Terahertz (THz) spectral range is situated between the infrared (IR) and microwave bands and spans the frequencies of ~0.1–3.0 THz, or the wavelengths of ~3 mm–100 µm¹. Advancement in THz spectroscopy and imaging is driven by a rapid progress in femtosecond laser technologies and novel efficient approaches for the THz-wave generation and detection^2-7. Since the first experimental observation of THz heat rays by Rubens and Nichols⁸, THz technology offers a variety of applications in fundamental and applied physics. Auston’s research on photoconductivity in semiconductor, excited by the visible / near-IR femtosecond laser pulses, leads to the development of pulsed THz photoconductive antenna-emitters and detectors, related opto-electronic components and principles of THz-wave generation and detection, THz pulsed spectroscopy (TPS) and THz pulsed imaging (TPI) systems at the turn of the XX and XXI centuries^{3, 4, 9}. Principles of THz pulsed spectroscopy and imaging allow one to develop quite portable, ergonomic, and cost-effective THz opto-electronic instruments for biomedical applications¹⁰.

Nowadays, a considerable attention is paid to fundamental and applied research in the areas of early non-invasive, least-invasive and intraoperative THz diagnosis of malignant and benign neoplasms with different nosologies and localizations (such as the skin, mucosa, colon, breast, stomach, liver, etc.)^11-26, diabetes mellitus²⁷, tissue injuries²⁸, hydration²⁹, and viability³⁰. Therapeutic applications of THz waves also attract considerable attentions, where both thermal and nonthermal exposure effects are observed, such as regulation of gene expression, biological membrane permeability, deoxyribonucleic acid (DNA) demethylation, etc.³¹.

The versatility of THz technologies is due to the unique features of THz-radiation – tissue interactions, including the strong sensitivity of THz waves to the content and state of tissue water¹⁷, that underlies the tissue biochemistry, largely determines the effective complex dielectric permittivity of hydrated tissues, and forms a highly sensitive endogenous marker for THz spectroscopy and imaging. For example, at the frequency of v=1.0 THz, the refractive index and absorption coefficient of fibrous connective tissues, with the water content as high as ≈60%−75% (by volume), are n≈1.95 and α=200 cm^–1, respectively, which are significantly higher than those of fat tissues (≈1.60 and 50 cm^–1, respectively), with the water content of ~10%^{10, 17, 32, 33}. Tissues of the brain have an exceptionally high content of water. Along with the tissue water, other classes of biomolecules (such as lipids and proteins) play a significant role in formation of the tissue dielectric response at THz frequencies. The assessment of all these biochemical compounds makes it possible to differentiate between normal and pathological tissues of the brain, as well as between different stages of a pathological process relying on the THz spectra and images. All these factors, along with the biochemical and structural neuronal and glial features, including microscopic variations of tissue properties (tissue microenvironments³⁴, cellular and tissue morphology³⁵, biomolecules³⁶ and their large-scale fluctuations^{25, 37}), make the brain an exciting subject for study using THz technology^{10, 38, 39}.

In Fig. 1, dimensions of the structural components in neural tissues (such as neurofibrils, neurons, glial cells, etc.), meninges (pia matter, arachnoid matter and dura matter) and blood vessels are shown at the scale posed by the THz wavelengths. Since traditional THz imaging and spectroscopy systems are based on diffraction-limited optics and obey the 0.5λ Abbe diffraction limit, the Rayleigh scattering approximation and the effective medium theory are usually applied to describe the THz-wave – neural tissues interactions. Such THz systems allow to study the effective dielectric response of tissues, which is spatially-averaged over the λ² beam spot area of an optical system¹⁷. However, dimensions of some structural features of neural tissues are comparable to the THz wavelengths, and modern high-resolution THz optical systems are capable of distinguishing them^{40, 41}. Studying the effects of Mie scattering on such structural elements of neural tissues and development of novel approaches to describe the THz radiation transport in heterogeneous tissues containing such elements are topical problems of THz biophotonics, which still remain unaddressed¹⁰. In the general landscape, most of the structural and molecular features of healthy and pathological neural tissues do not affect considerably their electrodynamic response at THz frequencies and, thus, are usually neglected during the THz measurements of tissues^{10, 17}. Hence, THz complex dielectric permittivities of water, biological liquids and tissues are commonly described by the Debye, Cole–Cole, Davidson-Cole, Havriliak–Negami, or other analytical models of relaxation-like complex dielectric permittivity^{21, 42-46}. Particularly, the double-Debye model found a wide variety of application in THz biophotonics^{10, 17, 21, 46}.

In this paper, we start with a brief review of THz opto-electronic systems and measurement principles that are widely applied in biophotonics. Then, we discuss applications of these systems in different branches of neurodiagnostic, such as the diagnosis of neurodegenerative diseases, myelin deficit, tumors and traumatic injuries of the brain. Finally, challenging problems of THz neurodiagnosis are summarized and discussed.

THz spectroscopy and imaging are considered as prospective medical tools due to non-ionizing nature of THz waves and their harmlessness to human body³¹, as well as endogenous character of contrast observed between healthy and pathological tissues in THz images and spectra^{10, 17}. Nowadays, a number of approaches for receiving THz images or spectra of biological tissues are widely applied, that are based on different principles of THz-wave generation and detection¹, rely on either continuous-wave (CW / narrow-linewidth) or pulsed (broadband) radiation, and implement either transmission or reflection measurement geometries.

On the one hand, only thin (100 µm) specimens of hydrated tissues ex vivo can be studied in the transmission mode due to a very high THz-wave absorption by tissue water. On the other hand, measurements in the reflection mode require complex calibration procedures in order to reduce the effect of sample displacements during the THz data acquisition, which is especially important when studying tissues in vivo⁴⁷. Finally, we notice that, due to the limited depth of THz wave penetration in biological tissues, in neurodiagnosis, only the intraoperative applications of THz technology are possible, when an open access to brain tissues in vivo is provided or the freshly-excised brain tissue specimens ex vivo are studied. Let us consider several THz experimental spectroscopic and imaging systems that were already applied to study tissues of the brain.

THz spectroscopy is a powerful tool of biophotonics^48-52, that allows analyzing molecular parameters of an analyte and (to some extent) its structural features. Subsequent acquisition and processing of the THz spectroscopic data make it possible to quantify the content and state of many biochemical components of biological tissues (such as water in cells and intercellular matrix), the organization of these components in the morphological plane, as well as kinetics of chemical reactions and physical processes in an analyte^{53, 54}.

The reason for such a widespread use of TPS systems in biophotonics is complete information collected by such systems from a single rapid measurement about the frequency-dependent amplitude and phase of a THz pulse in a wide spectral range⁵⁵. Thus, TPS allows reconstruction of the broadband complex dielectric response (or complex refractive index) of an analyte directly from the measured data, without any additional physical assumptions or the use of the Kramers-Kronig transform⁵⁶. In addition, THz spectroscopy makes it possible to analyze the physical response of an analyte either in the time domain (pulse response function) or the frequency domain (complex transmission or reflection coefficients and complex dielectric permittivity or refractive index)⁵⁷. In Fig. 2, we show examples of TPS systems operating in (a) transmission mode^58-63, (b) reflection mode^64-68, and (c) attenuated total reflection (ATR) configuration^{69, 70}; all of them are used to characterize biological objects. Transmission mode measurements are usually applied to study thin layers of biological liquids and tissues ex vivo sandwiched between the two THz windows^58-63, including their measurements at cryogenic temperatures⁵⁹ aimed at unmasking some (other than water) molecular properties of tissues. Reflection mode measurements are used to study freshly-excised (hydrated) tissues ex vivo and tissue in vivo^64-68, where bulk (thick) tissue specimens are considered. ATR geometry is applied to study both bulk tissue specimens and thin layers, down to monolayer of cells. ATR spectroscopy provides high sensitivity, but requires specific calibration procedures. During the TPS measurement, the THz beam path is usually purged by nitrogen gas⁶³ or even vacuumized⁷¹ to suppress distortion of the THz measurements data by water vapor absorption lines.

