Steven C. Gebhart

In vitro determination of normal and neoplastic human brain tissue optical properties using inverse adding-doubling

To complement a project towards the development of real-time optical biopsy for brain tissue discrimination and surgical resection guidance, the optical properties of various brain tissues were measured in vitro and correlated to features within clinical diffuse reflectance tissue spectra measured in vivo. Reflectance and transmission spectra of in vitro brain tissue samples were measured with a single-integrating-sphere spectrometer for wavelengths 400-1300 nm and converted to absorption and reduced scattering spectra using an inverse adding-doubling technique. Optical property spectra were classified as deriving from white matter, grey matter or glioma tissue according to histopathologic diagnosis, and mean absorption and reduced scattering spectra were calculated for the three tissue categories. Absolute reduced scattering and absorption values and their relative differences between histopathological groups agreed with previously reported results with the exception that absorption coefficients were often overestimated, most likely due to biologic variability or unaccounted light loss during reflectance/transmission measurement. Absorption spectra for the three tissue classes were dominated by haemoglobin absorption below 600 nm and water absorption above 900 nm and generally determined the shape of corresponding clinical diffuse reflectance spectra. Reduced scattering spectral shapes followed the power curve predicted by the Rayleigh limit of Mie scattering theory. While tissue absorption governed the shape of clinical diffuse reflectance spectra, reduced scattering determined their relative emission intensities between the three tissue categories.

Liquid-crystal tunable filter spectral imaging for brain tumor demarcation

Past studies have demonstrated that combined fluorescence and diffuse reflectance spectroscopy can successfully discriminate between normal, tumor core, and tumor margin tissues in the brain. To achieve efficient, real-time surgical resection guidance with optical biopsy, probe-based spectroscopy must be extended to spectral imaging to spatially demarcate the tumor margins. We describe the design and characterization of a combined fluorescence and diffuse reflectance imaging system that uses liquid-crystal tunable filter technology. Experiments were conducted to quantitatively determine the linearity, field of view, spatial and spectral resolution, and wavelength sensitivity of the imaging system. Spectral images were acquired from tissue phantoms, mouse brain in vitro, and human cortex in vivo for functional testing of the system. The spectral imaging system produces measured intensities that are linear with sample emission intensity and integration time and possesses a 1 in. (2.54 cm) field of view for a 7 in. (18 cm) object distance. The spectral resolution is linear with wavelength, and the spatial resolution is pixel-limited. The sensitivity spectra for the imaging system provide a guide for the distribution of total image integration time between wavelengths. Functional tests in vitro demonstrate the capability to spectrally discriminate between brain tissues based on exogenous fluorescence contrast or endogenous tissue composition. In vivo imaging captures adequate fluorescence and diffuse reflectance intensities within a clinically viable 2 min imaging time frame and demonstrates the importance of hemostasis to acquired signal strengths and imaging speed.

A probability-based spectroscopic diagnostic algorithm for simultaneous discrimination of brain tumor and tumor margins from normal brain tissue.

This paper reports the development of a probability-based spectroscopic diagnostic algorithm capable of simultaneously discriminating tumor core and tumor margins from normal human brain tissues. The algorithm uses a nonlinear method for feature extraction based on maximum representation and discrimination feature (MRDF) and a Bayesian method for classification based on sparse multinomial logistic regression (SMLR). Both the autofluorescence and the diffuse-reflectance spectra acquired in vivo from patients undergoing craniotomy or temporal lobectomy at the Vanderbilt University Medical Center were used to train and validate the algorithm. The classification accuracy was observed to be approximately 96%, 80%, and 97% for the tumor, tumor margin, and normal brain tissues, respectively, for the training data set and approximately 96%, 94%, and 100%, respectively, for the corresponding tissue types in an independent validation data set. The inherently multi-class nature of the algorithm facilitates a rapid and simultaneous classification of tissue spectra into various tissue categories without the need for a hierarchical multi-step binary classification scheme. Further, the probabilistic nature of the algorithm makes it possible to quantitatively assess the certainty of the classification and recheck the samples that are classified with higher relative uncertainty.

