The process of histopathology, comprising tissue staining and morphological pattern recognition, has remained largely unchanged for over 140 years. Although it is integral to clinical and research activities, histopathologic recognition remains a time-consuming, subjective process to which only limited statistical confidence can be assigned because of inherent operator variability. Although immunohistochemical approaches allow limited molecular detection, significant challenges remain in using them for quantitative, automated pathology. Vibrational spectroscopic approaches, by contrast, directly provide nonperturbing molecular descriptors, but a practical spectroscopic protocol for histopathology is lacking. Here we couple high-throughput Fourier transform infrared (FTIR) spectroscopic imaging of tissue microarrays with statistical pattern recognition of spectra indicative of endogenous molecular composition and demonstrate histopathologic characterization of prostatic tissue. This automated histologic segmentation is applied to routine archival tissue samples, incorporates well-defined tests of statistical significance and eliminates any requirement for dyes or molecular probes. Finally, we differentiate benign from malignant prostatic epithelium by spectroscopic analyses.

The recent development of Fourier transform infrared (FTIR) spectroscopic imaging has enhanced our capability to examine, on a microscopic scale, the spatial distribution of vibrational spectroscopic signatures of materials spanning the physical and biomedical disciplines. Recent activity in this emerging area has concentrated on instrumentation development, theoretical analyses to provide guidelines for imaging practice, novel data processing algorithms, and the introduction of the technique to new fields. To illustrate the impact and promise of this spectroscopic imaging methodology, we present fundamental principles of the technique in the context of FTIR spectroscopy and review new applications in various venues ranging from the physical chemistry of macromolecular systems to the detection of human disease.

Three different Raman microspectroscopic imaging methodologies using a single experimental configuration are compared; namely, point and line mapping, as representatives of serial imaging approaches, and direct or wide-field Raman imaging employing liquid-crystalline tunable filters are surveyed. Raman imaging data acquired with equivalent low-power 514.5-nm laser excitation and a cooled CCD camera are analyzed with respect to acquisition times, image quality, spatial resolution, intensity profiles along spatial coordinates, and spectral signal-to-noise ratios (SNRs). Point and line mapping techniques provide similar SNRs and reconstructed Raman images at spatial resolutions of approximately 1.1 microm. In contrast, higher spatial resolution is obtained by direct, global imaging (approximately 313 nm), allowing subtle morphological features on test samples to be resolved.

We characterize a visible reflectance hyperspectral imaging system for noninvasive, in vivo, quantitative analysis of human tissue in a clinical environment. The subject area is illuminated with a quartz-tungsten-halogen light source, and the reflected light is spectrally discriminated by a liquid crystal tunable filter (LCTF) and imaged onto a silicon charge-coupled device detector. The LCTF is continuously tunable within its useful visible spectral range (525-725 nm) with an average spectral full width at half-height bandwidth of 0.38 nm and an average transmittance of 10.0%. A standard resolution target placed 5.5 ft from the system results in a field of view with a 17-cm diameter and an optimal spatial resolution of 0.45 mm. The measured reflectance spectra are quantified in terms of apparent absorbance and formatted as a hyperspectral image cube. As a clinical example, we examine a model of vascular dysfunction involving both ischemia and reactive hyperemia during tissue reperfusion. In this model, spectral images, based upon oxyhemoglobin and deoxyhemoblobin signals in the 525-645-nm region, are deconvoluted using a multivariate least-squares regression analysis to visualize the spatial distribution of the percentages of oxyhemoglobin and deoxyhemoglobin in specific skin tissue areas.