Besides TPS, other THz spectroscopy modalities are vigorously explored nowadays and widely applied in different branches of science and technology, as well as have a potential in biophotonics and neurodiagnosis. Among them, we particularly notice

– the Fourier transform IR spectroscopy, that spans the IR and THz ranges and is used mainly in gas spectroscopy, condensed matter physics, and materials sciences^{72, 73};

– the backward-wave oscillator spectroscopy, that uses CW radiation, spans the GHz–THz range (<1.5 THz), has a narrow linewidth (10−5v), and is applied in condensed matter physics and physics of magnetism⁷⁴;

– the CW THz spectroscopy based on photomixers⁷⁵;

– the pulsed THz spectroscopy based on parametric generation and detection in MgO:LiNbO₃⁷⁶;

– the modern THz pulsed systems with high THz fields to study non-linear optical phenomena⁷⁷ and the optical pump – THz probe systems to study photoconductive materials⁷⁸.

Meanwhile, all these spectroscopic techniques mostly remain laboratory research tools and are rarely applied to study the THz electrodynamic response of biological tissues and liquids in biophysical laboratories or medical institutions.

Nowadays, THz imaging attracts increasing attention and offers novel opportunities in biomedicine^{48-52, 79}. Due to the lack of efficient THz detector arrays, majority of THz imaging systems use a single-pixel detector and, thus, involves point-by-point scanning of the sample surface with a focused THz beam of either pulsed or CW radiation followed by processing of thus collected 2D or 3D imaging data.

A widespread use of TPS systems leads to the development of different TPI methods⁸⁰. By studying the time- or frequency-domain response of an object, TPI yields 3D imaging data, that can be represented in form of the spatial distributions of the pulse response function R(t,r), broadband complex dielectric permittivity ϵ˜(ν,r), complex refractive index n˜(ν,r), and others, where r is a radius vector at the imaging plane. Such a comprehensive data about an imaged object open novel opportunities in analysis of the THz electrodynamic characteristics of tissues for subsequent differentiation between normal and pathologically-altered tissues. Various approaches for processing the TPI data were developed, aimed at dimensionality reduction and tissue identification and, thus, meeting the demands of different branches of medical diagnosis^{10, 17}. Despite such a rich information content, TPI possesses a limited operation rate due to the need of the time-domain TPS waveform collection at each point of the imaged object. This restrains capabilities of TPI systems in real-time biomedical applications².

The key advantage of CW THz imaging systems is a much higher operation rate than that of TPI, but such systems also suffer from lower information content⁴⁷. CW THz imaging can be based on a wide variety of quite-broadband or narrow-linewidth CW THz emitters and detectors and, thus, operating at different frequencies. For example, in refs.^{81, 82}, transmission-mode CW THz imaging system based on 3.4 and 3.7 THz quantum cascade lasers was developed, while in refs.^{83, 84}, a 2.52 THz gas laser was used as a key element during the transmission-mode CW THz imaging (Fig. 3). It is worth noting that reflection-mode and ATR-mode measurement geometries were combined with CW THz imaging principles. Particularly, in refs.^{58, 85-87}, reflection-mode CW THz imaging system based on a 2.52 THz gas laser was developed for biomedical purposes (Fig. 2(b)), while in refs.^{88, 89}, the ATR configuration was implemented for high-sensitive THz measurements of tissues (Fig. 2(c)). The common path CW THz reflection and ATR dual mode imaging system was developed⁹⁰. This dual-mode imaging of common optical path is realized by quickly switching between reflection window and total reflection prism. Finally, in ref.⁹¹, a hybrid CW THz imaging system was developed, that combines active mode (i.e., an object is illuminated by the THz radiation of an external source) and passive mode (i.e., self-emitted THz heat radiation of an object is detected) in order to retrieve maximum information about an object and is aimed at security applications.

Modern THz spectroscopy and imaging systems suffer from low spatial resolution, since they usually rely on the lens- or mirror-based optics, the resolution of which obeys the ≃0.5λ Abbe diffraction barrier of free-space focusing^{2, 10, 17}. Due to large THz wavelengths, resolution of such THz systems is a few hundreds of micrometers or few millimeters, even when working with a very wide-aperture lenses^{92, 93}. Overcoming the Abbe limit is of particular importance in THz biophotonics and, particularly, neurodiagnosis^{10, 17, 31}. In such applications of THz technology, the limited resolution does not allow to study subwavelength features of neural tissues, detect small-scale neoplasms, accurately delineate the tumor margins, and even locally expose tissues to THz waves.

In order to overcome the Abbe limit, several modalities of superresolution THz spectroscopy and imaging were proposed during the past two decades⁹⁴. Modern methods of THz image reconstruction rely on modeling of the point spread function of an imaging system, followed by the image reconstruction via the deconvolution or inverse filtering⁹⁵, or even based on the convolutional neural network approach⁹⁶. All these techniques help to boost the performance of almost any THz imaging system, with the resultant enhancement of resolution by tens of percent, but do not allow to overcome Abbe barrier. Here, we also notice the THz digital holography, synthetic aperture imaging, and computational imaging, which are capable of slightly sub-wavelength resolution, but still cannot overcome the Abbe barrier^97-101. All these methods have a potential in THz biophotonics, but require complicated techniques to resolve the inverse problems and are accompanied by specific image noises/distortions, which somewhat limits their practical utility.

Next, we consider different modalities of THz scanning-probe near-field optical microscopy (THz-SNOM), which can be classified into the tip- and aperture-based⁹⁴. The tip-based THz-SNOM systems detect the THz field, which is scattered by a metal or dielectric cantilever/tip handled in close proximity to an object, and provides the resolution as high as 10−2–10−4λ thanks to the strong confinement of the THz field on the tip end^102-105. Several emerging modalities of THz imaging, including wire media^{106, 107} and high-refractive-index optical fiber bundles⁷, can also be assigned to the tip-based THz-SNOM and has a potential for multipixel endoscopic imaging with the resolution down to 10−1λ. In turn, the aperture-based THz-SNOM uses subwavelength diaphragms to either illuminate an object or collect the scattered THz field at its shadow side, with the resultant resolution as high as ∼10−1–10−2λ determined by the size of the sub-wavelength aperture^108-110. As a subtype of the aperture-based THz-SNOM, it is possible to distinguish the innovative coded-aperture near-field THz microscopy¹¹¹ and the laser scanning-point THz source microscopy¹¹², both of which hold strong potential in THz biophotonics. Despite the superior resolution, all THz-SNOM systems suffer from low energy efficiency due to the use of subwavelength probes in an optical scheme. In this way, to achieve appropriate imaging quality, THz-SNOM systems require powerful emitters, sensitive detectors, and long image acquisition duration. Another disadvantage of THz-SNOM systems is a very small working distance between the probe and an object, due to which the probe may interact with an object, perturb its structure and distort the THz image, or even be damaged itself. Majority of the THz-SNOM systems have limited capabilities in imaging of amorphous media and soft biological tissues, which makes such systems laboratory research tools and restrains their applicability in THz biophotonics.

