SYSTEM AND METHOD OF DYNAMIC MICRO-OPTICAL COHERENCE TOMOGRAPHY FOR MAPPING CELLULAR FUNCTIONS
20230280271 · 2023-09-07
Assignee
Inventors
Cpc classification
G01N21/4795
PHYSICS
International classification
Abstract
An apparatus for obtaining image data and functional data from a biological sample, the apparatus including: an interferometer configured to acquire interferometric information at a plurality of time points along an imaging plane for which at least one axis of the plane is at least partially along a depth or axial dimension that is based on radiations provided from a reference interfered with by the biological sample; and a processor configured to receive the interferometric information from the interferometer and configured to: process the interferometric information to generate an image of the biological sample along the imaging plane; determine frequency information based on the plurality of time points of the interferometric information, the frequency information reflecting temporal modulations induced by dynamic functions of the biological sample; generate a spatial map of the frequency information, and the spatial map of the frequency information indicating the dynamic functions of the biological sample.
Claims
1. An apparatus for obtaining image data and functional data from a biological sample, the apparatus comprising: an interferometer configured to acquire interferometric information at a plurality of time points along an imaging plane for which at least one axis of the plane is at least partially along a depth or axial dimension that is based on radiations provided from a reference interfered with by the biological sample; and a processor configured to receive the interferometric information from the interferometer and configured to: process the interferometric information to generate an image of the biological sample along the imaging plane; determine frequency information based on the plurality of time points of the interferometric information, the frequency information reflecting temporal modulations induced by dynamic functions of the biological sample; generate a spatial map of the frequency information, and the spatial map of the frequency information indicating the dynamic functions of the biological sample.
2. The apparatus of claim 1, wherein the processor, when determining frequency information, is further configured to: identify temporal fluctuations in the interferometric information induced by the dynamic functions of the biological sample.
3. The apparatus of claim 2, wherein the processor, when identifying temporal fluctuations, is further configured to: conduct a power frequency analysis of the temporal fluctuations to identify fluctuations arising from intracellular motion of the biological sample.
4. The apparatus of claim 3, wherein the image comprises a plurality of subregions, and wherein the processor is further configured to individually perform the power frequency analysis on at least one of the plurality of subregions.
5. The apparatus of claim 4, wherein the interferometer is further configured to acquire the interferometric information at a plurality of time points during a longitudinal study of at least 24 hours; and wherein the processor is further configured to track dynamic functions of the biological sample during the longitudinal study.
6. The apparatus of claim 5, wherein a drug is applied to the biological sample during the longitudinal study.
7. (canceled)
8. The apparatus of claim 2, wherein the processor is further configured to determine the entropy of the frequency spectrum for quantifying the frequency content of signals.
9. The apparatus of claim 1, wherein the interferometer forms part of a .Math.OCT system.
10-11. (canceled)
12. The apparatus of claim 1, wherein the interferometric information comprises .Math.OCT frames, and wherein the processor, when processing the interferometric information to generate an image of the biological sample along the imaging plane, is further to: locally normalize and Gaussian filter the .Math.OCT frames to generate processed frames, compute an elastic unwarping transformation matrix for each of the processed frames, wherein the center frame is used as a reference, and apply the transformation matrices to the .Math.OCT frames.
13. The apparatus of claim 1, further comprising a galvanometer configured to scan at a scan rate; and wherein the interferometer, when acquiring interferometric information, is further configured to: acquire the interferometric information by repeatedly scanning the imaging beam laterally across a region of interest at a frequency set by the galvanometer scan rate.
14. The apparatus of claim 1, wherein the interferometer, when acquiring interferometric information, is further configured to: traverse the imaging beam across a lateral region of interest in a stepwise manner in a plurality of scans, and stop at equally-spaced positions in the lateral region during each scan of the plurality of scans to acquire a series of A-lines at a rate determined by an A-line rate.
15. The apparatus of claim 1, wherein the processor is further configured to: perform cross-correlation between at least two image frames to measure an amount of lateral shifting, and apply an image registration algorithm to the at least two image frames to correct for the lateral shifting.
16. The apparatus of claim 1, wherein the interferometer forms part of a .Math.OCT system, and wherein the processor, when processing the interferometric information to generate an image of the biological sample, is further configured to: correct for depth-dependent attenuation of .Math.OCT intensity in an axial direction.
17. The apparatus of claim 1, wherein the processor, when determining the frequency information, is further configured to: determine the frequency information on a subset of the plurality of time points.
18. The apparatus of claim 1, wherein the processor, when determining frequency information based on the plurality of time points, is further configured to: determine the frequency information by performing spectral analysis on the plurality of time points based on at least one of: a power spectrum, a standard deviation, a variance, or a Fourier entropy.
19. The apparatus of claim 1, wherein the processor, when generating an image of the biological sample, is further configured to: generate a super-resolution image of the biological sample.
20. The apparatus of claim 19, wherein the processor, when generating a super-resolution image of the biological sample, is further configured to: generate the super-resolution image of the biological sample based on performing Gauss-Newton non-linear least squares minimization with hybrid bidiagonalization regularization and whole-image affine transformation geometric warping.
21-24. (canceled)
25. The apparatus of claim 1, wherein the interferometer is further configured to acquire interferometric information comprising at least one of phase information or intensity information, and wherein the processor, when processing the interferometric information to generate an image of the biological sample, is further configured to: process the interferometric information to generate the image of the biological sample based on at least one of the phase information or the intensity information.