The Raman and infrared spectra for the C 4 ,C S ,C 12 and C 16 carboxylic acid series were obtained at liquid nitrogen temperatures, in the 2900 cm ~ 1 spectral region and were compared to spectra for selected branched and straight chain alkane systems. For hexadecanoic acid-d J the Fermi resonance interaction between the CD J symmetric stretching mode and the overtone level of the CD 3 asymmetric deformation vibration was examined as a function of physical state. In this example of a negative Fermi resonance perturbation, in which the lower frequency overtone component is more intense than the higher frequency fundamental, the coupling increases as the system passes from either the matrix isolated species or solution phase to the polycrystalline state. The spectra of the carboxylic acid series, as well as the spectra of 2,2,4-trimethyl pentane and n -octane, suggest that in the 2900 cm ~ 1 region an analogous Fermi resonance coupling exists between the CH 3 symmetric stretching vibration at -2938 cm~l and the 2871 cm~l overtone transition of the CH J asymmetric deformation. Unfortunately, the 2930-2940 cm-1 region for the solid systems is further complicated by transitions attributed to crystal interactions. Although the Fermi interaction for the methyl group has been recognized in the simpler halomethane system, for example, it has not been previously discussed in the more general context of significantly larger branched and straight chain molecules. Comparisons of the carboxylic acid spectra in the 2900 cm~l region have also led to assignments for the a-CH 2 (adjacent to the carbonyl group) and the w-CH 2 (adjacent to the methyl group) C-H stretching modes.

Vibrational spectroscopy allows a visualization of tissue constituents based on intrinsic chemical composition and provides a potential route to obtaining diagnostic markers of diseases. Characterizations utilizing infrared vibrational spectroscopy, in particular, are conventionally low throughput in data acquisition, generally lacking in spatial resolution with the resulting data requiring intensive numerical computations to extract information. These factors impair the ability of infrared spectroscopic measurements to represent accurately the spatial heterogeneity in tissue, to incorporate robustly the diversity introduced by patient cohorts or preparative artifacts and to validate developed protocols in large population studies. In this manuscript, we demonstrate a combination of Fourier transform infrared (FTIR) spectroscopic imaging, tissue microarrays (TMAs) and fast numerical analysis as a paradigm for the rapid analysis, development and validation of high throughput spectroscopic characterization protocols. We provide an extended description of the data treatment algorithm and a discussion of various factors that may influence decision-making using this approach. Finally, a number of prostate tissue biopsies, arranged in an array modality, are employed to examine the efficacy of this approach in histologic recognition of epithelial cell polarization in patients displaying a variety of normal, malignant and hyperplastic conditions. An index of epithelial cell polarization, derived from a combined spectral and morphological analysis, is determined to be a potentially useful diagnostic marker.

Vibrational spectroscopy, commonly associated with IR absorption and Raman scattering, has provided a powerful approach for investigating interactions between biomolecules that make up cellular membranes. Because the IR and Raman signals arise from the intrinsic properties of these molecules, vibrational spectroscopy probes the delicate interactions that regulate biomembranes with minimal perturbation. Numerous innovative measurements, including nonlinear optical processes and confined bilayer assemblies, have provided new insights into membrane behavior. In this review, we highlight the use of vibrational spectroscopy to study lipid-lipid interactions. We also examine recent work in which vibrational measurements have been used to investigate the incorporation of peptides and proteins into lipid bilayers, and we discuss the interactions of small molecules and drugs with membrane structures. Emerging techniques and measurements on intact cellular membranes provide a prospective on the future of vibrational spectroscopic studies of biomembranes.

Vibrational Raman spectroscopic experiments have been performed as a function of temperature on aqueous dispersions of synthetic DL-erythro-N-lignoceroylsphingosylphosphocholine [C(24):SPM], a racemic mixture of two highly asymmetric hydrocarbon chain length sphingomyelins. Raman spectral peak-height intensity ratios of vibrational transitions in the C-H stretching-mode region show that the C(24):SPM-H2O system undergoes two thermal phase transitions centered at 48.5 and 54.5 degrees C. Vibrational data for fully hydrated C(24):SPM are compared to those of highly asymmetric phosphatidylcholine dispersions. The Raman data are consistent with the plausible model that the lower temperature transition can be ascribed to the conversion of a mixed interdigitated gel state (gel II) to a partially interdigitated gel state (gel I) and that the higher temperature transition corresponds to a gel I----liquid-crystalline phase transition. The observation of a mixed interdigitated gel state (gel II) at temperatures below 48.5 degrees C implies that biological membranes may have lipid domains in which some of the lipid hydrocarbon chains penetrate completely across the entire hydrocarbon width of the lipid bilayer.