We particularly notice the solid immersion microscopy principles, that were recently translated from the visible–IR to the THz range^{94, 113} and, then, adapted for the CW THz imaging of amorphous objects and biological tissues with the resolution down to 0.15λ^{40, 114}. The essence of the solid immersion effect is a reduction in the dimensions of electromagnetic-beam caustic (focal spot), when the beam is focused in free space at a small distance (≪λ) behind the high-refractive-index optical element (i.e., the so-called solid immersion lens), with contribution of both ordinary-propagating and evanescence waves of the total internal reflection. The resolution of the THz solid immersion microscopy appears to be object dependent, but remains strongly subwavelengths in a wide range of the sample optical properties¹¹⁵. In ref.⁴¹, a method to resolve the inverse ill-posed problem of the THz solid immersion microscopy was developed. It allows to quantify the spatial distribution of the physical properties of an object over its aperture with superior resolution and, thus, to study sub-wavelength heterogeneities of biological tissues^{29, 116}.

Another approach similar to solid immersion effect is based on dielectric microspheres^{94, 117-121}. In this approach, a dielectric particle is introduced in the far field of the microscope (i.e., it is handled in front of the focal plane and can be placed either on the surface of an imaged object¹¹⁷ or in close proximity to it¹²²) in order to transfer near-field information and to magnify certain regions of the object. Such virtual images formed by each microsphere are collected with conventional objective lens of the microscope. In ref.¹²² white light microscope based on microsphere superlenses demonstrated resolution up to 10−2λ. Since this approach is usually used in transmission light it can be a perspective way for improvement of spatial resolution of THz microscopy of histological brain samples. Here we should also mention waveguide metasurface-based superlens, which allows to overcome the Abbe limit¹²³. Such superlens is formed by a gradient array of nanoscale slits perforated in a gold slab, and supports transverse-electric waveguide modes under linearly polarized illumination along the long axis of the slits. These waveguide modes can modulate not only optical phase but also evanescent waves which results in focusing of high-spatial-frequency waves, and thus to the quasi-far-field super-resolution focusing of light. Waveguide metasurface-based superlens developed in ref.¹²³ provide the resolution as high as 0.24λ.

Different modalities of THz spectroscopy and imaging have been already applied to study healthy and pathologically-altered brain tissues from laboratory animals and humans. In this section, we consider the process of THz technology translation to different branches of neurodiagnosis.

According to the literature data, THz technology was first applied to intact (healthy) brain tissue ex vivo. In refs.^{81, 82, 124} possibility of studying the anatomical structure of dehydrated brain tissue using CW THz imaging and near-field microscopy was demonstrated. A contrast between the THz response of gray matter and white matter was demonstrated and attributed to the different content of water, lipids, and proteins in these tissues. The listed factors are in descending order of their influence on the THz response of brain tissues. According to the well-known biochemical properties of brain tissues, in hydrated state (freshly-excised tissues ex vivo or tissues in vivo), gray matter has high water content and, thus, possesses higher THz refractive index n and absorption coefficient α than those of white matter¹²⁵. On the contrary, as follows from the images of dehydrated rat brain at 3.43 THz (Fig. 4(a–d)), dehydrated white matter has higher THz-wave absorption, which is predominantly due to an absence of water and higher content of lipids⁸¹.

Thus, we can conclude that in the native (hydrated) form, the effective THz dielectric response of brain tissue holds information about the content of tissues water. Therefore, THz spectroscopy and imaging can be used to map (qualitatively or quantitatively) this important parameter in healthy and pathological tissues of the brain. As a particular example of the water content estimation, in ref.⁵⁸, the authors quantified the fraction of water in freshly-excised intact brain and other tissues from rats using the TPS data at 0.5 THz, while the observed results showed good agreement with those measured by common tissue weighting technique (Fig. 4(e)). In turn, when working with dehydrated (paraffin-embedded or lyophilized) brain tissues, the effective THz optical properties reflect an information about the content of other molecular components of tissues (such as lipids and proteins) and have much lower values of n and α.

As shown in Fig. 1, brain tissues are usually assumed to be homogeneous isotropic in THz biophotonics. At the same time, some structural features of the brain and, particularly, of white matter might lead to the structural anisotropy of effective THz optical properties of such tissues. Optical anisotropy of white matter was earlier reported in other spectral ranges^{126, 127} and attributed to its fiber structure (i.e., the presence of packed bundles of axons with different orientations and degrees of order). The effective optical properties of such systems depend on the mutual orientation of the electromagnetic field polarization and wavevector towards the fibers. In the THz range, the structural anisotropy of white matter can be attributed to the anisotropic dynamic conductivity of electrolyte solution inside the ordered and oriented axons, as well as to the anisotropic character of Mie scattering of THz waves on axons (as mesoscale close-to-cylindrical multilayer dielectric particles). At the same time, the effect of brain tissue optical anisotropy still has not been properly studied in the THz range.

Knowledge about the THz optical anisotropy can be useful in studies of pathological changes in the white matter. The classical method for assessing the local anisotropy of the white matter is a magnetic resonance imaging (MRI)-based diffusion tensor method¹²⁷, which considers the anisotropy of water diffusion in tissues and features a limited sensitivity¹²⁸. THz evaluation of the white matter anisotropy can pave the way for more accurate study of the course of conducting fibers in different parts of the central nervous system, which is an important aspect of biomedical research. From the fundamental viewpoint, it is possible to study the connectome of various areas of the brain and the functional aspect (up to the cognitome). To one degree or another, many neurological and neurosurgical diseases have a direct or indirect effect on the properties and state of the white matter. White matter is the primary target of pathogenetic mechanisms in demyelinating diseases and in some forms of traumatic brain injury. At the same time, the white matter can be affected in neurotrauma, acute disorders of cerebral circulation, neurodegenerative, infectious and metabolic diseases¹²⁹. Moreover, tumors grow and infiltrate into various parts of the brain using the white matter tracts as pathways for spreading¹³⁰. Study of the structural and molecular features of white matter is of great interest for understanding the pathogenetic mechanisms, and THz methods might become the key for solving a number of specific problems in this area.

There is a demand on a better understanding of degenerative brain disorders. As an example of the neurodegenerative disease, we stress the Alzheimer's disease, which progresses to neurons death in different brain parts because of misfolding and accumulation of protein fractions that form deposits of β-amyloid and disrupting biochemical homeostasis and functional cell spectrum. Alzheimer’s disease is a leading cause of dementia, with loss of cognitive functions, memory, thinking skills, which is ultimately resulting in death^131-134. Cerebrospinal fluid analysis, neuroimaging (e.g., MRI), neuropsychological testing, blood testing, and some other methods have been used for Alzheimer’s disease diagnosis. These methods are either time consuming or high in cost¹³⁵, and their reliability may depend on the disease severity¹³⁶. Study of the origin, pathways, and contributing factors to neurodegeneration (such as oxidative stress and inflammation^137-139) is of crucial importance for understanding the progression of neurodegenerative disease and searching of possible therapeutic strategies. Recently, capabilities of THz methods in study of the neurodegenerative diseases were uncovered^{59, 61, 62, 68, 140-142}.