26. The apparatus of claim 1, wherein the processor, when determining frequency information, is further configured to: determine the frequency information based on providing a real-time estimate of the frequency information.
27. The apparatus of claim 26, wherein the processor, when providing a real-time estimate of the frequency information, is further configured to: provide a real-time estimate of the frequency information based on a running standard deviation or variance.
28. The apparatus of claim 1, wherein the interferometer is further configured to acquire interferometric information at a plurality of time points using a 2D scanning pattern, and wherein the processor, when processing the interferometric information to generate an image of the biological sample, is further configured to: process the interferometric information to generate a 3D image of the biological sample.
29. The apparatus of claim 1, wherein the interferometer is further configured to acquire interferometric information at a plurality of time points using a raster scanning pattern, and wherein the processor, when processing the interferometric information to generate an image of the biological sample, is further configured to: process the interferometric information to generate a 3D image of the biological sample.
30-58. (canceled)
Description
BRIEF DESCRIPTION OF THE DRAWINGS
[0022] Various objects, features, and advantages of the disclosed subject matter can be more fully appreciated with reference to the following detailed description of the disclosed subject matter when considered in connection with the following drawings, in which like reference numerals identify like elements.
[0023]
[0024]
[0025]
[0026]
[0027]
[0028]
[0029]
[0030]
[0031]
[0032]
[0033]
[0034]
[0035]
[0036]
[0037]
[0038]
[0039]
[0040]
[0041]
[0042]
DETAILED DESCRIPTION
[0043] Disclosed herein are embodiments of procedures that use high-resolution (e.g. 1-.Math.m axial resolution) micro-optical coherence tomography (.Math.OCT) to obtain cross-sectional images of intracellular dynamics with dramatically enhanced image contrast. Embodiments of these procedures, termed dynamic .Math.OCT (d-.Math.OCT), are accomplished by acquiring time series of .Math.OCT images and conducting power frequency analyses of the temporal fluctuations that arise from intracellular motion separately or individually within subregions of the sample, e.g. on a pixel-per-pixel or voxel-by-voxel basis. Results of d-.Math.OCT imaging of freshly excised human esophageal and cervical biopsy samples are disclosed herein. Depth-resolved d-.Math.OCT images of intact tissue show that intracellular dynamics provides a new contrast mechanism for .Math.OCT that highlights subcellular morphology and activity in epithelial surface maturation patterns. In another embodiment, d-.Math.OCT images are obtained by computing the changes in radiation phase at different time points at approximately the same location in the sample.
[0044] Despite significant advances in microscopy, many techniques currently in use obtain static images of cells. Imaging living cells opens up a new, functional dimension of evaluation, where organelle and intracellular molecular movements inform on pathophysiology in a manner that cannot be achieved by static snapshots in time. The use of fluorescence microscopy to track either fluorescently transfected organelles or actively perturbed microinjected fluorescent particles within cells are examples of recent techniques that probe intracellular dynamics. In addition to perturbing the biological system by adding exogenous labels, challenges related to the interrogation of a large ensemble of cells have hindered the scaling of these methodologies for studying intact tissue samples where heterogeneous cell populations exist.
[0045] Thus, in various embodiments the disclosed apparatus, systems, and methods provide improvements over current technology, including improvements in the fields of biological monitoring and diagnostics. The disclosed procedures facilitate monitoring of cellular/intracellular structures in live cells and tissues without having to introduce exogenous labels which, besides adding many practical hurdles, could perturb the cells or tissues and potentially alter the results.
[0046] Coherence-gating imaging techniques are microscopy approaches that overcome many of these limitations. Coherence-gating imaging techniques do not require the addition of labels, as they derive their image contrast from tissue reflectivity at refractive index interfaces within the sample. Yet, small variations in refractive indices within cells cause minute reflectivity changes that are difficult to ascertain in a single image; thus, images have relatively low contrast. By taking advantage of the differences in the motion of various subcellular compartments, temporo-spatial signal analysis can be used to significantly enhance intracellular visualization, for example in dynamic full-field optical coherence tomography (FFOCT). FFOCT employs a high-power microscope objective and low coherence optical interferometry to acquire high-resolution transverse (en face) images of natural tissue reflectance at a given depth. Through computation of the standard deviation or autocorrelation of time-dependent signals that arise from intracellular activity, en face dynamic FFOCT images can reveal clear intracellular features that may otherwise be obscured in static images.
[0047] While dynamic FFOCT can provide a new method for label-free cellular imaging, this technique is limited in that it acquires images in the transverse plane. It would be much more desirable to obtain depth-resolved, cross-sectional images, as many important tissues (e.g. epithelial tissues) mature vertically and aberrations in this maturation process are critical for disease understanding and diagnosis. Optical coherence tomography (OCT) is a 10-.Math.m-resolution coherence-gated imaging technique that obtains cross-sectional images of tissue. Similar to dynamic FFOCT, dynamic OCT measures changes in cross-sectional OCT images taken at the same location over time. Dynamic OCT has shown to inform on tissue viability, mucus viscosity, cell migration, and remodeling. Nonetheless, due to its relatively low spatial resolution, dynamic OCT cannot be used to see inside individual cells.
