1.

Introduction

Brain gliomas are the most common and deadly malignant brain tumors.¹ Gliomas are the leading cause of death by brain diseases in both children and adults,² and the median survival for patients with brain gliomas is a mere 12 to 15 months.³ An accurate, early clinical diagnosis of brain gliomas is crucial for patients’ treatment, in which microscopic pathology is the most commonly used technique for tissue analysis. Embedding brain gliomas in paraffin is a widely used first step in microscopic pathology analysis owing to its better storage and long-term tissue morphology preservation in clinical settings.⁴ In addition, there are a large number of paraffin-embedded brain gliomas stored in medical institutions, which also provide unique opportunities for later pathological studies. More importantly, by excluding water, paraffin-embedded brain gliomas allow us to reveal the intrinsic properties of brain gliomas.

The most common method for microscopic pathology analysis of paraffin-embedded brain gliomas requires pathological markers.⁵ The marker can be visualized using appropriate staining methods with an optical microscope. This method is called a gold standard. A final diagnosis is achieved based on histopathology, immunohistochemical stains, and other methods.^6,7 Consequently, it takes a significant amount of time and resources to obtain a reliable result. Unfortunately, experience is required to interpret paraffin-embedded brain samples, which is not an objective standard and is subject to interobserver variation or error.⁸

Therefore, an objective, label-free, quantitative analysis technique for paraffin-embedded brain glioma is required. The terahertz (THz) spectroscopic technique, which is a novel medical imaging modality,⁹ has received a attention for its benign nature (noninvasive and nonionizing) toward humans and its high sensitivity to biomolecules^10,11 and tissues.^12–14 In recent decades, the THz spectroscopic technique has been used to identify tumors on skin,^15–17 breast,^18–20 colon,^21–23 liver,²⁴ and other locations. All of the results show that the THz spectroscopic technique has potential for safely identifying disease in tissue samples.

Compared with normal brain tissues, brain gliomas have increased nuclear atypia, mitotic activity, florid microvascular proliferation, and necrosis,²⁵ all of which will change its THz refractive index, absorption, and scattering. Therefore, the THz spectrum of the brain glioma will be different from that of normal brain tissue. Previous studies on snap-frozen brain tissue²⁶ and fresh brain tumors²⁷ also provide preliminary experimental evidence that THz spectroscopy is able to identify brain tissue diseases. The brain tissues studied in these two studies contained an abundance of water. However, water greatly absorbs THz waves.⁹ The THz absorption reported in these experiments is primarily a result of the interstitial water content of the brain tumor and normal tissue. Most of the tissues studied by THz spectroscopy can be classified into three forms: fresh tissues,^12,13,16 frozen tissues,²⁶ and paraffin-embedded tissues.^23,24 The fresh tissues contain a large volume of water (THz absorption coefficient is about $240{\mathrm{cm}}^{-1}$ at 1 THz),⁹ which contribute most to THz absorption in fresh tissues, while frozen tissues²⁶ and paraffin-embedded tissues^23,24 have much lower THz absorption than that of fresh tissues by excluding water. Tumors cannot be accurately differentiated from normal brain tissue by using only the water content. As we stated previously, paraffin embedding brain gliomas is the standard method for brain pathology analysis. By excluding water, paraffin-embedded brain gliomas allow us to reveal intrinsic glioma properties in the THz spectrum.

In this study, we eliminated the impact of water and obtained the THz spectra (refractive index, absorption coefficient, and complex dielectric constant) of paraffin-embedded brain gliomas, which were established by standard surgical implantation of glioma models in mice. Furthermore, the THz spectral differences between brain gliomas and normal brain tissues were obtained and analyzed in this work. Based on these results, suitable parameters and frequency components for brain glioma identification are analyzed, which may serve as a new gold standard for pathology analysis.

2.

Samples and Methods

2.2.

Experimental Methods

The experimental setup used in this study is a transmission THz time-domain spectroscopy (TDS) system (Fig. 1), whose THz emitter is a photoconductive antenna with an electrode distance of 50 μm. The detector is based on a 1-mm-thick ZnTe crystal.

Fig. 1

Schematic representation of the terahertz time-domain spectroscopy system.

In our setup, a femtosecond fiber laser (MenloSystem Inc., Martinsried, Germany, F-780A), which operates with a 100-MHz repetition rate, is used to provide 100-fs laser pulses with a central wavelength of 780 nm. The laser pulses are split into a pump beam and a probe beam by a beam splitter. The pump beam is incident on the photoconductive emitter to produce electron–hole pairs, which are then accelerated by a biased electric field. According to Maxwell’s theory, these accelerated electron–hole pairs will generate a transient electric field and emit a transient electromagnetic wave. Because the duration of the excited laser pulse is 100 fs and the lifetime of the photoexcited carriers in the antenna substrate (low-temperature GaAs) is about 360 fs, the duration of the transient electromagnetic wave is on the picosecond order, lying in the THz band. The THz pulses are collimated to the sample by a pair of silver-coated off-axis parabolic mirrors, and then the pulses carrying information about the samples are guided to the THz detector by another pair of off-axis parabolic mirrors, changing the refractive index of ZnTe. This change will be encoded by the probe beam, which reaches the crystal synchronously, and will be detected by a balanced photodetector. After lock-in amplification, the signal will be recorded, and the waveform of the THz pulse will be measured by time-domain sampling. In order to remove the effect of water vapor, a sample chamber, which can be purged with nitrogen, was used to reduce the relative humidity down below 5% during the whole experiment. Moreover, each sampled data point in the THz pulse is measured five times and averaged to increase the signal-to-noise ratio.