Consider several representative examples of the THz technology applications in study of the Alzheimer's disease. In ref.⁵⁹, a difference between the THz absorption coefficient α of frozen intact human brain tissues ex vivo and those altered by the Alzheimer's disease was uncovered using the TPS system in conjunction with a closed-cycle cryocooler (Fig. 5(a–c)). It was explained by the different content of β-amyloid, lipofuscin, tau, glial fibrillary acidic protein in pathologically-altered tissues, as well as by the tissue atrophy processes. In ref.⁶¹, it was reported that Tryptophan plays an important role in formation of the THz dielectric response of the Alzheimer-s disease-affected tissues in the ex vivo mouse model. In ref.¹⁴⁰, a contrast between the intact animal brain and Alzheimer's disease models became even more clear in the distributions of THz refractive index n(r) obtained by TPI (Fig. 5(d–i)). In ref.⁶², TPS was combined with the nano-slot patterned metamaterials for highly-accurate probing of the hydration dynamics of β-amyloid aggregates in liquids (Fig. 5(j, k)). For this aim, β-amyloid aggregates-containing water nanodroplets were used for improved THz sensing. The integration of a metamaterial sensor and nanoconfined droplets allowed to enhance the sensitivity of THz measurements (up to 1 nM of β-amyloid aggregates in a buffer solution). The observed results are of high importance because the amyloid β peptide and its ordered aggregates are associated with the onset of the Alzheimer’s disease, and, thus, they are considered as biomarkers for diagnostic and therapeutic applications. In ref.¹⁴¹, the nature of the THz conductivity changes during biochemical evolution of the β-amyloid with different concentrations (from monomers to polymer fibrils) were studied using the near-field THz spectroscopy. It was found that the THz conductivity decreases with the evolving fibrillization states and increases with the elevating molar concentrations. To further identify markers underlying the contrast between intact and pathologically-altered brain tissues in the THz range (without taking into account the effect of water), the authors of ref.¹⁴² measured dehydrated tissue samples. In post-mortem tissues, a considerable difference was found between the THz reflectivity of normal brain fragments and Alzheimer's disease, and the contrast was especially pronounced for white matter. This difference was explained by the concomitant demyelination.

Another degenerative disease to be considered is a demyelinating disease that occurs most often in young patients as the most socially-active group of the population. In ref.⁶⁸, TPS was combined with chemometric analysis to study the models of the demyelination process in the paraffin-embedded brain of mice and rhesus monkeys ex vivo. From Fig. 5(l), we notice a contrast between the myelin deficit and normal mouse brain ex vivo represented in the n–α space.

THz technology is considered as a prospective tool of the intraoperative diagnosis of human brain tumors³⁸, with an emphasis on gliomas, which are among the most common and deadly pathologies of the brain, constituting ≃26% of all primary brain tumors and ≃81% of malignant primary brain tumors¹⁴³. Gliomas are classified by the World Health Organization (WHO) into the WHO Grades I to IV, where Grades I, II and III, IV stand for the low- and high-grade gliomas, respectively¹⁴⁴. Glioblastoma (WHO Grade IV glioma) is the most dangerous tumor of the brain, with the five-year relative survival rate of only ≃6.8%¹⁴³. Surgery is the mainstay of the glial tumors’ treatment, the main goal of which is a gross-total resection of a tumor with maximal preservation of surrounding intact tissues¹⁴⁵. Gliomas usually have unclear margins, that complicate their gross-total resection. In many cases, accurate delineation of glioma margins can be provided only by the ex vivo histopathological examination of the excised tissues, using the hematoxylin and eosin (H&E)-stained histology aided by the molecular sensing and genetics¹⁴⁶. Such examination of excised brain tissues can be performed either intraoperatively (thus, extending the terms of surgery) or postoperatively (aimed at making a definitive molecular pathological diagnosis). Despite histopathology remains a gold standard in tumor diagnosis, we notice an increasing demand for the novel methods of the rapid intraoperative detection of tumor margins.

Common methods of the pre-operative imaging (MRI, computed tomography, and positron emission tomography) do not provide reliable accuracy of the tumor margins detection due to the brain shifts, caused by the dura mater opening, tumor removal, brain tissue edema, and cerebrospinal fluid losses¹⁴⁷. Thus, real-time imaging techniques (intraoperative ultrasound and MRI) were developed¹⁴⁸. However, these instruments suffer from low spatial resolution, while their integration into modern neurosurgical workflows remains labor intensive and expensive. Other intraoperative neuroimaging methods are fluorescent-based techniques involving the use of the 5-aminolevulinic acid-induced fluorescence of protoporphyrin IX¹⁴⁹, fluorescein sodium¹⁵⁰, or other fluorophores. They are inexpensive and provide high sensitivity for high-grade gliomas and meningiomas, but their sensitivity drops for the pediatric and low-grade tumors¹⁵¹. Optical coherence tomography^{152, 153}, Raman spectroscopy and imaging¹⁵⁴, confocal and polarization-sensitive microscopy¹⁵⁵, visible and near-IR spectroscopy¹⁵⁶, as well as photoacoustic imaging¹⁵⁷ are vigorously explored, as tools for the intraoperative brain tissue imaging, but still remain far from clinical practice. Efficiency of these methods is restrained by a number of factors, such as the aforementioned application of contrast agents, limited sensitivity and specificity, and high labor intensity.

A potential of THz technology in the intraoperative diagnosis of brain tumors was studied by the THz spectroscopy and imaging of different tumor models from rats and mice³⁸. In ref.³⁵, orthotopic glioma model from rats was imaged ex vivo using the reflection-mode TPI system. The observed THz images revealed a difference between intact tissues and a tumor in both freshly-excised and paraffin-embedded (dehydrated) states. In Fig. 7(a–i), visible, THz, and MRI images of the freshly-excised orthotopic glioma samples are compared. For the freshly-excised tissues, the difference between intact ones and a tumor was attributed to the increased content of tissue water in a tumor, which is due to the abnormal microvascularization, edema, and body fluids around necrotic debris. For the paraffin-embedded tissues, the contrast is less pronounced and originates reportedly from changes in cell density in a tumor. In ref.⁶⁰, TPS was applied to study paraffin-embedded intact tissues and glioma model (GL261 cell line) from mice ex vivo. The observed THz dielectric spectra justified a contrast between paraffin-embedded intact tissues and a tumor in the THz range. Next, we stress remarkable results of ref.⁶⁷, the authors of which applied TPS to study the freshly-excised and dehydrated intact tissues and glioma model (C6 cell line) from rats ex vivo. The measured THz dielectric spectra were analyzed using a linear spectrum decomposition approach, which revealed an increased water content in a tumor and, thus, justified water as the main reason of endogenous contrast between healthy tissues and a tumor. Finally, in refs.^{66, 85, 86}, different methods of THz spectroscopy, reflectometry and imaging were applied for discrimination between intact brain tissues and tumors, involving ex vivo and in vivo glioma models from mice and rats (eGFP+GSC-11, C6, and U87-MG cell lines), as well as few samples of high-grade gliomas from humans.

Recently, different approaches of boosting the THz imaging of brain tissues were proposed. First, in ref.⁸⁹, temperature-dependent transmission-mode TPS and 2.52 THz CW ATR imaging were used to study orthotopic glioma model from mice ex vivo. In Fig. 6(j–o), representative THz images are shown, that were collected in the temperature range from −20 to 35 °C, compared with a visible photo, and verified by the H&E-stained histology. It was found that THz response varies, as a function of temperature, in a different manner for intact tissues and a tumor. Such a temperature-dependent character of the tissue response at THz frequencies can be used for the improved differentiation between intact tissues and a tumor. Furthermore, some optimal temperatures of a tissue specimenex vivo can be selected to improve a contrast between healthy and pathological brain tissues in the THz range (similar effect was observed earlier for the TPS and TPI measurements of mucosa cancer²⁶). Second, to enhance the accuracy of the tumor margins delineation, in ref.⁸⁷, a region-of-interest (ROI) segmentation method was developed for work with the 2.52 THz CW imaging system. This method combines a variety of signal processing techniques, including the block matching 3D denoising, fuzzy c-means clustering, morphology operation, and canny edge detection. It was first verified involving test objects with sharp reflectivity changes and then applied to detect margins of rat glioma model ex vivo with quite smoothed margins. The observed results highlight a potential of the CW THz imaging and the proposed hybrid ROI segmentation method in brain tumor diagnosis.

