Bone Trabecular Microstructural Properties Evaluation in Human Femur using Mechanical Testing, Digital X-ray and DXA

Sapthagirivasan V1, Anburajan M2

1Research Scholar, 2ProfessorDepartment of Biomedical Engineering,

SRM University, Chennai, Tamilnadu, India. sapthagiri.ece@gmail.com1, anbufelix@yahoo.com2

Venkatesh Mahadevan3,Information Systems & eBusiness,

Swinburne University of Technology,Melbourne, Australia.

vmahadevan@swinburne.edu.au3

VI-1 978-1-4244-8318-1/10$26.00 c2010 IEEE

Abstract— Early detection of fracture risk is important for initiating treatment and improving outcomes from both physiologic and pathologic causes of bone loss. While bone mineral density has traditionally been used for this purpose, alternative structural imaging parameters (quality measures) are proposed to better predict bone’s true mechanical properties. To further elucidate this, trabecular bone from human femur were used to evaluate the interrelationship of mechanical and structural parameters using mechanical testing, dual energy X-ray absorptiometry (DXA) scanning and digital X-ray imaging. Directional specific structural properties were assessed and correlated to mechanical testing, texture feature analysis and DXA. Our study was performed on 20 post menopausal women with or/and osteoporotic fractures and 21 healthy pre menopausal women. For all subjects radiographs were obtained of the femur with a new high-resolution X-ray device. The bone mineral density measurements were assessed by DXA. Furthermore, the results of this study show that a loss of bone primarily affects the connectedness and overall number of trabeculae. Thus, this method seems to be a promising routine technique for the determination of osteoporosis fracture risk from radiographs of the femur independently from BMD.

Keywords-BMD; DXA; osteoporosis; X-ray; trabecular

I. INTRODUCTION

With the better of life expectancy the risk of facing diseases causes by aging process is increasing. One of those diseases is the loss of bone mass or osteoporosis. Osteoporosis is a worldwide medical condition affecting middle-aged and older populations, especially women. In the INDIA, one in three women and one in eight men over the age of 50 will suffer a fracture attributed to osteoporosis [1]. Fractures come at a great personal and socioeconomic cost. Twenty percent of people who have a hip fracture die within 12 months, half of those who survive can no longer live independently, almost half of those who could walk unaided are no longer able to do so and a quarter are still unable to prepare their own dinner [2-4]. Earlier detection of osteoporosis can be done by using Bone Mineral Densitometry (BMD) technique using various modes such as ultrasound or Dual Energy X-ray Absorptiometry (DXA). For the time being DXA is considered as a gold standard for detection of Osteoporosis. Osteoporosis is defined by the World Health Organization as a BMD lower by more than 2.5 standard deviations than the mean value for young adults. As people get older, particularly

Postmenopausal women, they lose bone mass and become more susceptible to osteoporosis and fracture [5]. BMD alone cannot accurately predict the fracture, as other factors such as the shape and structure of bone and the risk of falling are also important. The architecture of the bone is composed of the cortical bone shell and trabecular bone core. Trabecular bone is a spongy, porous type found at the ends of all bones, such as pelvis and spine [6]. In proximal femur, trabecular bone forms a pattern of net-like strands varying in thickness and numbers [7] shown in Fig 1.1. It has a complex three dimensional structure consisting of struts and plates.

Fig 1.1 Femur trabeculaeMany lines of evidence indicate that the decreased bone

strength characteristic of osteoporosis is dependent not only on BMD, but also on trabecular bone microarchitecture and mineralization. The correlation between bone strength and bone mass is well established but the relationship between trabecular microarchitecture and biomechanical properties are less explored.

Trabecular patterns appearing on digital X-rays contain rich information about the bone status. The observation of trabecular pattern change for diagnosis of osteoporosis was first proposed in the 1960s using radiographs of proximal femur. The diagnosis was known as Singh Index grading system [3]. A number of physicians, due to the lack of diagnosis equipment like DXA, observe the trabecular change visualize in proximal femur recorded in radiographs to assess osteoporosis. On radiographs, femur (trabecular) bone structure appears as a distinct pattern. Fig 1.1 shows radiograph of proximal femur and its groups of trabeculae.

