1.

Introduction

Near-infrared (NIR) spectroscopy (650 to 950 nm) has become an important modality for imaging and medical diagnostics, with applications that include cancer detection, assessing changes in tumors, measurement of blood flow to the brain, and evaluation of patients with Paget’s disease of bone.^1–7 Short wavelength infrared (SWIR) spectroscopy, with wavelengths from 1000 to 2500 nm, may offer better penetration into turbid media, while minimizing radiation exposure. Biomolecules in tissue, such as oxyhemoglobin, hemoglobin, and water, have distinct absorption spectral patterns, thus they give optical fingerprints. The SWIR optical windows, which correspond to second, third, and fourth windows in the 1000- to 2500-nm wavelength range, occur between water absorption peak maxima and can reveal the presence of lipids, fats, glucose, and collagen, as well as of many other molecules, due to their distinct optical absorption spectral profiles.

Water absorption, as well as light scattering, limits the information collected by a detector. Rayleigh scattering is known to decrease with longer wavelength, varying as $1/{\lambda }^4$. Mie scattering will also decrease [varying as $1/{\lambda }^n$ for $n\ge 1$, where $n$ varies with particle size for ${\mu }_{\mathrm{s}}$, not for reduced scattering coefficient ${\mu }_{\mathrm{s}}^{\prime }={\mu }_{\mathrm{s}}(1-g)$]. Details on Mie scattering are available elsewhere.⁸ Because of this, the utilization of SWIR wavelengths could present the opportunity to study chromophores beyond the conventional, first, NIR window.^9–11 Utilizing SWIR wavelengths can result in decreased scattering, might allow for less blurring, and result in better penetration into the tissue media; therefore, better contrast images might be obtained. Recently, NIR windows II (located between water peak maxima at 970 and 1444 nm) and III (located between water peak maxima of 1444 and 1950 nm) have been used for medical applications, such as in vivo imaging of small animals and the detection and monitoring of intestinal ischemia in vivo in a porcine model.^9–13 Window II occurs between 1100 and 1350 nm and window III between 1600 and 1870 nm. Window III has been referred to as the “golden window.”¹²

A fourth NIR window, occurring between 2100 and 2350 nm (shown in Fig. 1), has been mostly ignored because of greater absorption by water at these wavelengths. However, Yoo et al. have shown that a small amount of absorption can limit the detection of diffusive photons, which may cause a blurring of images, and can highlight the ballistic and snake photons, which are responsible for producing clearer images.¹⁴ Indeed, water absorption is less at shorter NIR wavelengths, but, due to the inverse wavelength dependence, scattering is minimal; this may result in a balance between absorption and scattering events. The fourth NIR window has also been discounted due to a lack of sensitive detectors. The advent of new photodetectors and laser sources now makes window IV a viable spectral domain (Table 1). The photodetector of choice in window IV is InSb (either as single and array photodiodes). Ultrafast lasers that utilize wavelengths from window IV include the supercontinuum laser, with spectral range from 400 to 2500 nm, the Cr 2+:ZnSe/S lasers, and GaSb laser diodes.^15,16 Femtosecond Cr2+-doped ZnS and ZnSe lasers can be used at wavelengths from 1800 to 3400 nm, with Kerr mode locking that can generate ultrashort (41 fs) pulses. Room-temperature Cr2+-based lasers can provide tens of Watts of power and ultrabroad tunability ($\sim 1400\mathrm{nm}$ between 2 and $3.4\mu \mathrm{m}$).¹⁶ GaSb laser diodes are available from 1800 to 3000 nm. Brolis semiconductors now offer 2100 and 2200 nm GaSb diode lasers for SWIR microscopes. In window III, the 200-fs laser from Calmar [called Bodega femtosecond optical parametric laser (FLOPA)] is available at $\sim 1700\mathrm{nm}$. Recently, Sordillo et al. reported on the use of the fourth NIR window for the assessment of bone, at the 2016 SPIE Photonics West Conference.¹⁷ As an extension of our work, we have measured the absorbance and have calculated the total attenuation lengths from lipid or collagen rich tissues, including breast and prostate cancer, bone, lipids, and intralipid solutions, at the first NIR and SWIR windows. The molecules that absorb by excitation of combination vibrations in the SWIR are presented.