[0048] Disclosed herein is a new technology termed dynamic micro-optical coherence tomography (d-.Math.OCT). .Math.OCT is a very high-resolution form of OCT that has resolutions of approximately 2 × 2 .Math.m (lateral) × 1 .Math.m (axial) or alternatively approximately 4 × 4 .Math.m (lateral) × 2 .Math.m (axial). .Math.OCT has been shown to be capable of visualizing cross-sectional images of cells at an unprecedented level of detail. With d-.Math.OCT, new microscopic information and enhanced contrast emerge by conducting power frequency analysis of the temporal fluctuations that arise from intracellular motion on a pixel-per-pixel basis. Because d-.Math.OCT has a depth scan priority, it can uniquely probe subcellular dynamics over a very wide (0-100, 0-1000, 0-10000, 0-100000 Hz) frequency range. Results using d-.Math.OCT are demonstrated using biopsy samples, spheroids, excised mouse tumor tissue, etc. showing that it highlights depth-resolved, intracellular dynamics of intact tissues, a potentially impactful capability for the biomedical sciences and clinical diagnosis.
[0049] In various embodiments, the methods disclosed herein may be carried out using a .Math.OCT apparatus such as that shown in
[0050]
[0051] Thus, in certain embodiments there is provided an apparatus for obtaining image data and functional data from biological tissue. The apparatus may include an interferometer to acquire interferometric information along an imaging plane that is based on radiations provided from a reference interfered with by the biological tissue. The apparatus may also include a processor configured to receive the interferometric information from the interferometer. The processor may be configured to perform various steps, including processing the interferometric information to generate a morphological or phase image of the biological tissue along the imaging plane, determining frequency information from temporal modulations of the interferometric information induced by dynamic functions of the biological tissue, and generating a report or image that spatially maps the dynamic functions of the biological tissue. This report or image may be combined with the morphological image.
[0052] In various embodiments, the processor may be further configured to identify temporal fluctuations in the interferometric information induced by dynamic functions of the biological tissue. In other embodiments, the processor may be further configured to conduct a power frequency analysis of the temporal fluctuations to identify fluctuations arising from intracellular motion of the biological tissue. In still other embodiments, the processor may be further configured to perform the power frequency analysis on a pixel-by-pixel basis. In yet other embodiments, the processor may be further configured to track dynamic functions of the biological tissue across longitudinal studies. In still further embodiments, the longitudinal studies may include drug delivery to the biological tissue. In yet other embodiments, the drug may be a chemotherapeutic drug. In other embodiments, the interferometer may form part of a .Math.OCT system. In still other embodiments, the interferometric information may be resolved to at least 2 .Math.m by 2 .Math.m laterally or 1 .Math.m axially. In yet another embodiment, the interferometric information may be resolved to at least 4 .Math.m by 4 .Math.m laterally or 2 .Math.m axially (depth). In certain embodiments, the biological tissue may be in vivo and in other embodiments, the tissue may be ex vivo, or three-dimensional cell cultures in vitro. In some embodiments, processing the interferometric information to generate a morphological image of the biological tissue along the imaging plane may further include steps to compensate for motion artifacts, including locally normalizing and filtering the intensity or phase values of pixels in the .Math.OCT frames to generate processed frames, computing an unwarping transformation matrix (e.g. elastic unwarping) for each of the processed frames, with or without a reference frame, and applying the transformation matrices to the .Math.OCT frames.
PCA Dimension Reduction and Machine Learning
[0053] In various embodiments there is provided a method for obtaining image data and functional data from biological tissue. The method may include various steps including: acquiring interferometric information along an imaging plane that is based on radiations provided from a reference interfered with by the biological tissue, processing the interferometric information to generate a morphological image of the biological tissue along the imaging plane; determining frequency information from the temporal modulations of the interferometric information induced by dynamic functions of the biological tissue, and generating a report that spatially maps the dynamic functions of the biological tissue with the morphological image.
[0054] In various embodiments disclosed herein, spectral analysis (e.g., power spectrum, standard deviation, variance, Fourier entropy) of temporal fluctuations found in .Math.OCT videos was used to generate data that allows cross-sectional cellular dynamics in whole tissue to be studied. By appropriate binning of the spectral data to create pseudo-colored composite images, it is demonstrated that the presently-disclosed methods enhance our ability to delineate cellular/subcellular features in the cross-sectional imaging plane and characterize intracellular dynamics within tissues without any need for exogenous labeling. The results also emphasize the capabilities of d-.Math.OCT to interrogate cross-sectional subcellular structure and concomitant variations in cell dynamics and activity from the basal to superficial epithelial layers. The significance of this capability is high, as depth-dependent changes in cellular and architectural maturation patterns are critical for the diagnosis of many epithelial diseases, including dysplasia and cancer.
[0055] In certain embodiments, d-.Math.OCT is utilized in freshly excised biopsy samples that are kept alive in culture media. Images from these biopsies were difficult to register with histological images at the subcellular level and thus some of the interpretations of subcellular behavior are preliminary. There is also much to learn about the mechanistic origins of the cellular motions measured by this technique. In the literature, there is evidence that supports the notion that these dynamical fluctuations, while random, are distinctly different from thermal-driven Brownian motion. Instead, they are a consequence of an aggregate of ATP-dependent random forces that orchestrate cell motility, among other biomechanical effects. Additionally, an improved understanding of the relationship between intracellular dynamics and underlying physiological and pathobiological conditions still needs to be obtained. Numerous studies, mainly with in vitro models, have investigated the origins of intracellular dynamics. In some of these studies, dynamic signals have been associated with contractile protein filaments in organelle transport and cell motility, which have shown to be modulated by pharmaceutical agents. These results indicate that the physiological origins of the dynamical behavior being probed can be used to inform of pathologies and their response to treatment. Owing to its high resolution in all three dimensions, d-.Math.OCT is well suited to perform these investigations. Yet, larger studies of various tissue/cell types in different states with identification/modulation of specific intracellular molecular motion-dependent mechanisms are warranted to fully understand the wealth of information that d-.Math.OCT provides.