Glioma models from animals only partially mimic structural and biophysical properties of human brain gliomas. Therefore, in ref.⁶⁴, TPS was applied to study the THz dielectric response of gelatin-embedded intact tissues, edematous tissues, and WHO Grades I–IV gliomas of the human brainex vivo. To preserve tissues from hydration/dehydration, retain their original molecular properties during the THz measurements, and, thus, sustain their THz response unaltered (as compared to the freshly-excised tissues), gelatin slabs were used for the tissue fixation¹⁵⁸. In Fig. 7, the measured THz optical properties are represented in form of the tissue refractive index n and absorption coefficient α, where statistical differences between intact tissues and gliomas of all WHO grades are evident. However, the THz response of edematous tissue appears to be very similar to that of a tumor, while there is also no significant contrast between the response of gliomas of the different grades. Thus, the THz response of human brain tissues ex vivo from ref.⁶⁴ confirmed earlier results of the aforementioned studies, that involved THz measurements of gliomas models from mice and rats (Section Glioma models from mice and rats). Furthermore, the data from ref.⁶⁴ were qualitatively verified in recent research work, where the TPS ATR measurements and the 2.52 THz CW ATR imaging of intact tissues and WHO Grade I–IV gliomas from humans were performed^{69, 88}. Indeed, THz methods hold a potential in the intraoperative differentiation between intact tissue and gliomas of the human brain.

High dispersion of the experimental n- and α-spectra (Fig. 7) was attributed to heterogeneous character of tissues. Particularly, in ref.⁶⁴, intact brain tissues were not differentiated into the white matter and gray matter, which might be the reason of higher dispersion for the intact tissues. In turn, the dispersion of glioblastoma optical properties reportedly originates from the presence of tissue heterogeneities due to necrotic debris, which is inherent to this type of tumor. Such tissue heterogeneities might be specific for various human brain tumors, forming a subject of additional comprehensive research work (it will be addressed below, to some extent). Nonmonotonic dispersion at lower (at <0.3 THz) and higher (>0.8 THz) frequencies is due to the measurement errors originating from fluctuations of humidity along the THz beam path, limited resolution of TPS, as well as the drops of the TPS system sensitivity at low and high frequencies.

Next, in ref.⁶⁵, THz dielectric response of freshly-excised intact tissues, edematous tissue, and WHO Grades I–IV gliomas of the human brain ex vivo was analyzed using relaxation models of complex dielectric permittivity. Particularly, the double-Debye (DD) model, which is based on a pair of the Debye relaxators, and the double-overdamped-oscillator (DO) model, which includes a pair of the Lorentz oscillators with a high damping constant, were considered. The DD model has the form⁴⁶

1ϵ˜=ϵ∞+Δϵ11+iωτ1+Δϵ21+iωτ2,

where ω=2πν is an angular frequency, Δϵ1, Δϵ2 are magnitudes that regulate a contribution of the “slow” and “fast” Debye relaxations with their time constants τ1, τ2, respectively, ϵ∞ is a high-frequency constant dielectric permittivity at ν≫(2πτ1,2)−1. The DD model is common for THz biophotonics. It allows accurate parameterization of the THz spectroscopic data for water, water solutions, biological liquids, and tissues^{10, 17}, for which only 5 independent parameters are required – namely, ϵ∞, Δϵ1, Δϵ2, τ1, and τ2.

In turn, the DO model is novel for THz biophotonics. It has the form¹⁵⁹

2ϵ˜=ϵ∞+Δϵ11−ω2ω012+iωγ1ω012+Δϵ21−ω2ω022+iωγ2ω022,

where Δϵ1, Δϵ2, ω01, ω02, and γ1, γ2 stand for magnitudes, quasi-resonant frequencies (given by the restoring forces) and damping constants, respectively, of the “slow” and “fast” relaxations, which are equivalent to the “slow” and “fast” Debye relaxations ( Eq. (1)). A constant ϵ∞ and a magnitude Δϵj of j-th Debye relaxation are equal to those of a corresponding overdamped oscillator. One can also calculate other parameters of j-th overdamped oscillator based on corresponding Debye relaxator:

3γj=ω0,j2τj,γj=Cω0,j,C≫1,

here, C is a constant, the particular value of which is not so important, unless it is much larger than unity. In ref.⁶⁵, this constant was C=102. The DD and DO models provide almost equal values of complex dielectric permittivity ϵ˜=ϵ'−iϵ'' (or complex refractive index n˜=n−ic0α/(2πν), where c0≈3×108 m/s is the speed of light in free space, and α is the absorption coefficient, by field) at low-frequencies up to ω∼(2πτi)−1<ω0,i≪γi. At the same time, at higher frequencies ω>(2πτi)−1, the DD model gives non-physical high loss, due to which this model does not satisfy the sum rule and predict infinite number of charged carriers underlying the dielectric response of matter. In turn, the DO model satisfies the sum rule and, thus, seems to be more physically rigorous⁶⁵.

Based on the THz dielectric spectra from Fig. 7 and ref.⁶⁴, parameters of the DD and DO models were calculated in ref.⁶⁵. They are shown in Tab. 1, where, for all the considered tissues, the “slow” and “fast” relaxation times τ1, τ2 were set equal to those of free water from ref.⁴⁶. The observed higher magnitudes of the “slow” and “fast” relaxations Δϵ1, Δϵ2 for all glioma grades (Table 1) are attributed to the higher content of polar H₂O molecules in a tumor than that in intact tissues¹⁶⁰. The developed DD and DO models can be used to describe the THz-wave – brain tissue interactions in different branches of THz biophotonics using analytical or numerical methods of electrodynamics. Moreover, parameters ϵ∞, Δϵ1 and Δϵ2 of these models can serve as physically-reasonable principal components for the differentiation between normal and pathologically-altered brain tissues, as detailed in ref.¹⁶¹.

Next, water content in intact tissues and WHO Grade I–IV gliomas of the human brain ex vivo was estimated relying on the sum rule⁶⁵

4f=NtissueNwater=∫ωminωmaxωϵtissue''dω∫ωminωmaxωϵwater''dω,

where integration of imaginary part ϵ'' of the complex dielectric permittivity within the considered spectral range (ωmin,ωmax) is involved for tissues, in the numerator, and for a liquid water, in the denominator. The obtained water content estimates are shown in Fig. 8, from which we notice that the water content is 71.1±6.8% for the intact tissues, while that in edema and WHO Grade I–IV gliomas is ∼5–10% higher. From Fig. 8, we also notice that the obtained values of water content overall agree with the data from ref.^162-165, that were measured using TPS or other experimental techniques.

Despite water remains the main endogenous marker of neoplasms in the THz range¹⁰, establishing a relationship between the tissue response at THz frequencies and the well-known molecular markers of a tumor (other than water) is a challenging problem of THz biophotonics, since the role of such markers in modern oncology increases dramatically and even become decisive. Consider several important pathogenetic and clinical molecular events that are increasingly used in brain tumor diagnosis along with the common classification of gliomas into the abovementioned WHO Grades I–IV:

Ki-67 labeling index. Proliferation is the most important function of cells of any neoplasm, which determines the malignant potential of a tumor. In the pathological diagnosis of tumors, the Ki-67 marker is used to assess this function^{166, 167}. This protein is involved in direct preparation for cell division and, with a high degree of accuracy, marks cells preparing to enter the process of mitosis. This marker has confirmed its high predictive value, independent of other histological criteria, for a number of tumors, including diffuse gliomas^{168, 169}. The assessment of the Ki-67 labeling index level plays a decisive role in formulation of the pathohistological diagnosis and determining the tactics of further treatment^{168, 169}. The activation of the cells’ proliferative program affects inevitably their metabolic profile. One of its biochemical consequences is a change in the amount and state of tissue water, which is especially noticeable under conditions of accelerated pathological cell division in the framework of carcinogenesis processes¹⁷⁰.