Fig. 1

(a) Optical density (O.D.) spectrum of water molecules in the NIR spectral range; (b) OD spectra of collagen and water at the first NIR and the three SWIR windows.¹¹

Table 1

Ultrafast and continuous wave (CW) laser light sources for window III and IV.

2.

Methods

The absorbance spectra from the samples were acquired using the UV–vis–NIR Cary 500 spectrophotometer (Varian, Inc., Palo Alto, California). The Cary 500 is a commercially available, research grade, spectrophotometer with a PbS detector, a deuterium arc UV lamp, and a tungsten halogen visible light. To produce a narrow band, nearly monochromatic, light beam, light from these sources is diffracted using a grating. This light is then allowed to hit the sample, which is placed at a fixed position in the sample compartment. Finally, optical density (OD) or percent transmission of light through the sample is recorded on the computer. These conditions were kept constant for all samples. The wavelength range was between 600 and 2500 nm, covering the first NIR and the SWIR ranges. Total attenuation lengths ($l_{\mathrm{t}}$) were measured in windows I ($\lambda =750\mathrm{nm}$), II ($\lambda =1200\mathrm{nm}$), III ($\lambda =1700\mathrm{nm}$), and IV ($\lambda =2200\mathrm{nm}$). A sample of cleaned human fibula was obtained from a cadaver and cut to $\sim 150\mu \mathrm{m}$. 20% intralipid was diluted to concentrations of 0.5%, 1%, 1.5%, and 2%. Tissue samples were purchased from National Disease Research Interchange and Cooperative Human Tissue Network after approval by an institutional review board. The samples were neither fixed nor treated chemically. Samples were cut to thicknesses of $\sim 50$, 100, and $200\mu \mathrm{m}$.

3.

Theoretical Background

Light through turbid media is described by the trajectories of three components: diffusive, ballistic, and quasiballistic (snake) photons.^8,18 When the media is thin, the decay of ballistic photons ($I_{\mathrm{b}}$) is described by the differential equation:¹⁹

For thicker media, the diffusive photons ($I_{\mathrm{d}}$) are governed by the diffusion equation, which can be approximated by¹⁹

When the solid angle $\delta \mathrm{\Omega }$ is small, the total transmitted light intensity $I$ can be written as the sum of the number of ballistic photons ($I_{\mathrm{b}}$) and of a fraction of the diffusive photons ($I_{\mathrm{d}}$)²²

Yoo et al. have studied photon transport through random media utilizing ultrafast laser pulses and time-resolved detection.^14,24 The authors showed that for $z/l_{\mathrm{tr}}<10$, the diffusion approximation deviated monotonically from the measured scattered pulse, and that scattered photons arrived earlier than the diffusion approximation had predicted. Zhang et al. found that, at $z/l_{\mathrm{tr}}\sim 3$, there is a sharp transition from ballistic to diffusive behavior.²⁵ The thickness $z_{\mathrm{c}}$ where the ballistic and diffusive photons are equal is given as¹⁹

4.

Results

The measured percent transmission intensities and corresponding OD through a sample of human fibula using the four windows are plotted in Figs. 2(a) and 2(b). From these plots, the total attenuation coefficient (${\mu }_{\mathrm{t}}$) obtained from the bone sample is calculated using Beer–Lambert’s law

Fig. 2

(a) Percent transmission (T) and (b) OD spectra from a bone sample.

Figures 3(a) and 3(b) show the spectrum of the total attenuation coefficients (${\mu }_{\mathrm{t}}$) and the spectrum of total attenuation lengths ($l_{\mathrm{t}}$) from bone. Longest relative total attenuation lengths occur at wavelengths within the third and fourth windows. A salient feature is an oscillating structure about window IV. These spectra reveal small oscillations at wavelengths above 1800 nm, which are most likely due to the vibrational combination modes from collagen.