[0056] While intact tissue is an excellent substrate for d-.Math.OCT, embodiments of the disclosed procedures could also be of great use for additional assays including two- and three-dimensional cell culture, spheroids, organoids, and organs-on-chips, among others. The advantage of d-.Math.OCT as a cell viability assay would be the ability to determine the metabolic or pathobiologic state of cells in these platforms without destroying the sample for viability staining or cell-type characterization. Such an application of d-.Math.OCT could improve the efficiency of many multicellular assays being developed today.
[0057] Another application of d-.Math.OCT is its performance in vivo. .Math.OCT has now been demonstrated in living human patients in the nasal cavity, showing a unique capacity to interrogate ciliary and mucus dynamics. d-.Math.OCT will be more difficult to implement in vivo because of the need for patient stabilization at the subcellular level during an extended imaging period (seconds). Technologies such as tight coupling of the tissue with a .Math.OCT probe could overcome this problem, opening up a new label-free option for high cellular contrast and functionally informed optical biopsy.
[0058] Following is a description of materials and procedures for carrying out various embodiments along with description of exemplary results.
Tissue Preparation and Imaging
[0059] In certain embodiments, human upper gastrointestinal biopsies and cervical biopsies were obtained from study participants. Each of the biopsies was immersed in cell culture medium immediately after excision to preserve tissue viability. In preparation for d-.Math.OCT imaging, the biopsies were placed on a glass slide and, by means of a dual-axis goniometer, the luminal side of the tissue was tilted with respect to the imaging beam to avoid specular reflection. All imaging was performed at room temperature (25° C.) and samples were kept moist over the course of an imaging session by the addition of small amounts of cell culture medium on the tissue as necessary. Despite efforts to keep the sample well-moistened, image artifacts arising from moisture evaporation were occasionally noticeable in the real-time display of the images, especially for small biopsy samples. Therefore, certain samples were performed with the luminal side of the tissue placed in contact with a glass slide to minimize moisture loss during imaging. After d-.Math.OCT imaging, the imaged region was marked with a dye before being fixed in formalin. The samples were then processed to obtain 5 .Math.m thick hematoxylin and eosin (H&E) stained histology slides corresponding to the region imaged by d-.Math.OCT.
Embodiments Providing High Spatial and Temporal Resolution Dynamic .Math.OCT Imaging
[0060] In various embodiments herein are provided systems and methods that probe subcellular dynamics in a cross-sectional plane. Such a cross-sectional imaging system may have resolutions less than 2 .Math.m axial resolution and less than 5 .Math.m lateral resolution. The methods may involve acquiring multiple images, regions, lines, or points over time. A power spectral density measurement may be made for each pixel in the image. Exemplary embodiments include time frequency analysis (e.g. short-time Fourier transform, Wigner transform, Wavelet transform, etc.) or power spectrum estimation techniques (e.g. Welch’s method) at every pixel of the image, or other methods for computing frequency information over time locally. An alternative embodiment is to determine the entropy of the frequency spectrum or other methods for quantifying the frequency content of signals.
[0061] Two beam scanning embodiments are disclosed. One scheme enables the probing of frequency content up to half of the galvanometer mirror scan rate, in one embodiment the rate ranges from about 1-40 Hz. In another embodiment, the mirror stops at a single spot, acquires data at the Aline rate, in one case approximately 1-20 kHz, but alternatively could be approximately 1 kHz - 3 MHz.
Microscopy, Data Acquisition, and Data Processing
[0062] A benchtop .Math.OCT system microscope was used to provide cross-sectional images of intact tissues with a resolution of 2 × 2 .Math.m (lateral) × 1 .Math.m (axial) to a depth of 300 .Math.m. Imaging was performed using an A-line (depth-dependent reflectivity profile) rate of 20.48 kHz, and a 1 mm lateral scan was achieved by scanning the beam across the sample using a galvanometer mirror. A custom-written data acquisition program capable of real-time display of image contrast based on a pixel-by-pixel standard deviation approximation was used to facilitate the identification of viable portions of the tissue samples. Two different imaging schemes that probed the dynamics of distinct frequency ranges were utilized to yield d-.Math.OCT images. The first method, termed multi-scan (MS) d-.Math.OCT, included repeatedly scanning the imaging beam laterally across region of interest at a frequency set by the galvanometer scan rate over a time period. To probe frequencies up to 20 Hz, the imaging beam was scanned repeatedly across a 1 mm region of interest at 40 Hz (to avoid aliasing effects), typically over a duration of 25 s. This yielded 1000 cross-sectional images of the same location, each composed of 512 A-lines. In the second imaging scheme, termed single-scan (SS) d- .Math.OCT, the imaging beam traversed across a lateral range of interest in a stepwise fashion, stopping at equally-spaced positions each time to acquire a series of A-lines at a rate determined by the A-line rate. To match the imaged region of interest as the first scheme, 25 s (i.e. same amount of time as the previous scanning scheme) is required to acquire 1000 A-lines at 20.48 KHz at 512 discrete positions across the 1 mm range. SS d- .Math.OCT enables the probing of cellular dynamics greater than two orders higher than the first, albeit at a lower frequency resolution. A higher frequency resolution can be achieved with acquisitions of longer periods of temporal data. To probe frequencies up to 10.24 kHz, the imaging beam traversed across the same 1 mm lateral range in a stepwise fashion, stopping at 512 equally-spaced positions where 1000 A-lines were acquired at a rate of 20.48 kHz. Both scanning protocols took the same amount of time (25 s) to complete. In various embodiments, frequencies are probed in a range of 0 Hz to 50 Hz based on time points collected at each location or subregion (e.g. pixels and/or voxels) within the sample of up to 100 Hz, where the frequencies are generally probed at half the frequency of the collected data to avoid aliasing effects.