To solve the variability problem of Singh index grading system, we proposed the texture analysis system for osteoporosis assessment by observing trabecular change in proximal femur. Gabor filter, wavelet transforms and fractal analysis algorithms will be applied to extract the features of trabecular pattern recorded on proximal femur radiographs. Initial research in features extraction of proximal femur trabecular pattern using Gabor filter and discrete wavelet transform (DWT) has been performed with quite promising result [7-9]. The extracted features will represent the quality or structure of the bone, better quality represents better bone strength, lower quality leads to low bone strength and could be suspected as osteoporotic. The extracted features of the samples, in the form of energy, are then compared with their corresponding BMD obtained by DXA.

II. MATERIALS AND METHOD

A. Anatomy of the human hip

In human anatomy, the femur is the longest and largest bone shown in Fig 2.1. Along with the temporal bone of the skull, it is one of the two strongest bones in the body. The average adult male femur is 48 centimeters (18.9 in) in length and 2.34 cm (0.92 in) in diameter and can support up to 30 times the weight of an adult [11] and which forms part of the hip joint (at the acetabulum) and also part of the knee joint, which is located above. There are four eminences, or protuberances, in the human femur: the head, the greater trochanter, the lesser trochanter, and the lower extremity. They appear at various times from just before birth to about age 14. Initially, they are joined to the main body of the femur with cartilage, which gradually becomes ossified until the protuberances become an integral part of the femur bone, usually in early adulthood. The shaft of femur is cylindrical with a rough line on its posterior surface (linea aspera).

Fig 2.1 Anatomy of the human femur

B. Subjects and Sources

Right femur X-ray images of 41 Indian women were obtained with slight variations of X-ray exposure condition. The protocol was a uni-center case-control study of Indian postmenopausal women recruited from the private scan center, chennai, India. It was performed on 20 healthy pre menopausal women (n=20, 34.67 ± 6.8 years) with no fracture and 21 post menopausal women without fractures (n=21, 55.77 ± 14.9 years). Bone mineral density (BMD) (g/cm2) from the femoral neck (FN-BMD), femoral shaft (FS-BMD), femoral wards (FW-BMD), femoral trochanter (FTr-BMD) and total Femur (FT-BMD) were measured by dual-energy X-ray absorptiometry (GE Healthcare Lunar DPX). For all subjects the radiograph of the right femur was obtained with a new high-resolution X-ray device with direct digitization (BMATM, D3A®, Medical System, Orleans, France).

C. Bone trabecular feature extraction

Typically, there is wide variation in the intensity of digital x-ray image from different patients. This variation is strongly correlated to the person’s skin pigmentation and bone. Hence, it is necessary to identify a reference frame and normalize the intensities of all other images against it. We performed this intensity normalization step using histogram specification [10]. This modifies the image values through a histogram transformation operator which maps a given intensity distribution a(x,y) into a desired distribution c(x,y) using a histogram equalized image b(x,y) as an intermediate stage. This process is applied independently to each individual sub blocks of 3 x 3.

After normalization, we were applied a local contrast enhancement method to improve both the contrasting attribute of bone and the overall intensity in the image. This operation was performed on the intensity channel of the image. The aim is to apply a transformation of the values inside small windows in the image in a way that all values are distributed around the mean and shows all possible intensities. Hence, given each pixel p in the initial image and a small running window W then the image is filtered as follows to produce the new image I

(1)

From the equation (1)(2)

where n is number of bits per pixel in the given image and

w is sigmoid function.

The max and min refer to the maximum and minimum intensity values in the whole image, while mean and standard deviation within each window produces significant contrast enhancement when standard deviation is small, i.e. the contrast is low, and little enhancement when standard deviation is large, i.e. contrast is high.

Once the image has been pre-processed as described above, we perform Gabor and wavelet (4 level decomposition) operation to the classification of osteoporosis based on the change of trabecular pattern. The classification was based on features extracted by Gabor and wavelet in the form of energy based on the following equation (3).

(3)

for image I(m,n) with 1 ≤ m ≤ M and 1 ≤ n ≤ N.Gabor and wavelet features and mechanical properties

were calculated on certain region of interest (ROI) of proximal femur known as Ward’s triangle, femoral neck, femoral head, shaft and greater trochanter. The Ward’s triangle is the region that is most sensitive to bone mass lost. Fig 2.2 shows the ROI used in this paper.