Fig. 3

(a) Total attenuation coefficients (${\mu }_{\text{total}}$) and (b) total attenuation lengths ($l_{\text{total}}$).

Figure 4 shows the spectra obtained from different concentrations of intralipid solution. For 0.5% and 1% intralipid solution, longest total attenuation lengths occur at window II. For 1.5% and 2% intralipid solution, longest total attenuation lengths are seen at window III. For 20% intralipid, longest total attenuation lengths occur at the third and fourth windows. Similarly, longest total attenuation lengths for both prostate and breast tissues are seen at window III. The oscillation in absorption plots beyond 1800 nm is attributed to collagen and lipids [see Fig. 1(b)]. Table 2 shows $l_{\mathrm{t}}$ for tissues with high collagen. Table 3 shows total attenuation lengths for chicken fat and for intralipid solution of different concentrations.

Fig. 4

$l_{\mathrm{t}}$ for 0.5%, 1%, 1.5%, 2%, and 20% concentrations of intralipid solution.

Table 2

Total attenuation lengths from high collagen tissues at the first NIR and the SWIR optical windows.

Table 3

Total attenuation lengths of chicken fat and of solutions of intralipid at the first NIR and the SWIR optical windows.

5.

Discussion

Limited studies have been done utilizing wavelengths in the SWIR range, especially wavelengths above 2000 nm. Using samples of human skin, Troy et al. have reported the absorption and reduced scattering coefficients obtained at wavelengths between 1000 and 2200 nm. However, at wavelengths from 1900 to 2050 nm, the coefficients were not calculated because of increased water absorption (peak at $\lambda =1920\mathrm{nm}$).²⁶ These authors also determined that, for wavelengths less than 1800 nm, a 2% intralipid solution was an appropriate skin tissue phantom. For wavelengths greater than 2100 nm, a 2% to 4% intralipid solution would be an appropriate phantom. Zamora-Rojas et al. investigated porcine skin in the SWIR range from 1150 to 2250 nm.²⁷ Due to a large water band, the authors could not obtain absorption coefficients in the spectral range from 1880 to 2040 nm but otherwise showed that scattering decreased with increasing SWIR wavelengths. Wang et al. looked at samples of human carotid atherosclerotic plaque at several SWIR regions (1130 to 1260 nm, 1620 to 1820 nm, and 2200 to 2330 nm). Ratios of lipid to protein were used to distinguish stable, vulnerable, and intermediate plaque.²⁸

Bone has not been studied previously at wavelengths within window IV. The reduced scattering and absorption coefficients of temporal bones were studied at wavelengths of 635, 690, 780, and 830 nm, but the authors did not go beyond these wavelengths.²⁹ Likewise, optical properties of the bones in the skull in the range from 650 to 950 nm were investigated, but the authors did not extend their studies to higher wavelengths.³⁰ In another study, changes in bone density at wavelengths 520 to 960 nm were reported. Again, they did not reach wavelengths in the SWIR range.³¹ Pifferi et al. looked at the human calcaneus at different wavelengths between 650 and 1000 nm. They were able to estimate the concentration of key tissue constituents, such as bone mineral, water, lipids, oxy-, and deoxyhemoglobin in human calcaneus, using the absorption spectral profiles of these constituents. It was concluded that the amount of bone mineral matrix decreased in older patients while lipids increased.³² Others have studied cranial bones at wavelengths extending from 800 to 2000 nm. The absorption spectra showed major absorption peaks located at 978, 1192, 1745, and 1930 nm. These investigators felt the peaks were secondary to increased lipid content.³³ Although the authors were able to reach wavelengths at the second and third NIR optical windows, they were not able to study window IV.