[0063] In both imaging schemes, the fluctuation of the .Math.OCT signal intensity as a function of time is analyzed at every pixel of a cross-sectional image. In yet another embodiment, neighboring pixels can be evaluated and averaged to reduce noise. With an assumption that the fluctuations are ergodic processes, by employing short-time Fourier transform (STFT) with Welch’s method, an estimated power spectrum for each pixel in the .Math.OCT dataset was obtained. In this approach, time dependent data was analyzed in shorter segments instead of in its entirety.
[0064] Even though the d-.Math.OCT imaging time was reasonably short (25 s), images of tissue that underwent the first beam scanning scheme suffered nonlinear and spatially dependent motion drift on the order of tens of micrometers. Tissue motion was likely due to evaporation and/or thermal expansion, causing tissue settling during imaging. To compensate for these motion artifacts, .Math.OCT frames were locally normalized and Gaussian filtered. An elastic unwarping transformation matrix was computed from these processed frames using the center frame (500) as the reference. These transformation matrices were then applied to the original .Math.OCT data. Subsequently, with the assumption that the fluctuations are ergodic processes, an estimated power spectrum for each pixel in the .Math.OCT dataset was obtained by employing a short-time Fourier transform with Welch’s method. With this approach, time-dependent data was analyzed in shorter segments instead of in its entirety. Multiple segments, each a fifth in length (L = 200) as compared to the full temporal data and each overlapped by 50% with the next, were processed to derive multiple modified periodograms, I.sub.k. For a given length of recorded data, the length of each segment and the amount of overlap determined the resultant spectral resolution and variance of the estimated power spectral density,
[0065] Based on these data acquisition parameters, the estimated power spectrum was computed from nine averages, thus effectively reducing the variance in the estimated power spectrum by that equivalent factor.
[0066] In both imaging schemes, similar processing methods were applied to the resultant time-dependent data to obtain 256 equally-spaced frequency bins that ranged from either 0-20 Hz or 0-10.24 kHz. The latter imaging scheme enabled the probing of cellular dynamics two orders of magnitude higher than the former, albeit at a lower frequency resolution.
[0067] To enable direct visualization of the distribution of the dominant fluctuation frequencies within the tissue, P̂ was further binned into three frequency ranges and was each assigned a color channel within an RGB image (e.g. red: 0-0.08 Hz, blue: 0.55-0.78 Hz, green: 4.00-20.00 Hz) to create a final d-.Math.OCT image representation. In certain embodiments, the frequency range color assignment was selected empirically to accentuate subcellular and cellular features of the tissue and, thus varied between samples. All .Math.OCT and d-.Math.OCT images presented here were scaled to achieve an isotropic pixel aspect ratio, assuming a tissue refractive index of 1.4, to ensure accurate representation and to facilitate comparison with corresponding histological images. All image reformatting and analysis were conducted using imageJ and Matlab (Mathworks Inc).
[0068] In various embodiments, the fluctuation frequency information can be binned or divided into any number of subgroups (e.g. from 2 subgroups up to 10 or more subgroups) and each of the subgroups may be displayed differently on a map of the tissue (e.g. as shown in
[0069] The identities and maximum number of subgroups of frequency ranges or peaks and the most meaningful frequency ranges or peaks can be determined using dimension reduction techniques such as principal component analysis (PCA), kernel PCA, linear or generalized discriminant analysis, UMAP, t-SNE, or the like. Principal component images, corresponding to the principal components that describe most (e.g. 80% or 90%) of the variance in the power spectrum for example, can correspond to different frequency patterns that indicate different regimes of motion in the image. As with the frequency bins described above for the R, G, B channels of a color image, PCA principal component weight images can also be used to create a multicolor image. Similar approaches may be used for other dimension reduction methods to create multidimensional images.
Image Registration
[0070] As spectral analysis for d-.Math.OCT processing is done on a pixel-by-pixel basis, bulk motion or motion not related to intracellular and membrane fluctuation motion that occurred during image acquisition should be corrected. An image registration algorithm was applied to correct for in-plane bulk motion artifacts as necessary. Two-dimensional cross-correlation was performed between or among successive image frames to measure the amount of lateral shifts required to register the images. After conducting bulk motion correction, localized unwarping with affine transformation can refine the motion correction further.
Attenuation Correction
[0071] Depth-dependent attenuation of .Math.OCT intensity occurs along the optical axis, causing information below the surface to gradually drop to noise level. An algorithm was developed to perform attenuation correction and amplify .Math.OCT signals along depth during image post-processing. Algorithms for such depth dependent correction include exponential fitting, reslicing->image normalization or equalization->reslicing, or time-gain or depth-gain adaptive compensation.
