Novel Assessment of Bone Using Time-Resolved Transcutaneous Raman Spectroscopy


  • The authors have no conflict of interest.


With fragility fractures increasing as the population ages, there is a need for improved means to estimate risk of fracture. We recorded Raman spectra of both the mineral and organic phases of bone transcutaneously, a technology with potential to enhance bone quality and fracture risk assessment.

Introduction: The current “gold standard” assessment of bone quality is BMD determined by DXA. However, this accounts for only 60–70% of bone strength. X-rays are absorbed by the mineral phase of bone, whereas the organic phase remains essentially invisible; however, bone strength is critically dependent on both phases. We report, for the first time, a Raman spectroscopic technique that analyses both phases of bone beneath unbroken skin by eliminating spectral components of overlying tissues.

Materials and Methods: We used an 800-nm laser (1-kHz, 1-ps pulses) with a synchronized 4-ps Kerr gate with variable picosecond delay that effectively shuttered out photons from overlying tissues. We measured bone Raman spectra at a point 2 mm above the carpus from two mouse genotypes with extreme differences in bone matrix quality: wildtype and oim/oim (matched for age, sex, and weight). Typical depth was 1.1 mm. We repeated the measurements with overlying tissues removed down to bone. Oim/oim mice produce only homotrimeric collagen, which results in poorly mineralized bone tissue.

Results: The main spectral features were present from both bone phases. The spectral bands were in similar ratios when measured through the skin or directly from bone (in both genotypes). The band of the mineral phase (phosphate ν1) was smaller in oim/oim mice when measured directly from bone and through skin. The band associated with a particular vibrational mode of organic phase collagen (CH2 wag) showed a frequency shift between the genotypes.

Conclusions: This novel technique allowed us, for the first time, to make objective transcutaneous spectral measurements of both the mineral and the organic phases of bones and distinguish between normal and unhealthy bone tissue. After further optimization, this technology may help improve fracture risk assessments and open opportunities for screening in anticipation of the predicted increase in fragility fractures.


THE MAJOR CLINICAL demand for bone quality assessment arises from the diagnosis and treatment of osteoporosis, the devastating disease that leads to 1.5 million fractures per year in the United States,(1) with high morbidity and reduction of quality of life. Forecast demographic changes in age distribution will lead to a more aged population and will result in a further increase in the incidence of fragility fractures, with the inevitable increase in pain and suffering. Fragility fractures of the hip, for instance, are expected to rise worldwide from 1.3 million in 1990 to 2.3 million in 2020.(2) Preventative strategies, as well as treatment of fractures once they have occurred,(2) will be needed to meet this escalating problem.

DXA, although the current gold standard screening modality, is not able to predict bone strength fully. DXA-measured BMD is only able to account for 60–70% of the variation in bone strength.(3) Recent evidence has shown unambiguously that fragility fractures of the wrist have a clear genetic etiology, yet there is a low genetic correlation between these fractures and BMD at the forearm.(4) It is clear that BMD is not the sole predictor of fracture risk(5,6); important factors are being missed. Factors such as body size, environmental factors, trabecular microarchitecture, genetics, and increased bone metabolism are known to have their effects.(7) Bone is a composite material, and X-ray techniques rely on absorption only from the mineral phase(8) (carbonated apatite); the organic phase (primarily collagen I) remains essentially invisible, yet resistance to fracture in bone is known to be critically dependent on both its phases,(9) with strong evidence that the collagen in bone undergoes changes in osteoporosis.(10–17) There is a clear need for an enhanced noninvasive technology that, in addition to measuring the properties of the bone mineral phase, safely measures bone collagen quality.

Both infrared and Raman spectroscopies have provided this information; they have been used to measure mineral/matrix ratio (related to bone material density), mineral crystallinity (related to mineral quality), and cross-linking within the organic phase (related to collagen quality).(18–20) However, until this study, it has not been possible to perform vibrational spectroscopy at useful depths. To date, these techniques have been applied, almost without exception, to excised tissues. There are three reasons for not being able to extract bone spectra through skin: (1) poor penetration of the probe laser; (2) high elastic scattering preventing direct imaging; and (3) high levels of collagen I in skin that effectively mask that of the underlying bone tissue, whose collagen is also primarily type I.

We overcame these shortcomings by using a pulsed laser at a wavelength with good penetration (800 nm) with an ultrafast shutter (Kerr-gate(21,22)) that selectively blocked the Raman photons from the superficial tissues. Although this technology has been used with success with finely divided powders,(23,24) this is the first time it has been used on biological tissues. The aim of this study, therefore, was to test the hypothesis that spectral features of both the mineral and organic phases of bone specimens with known differences in material properties can be measured objectively through the unbroken skin using time-resolved Raman spectroscopy.