Fig 2.2 Various ROI in right femur digital x-ray image

III. RESULT AND DISCUSSION

In this paper we used four different trabecular pattern recorded in 41 patient’s radiograph were extracted. The feature extracted from wavelet features by energy computation and then compared to trabecular energy computation predetermined trabecular energy. The capability of wavelet features to assess the degree or level osteoporosis. The result of features extraction and energy computation by applying DWT in four scales or levels decomposition using Gabor wavelet for one radiograph sample is shown in Table 1 and from Table 2, 3 and4, it is clear that significant energies were obtained from level 1st and 4th level decomposition for approximation coefficient. Therefore for the rest of the radiographs samples, features extraction will be computed for approximation coefficient at 1st and 4th level decomposition. The results of energy computation for 41 radiographic samples are plotted and shown in Fig 3.2. The energy computed from trabecular pattern of normal bone samples appear to be higher than the energy from samples of the osteopenia and osteoporosis. The healthiest bones, which are having the highest energy as shown in Fig 3.2 and the osteoporotic bones, which are having the lowest energy.

Fig 3.1 (a) Age Vs BMD, (b) Age Vs BMI, (c) BMI Vs Age , (d) BMI VsBMD

(a) (b)

(c)Fig 3.2 plotting of energy Vs Approximation Co.eff. (a) Normal, (b) Osteopenia, (c) Osteoporosis.

Fig 3.3 Implementation results (a) original femur x-ray image, (b) Marked Region of Interests,(c) Trabecular enhanced ROIs, (d) Original ROI of femur head, (e) Enhanced Trabecular ROI pattern

Table 1 Approximation Coefficient values up to 4 levels

Region A1 A2 A3 A4Gr.Troc 0.496 0.621 0.589 0.557Lsr.Troc 0.643 0.647 0.554 0.549Neck 0.657 0.665 0.584 0.583

Head 0.651 0.670 0.595 0.534

Table 2 DXA BMD of Right femur

Group BMI

Proximal Femur BMD by DXA (g/cm2)

T-scoreTroc. Neck Ward Shaft Total

Normal27.5 ± 4.5 0.76 ± 0.09 0.91 ± 0.11 0.73 ± 0.1 1.19 ± 0.08 0.98 ± 0.08 -0.21 ± 0.63

n=8, 55 ± 3.1

Osteopenia21.41 ± 2.65 0.61 ± 0.07 0.76 ± 0.03 0.61 ± 0.08 0.8 ± 0.36 0.81 ± 0.05 -1.58 ± 0.38

n=8, 46 ± 15.7

Osteoporosis21.92± 2.62 0.39 ± 0.11 0.59 ± 0.08 0.37 ± 0.09 0.64 ± 0.17 0.53 ± 0.12 -3.8 ± 0.94

n=5, 72.2 ± 11.9

Young Adult24.16± 3.7 0.83 ± 0.12 1.02 ± 0.17 0.87 ± 0.21 1.23 ± 0.18 1.05 ± 0.16 0.34 ± 1.25