[0072] Image registration and attenuation correction were employed as necessary before the power spectrum at each pixel of the cross-sectional image was calculated.
[0073] Spectral information is binned into multiple frequency ranges and the result is color-coded to create a pseudo-color composite. This facilitates visualization of cellular structures that were not discernible in static .Math.OCT images.
[0074] Image registration is another embodiment added to the method to improve quality of results that corrects for motion of the specimen during the acquisition sequence.
[0075] Attenuation-correction of the .Math.OCT signal along the optical axis is a further embodiment that improves the quality of the results by correcting for intensity fluctuations that are due to overlying optical properties.
[0076] During data acquisition, real-time running standard deviation estimation was used to identify viable regions that exhibit variation in cell dynamics in samples.
Report Generation
[0077] After a morphological image and/or a frequency map based on a morphological image is generated, a report may be generated and transmitted based on one or more of the image, the map, a diagnosis based on the image or map, or a characteristic that is determined based on the image, map, or diagnosis. The report may include image data such as the morphological image or frequency map and also may include information regarding the diagnosis or characteristic that is determined based on the image, map, or diagnosis. The report may be transmitted in various forms including electronically and/or in a hard (e.g. paper) copy. The report may be transmitted to a clinician (e.g. a physician, nurse, technician, or other medical professional), a researcher, a patient, or an entity such as a healthcare provider.
[0078] Functional imaging with d-.Math.OCT was performed on biopsies obtained from subjects who were diagnosed with various esophageal disorders.
[0079]
[0080] The ability of d-.Math.OCT to accentuate cellular structures that were otherwise not visible with .Math.OCT is further illustrated in an example of a biopsy taken from the gastroesophageal junction of a human subject (
[0081] d-.Math.OCT provided additional findings when imaging cervical squamous epithelium (
EXAMPLES
[0082] Following are non-limiting examples of procedures that may be performed using one or more of the disclosed apparatus, methods, or systems:
Imaging and Identification of Cells
[0083] d-.Math.OCT imaging of freshly excised human tissues may be conducted and registered to histologic slides on a per-cell basis. Available d-.Math.OCT machine learning algorithms may be trained/validated to detect distinct human cell and tissue types, using corresponding H&E and immunohistochemistry (IHC) as ground truth.
[0084] The capability of d-.Math.OCT to differentiate live/apoptotic/dead cells in 3D cultures may be accomplished by treating spheroids from multiple different human and murine melanoma cell lines. d-.Math.OCT may be conducted on control and treated spheroids and results compared to gold standard, fluorescence-based analysis of cell death with Hoechst-33342, annexin V-FITC, and propidium iodide (PI) staining. In such cases, we have found that live spheroid tumor cells exhibit medium frequency motion content, apoptotic cells exhibit high frequency motion content, and dead cells exhibit low or zero frequency motion signatures.
[0085] Dynamic .Math.OCT imaging may be performed on freshly excised, implantable microdevice-treated human tumor mouse models and compared to a panel of established IHC markers associated with drug efficacy. The data will define drug- and tumor-dependent d-.Math.OCT signatures associated with apoptosis and cell death in cancerous tissues.
[0086] Human tumor mouse model tissue may be modified by well-characterized pharmacological inhibitors of molecular processes related to cell proliferation (cytoskeletal, metabolic, growth). The change in the cell/tissue frequency content as a function of drug mode of action informs on the mechanisms that underlie the d-.Math.OCT signal.
Image Quality Improvements and Characterization of Intracellular Motion
[0087] Imaging system modifications, including the use of a broader bandwidth light sources, larger numerical aperture (NA), and extended depth of focusing optics may increase spatial resolution two-fold along all dimensions. Subpixel super-resolution and temporal extrapolation algorithms may be used to further increase spatial and frequency resolutions.
[0088] Time-frequency analysis and machine learning algorithms may be used to mine the d-.Math.OCT data more extensively to discover additional biomedically-relevant information in the motility signal.
[0089] Stabilization, out-of-plane 3D scanning, and high-speed, phase-sensitive d-.Math.OCT methods may be developed to reduce motion artifacts encountered when imaging living systems. Two probes (handheld and endoscopic) may be created for d-.Math.OCT imaging in patients; feasibility of using these devices in vivo may be tested in skin and upper gastrointestinal tract clinical pilot studies.
[0090] In various embodiments, dynamic .Math.OCT was used to measure intracellular motion within cells in cross-sectional images of freshly excised human tissues. With dynamic .Math.OCT (d-.Math.OCT), the pixel-by-pixel temporal power spectrum of successively acquired .Math.OCT images was computed and different frequency bands were encoded in distinct RGB color channels (
[0091] Embodiments of the disclosed procedures may be used to establish a label-free, depth-resolved, microscopic imaging method for phenotyping cells in tissues based on intracellular motility. This has been accomplished by growing an understanding of the biological and clinical relevance of d-.Math.OCT and by developing and validating advanced new d-.Math.OCT technology that dramatically increase its capabilities. The potential impact of d-.Math.OCT on a variety of biomedical fields is described below.