Numerous studies have utilized noninvasive, NIR, optical techniques to distinguish and to assess diseased versus healthy tissues. One study used light between 610 and 1040 nm to obtain the absorption spectra of benign and malignant breast tissues. Based on the absorption profiles of these tissues, it was found that the collagen and lipid content varied significantly among the patients. Women with highly adipose breasts were found to have relatively little collagen in their breasts, while relatively more collagen was found in women with very small breasts.³⁴ Another group used light from the visible to the NIR range (500 to 1600 nm) to distinguish breast cancer, benign tumors, and normal tissues, by utilizing comparisons of the relative fractions of collagen, water, and fat. Average lipid volume (%) at areas of normal tissue was 80%, whereas it was only $\sim 10\%$ for carcinoma in situ (a very early stage of cancer) and $<5\%$ for fibroadenoma or for carcinoma, whereas collagen volume was higher for carcinoma than for fibroadenoma or for adipose tissues.³⁵ These results were consistent with those we have previously reported, which showed increased tryptophan emission in breast cancer samples compared to their paired normal samples, and revealed higher levels of tryptophan in grade III cancers compared to grade I cancers.^36,37 Since carcinomas of the prostate and breast demonstrate decreased fat content and increased protein content in comparison to their adjoining nonmalignant tissues, the minimal water absorption occurring at window IV may allow better assessment of diseased versus healthy tissue, based on the relative fractions of collagen, water, and fat that are detected.

Wilson et al. have recently reported that the absorption peaks of lipids occur at 920, 1040, 1210, 1730, and 1760 nm.⁹ Our study shows that the total attenuation length ($l_{\mathrm{t}}$) from all tissue samples were significantly longer in the three SWIR windows compared to the conventional first NIR window, despite increased collagen or lipid content. Scattering losses due to the solid angle were the same for all specimens. A single scattering event is likely when $l_{\mathrm{t}}$ is greater than the sample thickness. Total attenuation measurements were evaluated based on the relative absorption measurements from the first NIR window and the SWIR windows for each tissue. Attenuation lengths obtained from studying lipid rich tissues were longer in the SWIR optical windows (specifically NIR windows II and III) compared to the first NIR window. It is possible that the second and third NIR optical windows are most valuable in studies of tissues with decreased collagen and increased lipid content, whereas windows III and IV might be particularly useful for diagnosis and evaluation of bone diseases or cancer (Table 4).

Table 4

Select chromophores at the first and SWIR optical windows.

Attenuation lengths obtained from studying bone tissues were greatest at the longest wavelength range in the SWIR (namely at third and fourth windows). The composition of bone is approximately one-third collagen, and, importantly, bone has very little fat content.^38,39 Collagen has an absorption peak at 2300 nm. It may be that minimal absorption at the fourth window and negligible scattering provides the ideal combination for deep tissue imaging of bone and other collagen-rich tissues. Window IV may also be most valuable for evaluation of other tissues, including cartilage, teeth, blood vessels, intervertebral discs, and skin.

6.

Conclusion

Total attenuation lengths ($l_{\mathrm{t}}$) from cancers, bone, chicken fat, and solutions of intralipid were calculated at the first NIR and SWIR optical windows. $l_{\mathrm{t}}$ results from all tissue samples reveal significantly greater lengths in the three SWIR windows compared to the conventional first NIR window, despite increased collagen or lipid content. Moreover, we note that NIR windows II and III appear to be the optimal SWIR range for studying tissues with a higher fat content, for example, normal breast, brain, and normal prostate. Because of their increased collagen content, cancer and diseases of bone may be best studied at windows III and IV. Among potential medical uses of the fourth window are the evaluation of skin cancers, studies of brain, atherosclerosis and vulnerable plaque, aneurysms, and evaluation of disorders of bone, including Paget’s disease and osteogenesis imperfecta. Imaging using the SWIR windows with a compact supercontinuum laser light system (with advantages including high power, broadband spectrum, and the ability to span from the ultraviolet to the infrared) may be utilized in combination with many optical applications including label-free microscopy, fluorescence lifetime imaging, and may also be useful during endoscopy to reveal early and in situ cancers.