Mutations in the isocitrate dehydrogenase genes (IDH1andIDH2). For diffuse glial tumors, the mutational status of the IDH1 and IDH2 genes became the defining diagnostic parameter^{171, 172}. These mutations act as the earliest molecular events in diffuse gliomas and determine the further course of carcinogenesis. The presence of these mutations has an extremely pronounced positive effect on the prognosis of patients with diffuse gliomas^{171, 172}. The influence of IDH1 and IDH2 mutations on the processes of nucleation and development of tumors is mediated by metabolic factors. An indirect feature of these modifications is the change in the state and content of water in tumor cells.

Methylation of the O⁶-methylguanine–DNA methyltransferase (MGMT) gene promoter. A molecular parameter of high diagnostic value is the methylation status of the promoter region of the MGMT gene. This molecular event refers to epigenetic modifications that regulate genome expression. Methylation of the MGMT gene promoter is associated with profound changes in the molecular and functional status of tumor cells. It serves as an independent favorable prognostic factor for survival of patients with diffuse gliomas^{173, 174}. Methylation of the MGMT gene promoter is accompanied by deep metabolic rearrangements in tumor cells associated with a change in the protein composition of tumor cells, which entails modifications in the content and state of intracellular water due to the polarity of most protein molecules.

There are some prerequisites to study the listed molecular markers of gliomas using the THz methods and other modern optical modalities of tissue spectroscopy and imaging, for which either freshly-excised or fixed tissues ex vivo can be considered. In ref.⁶⁹, it was demonstrated that the IDH mutant and wild-type glioma tissues can be distinguished by TPS arranged in the high-sensitive ATR configuration. From Fig. 9, we notice that THz spectroscopy has a potential for a rapid molecular typing of glioma tissues for this important diagnostic factor. In ref.¹⁷⁵, THz absorption spectra α of 2-Hydroxyglutaric acid disodium salt (2HG) are investigated. 2HG is another biomarker existing in glioma, which can serve for differentiation between tumor and normal tissue. Authors revealed correlation between THz absorption spectra and molecular structure of the 2HG isomers.

In order to study heterogeneities of healthy and pathological brain tissues at the THz-wavelength scale, in refs.^{41, 116}, TPS measurements and quantitative superresolution 0.6 THz CW solid immersion microscopy (with the spatial resolution as high as 0.15λ)⁹⁴ of freshly-excised intact tissues and glioma model 101.8 from rats ex vivo were performed. The homograft glioma model 101.8¹⁷⁶ involves injection of rat brain glioma tissues into the brain of another rat. This model was obtained and is kept in the collection of the Research Institute of Human Morphology (Moscow, Russia). It is compliant with glioblastoma (WHO Grade IV glioma)¹⁴⁴ and widely applied in experimental neurooncology. Glioma 101.8 is heterogeneous in nature and mimics unclear (diffuse) margins and tissue heterogeneities (such as areas of necrosis, hemorrhages, and microvascularity), that are usual to the human brain gliomas. Therefore, glioma 101.8 is a perfect candidate for evaluating the performance of novel modalities of brain tissue spectroscopy and imaging^{152, 153}.

In Fig. 11(a–c), THz refractive index n and absorption coefficient α (by field) are shown for the freshly-excised intact rat brain and glioma ex vivo, measured by a diffraction-limited TPS system and verified by the H&E-stained histology¹¹⁶. Mean THz optical properties and their spatial dispersion over the rat brain are shown by markers and error bars, respectively. Due to the limited TPS resolution, intact tissues were not differentiated into the white matter and grey matter, and no subwavelength neurovascular structures and heterogeneities of the brain were identified. TPS data from Fig. 11(a, b) confirmed endogenous contrast between intact tissues and glioma 101.8 from rats ex vivo in the THz range, which is similar (but less pronounced) to that observed earlier for distinct xenograft glioma models from laboratory animals (Section Glioma models from mice and rats)^{60, 66, 67, 85, 86} and humans (Section Human brain gliomas)^{64, 65, 69, 88}. Also, from Fig. 11(a, b), considerable dispersion of the measured THz optical properties across the sample is notable for both the intact tissues and a tumor¹¹⁶, which is similar to observations of glioma model (Fig. 7(a–i))³⁵ and human brain gliomas (Fig. 7)⁶⁴. As mentioned above, such heterogeneity is assumed to originate from various neurovascular structures, fluctuations of the water content and cell density over the brain, but the diffraction-limited resolution of common TPS systems fails to justify the nature of such tissue features.

This problem was addressed using the quantitative superresolution 0.6 THz (λ≈500 µm) microscopy in refs.^{41, 116}. Figure 10(d–o) show results of studying intact tissues and glioma 101.8 in form of the visible photo, THz intensity image I(r), refractive index distribution n(r), absorption coefficient distribution α(r), water content distribution C(r), and H&E-stained histology, as detailed in ref.⁴¹. The observed subwavelength distributions of the THz optical properties n(r), α(r) and tissue water content C(r) agree with the average values measured by the diffraction-limited TPS systems (Figs. 8 and 11(a, b))^{64, 65, 67, 116}. Meanwhile, thanks to the superior resolution of the THz microscope, considerable heterogeneity of the brain tissues is evident.

From THz images in Fig. 10(d–o), a pronounced difference is observed between the white matter and gray matter, which is consistent with previous observations^{35, 68}. Particularly, white matter possesses higher THz refractive index n and absorption coefficients α due to an increased content of both tissue water C (in the form of an electrolyte solution inside axons) and myelin (a lipid-rich substance that surrounds the axons). Tumor also possesses higher n- and α-values than those of intact tissues owing to higher water content C in a tumor^{65, 67}. A considerable mesoscale heterogeneity of a tumor is also notable, originating from the tumor cells’ accumulations, vessels, necrotic debris, and hemorrhages^{41, 116}.

On the one hand, the observed results of the THz microscopy of the rat brain ex vivo justify a label-free contrast between intact tissues and a tumor. On the other hand, the THz images highlight considerable heterogeneity of brain tissues at the THz-wavelength scale. This poses important problems of studying the effects of Mie scattering of THz waves in brain tissues and adapting the radiative transfer theory for the needs of THz biophotonics and neurodiagnosis. Indeed, both absorption and scattering effects can considerably impact the THz-wave transport in healthy and pathological tissues of the brain. These scattering effects might either complicate the THz diagnosis of tumors (they can lead to strong variability of the effective THz response of tissues) or become a source of additional useful information for the tissue differentiation. Particularly, such structural features of a tumor, as areas of necrosis and hemorrhages, are typical for glioblastoma and, thus, can be used as a source of additional information for the differentiation between glioblastoma and other benign and malignant gliomas.

Although most research papers focus on THz spectroscopy and imaging of brain gliomas, as the most common and dangerous tumors of the brain, THz technology has a potential in the intraoperative diagnosis of other types of primary and secondary brain tumors. For example, in ref.¹⁷⁷, an ability for the differentiation between intact tissues and human brain meningioma of the WHO Grade I was uncovered using TPS, but this study involved only one sample of a tumor. Attractiveness of THz technology in diagnosis of other types of human brain tumors should be further confirmed, since it requires collections and analysis of the THz response of large number of tumor samples.

The study and characterization of cell cultures by THz technology are also worth mentioning. In ref.¹⁷⁸ the THz dielectric responses of living glial-like cells (PC12, SVG P12 and HMO6) have been demonstrated based on the combination of the single-interface and two-interface ATR models without cell thickness. Moreover, the glioma cells (C6 and U87) exhibited different dielectric properties compared with the glial-like cells, which could be one reason for the glioma tissue diagnosis using THz wave.