Laser setup

The description of the system used for Kerr-gated picosecond Raman spectroscopy has been described previously.(21) In brief, a femtosecond mode-locked Ti:sapphire laser produced a 1-kHz pulse train, which was amplified by a Nd:YLF-pumped regenerative amplifier to produce 1-ps, 800-nm pulses at 2 mJ at 1 kHz. From this, a 500-μJ pulse was isolated by a beam splitter to drive a 4-ps Kerr shutter. The time interval between the two pulses was controlled with an optical delay line; our experiments used delays between 0 and 20 ps. The pulse energy at the specimen had an average power of 340 mW at the 1-mm spot size. We used a conventional single-stage Raman spectrograph (Spex; Horiba Jobin Yvon) and back-illuminated deep depletion CCD (DU420BR-DD; Andor Technology) with the probe laser at an angle of 180° to the collection path (backscattering geometry). The overview of this setup and the principle by which it is possible to measure Raman spectra through overlying structures is summarized in Fig. 1.

Figure FIG. 1..

Using a pulsed laser to probe the surface of a mouse forelimb and adjusting the picosecond delay of synchronized pulses to an ultra-fast Kerr gate, we were able to shutter out the Raman photons from the overlying tissues and record only the Raman spectrum from bone lying deep to the skin.

Of the various spectral bands recorded, we chose to study one from the inorganic phase of bone (phosphate ν1 band at ∼958 cm−1) and one from the organic phase (the CH2 wag of the collagen at ∼1451 cm−1 (CH2 wag refers to a vibrational mode in which two hydrogen atoms covalently bonded to the same carbon atom move symmetrically in a manner similar to the oars of a rowed boat).

Mouse samples

Although our long-term aim is to develop a system to help in diagnosing and monitoring conditions such as osteoporosis, animals with osteogenesis imperfecta (OI) were selected for this initial study for two reasons: (1) the difference between Raman spectra of normal and osteoporotic bone can be quite subtle, whereas the Raman spectra of OI bone are known to be considerably different from normal and (2) bone from mice with the oim mutation exhibit both mineral and collagen abnormalities and thus provide a model in which differences in both these phases can be assessed.

Two genotypes, B6C3Fe Col1a2oim/Col1a2oim mice or oim/oim (a strain of mice found to suffer from OI), and their wildtype controls were used. Mice homozygous for the oim mutation are deficient in proα2(I) collagen, and hence only homotrimeric [α1(I)3] collagen is deposited in their extracellular matrix.(25) This is known to affect not only the collagen but also the level of mineralization and results in a weaker bone.(26) We used forelimbs from three skeletally mature mice from each genotype. The mice were matched for age, sex, and weight. Each limb was shaved and mounted in a purpose-built jig with cranial surface facing the probe laser. The time taken to complete a spectral measurement was between 7 and 15 minutes. Heating of the specimens by the laser was prevented by slow water irrigation (this would not be necessary in the clinic once the technique has been optimized down to a power level that is safe for living skin). We recorded spectra at a point 2 mm above the carpus for all transcutaneous tests, the depth of the bone from the skin surface at this point typically being 1.1 mm. The three spectra from the mice of each genotype were summed for display in Fig. 2 and analyzed separately for the comparative studies.

Figure FIG. 2..

Raman spectra of bone measured from two genotypes: wildtype mice and oim/oim mice. The spectra were recorded both through the overlying skin and directly on bare bone (the two spectra are plotted on the same graph with an arbitrary vertical offset for comparison). (A) Raman spectra of wildtype mice. (B) Raman spectra of oim/oim mice. (C) The difference Raman spectra obtained by subtracting the spectra from the two genotypes. The two features highlighted in each graph indicate features from both the mineral phase (phosphate ν1 at about 958 cm−1) and the organic phase (CH2 wag of collagen at about 1451 cm−1).

As a control, after transcutaneous measurements, the overlying soft tissues were carefully removed, and the exposed bone was remeasured at the same site and at the midpoint of the bare humerus.

Statistical methods

The comparison of the mean area ratio of the phosphate ν1 band at ∼958 cm−1 to the band associated with the CH2 wag of the collagen at ∼1451 cm−1 was performed using a two-factor ANOVA test (genotype, with/without intervening soft tissues) to detect any significance in means in any of the groups. Student's t-test with pooled variance was used to define the p values between the means of two specific groups.


