Texture Analysis for Postmenopausal Osteoporosis

The study proposed in this application is a part of a larger project entitled "Clinical
utility of radiographic texture analysis in diagnosing and treating osteoporosis". The
overall goal of the larger project is to determine whether computerized texture analysis of
digitized high-resolution images of trabecular bone (texture analysis) improves our ability
to diagnose bone fragility and follow natural history and/or response to pharmacological
therapy of osteoporosis. In the study proposed here we plan to examine changes in the results
of texture analysis during two years of pharmacological therapy for osteoporosis.

Role of densitometry in osteoporosis:

Measurement of bone mineral density is the principal diagnostic method used in clinical
practice and in research studies, both to identify patients who have the disease and to
follow their response to therapeutic agents. The technique used most widely is dual-energy
X-ray absorptiometry (DXA), which has advantages of low cost and radiation exposure, and high
precision and accuracy of 1-2% and 4-8%, respectively [Garner, 1996 and Melton, 1990]. Based
on the association between the low BMD and increased risk of fracture, BMD-based treatment
guidelines have been developed [Melton, 1993 and National Osteoporosis Foundation, 1999].
There is, however, a considerable overlap between BMD of patients who sustain fragility
fractures and those who do not [Cummings, 1993; Marshall, 1996; Melton, 1989; Ross, 1990 and
Wasnich, 1990]. The problem arises because the fragility is determined not only by the
quantity of the bone (measured as bone density), but also by its "quality" which is believed
to be related to the preservation of the normal trabecular pattern [Parfitt, 1987]. Bone
quality is not specifically assessed using current diagnostic methods. Information about bone
quality, however, would be of substantial clinical and scientific value, as it would
complement the BMD measurement when selecting patients for therapy and when studying bone
loss or assessing effects of therapeutic agents.

Texture analysis:

A novel approach to noninvasive and practical assessment of bone structure is to analyze the
texture of high resolution radiographs of trabecular bone [Link, 1999]. Dr. Giger has
developed a method for characterizing bone structure by computerized texture analysis of
digitized high-resolution radiographs [Jiang, 1999; Caligiuri, 1993; Caligiuri, 1994;
Chinander, 1999 and Chinander, 2000]. In this approach, the texture is analyzed in several
ways, including Fourier based analysis, which yields root mean square (RMS) as a measure of
magnitude of trabecular bone texture pattern, and the first moment of power spectrum (FMP)
which characterizes the texture pattern's frequency; and Minkowski dimension fractal analysis
[Caligiuri, 1993; Chinander, 1999; Chinander, 2000; Benhamou, 1994; Jiang, 1999; Majumdar,
1993 and Maragos, 1994]. Radiographic texture analysis has been studied in vivo, on lumbar
spine radiographs and found to predict presence of vertebral fractures elsewhere in the spine
more reliably than did the BMD of the spine [Caligiuri, 1993 and Caligiuri, 1994;]. In
addition, in an in vitro study texture features as well as BMD were analyzed in femoral neck
specimens obtained during surgical hip replacement. Mechanical loading (crush test) was then
performed on cubes of trabecular bone machined from these specimens to determine their bone
strength. It was found that the combination of BMD and texture analysis predicted bone
strength better than BMD alone [Jiang, 1999; Chinander, 1999 and Chinander, 2000].

Biochemical markers of bone turnover:

In studies of osteoporosis, the bone mass is assessed by measuring BMD while the metabolic
activity of the bone is assessed by measuring the biochemical markers of bone turnover
[Looker, 2000]. These markers have limited utility in individual patients because they have
high within-person variability (low precision), and because it is not clear which markers are
useful in which clinical situation [Looker, 2000 and Bauer, 1999]. In contrast, comparing
biochemical markers between groups of patients in clinical studies has been found to be
useful in two settings. Firstly, it has been found that high levels of biochemical markers of
bone resorption predict fractures independent of BMD [Garnero, 1996 and van Daele, 1996].
Secondly, early changes in bone markers (at 3-6 months) during anti-resorptive therapy
predict later changes in BMD and fracture rates [Ravin, 1999; Greenspan, 1998; Chesnut, 1997
and Bjarnason, 1997]. The mechanisms underlying these observations have not been elucidated
to date. It is speculated that increased bone resorption, which is reflected in elevation of
biochemical markers of bone turnover, increases fragility by weakening trabecular structure
prior to or independent of measurable BMD changes. Similarly, decreased bone resorption
during pharmacological therapy is likely to improve the trabecular structure before or
independent of its effects on BMD. Since the aim of our research is to (indirectly) examine
the trabecular structure by performing the radiographic texture analysis, we plan to
determine whether the changes in biochemical markers of bone turnover during antiresorptive
therapy will correlate with changes in the results of texture analysis.

Rationale for the study:

Anti-resorptive therapy reduces bone fragility and increases bone density. It is likely that
the trabecular structure of the bone also changes during treatment. Peripheral densitometry
has not been used so far to monitor response to therapy. If the combination of texture
analysis and peripheral BMD change reproducibly during treatment it may be possible to employ
this combination to monitor therapeutic response. In so doing, one could avoid the need to
use the central densitometry and biochemical markers of bone turnover since the former is
cumbersome while the latter suffers from low precision.

Potential advantages of using a portable peripheral densitometer: The texture analyses
described above were developed for high-resolution radiographs, which were digitized and
subjected to computer analysis. The new DXA imaging systems such as GE/Lunar PIXI which will
be used in our research, provide digital images with resolution sufficient for computerized
texture analysis (200 micron pixels). Furthermore, PIXI can generate the image in a shorter
time (seconds vs. minutes) and at a fraction of radiation dose of conventional radiographs.
Finally, since this is a portable densitometer, the methodology developed in this proposal
has the potential to be widely applicable to large segments of the population, including
frail elderly who have limited mobility and high prevalence of osteoporosis.

STUDY PROCEDURES

The studies will be performed in the outpatient facility of the University of Chicago. Every
3 months for the first 6 months and every 6 months for the remainder of 2 years, the subjects
will come in the morning in the fasting state, provide a urine sample (second morning void)
and blood sample for measurement of biochemical markers of bone turnover. Height and weight
will be recorded at each visit, and any change in health status, including fractures
ascertained. We will also assess other factors known to influence bone turnover, such as diet
and physical activity. Every 12 months, the subjects will fill out Block food frequency
questionnaire from Berkley Nutrition Services. In addition, every 6 months they will fill out
a calcium intake questionnaire, which will be analyzed by the nutritionist and a short
physical activity questionnaire, which was used in PEPI trial for assessment of physical
activity. Medication compliance will be assessed by questioning the patients and counting the
number of calcium and alendronate tablets remaining from the previous visit.

After these tests are completed, the subjects will go to the densitometry suite of the
Endocrinology clinic where BMD will be measured and heel images obtained for texture
analysis. The left heel will be scanned twice using the PIXI densitometer (GE/Lunar
corporation) for measurement of BMD of the heel and texture analysis. (If there is a
deformity of the left heel, right heel will be used for all examinations.) In addition, every
6 months, BMD of the lumbar spine and proximal femur will be measured using the central
densitometer Prodigy (GE/Lunar corporation). The same instrument will be used for lateral
vertebral assessment (a method used for detecting vertebral deformities on images of the
lateral spine from the densitometer), which will be performed every 12 months.

Inclusion Criteria:

- The study will enroll 40 postmenopausal women with a T score < -2 either at the lumbar
spine or the femoral neck: 20 who decide to begin anti-resorptive therapy (treated
group), and 20 women who decline such therapy (control group). We will attempt to
match the patients and the controls for T score (within 0.3) and age (within 5 years).

- All study participants will be:

- at least 3 years past the last menstrual period,

- not on HRT, Raloxifene or calcitonin for at least 6 months.

Exclusion Criteria:

- All study participants will not be on bisphosphonates during the previous 12 months.

- Women with secondary causes of osteoporosis will be excluded.