To obtain the sample spectra, both the reference pulse, which is measured in the absence of the sample, and the sample pulse, which is measured when a sample is positioned in the focal point of the THz wave (Fig. 1), were measured as shown in Fig. 2(a). Then, a fast Fourier transform is used to obtain the frequency-domain electric field, which is shown in Fig. 2(b) demonstrating that the maximum signal-to-noise ratio of the system surpasses 2000. As the THz wave propagates from the THz emitter to the detector, it could be described in the frequency-domain as

Eq. (1)

$$E_{\mathrm{ref}}(\omega )=E_0(\omega )\exp (-i\omega t+i\frac{\omega }{c}l),$$

Eq. (2)

$$E_{\mathrm{sam}}(\omega )=E_0(\omega )\frac{2}{N(\omega )+1}\exp [-i\omega t+i\frac{n(\omega )\omega }{c}d+i\frac{\omega }{c}(l-d)]\times \exp [-\frac{\kappa (\omega )}{c}d]\frac{2N(\omega )}{N(\omega )+1}\times \sum _{\mathrm{k}=0}^{\infty }{\{{\left[\frac{N(\omega )-1}{N(\omega )+1}\right]}^2·\exp [i\frac{2n(\omega )\omega }{c}d-\frac{2\kappa (\omega )}{c}d\left]\right\}}^k,$$

where $E_0$ is the original THz pulse emitted by the photoconductive antenna, $l$ is the full path that the THz wave travels, $d$ is the thickness of the sample, and $N(\omega )=n(\omega )+i\kappa (\omega )$ is the complex refraction of measured sample, in which $n(\omega )$ is the refractive index, and $\kappa (\omega )$ is the extinction coefficient.

Fig. 2

Measured reference pulse and sample pulse: (a) time-domain and (b) frequency-domain.

To calculate the refractive index and the absorption coefficient, the following division will be performed

Eq. (3)

$$S(\omega )=\frac{E_{\mathrm{sam}}(\omega )}{E_{\mathrm{ref}}(\omega )}.$$

Our paraffin-embedded samples are made by a precise medical-standard slicer, whose thicknesses were then measured by a vernier micrometer. The refractive index and the extinction coefficient can be derived using numerical methods.³⁰ When calculating refractive index and absorption coefficient, we also considered samples’ thickness variation, which contribute to the error bar in THz refractive index and absorption coefficient (see Figs. 3 and 4). Then, the absorption coefficient can be calculated as follows:

Eq. (4)

$$\alpha (\omega )=\frac{2\kappa (\omega )\omega }{c}.$$

Fig. 3

Refractive indices of glioma and normal tissue.

Fig. 4

Absorption coefficient of glioma and normal tissue (one order of magnitude lower absorption coefficient in paraffin-embedded brain tissues than these of fresh tissues^12,13,16 is mainly because water is excluded in the former samples, which is the intrinsic properties of brain tissues in THz range).

3.

Results and Discussion

Twenty paraffin-embedded brain samples, which include 10 brain gliomas with the same illness and 10 normal brain tissues, were tested in our experiments, and three different spots on each sample were measured. The preparation processes, especially the lesion severity, of the 10 pieces of brain gliomas are the same. To this extent, all the samples and measured spot are “identical” and of the same refractive index and absorption coefficient in the THz range. But biology is diverse; the statistics diversity leads to difference in our experimental results. Also, the statistics diversity is shown as statistics error bars in the THz spectrum of refractive index and absorption coefficient (see Figs. 3 and 4). The error bars are derived from the standard error of the mean. Therefore, Figs. 3 and 4 display 95% confidence intervals.

As shown in Fig. 3, there is a clear difference in the refractive indices between the brain gliomas and the normal brain tissues. The refractive indices of glioma are stable at about 1.6 THz over the measured THz range from 0.2 to 2.0 THz. On the other hand, the refractive indices of normal tissue display a slight increase in the low-frequency range before 1.5 THz and a slight decrease for higher frequencies. The mean values of the refractive indices for normal tissue are between 1.2 and 1.5 THz.

Regarding the absorption coefficient, a distinct difference between the brain gliomas and the normal brain tissues occurred only at frequencies below 1 THz, as shown in Fig. 4. At higher frequencies, the error bars overlapped, and it was difficult to distinguish the difference between gliomas and normal tissues.