Typical Raman spectra from bone are well established,(27) with different specific regions being associated with the mineral and the organic phases. We have studied two features: (1) the phosphate ν1 band of the apatite of the mineral phase and (2) the band associated with the CH2 wag of collagen of the organic phase. We report two forms of comparisons with Fig. 2, the first between spectral features of the two genotypes of mice and the second between spectra from the same genotype with and without overlying soft tissues.

It can be seen from the spectra recorded from the exposed bone that there are considerable differences between the two genotypes (Figs. 2A and 2B). These are consistent with the findings reported using Fourier transform infrared (FTIR) spectroscopy of the oim/oim mouse.(25) The difference spectra in Fig. 2C clearly indicate the changes between the genotypes. The heights of some of the spectral bands have changed, and the wavenumber of the band of the CH2 wag has shifted.

By examination of the ratios of the band areas associated with phosphate ν1 and the CH2 wag between the two genotypes, a significant difference of the mean ratio between the genotypes can be seen both measured through the skin (p = 0.007) and directly from the bone (p = 0.005; Fig. 3). Furthermore, there is no significant difference in mean area ratio for the same genotype whether measured through skin or from bare bone (p > 0.05).

Figure FIG. 3..

Ratio of the area of the phosphate ν1 band of the mineral phase (958 cm−1) to the area of the band of the CH2 wag of collagen of the organic phase (1451 cm−1) for each genotype (mean and CV).


Recent work exploring changes of trabecular bone collagen using vibrational spectroscopy has clearly shown that there is a detectable difference between normal and osteoporotic bone associated with changes in collagen cross-links.(14,28–30) These studies showed that distinctions could be made by examination of bone harvested from the iliac crest, quite remote from the common sites of fragility fractures, indicating that the reported phenomenon was systemic. However, such techniques require the analysis of biopsy specimens, making it unsuitable for mass screening.

Our study clearly shows that major components of the Raman spectrum of bone can be detected without excision of bone or even cutting the skin. Major bands associated both with the mineral and with the organic phases(31) are clearly visible in the spectra measured through skin in Fig. 2. Interestingly, our technique through the skin was sufficiently sensitive to record a slight variation of the wavenumber of the peak associated with collagen between the genotypes (Fig. 2C). Positions of Raman bands are known to be sensitive to both the environment and chemical make-up of the matrix. It is these subtle features that make Raman such a useful analytical tool; the spectral differences observed here show this and its potential for bone analysis.

Although these features are apparent, there is noise in the upper spectra in Figs. 2A-2C. There are many possible ways to optimize the system to reduce this. The maximum tissue penetration in adipose tissue or tissues of varying water content is yet to be determined. We can, however, already reach depths at which a subcutaneous bone surface can normally be found in the human; further optimization will increase the number of potential sites that can be measured noninvasively (not to mention the sites available with minimally invasive approaches for which this technique can be developed).

Because the overlying tissues will elastically scatter the laser light, it is not possible, without further work, to be precise as to the area of illumination of the bone or to the depth of penetration of the bony tissue. However, it is reasonable to assume that the recorded spectra will have been “averaged” over a small volume of bone lying immediately below the periosteal surface at the site of illumination. Until we have an understanding of the depth of penetration of transcutaneous laser light into bony tissue, it is going to be difficult to predict the limitations of the technology in the assessment different types of bone. It may be that the thin cortical shells surrounding cancellous bone will illicit subtle differences in spectra from those of sites where the bone is more compact, because these thin shells may be penetrated fully by the laser. It is clear that further work is required to explore and interpret this exciting new technology to the full.

We conclude that the time-gated Raman spectroscopic technique has allowed for the first time vibrational spectral measurements of both the mineral and organic phases of bone tissue through the unbroken skin. The measurements also highlight that the technique is able to show a significant difference between normal and unhealthy bone, thus validating our approach. We are embarked on the optimization of the instrumentation to detect more subtle differences of specific spectral features. If this system development is successful, we anticipate that this technology will help improve fracture risk assessments, probably in conjunction with other measurements such as DXA.


The authors thank the Council for the Central Laboratory of the Research Councils (CCLRC), UK, for granting access to their Central Laser Facility to perform this work. We acknowledge support in part by NIH Grant R01 AR47969 and the University of Michigan Core Center for Musculo-Skeletal Research through NIH Grant P30 AR46024 (MDM) and NIH Grant R01 AR48337 (NPC). We also thank Andor Technology for the loan of IR deep depletion CCD camera.