Date Published: August 29, 2017
Publisher: Springer Berlin Heidelberg
Author(s): James A. Kimber, Sergei G. Kazarian.
Spectroscopic imaging of biomaterials and biological systems has received increased interest within the last decade because of its potential to aid in the detection of disease using biomaterials/biopsy samples and to probe the states of live cells in a label-free manner. The factors behind this increased attention include the availability of improved infrared microscopes and systems that do not require the use of a synchrotron as a light source, as well as the decreasing costs of these systems. This article highlights the current technical challenges and future directions of mid-infrared spectroscopic imaging within this field. Specifically, these are improvements in spatial resolution and spectral quality through the use of novel added lenses and computational algorithms, as well as quantum cascade laser imaging systems, which offer advantages over traditional Fourier transform infrared systems with respect to the speed of acquisition and field of view. Overcoming these challenges will push forward spectroscopic imaging as a viable tool for disease diagnostics and medical research.
As the understanding of disease progression improves, there is a need to put this knowledge to use in diagnostics, yet acquiring the necessary information may be beyond the scope of routine analytical approaches. The need for rapid and accurate disease diagnostics from biopsy samples, and the effect of newly developed formulations, such as drugs, on cells are of particular recent interest. Current approaches for biopsy sample analysis often rely on the use of staining, which adds a degree of subjectivity from histopathologists, and for studies of cells, molecular labelling can cause misleading results because of changes in the structure or affinity of the target molecule [1, 2]. Vibrational spectroscopic approaches such as Fourier transform infrared (FTIR) spectroscopy are label-free as their chemical specificity allows the differentiation of substances on the basis of chemical content and therefore allows the study of biomaterials and cells without interference from added chemical labelling species . As such, vibrational spectroscopy has attracted much interest for its applicability to study biomaterials and cells, where spectroscopic imaging can offer the possibility of more accurate and faster disease diagnostics, particularly for cancer detection [4–7]. In addition, FTIR spectroscopy can provide new insight into the differentiation of stages of disease, determine tumour margins and enhance understanding of chemical changes at an intracellular level, as well as provide information about the effects of drugs on cells. The wealth of information contained within spectra can also aid in machine learning and classification, leading to statistically rigorous, objective diagnostics [8–10]. This article reports the current state of the art and the challenges within the field of infrared spectral histopathology , focusing on specific trends and issues associated with a push for a higher spatial resolution and improving quality of spectral data for analysis of relevant samples.
Since the availability of focal plane array (FPA) detectors in the last two decades, FTIR spectroscopic imaging of biological tissues and cells has become feasible [11, 12]. Imaging has distinct advantages over point mapping, as acquiring an array of spectra over an area takes orders of magnitude less time, and thus also allows the study of dynamic processes. This is particularly advantageous for studies of live cell cultures as there are changes in the morphology and position of cells over time, and it also allows the mapping of large areas (i.e. several square millimetre) of biopsy samples at high resolution (approximately 10 μm) within reasonable timeframes [13, 14]. Infrared microscopes are needed to obtain high-resolution spectroscopic images, and biological samples and systems are typically measured either in transmission mode  or in attenuated total reflection (ATR) mode, both of which have distinct advantages and limitations . The quality of the data obtained can be assessed with three metrics: spatial resolution, signal-to-noise ratio (SNR) and presence of spectral artefacts (defined here as changes in absorbance or position of spectral bands due to non-chemical effects). The latter two significantly influence the limit of detection of particular chemical signatures, but the spatial resolution also plays a role as the influence of small regions of interest in the spectra obtained may be averaged out if insufficient spatial resolution is used. Furthermore, in the context of this article, SNR refers to the absorbance noise rather than that of transmittance, as it is the absorption spectrum which contains the spectral information of interest. As such, the absorbance SNR depends on the difference between transmittance spectra, which is in part controlled by sample thickness and not just detector sensitivity or source intensity . Therefore, we discuss here recent trends in achieving high spatial resolution in spectroscopic imaging as well as recent advances for improvement of quality of measured spectra.
Infrared spectroscopy based on quantum cascade lasers (QCLs) as a source of infrared radiation use discrete-frequency tuneable lasers which produce mid-infrared light several orders of magnitude brighter than that from a silicon carbide heating element or synchrotron. QCL spectrometers are also more compact and less sensitive to vibrations than benchtop FTIR spectrometers with moving mirror interferometers. The high spectral power density of QCLs is ideal for use in microscopes, where samples need to be illuminated by a concentrated beam, and allow thermoelectrically cooled bolometer array detectors (480 × 480 pixels with 17.5-μm2 pixel size) to be used, compared with liquid-nitrogen-cooled FPA detectors with 128 × 128 pixels and 40-μm2 pixel size that are used for FTIR imaging. These factors make QCLs attractive for spectroscopic imaging within clinical environments, and they have proven to be capable of measuring large areas of a single biopsy sample or tissue microarrays within a relatively short time. For example, spectroscopic imaging of a 1mm-diameter tissue sample would take between 5 and 6 h with a comparable FTIR imaging microscope system , whereas by use of the discrete nature of a QCL microscope system to measure only a small number of frequencies needed for disease classification (e.g. 31), a similar sample of the same diameter can be measured in less than 6 min. With an FTIR spectrometer, increasing the moving mirror speed, changing the interferogram sampling interval and lowering the spectral resolution are methods to decrease image acquisition time, but this still produces a large amount of data in the form of interferograms that must be recorded and processed. As such, the scope for data reduction without significant loss of classification performance when discrete-frequency QCL systems are used allows much larger spatial areas to be analysed than previously feasible because of fast data acquisition, facilitating larger high-throughput studies of multiple samples.
Advances in spectroscopic imaging of biological systems and samples and the focus on disease diagnostics make compelling cases for its use in spectral histopathology within a clinical environment and the study of drugs within cells. The wealth of information available in spectra and the extraction of key spectroscopic changes from samples in a non-destructive way allows classification between cells and disease states, and has the potential to enhance the sensitivity and specificity above those of currently used methods. The challenges of obtaining quality data from biopsy samples and cells in terms of good SNR, artefact-free spectra and high acquisition speed are being addressed in a variety of ways, from computational algorithms and novel optics to more powerful sources of infrared and the development of compact QCL-based imaging spectrometers. Combining these developments in spectroscopic imaging into the design of new microscopes and instruments, such as the use of a QCL source with a high-magnification microscope equipped with added lenses and/or an ATR crystal, would allow the acquisition of higher-resolution, higher-quality spectroscopic images than the current state of the art. This, along with larger data sets and applications of machine learning, brings closer the prospect of routine ‘digital histopathology’  and spectral cytopathology, and the formation of multitype relational data sets comprising data from a combination of approaches, such as FTIR, Raman and fluorescence imaging, would also significantly advance these areas of research. Nevertheless, spectrometers based on the recent advances discussed here are still some time away from being used in clinics, due in part to the need for validation of the practical approach and data processing algorithms. This is why research in this area going beyond the state of the art is needed, bringing its outcomes for the benefits of improved diagnostics and treatment in health care. In summary, label-free FTIR spectroscopic imaging and QCL-based imaging offer the possibility of addressing clinical challenges by providing enhanced disease diagnostics, particularly for cancer detection and analysis, as spectroscopy has a potential to provide new insight into the differentiation of stages of disease and enhance our understanding of chemical changes at an intracellular level.