In our experimental results, both the refractive indices and the absorption coefficient of gliomas are larger than those of normal tissue. Because our specimens are dehydrated, these differences are not caused by the water content. Consequently, our results reveal differences in the cell components and morphology between gliomas and normal brain tissues. Different cell components may cause different THz responses by absorption, dispersion, and reflection, while different morphologies may cause different THz responses by scattering. Identifying the contribution of each mechanism requires an additional medical analysis.

To be clear, we calculated the differences in the refractive indices and absorption coefficient spectra between gliomas and normal brain tissues (Fig. 5). In the frequency range from 0.2 to 1.5 THz, the difference in the refractive indices decreases as the frequency increases, while the difference increases for increasing frequency in the 1.5- to 2.0-THz range. On the other hand, the difference in the absorption coefficient spectra is only obvious below 1 THz, especially at frequencies of 0.3, 0.55, and 0.76 THz as shown in Fig. 5(b).

Fig. 5

(a) Refractive index and (b) absorption coefficient difference between glioma and normal brain tissue.

The dielectric constant is also an important physical parameter and can be derived by $\epsilon ={\epsilon }_r+i{\epsilon }_i={(n+i\kappa )}^2$. Figure 6 shows the calculated dielectric constant results. As we can see, differences between gliomas and normal brain tissues are apparent for both the real and imaginary parts of the dielectric constant.

Fig. 6

Dielectric constants of glioma and normal brain tissue: (a) real part and (b) imaginary part.

For imaging applications, the imaging contrast is an important parameter, which depends on the relative spectral difference rate between gliomas and normal tissues. Higher relative spectral difference rates produce clearer images. The relative rate can be calculated from the difference between the glioma spectra and the normal tissue spectra divided by the normal tissue spectra. For unstable biomaterial parameters, such as those shown in Figs. 3–6, the error bars cannot be ignored, and the confidence intervals must be considered when calculating the difference between spectra. In this study, we employ a strict standard for evaluating the imaging contrast. The differences between spectra were obtained by subtracting the spectral value of the upper edge of the normal tissue confidence interval from the spectral value of the lower edge of the glioma confidence interval. Furthermore, the relative rates of all parameters discussed were calculated as shown in Fig. 7.

Fig. 7

Relative rates of the refractive index, absorption coefficient, and dielectric constant. (a) The full curves and (b) the positive portions of the curves.

To define the imaging contrast, Fig. 7(a) shows the relative rates for the refractive index, absorption coefficient, and dielectric constant. As shown in Fig. 7(a), the rates may be negative, which is caused in our method when the error bars overlap. In this situation, it is impossible to distinguish glioma from normal tissue. We are only concerned with positive relative rates. To be clear, the positive portions of the curves are illustrated in Fig. 7(b). Fore frequencies below 1.0 THz, the absorption coefficient and the imaginary part of the dielectric constant show better contrast. On the other hand, the refractive index and the real part of the dielectric constant show better contrast in the 1.7- to 2.0-THz range.

Of all four parameters, the imaginary part of the dielectric constant provides the largest contrast below 1.0 THz and reaches a relative rate of about 26% at 0.55 THz. In the 1- to 2-THz frequency range, the real part of the dielectric constant provides a better standard for distinguishing gliomas from normal tissues. According to these results, imaging based on the complex dielectric constant would provide the most reliable method. THz pulse imaging, based on THz-TDS spectroscopy, is the most powerful THz imaging technology, capable of obtaining all four parameters mentioned above. According to our results, THz pulse imaging can potentially obtain a clear image of paraffin-embedded brain gliomas.

On the other hand, when it is not necessary to obtain all information for a particular sample, continuous-wave THz imaging is a time-saving diagnostic technique, which has been widely studied in biomedical applications.^31,32 Moreover, if suitable frequency components are utilized, continuous-wave imaging may provide a useful result.³³ Our results should also be valuable for distinguishing between gliomas and normal brain tissues in continuous-wave THz imaging. There are two main methods of continuous-wave imaging: coherent imaging, which depends on the difference of the refractive indices, and intensity imaging, which depends on the difference of the absorption coefficients. As shown in Fig. 7(b), the relative rate of the refractive index is stable in the low-frequency range and shows some increase in the 1.7- to 2.0-THz frequency range. Both of these frequency ranges are suitable for coherent imaging. Moreover, the absorption difference shows two peaks at 0.55 and 0.76 THz, and these two frequency components are suitable for intensity imaging.

4.

Conclusion

In summary, we successfully measured the refractive index, absorption coefficient, and complex dielectric constant of paraffin-embedded brain gliomas and normal brain tissues in the 0.2- to 2.0-THz frequency range. According to our results, gliomas had a higher refractive index, absorption coefficient, and dielectric constant than normal brain tissues. We further discussed the imaging contrast. From our analysis, the dielectric-constant-based imaging may give the best contrast, and it is very suitable for THz pulsed imaging. THz waves with higher frequencies, such as those in the 1.7- to 2.0-THz range, are suitable for coherent imaging, while waves with lower frequencies, especially those at 0.55 and 0.76 THz, are suitable for intensity imaging. These results demonstrate that THz technology has the potential to distinguish gliomas from normal brain tissues and are valuable for THz imaging of paraffin-embedded brain glioma.