. Gonzalez RC, Woods RE, editors. Digital image processing. Upper Saddle River, New Jersey: Prentice-Hall; 2002.
  27. Resnick D. Patterns of migration of the femoral head in osteoarthritis of the hip. Roentgenographic-pathologic correlation and comparison with rheumatoid arthritis. AJR Am J Roentgenol 1975;124:62–74.[Abstract]
  28. Goker B, Sancak A, Arac M, Shott S, Block JA. The radiographic joint space width in clinically normal hips: effects of age, gender and physical parameters. Osteoarthritis Cartilage 2003;11:328–34.[CrossRef][Medline]




This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by BONIATIS, I
Right arrow Articles by PANAYIOTAKIS, G
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by BONIATIS, I
Right arrow Articles by PANAYIOTAKIS, G

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
BJR DMFR IMAGING  ALL BIR JOURNALS