Comparison of spectral variation from spectroscopy to spectral imaging

Optical biopsy has been shown to discriminate between normal and diseased tissue with high sensitivity and specificity. Fiber-optic probe-based spectroscopy systems do not provide the necessary spatial information to guide therapy effectively, ultimately requiring a transition from probe-based spectroscopy to spectral imaging. The effect of such a transition on fluorescence and diffuse reflectance line shape is investigated. Inherent differences in spectral line shape between spectroscopy and imaging are characterized and many of these differences may be attributed to a shift in illumination-collection geometry between the two systems. Sensitivity of the line-shape disparity is characterized with respect to changes in sample absorption and scattering as well as to changes in various parameters of the fiber-optic probe design (e.g., fiber diameter, beam steering). Differences in spectral line shape are described in terms of the relative relationship between the light diffusion within the tissue and the distribution of source-detector separation distances for the probe-based and imaging illumination-collection geometries. Monte Carlo simulation is used to determine fiber configurations that minimize the line-shape disparity between the two systems. In conclusion, we predict that fiber-optic probe designs that mimic a spectral imaging geometry and spectral imaging systems designed to emulate a probe-based geometry will be difficult to implement, pointing toward a posteriori correction for illumination-collection geometry to reconcile imaging and probe-based spectral line shapes or independent evaluation of tissue discrimination accuracy for probe-based and spectral imaging systems.

Experimental and simulated angular profiles of fluorescence and diffuse reflectance emission from turbid media

Given the wavelength dependence of sample optical properties and the selective sampling of surface emission angles by noncontact imaging systems, differences in angular profiles due to excitation angle and optical properties can distort relative emission intensities acquired at different wavelengths. To investigate this potentiality, angular profiles of diffuse reflectance and fluorescence emission from turbid media were evaluated experimentally and by Monte Carlo simulation for a range of incident excitation angles and sample optical properties. For emission collected within the limits of a semi-infinite excitation region, normalized angular emission profiles are symmetric, roughly Lambertian, and only weakly dependent on sample optical properties for fluorescence at all excitation angles and for diffuse reflectance at small excitation angles relative to the surface normal. Fluorescence and diffuse reflectance within the emission plane orthogonal to the oblique component of the excitation also possess this symmetric form. Diffuse reflectance within the incidence plane is biased away from the excitation source for large excitation angles. The degree of bias depends on the scattering anisotropy and albedo of the sample and results from the correlation between photon directions upon entrance and emission. Given the strong dependence of the diffuse reflectance angular emission profile shape on incident excitation angle and sample optical properties, excitation and collection geometry has the potential to induce distortions within diffuse reflectance spectra unrelated to tissue characteristics.

Characterization of a Spectral Imaging System

Complete infiltrating brain tumor margin resection continually eludes neurosurgeons due to inherent limitations of current margin localization techniques. A need exists for an objective, on-site, real-time imaging system which can accurately localize brain tumor margins and therefore be used as a basis for image-guided surgery. Optical biopsy methods are a proven means for successful brain tissue discrimination, indicating promise for spectral imaging to fill such a need. Before testing spectral imaging for surgical guidance, various spectral imaging modalities must be systematically compared to determine the modality most conducive to the clinical setting. A liquid crystal tunable filter spectral imaging system was characterized for field of view, spatial and spectral resolution, and ability to retain spectral features acquired from a clinical single-pixel spectroscopy system. For a 35-mm diameter field of view, the system possessed a spatial resolution of 50 μm in both image dimensions and a spectral resolution which monotonically increased from 10 to 30 nm over the tuning range of the filter. Differences between imaging and single-pixel spectra for location and FWHM of fluorescence peaks from two fluorescent dye targets were summarily less than 3 nm. However, two remediable artifacts were introduced to imaging system spectra during spectral sensitivity correction.