[0092] Biomedical imaging. .Math.OCT is a rapidly growing technology that is well positioned to redefine the field of in vivo microscopy due to its capacity to obtain 1-.Math.m-resolution, cross-sectional images of tissue and its relatively straightforward implementation in living patients. .Math.OCT has now been implemented using small diameter probes in vivo, different technology implementations are emerging, and a variety of organ systems/diseases are under investigation. Tissue dynamics phenotyping is also expanding, in conjunction with an increasing need for label-free cell assessment for drug development and individualized therapeutic assays. By merging .Math.OCT with intracellular motility phenotyping, d-.Math.OCT will accelerate both fields, while delivering a powerful new biomedical imaging capability with significant impact.
[0093] In vivo microscopy. d-.Math.OCT is well situated to overcome many of the limitations of tissue excision and histopathology. It has cellular resolution, enhanced contrast, and cross-sectional imaging that are close to that of the H&E gold standard (
[0094] 3D cell models. New technologies that grow a patient’s cells into 3D cell models (organoids, spheroids, organ on a chip, etc.) are playing an increasing role in preclinical drug screening, as they better recapitulate the cellular heterogeneity and function of the original tissue than 2D culture. Commercial assays commonly require the cells to be killed, stained for viability, and counted. As with other high-resolution optical metabolic imaging technologies, d-.Math.OCT may be used to quantify viability, cell type, and cell state longitudinally, without killing the cells. d-.Math.OCT also has a large field of view and 1-.Math.m depth resolution, features that may enable accurate and efficient 3D cellular imaging in organotypic culture platforms.
[0095] Individualized therapy. Many different classes of drugs with vastly different mechanisms of action are available to treat patients or are under development. Yet, existing approaches do not provide a good method to measure a drug’s efficacy in the living tissue in terms of its specific mechanism of action (e.g. cell viability, the activation of immune cells, a change in metabolic rate, or cell division and proliferation). Some clinical trials perform repeat biopsies to obtain such information but are restricted to single time points with a static readout. d-.Math.OCT, conducted in vivo, may be used for optimizing individualized therapy, to enable diseased tissue to be monitored continuously, providing an earlier and more direct means for determining whether the compound is having the desired effect.
[0096] Implantable microdevices (IMD). The IMD is a new technology for monitoring the effects of multiple therapies simultaneously in patients. The device includes separate wells, each containing a different drug/cocktail. After implantation and incubation in a patient’s tumor, the devices and adjacent tissue are surgically removed and interrogated with IHC histopathology to quantify necrosis. IMDs may instead be regularly interrogated by d-.Math.OCT in vivo, affording critical cross-sectional imaging information with a functional microscopic readout. As with systemic therapies, real-time continuous monitoring of drug effects significantly increases the impact of this technology versus currently available single time point methods.
[0097] .Math.OCT is enabled by the combination of a broad bandwidth supercontinuum source, common path interferometry, and unique extended depth of focus imaging. d-.Math.OCT is a novel extension of conventional .Math.OCT, which imparts unprecedented cellular, cross-sectional, label-free images with more meaningful contrast.
[0098] Biological significance of the dynamic signal: A key innovation in this disclosure is the creation of a d-.Math.OCT human cell and tissue atlas, matched to histology (IHC) that spans a variety of cells, tissue types, and pathologies. Analysis of this dataset may determine d-.Math.OCT’s diagnostic accuracy for many important diseases. Studies done in spheroids and human mouse tumor models may inform on d-.Math.OCT’s capability to quantitate response to therapy and characterize different mechanisms that contribute to intracellular motility. Taken together, this set of experiments may significantly advance the boundaries of knowledge in the field of intracellular motility imaging.
[0099] Extending d-.Math.OCT’s capabilities: New techniques for increasing spatial and temporal resolution have been developed and validated, extracting maximal information from the intracellular motility signal, and imaging in vivo. These cutting-edge technologies will make d-.Math.OCT far more capable and broadly applicable than it is today. These innovative developments are described in detail below.
[0100] d-.Math.OCT enables high contrast, subcellular, cross-sectional motility imaging of intact human and animal tissues ex vivo. d-.Math.OCT data (
[0101] 1. d-.Math.OCT provides cross-sectional, subcellular images of motility with increased cellular contrast compared to standard .Math.OCT for a wide variety of tissues (
[0102] 2. Intracellular features can be visualized based on modulation frequency discrimination (
[0103] 3. Spatiotemporal information can be used to encode cell type and/or state (
[0104] 4. The cross-sectional imaging and frequency encoding capability of d-.Math.OCT enables the observation of the functional maturation of epithelial cells from the basal layer to the surface (
[0105] 5. d-.Math.OCT enables the interrogation heretofore unrecognized high-frequency content in the kHz range (
[0106] 6. d-.Math.OCT shows a distinct change in intracellular modulation indicative of cell viability after application of a therapeutic agent in spheroids (
[0107] 7. d-.Math.OCT highlights inflammation influx (
[0108] 8. d-.Math.OCT enables high-contrast, functional imaging in human skin in vivo (
[0109] Determining the biological significance of d-.Math.OCT in cells and tissue.
[0110] Determining the accuracy of d-.Math.OCT for discriminating clinically relevant human cell/tissue types.
[0111] Initial data (
[0112] Type and number of specimens. Tissues from biopsies and human surgical specimens of normal, dysplastic, and malignant epithelia, solid tumors, and lymph nodes (Table 1). These tissue types may be used based on their availability, clinical relevance, and diversity of relevant cell types. Sample sizes of various numbers (e.g. n = 10 specimens per tissue type) may be used from different patients. Approximately 10 distinct d-.Math.OCT sections per specimen may be imaged, processed, and analyzed.