n=20, 34.67 ± 6.8

Table 3 Energy calculated by Gabor Wavelet for original images

Group BMI

Proximal Femur Energy by Wavelet for original images

Troc.Gr Troc.Lsr Neck Head

A1 A4 A1 A4 A1 A4 A1 A4

Normal

27.5 ± 4.50.5 ± 0.06 0.56 ± 0.07

0.64 ± 0.01 0.55 ± 0.12

0.66 ± 0.28 0.58 ± 0.17

0.65 ± 0.18 0.53 ± 0.21n=8, 55 ± 3.07

Osteopenia21.41 ±2.65

0.54 ± 0.17 0.58 ± 0.03

0.51 ± 0.02 0.58 ± 0.17

0.6 ± 0.02 0.56 ± 0.12

0.68 ± 0.16 0.56 ± 0.15n=8, 46 ± 15.7

Osteoporosis

21.92±2.620.66 ± 0.11 0.57 ± 0.01

0.61 ± 0.16 0.54 ± 0.21

0.64 ± 0.09 0.58 ± 0.11

0.68 ± 0.17 0.54 ± 0.13

n=5, 72.2 ± 11.95

Table 4 Energy calculated by Gabor Wavelet for trabecular enhanced images

Group BMI

Proximal Femur Energy by Wavelet for trabecular enhanced images

Troc.Gr Troc.Lsr Neck Head

A1 A4 A1 A4 A1 A4 A1 A4

Normal

27.5 ± 4.50.36 ± 0.06 0.89 ± 0.17

0.41 ± 0.01 0.82 ± 0.2

0.35 ± 0.22 0.81 ± 0.17

0.36 ± 0.21 0.69 ± 0.21n=8, 55 ± 3.07

Osteopenia21.41 ±2.65

0.34 ± 0.14 0.83 ± 0.13

0.42 ± 0.06 0.57 ± 0.17

0.35 ± 0.05 0.6 ± 0.21

0.28 ± 0.16 0.91 ± 0.15n=8, 46 ± 15.7

Osteoporosis

21.92±2.620.33 ± 0.18

0.0.88 ± 0.09

0.41 ± 0.19 0.87 ± 0.21

0.38 ± 0.19 0.87 ± 0.11

0.37 ± 0.17 0.83 ± 0.14

n=5, 72.2 ± 11.95

IV. CONCLUSION

Gabor filter and discrete wavelet transform has successfully applied to texture analysis of the trabecular pattern recorded in the radiograph of proximal femur. The extracted features from trabecular pattern in the form of energy were able to give information about the quality of the bones for the assessment of osteoporosis. The fractal dimension using box counting algorithm also has a significant correlation to bone quality assessment using trabecular energy with BMD and these computerized algorithms it is possible to use them for screening purpose as radiographic facilities are available all over the country. The early detection of osteoporosis by the change in trabecular bone assessed using these algorithms will become a significant contribution to improve the quality of healthcare.

ACKNOWLEDGMENTS

The authors wish to thank Department of Orthopedic and Traumatology, Faculty of Medicine, SRM University, for suggesting radiographic guidance and also thankful to Mr. Mariadas and Mr.Dulip Arthi Scans Chennai, India for their support for providing radiographic images.

REFERENCES[1] Osteoporosis Society of India. Action Plan Osteoporosis:

Consensus statement of an expert group. New Delhi, 2003.[2] Tony M K, Elise F M, Glen L N and Oscar C Y.

“Biomechanics of Trabecular Bone”. Annual Review of Biomedical Engineering, vol 3, pp. 307-333, 2001.

[3] Smyth P P, Adams J E, Whitehouse R W and Taylor C J. “Application of Computer Texture Analysis to the Singh Index”. The British Journal of Radiology, vol 70, pp. 242-247, 1997.

[4] Sangeetha S, Christopher J J, and Ramakrishnan S. “Wavelet Based Qualitative Assessment of Femur Bone Strength using Radiographic Imaging,” International Journal of Biological and Life Sciences, vol 3, issue 1, pp. 276-280, 2007.

[5] Anburajan M, Rethinasabapathi C, Paul Korath M, Ponnappa B G, Govindan A, Jagadeesan K. “Age Related Proximal Femur Bone Mineral Loss in South Indian Women: A Dual Energy X-Ray Absorptiometry (DXA) Study,” Journal of Association of Physicians of India, vol 49, pp. 442-445, 2001.

[6] Sooyeul L, Seunghwan K, Pyo H B, Park S H. “Bone Mineral Density Estimation from X-ray images,” Proc. of the First Joint BMESEMBS Conference Serving Humanity, Advancing Technology, vol 2, pp. 1040, 1999.

[7] Pramudito J T, Soegijoko S, Mengko T R, Muchtadi F I and Wachjudi R G. “Trabecular Pattern Analysis of Proximal Femur Radiographs for Osteoporosis Detection,” Journal of Biomedical and Pharmaceutical Engineering, vol 1, pp. 45-51, 2007.

[8] Mengko T R and Prarnudito J T. “Implementation of Gabor Filter to Texture Analysis of Radiographs in the Assessment of Osteoporosis,” MVA2002 IAPR Workshop on Machine Vision Applications, vol 2, pp. 251-254, 2002.

[9] Mengko T R and Pramudito J T. “Texture Analysis of Proximal Femur Radiographs for Osteoporosis Assessment”, WSEAS Transaction on Computers, vol 3, pp. 92-97, 2004..

[10] Gonzalez R C and Woods R E. Digital Image Processing. 3rd

Edition. Prentice Hall, Upper Saddle River, 2008. [11] Jour A and Gray. (2010), “The Femur (Thigh Bone)”, Yahoo

Educations [Online]. Available From: http://education.yahoo.com/reference/gray/subjects/subject/59[Accessed 24/08/2010].