THz waves can be used to access viability of tissues and cells. In this way, in ref.¹⁷⁹, an ability of TPS in the ATR configuration to evaluate (in a label-free, non-invasive, and fast manner) the effects of bioactive constituent on living glioma cells was demonstrated. Culture of glioma cells (U87) was exposed to ginsenoside Rg3. THz optical properties of cells were measured using TPS, while the cell growth inhibition rate was studied using conventional cell viability test kit or by the cellular morphological changes observed with fluorescence microscopy. The results of this study verify the TPS effectiveness in detecting the effects of glioma cells exposure to ginsenoside Rg3 (G-Rg3)in vitro. This highlight other important applications of THz technology in neurooncology, that is associated with assessing the efficiency of bioactive constituents and other exposure factors on living cells and tissues of the brain.

Traumatic brain injury (TBI) is the damage caused to brain by external mechanic forces, that exceeds the protective capacity of the brain and occurs due to such reasons as crush, falls, blast waves, penetration by a projectile, etc. TBI is the most common disease with high mortality and disability, while its severity is graded into mild, moderate, and severe¹⁸⁰. TBI pathophysiology is complicated with diverse manifestations, including the immediate and delayed mechanisms¹⁸¹. The initial brain injury triggers a variety of molecular and biochemical events: cerebral blood flow changes, axonal shearing, metabolic imbalance, etc. The delayed secondary damage might result into neuronal and glial damage, including brain edema, neuro-inflammation, and delayed neuronal death. TBIs are responsible for 10 million deaths and hospitalizations annually. Therefore, TBI diagnosis and analysis of their severity are of crucial importance for selection of the treatment and rehabilitation strategies, as well as prognosis of the patient’s survival and quality of life.

Nowadays, the advanced imaging modalities based on different physical effects are applied for the TBI diagnosis, such as computed tomography¹⁸², diffusion tensor MRI¹⁸³, positron emission tomography¹⁸⁴, laser-induced photoacoustic imaging¹⁸⁵, fluorescence molecular imaging¹⁸⁶, and others. Nevertheless, all these conventional modalities of TBI diagnosis possess limited resolution, sensitivity, and specificity, as well as fail to adequately depict brain injury in mild TBI, being insensitive to diffuse axonal injuries. Hence, development of a rapid and label-free imaging technology to detect margins of TBI and distinguish between its different degrees is a challenging problem of modern neurosurgery, especially for the mild TBIs.

Most recently, THz technology was considered as an innovative tool for the intraoperative diagnosis of TBIs with different degrees. Particularly, in ref.⁸³, TBI models of the mild, moderate, and severe degrees in rat brains ex vivo were studied using TPS and THz CW imaging systems, where both freshly-excised and paraffin-embedded (dehydrated) tissues were considered. In Fig. 11(a), THz images of freshly-excised intact rat brain ex vivo are compared with visible and MRI images, while in Fig. 11(b), THz absorption coefficient α are shown for the paraffin-embedded intact rat brain and TBIs of the different degrees. The observed results revealed that THz spectroscopy and imaging of both freshly-excised and paraffin-embedded tissues allow for clear discrimination between the intact tissues and TBI, as well as between TBIs of different degrees. THz data somewhat correlate with the visible and MRI imaging. THz absorption coefficient α increases with the TBI degree, as compared to the intact brain. This was attributed to the edema, increased water content and decreased cell density in the injured brain. It is worth noting that the origin of contrast between TBIs and intact tissues in THz images and spectra is similar to that observed for brain tumors to some extent (Section Tumors).

In the followed-up ref.⁸⁴, the authors studied the intact rat brain ex vivo and TBI models of the different degrees using the transmission-mode 2.52 THz CW imaging. For this aim, the freshly-excised 40-μm-thick slices of brain tissues were imaged. Using the obtained data, machine learning algorithm was developed for the automatic and rapid classification between intact tissues and different degrees of TBI. Next, in ref.¹⁸⁷, TPS was applied for the early diagnosis of blast-induced TBI via measurements of the serum and cerebrospinal fluid. Moreover, the THz spectra of total protein in the hypothalamus and hippocampus at various time points after blast exposure were studied and analyzed, from which clear differences were observed between the THz response of an analyte at distinct time steps. Finally, the principal component analysis and machine learning algorithms were used to automatically identify the degree of blast-induced TBI. In this way, the discussed pilot research works highlight strong potential of THz spectroscopy and imaging in diagnosis of TBIs.

Despite the discussed potential applications of THz technology in different branches of neurodiagnosis, a variety of fundamental and applied problems still restrain transfer of THz technology to a medical practice. In this section, we stress the following important issues.

THz-wave – tissue interactions & tissue heterogeneity. In the visible and IR ranges the interactions between electromagnetic-waves and tissues have been studied comprehensively since the middle of the XX century¹⁸⁸. Compared to these frequency ranges, THz biophotonics is quite novel research area, with much less data on the THz-wave – tissue interactions accumulated up to date^{10, 17, 31}. Therefore, study of the THz optical properties of tissues, the physical effects underlying the tissue response at THz frequencies, and the origin of contrast between healthy and pathologically-altered tissues in THz range are required in order to uncover advantages of THz technology over other tools of medical spectroscopy and imaging. Particular attention must be paid to the effects of Mie scattering of THz waves on the tissue heterogeneities (Fig. 1). Namely, structural heterogeneities of intact and pathologically-altered tissues lead to the mesoscale spatial fluctuations of the THz optical properties of tissues, which gives rise to the Mie scattering of the THz waves^{189, 190}. These scattering effects can either hamper the THz diagnosis or provide additional useful information for the tissue quantification^{41, 116}.

A question arises about applicability of the standard effective medium theory and the related complex dielectric permittivity models in the THz range^{10, 17, 31}. The effective medium theory assumes tissues to be homogeneous at the THz-wavelengths scale, and, thus, it fails to account for the non-Rayleigh scattering effects. In addition to the discussed heterogeneities of rat brain tissues (Section Brain tissue heterogeneity), mesoscale THz-wave scatterers were observed earlier in the tissues of the human breast and tongue using the THz solid immersion microscopy^{40, 191}. At the same time, the effects of THz-wave scattering on such tissue heterogeneities are still to be comprehensively analyzed. Development of a correct physical model to describe the THz-wave transport in tissues with such scatterers remains an important problem of THz biophotonics. For this aim the Mie scattering theory and the radiation transfer theory can be adapted^{189, 190}.

Heterogeneous character of tissues can complicate the intraoperative THz delineation of tumor margins. Indeed, common effective medium theory formalism involves analysis of the effective THz optical properties of tissues, which are averaged within the diffraction-limited THz beam spot and which probably would not provide reliable data for the intraoperative delineation of the tumor margins. In this way, to make the THz diagnosis possible, one should resort to novel modalities of THz spectroscopy and imaging of tissues, that overcome the Abbe diffraction limit and improve the tissue characterization^{2, 94}. A comprehensive analysis of the brain tissue response at THz frequencies with subwavelength spatial resolution (with an emphasis on the perifocal region of a tumor) is in order to better understand the origin of tissue heterogeneity, as well as to introduce advanced approaches for the tissue characterization and differentiation, that go far beyond a simple comparison of the effective THz optical properties of healthy and pathologically-altered tissues. In such case, being comprehensively analyzed and modeled, tissue heterogeneities can serve as a source of additional useful information for THz diagnosis.