Instrumentation considerations in spectral imaging for tissue demarcation: comparing three methods of spectral resolution

Multiple methodologies exist to implement spectral imaging for tissue demarcation and disease diagnosis. In this paper, benchtop acousto-optic tunable filter (AOTF), liquid-crystal tunable filter (LCTF) and Fourier interferometric spectral imaging systems were quantitatively compared in terms of imaging speed of soft tissue autofluorescence. Optical throughput, image signal-to-noise ratio (SNR), and collagen autofluorescence imaging in chicken breast were assessed. Within this comparison, the Fourier system possessed the largest optical throughput (~50%) relative to the tunable-filter imaging systems; however, its throughput advantage failed to correlate to improved image SNR over the LCTF system. Further, while the autofluorescence imaging capability of the Fourier system exceeded that of the LCTF system for comparable total image integration times, the LCTF is capable of producing equivalent autofluorescence SNR with superior SNR when interrogations at only a few wavelengths are required and the random access filter tuning of the LCTF can be exploited. Therefore, the simple, rugged design and random-access filter-tuning capability of LCTF-based spectral imaging makes it best-suited for clinical development of soft tissue autofluorescence imaging.

Dynamic, three-dimensional optical tracking of an ablative laser beam.

Surgical resection remains the treatment of choice for braintumors with infiltrating margins but is currently limited by visual discrimination between normal and neoplastic marginal tissues during surgery. Imaging modalities such as computed tomography, magnetic resonance, positron emission tomography, and optical techniques can accurately localize tumor margins. We believe coupling the fine resolution of current imaging techniques with the precise cutting of midinfrared lasers through image-guided neurosurgery can greatly enhance tumor margin resection. This paper describes a feasibility study designed to optically track in three-dimensional space the articulated arm delivery of a noncontact ablative laser beam. To enable optical tracking of the laser beam focus, infrared-emitting diodes (IREDs) were attached to a handpiece machined for the distal end of the articulated arm of a surgical carbon dioxide laser. Crosstalk between the ablative laser beam and the tracking diodes was measured. The geometry of the adapted laser handpiece was characterized to track an externally attached passive tip and the laser beam focus. Target localization accuracies were assessed for both instrument points-of-interest and the sources of tracking errors were investigated. Stray infrared laser light did not affect optical tracking accuracy. The mean target registration errors while optically tracking the laser handpiece with a passive tip and the laser beam focus were 1.31 ± 0.50 mm and 2.31 ± 0.92 mm , respectively, and were equivalent to the errors tracking a 24-IRED pen probe from Northern Digital in a side-by-side comparison. The majority of error during ablation tracking derived from registration accuracy between physical space and the defined space of the ablation phantom and from an inability to freehand align the laser focus with the target in a consistent manner. While their magnitudes depend on spatial details of the tracking setup (e.g., number and distribution of fiducial points, working distance from the camera, etc.), these errors are inherent to any freehand laser surgery.

Brain tumor demarcation with liquid-crystal tunable filter spectral imaging

Past studies have demonstrated that combined fluorescence and diffuse reflectance spectroscopy can successfully discriminate between normal, tumor core, and tumor margin tissues in the brain. To achieve efficient surgical resection guidance with optical biopsy, probe-based spectroscopy must be extended to spectral imaging to spatially demarcate the tumor margins. This paper describes the design and testing of a combined fluorescence and diffuse reflectance imaging system which uses liquid-crystal tunable filter technology. Experiments were conducted to quantitatively determine its linearity, field of view, spatial and spectral resolution, and wavelength sensitivity. For functional testing, spectral images were acquired from tissue phantoms, mouse brain in vitro, and rat brain cortex in vivo. The spectral imaging system is characterized by measured intensities which are linear with sample emission intensity and integration time, a one-inch field of view for a seven-inch object distance, spectral resolution which is linear with wavelength, spatial resolution which is pixel-limited, and sensitivity functions which provide a guide for the distribution of total image integration time between wavelengths. Functional testing demonstrated good spatial and spectral constrast between brain tissue types, the capability to acquire adequate fluorescence and diffuse reflectance intensities within a one-minute imaging timeframe, and the importance of hemostasis to acquired signal strengths and imaging speed.

Spectral Imaging for Brain Tumor Margin Demarcation

The transition from probe-based spectroscopy to spectral imaging induces lineshape changes in fluorescence and diffuse reflectance. Correcting these effects enables discrimination algorithms determined from probe-based spectroscopy to be applied for surgical guidance with spectral imaging.