Novel modalities of near-field THz microscopy with the resolution as high as ∼10−2–10−4λ^{94, 103, 104, 192} can open opportunities in neurosciences and brain tumor diagnosis. Applications of the THz microscopy for detecting individual cells and their populations within the glioblastoma tissue seems to be an extremely important task that allows for solving a number of fundamental and clinical tasks. Internal heterogeneity of glioblastomas reflects the key mechanisms of their pathogenesis. Assessment of the cell-population heterogeneity allows to accurately determine the degree of biological aggressiveness of a tumor and then select the most effective way of its treatment^{193, 194}. Particularly, it is important to identify populations of glioma stem cells, which are able to spread widely in the brain beyond the main tumor node¹⁹⁵. Glioma stem cells also form the main sources of the tumor cell mass repopulation and, thus, a key etiotropic cause of the tumor recurrence¹⁹⁵. Moreover, these cells, in the form of a mesenchymal variant, provide the formation of tumor resistance to the chemotherapy and radiation therapy¹⁹⁶. Detection and resection of brain tissue infiltrated with these cells is a critical task, aimed at increasing the surgical treatment radicality and improving the patient’s survival. At the moment, such features of the tumor cell-population composition can be detected only by the single-cell sequencing, that remains extremely expensive and inapplicable in the intraoperative conditions¹⁹⁷. The development of new tools for the intraoperative cell-population mapping in brain tissue based on the THz microscopy principles is an extremely important fundamental and clinical task.

Studying brain tissue with superresolution THz microscopy and polarization-sensitive THz imaging might be useful for understanding the course of conducting fibers (axons and their bundles) in different regions of the central nervous system. This can pave the ways for solving important problems of biomedical research, studying the connectome of various brain areas, and evaluating the functional aspects of the brain (up to the cognitoma). However, such opportunities have not been addressed yet.

Advanced and cost-effective opto-electronic systems for the THz biophotonics. THz instruments are still rare, cumbersome, non-ergonomic, and expensive. It will take considerable research and engineering efforts to develop portable, cost- and energy-effective THz components and systems for applications in a clinical environment. We should emphasize the need for the novel THz optical materials and related fabrication techniques, aimed at producing open-space THz optical elements^198-201, as well as hard waveguides, flexible fibers, and optical fiber bundles^{5-7, 202, 203}, aimed at the THz-wave delivery to hard-to-access tissues and internal organs for diagnostic and therapeutic applications.

We also mention the need for the THz emitters and detectors with improved performance, that can bring THz spectroscopy and imaging closer to a real-time operation, as well as improve sensitivity of THz measurements. For this aim, an increasing number of tools appear annually. Among them, we particularly mention innovative plasmonic THz photoconductive emitters and detectors^204-206, THz photoconductive antenna arrays²⁰⁷, fast rotary optical delay lines^{208, 209} and THz devises based on novel two-dimensional semiconductors or principles of spintronics^{78, 210, 211}. Despite a variety of superresolution near-field THz imaging modalities are vigorously explored nowadays (Section Superresolution THz microscopy)^{2, 40, 41, 94, 103, 108, 109, 111, 113, 115, 116}, they are still to be adapted for the needs of THz biophotonics, neuroscience, and neurodiagnosis.

Improving the THz-wave penetration in tissues. A limited depth of the THz-wave penetration into hydrated biological tissues (10–100 µm), considerably restrains the range of THz technology applications in biophotonics, neurosciences, and neurodiagnosis, while improvement of the THz-wave penetration depth can broaden a range of these applications. For this aim, modern methods of immersion optical clearing of tissues were recently adapted from the visible and IR ranges^{188, 212, 213}. This method is based on the application of specific liquid chemical agents to a tissue. This leads to water diffusion from the tissues and agent diffusion into them and, thus, results in substitution of tissue water by an agent and provides temporal and reversible reduction of the tissue water content (tissue dehydration), refractive index matching, tissue shrinkage and better ordering²¹⁴. In the visible–IR ranges, immersion optical clearing significantly suppresses the light scattering in tissues, thanks to the achievement of tissue spatial homogeneity. In turn, in the THz range, where scattering effects are not so dominating and only the tissue water strongly absorbs THz waves, the reduction of water content is more important, thus, decreasing the THz-wave absorption by tissue water and increasing the depth of THz-wave penetration into tissues²¹⁵. In ref.⁶³ optimal hyperosmotic agent for the immersion optical clearing of tissues in the THz range was selected. For this aim authors measured THz absorption spectra of most common agents and estimated coefficients of their diffusion into the freshly-excised rat brain tissues. It was shown that glycerol, as an agent, provides simultaneously high penetration depth and diffusion rate. Thus, it seems to be optimal for immersion optical clearing in the THz range. In ref.²¹⁶ example of the THz-wave penetration depth enhancement were demonstrated by applying glycerol to a 224-μm-thick abdomen skin tissue ex vivo from a mouse. At the same time, research efforts are still required to form immersion optical clearing protocols for different branches of THz biophotonics, including the THz neurodiagnosis.

Exogenous markers and nanoparticles for THz diagnosis. Exogenous markers also possess high potential for THz diagnosis. Their qualitative and/or quantitative characterization can be considered as additional evaluation and connected with molecular factor for diagnosis. Most often, these markers are a complex of a specific molecule, such as an antibody or a nucleic acid fragment, or, sometimes, a metabolite, that can bind highly selectively to the desired molecular factor or accumulate specifically in pathologically-altered cells due to the peculiarities of their metabolism, and a carrier that, being accumulated in cells and tissues, can manifest itself quite clearly when the appropriate research method is used. Today, exogenous markers are increasingly being used in neurodiagnostics involving such approaches like MRI^{217, 218}, computed tomography^{219, 220}, and fluorescent microscopy in visible and IR regions²²¹. For example, the ingestion of 5-aminolevulinic acid results in the accumulation of protoporphyrin IX (the product of its biochemical transformation) specifically in tumor cells due to the gradient of their metabolic activity. This induces the exogenous fluorescence of the tumor and underlies the imaging approach for identification of the boundaries of brain tumor²²¹. The first implementation of the same principle for THz diagnosis was described in refs.^{222, 223}. In particular, the use of specific nanoparticle-based biosensors^{224, 225} enables accurate detection of the content of various miRNAs, including miRNA-21²²⁶. Moreover, as it was demonstrated in ref.²²⁷, biosensors based on gold nanoparticles can be applied for proteomic profiling, in particular, for finding molecules of epidermal growth factor receptors, as well as avidin, which plays a significant role in the pathogenesis of a number of neoplasms, including glioblastomas, and are included in clinical and diagnostic criteria. Some research also applied Gadolinium oxide as a contrast agent in THz imaging, to diminish the deteriorated imaging contrast caused by strong THz absorption by water in vivo^228-230.

Multimodal diagnosis. The combination of several diagnostic approaches, integrated together in one complex diagnostic system, possesses high efficiency in clinical practice²³¹. For instance, the combination of TPI, MRI, OCT, fluorescence imaging relying on green fluorescence protein (GFP) and protoporphyrin IX (PpIX), white light imaging and H&E-stained histology make it possible to more clearly define the boundaries of brain tumors, perifocal zone, as well as to evaluate the molecular properties and the degree of pathohistological malignancy of the neoplasm⁸⁵. In ref.²³², the application of TPS and OCT together for intraoperative neurodiagnosis offset the drawbacks of these methods; it yields differentiation of gliomas from normal tissues, performed by TPS, and discrimination between high-grade and low-grade gliomas, performed by OCT. We also should mention that not only OCT^{233, 234}, but other methods, such as thermography²³⁵, multi-spectral imaging^{236, 237}, high frequency ultrasound imaging²³⁸, Raman spectroscopy²³⁹, polarization imaging^{19, 155, 240} and multi-photon microscopy²⁴¹, being combined with either THz spectroscopic or imaging modalities, can be rather promising for intraoperative neurodiagnostics.

In this review, modern research status in THz neurodiagnostics is discussed, including diagnosis of neurodegenerative disease, myelin deficit, tumors of the central nervous system (with an emphasis on brain gliomas), and traumatic brain injuries. Fundamental and applied challenges in study of the THz-wave – brain tissue interactions, development of THz biomedical tools and systems for neurodiagnostics, as well as strategies to accelerate their future clinical applications are also overviewed.