TABLE-US-00001 Skin Exocervlx Normal Normal Benign Pigmented Lesions LSIL Dysplastic Lesions HSIL Melanoma Basal Cell Carcinoma Squamous Carcinoma GI tract Lung Normal Esophagus Normal Barrett’s Esophagus Squamous Cell Cancer Dysplastic BE Adenocarcinoma Esophageal Adempcarcinoma Large Cell Carcinoma Normal Stomach Normal Duodenum Normal Colon Adenomatous Polyps Colon Carcinoma Breast Normal DCIS Ductal Carcinoma LCIS Lobular Carcinoma Lymph Nodes
[0113] d-.Math.OCT imaging. Imaging may be conducted initially with conventional bench top d-.Math.OCT, as disclosed above, although advanced d-.Math.OCT technology may be incorporated. Imaging may be performed on fresh tissues, less than one hour after resection. Specimens may be cut into 200 .Math.m-thick sections, placed in a specimen holder that facilitates diffusion through the cut surfaces, and immersed in culture media. Tissues may be imaged at room temperature or at 37° C., with the epithelial surface facing upwards (when relevant) to facilitate imaging. After imaging, the location of the d-.Math.OCT cross-section may be marked on the tissue via fine laser cautery on the top and bottom of the sample. Laser cautery may be overmarked with a fluorescent and/or absorbing ink that may be visible by histology.
[0114] Histological processing. Tissue may be formalin fixed and paraffin embedded (FFPE). 3D serial sections, separated by 4 .Math.m, may be cut. Virtual H&E (VHE) stains may be computed from autofluorescence images input into a convolutional neural network (CNN) trained on a generative adversarial network (GAN). Sections may then undergo DAPI staining (to label nuclei) and tissue-based cyclic immunofluorescence (t-CyCIF). Cell types of interest may be identified by specific IHC markers, examples of which are in Table 2. 3D datasets may be registered, warped, and reformatted so that top and bottom marks are in matching d-.Math.OCT and VHE/IHC images. Additional dataset warping may be conducted so that the same cells are seen in both the VHE/IHC and d-.Math.OCT images.
TABLE-US-00002 Examples of cell types and IHC markers used in studies T cells CD3+ B cells CD79a CD20 NK cells CD3-, CD56+ Monocytes CD11b+ Macrophages CD68/F480 Neutrophils CD11b+, Ly6G+ Fibroblasts Virnentin+ Epithelial cells BER-EP4, pan-cylokeratin, EpCAM Malignancy Epcam, p53, panCK (organ & model specific: e.g. PyMT. CD45)
[0115] Cell type analysis. d-.Math.OCT periodogram images may be computed using standard procedures or by more advanced algorithms and input into custom CellProfiler machine learning pipelines. Training and refinement may generate rules via CellProfiler Analyst, using t-CyCIF and VHE-determined cell type as ground truth. Accuracy of classification may be conducted prospectively.
[0116] Tissue diagnosis analysis. Blinded pathologists may render consensus diagnoses of d-.Math.OCT sections, based on 1) d-.Math.OCT images and 2) d-.Math.OCT images with overlaid cell type classification. Sensitivity/specificity for each tissue/disease type may be measured using VHE diagnosis as the gold standard.
[0117] Statistical rationale for number of specimens. Given 100 images per tissue type and a minimum of 100 cells per image, it is anticipated that there may be more than 10,000 cells per tissue type identified by d-.Math.OCT. Considering a least-abundant class of 5%, we may have a minimum of 500 positive cells to train, refine, and validate. Anticipating a sensitivity and specificity of 90%, and assuming 1:1:1 distribution of samples for training, refinement, and validation, 95% confidence intervals of ±5% for sensitivity and ±1% for specificity may be attained.
[0118] Additional approaches - image registration. Additional methods to improve dataset registration may be implemented, including the use of high precision tape sections and 3D OCT or Micro-CT imaging of the PPFE block to provide 3D warping boundary conditions. Frequency content may vary with cell state. Since d-.Math.OCT provides a measure of cell activity, there may be multiple distinct d-.Math.OCT periodograms that correspond to the same cell type by IHC.
[0119] Number of samples is insufficient for rare cell types. If the number of samples is insufficient for proper analysis of rare cell types, one may enrich by selecting alternative tissues (e.g. tonsils for lymphoid cells), increasing the number of samples, and/or utilizing other techniques such as k-fold cross validation.
[0120] Virtual H&E. VHE generated through autofluorescence is used here to ensure that all slides and images are from the same plane in a manner that does not interfere with antigen retrieval. Should VHE be inadequate for pathologic diagnosis, VHE could be generated using other fluorescence agents (e.g. DAPI and eosin).
[0121] Use of d-.Math.OCT to discriminate live/apoptotic/dead cells in spheroids.
[0122] Methods which are capable of rapidly detecting live/apoptotic/dead cells in 3D culture models without altering the specimen are expected to increase efficiency in this growing area of biomedical research. Preliminary data (
[0123] Type and number of specimens. Cell types may be utilized that are representative of those used in 3D culture research to ensure that the results are consistent across a spectrum of biological and pharmacological conditions. Initial studies may use murine melanoma cells (e.g. B16.F10). Subsequent experiments may be conducted using differentiated (e.g. A375) and de-differentiated (e.g. RPMI7951) human melanoma cells.
[0124] Spheroid growth and treatment. Details on the generation of spheroids is described in detail in the literature. Briefly, tumor cells (1 × 10.sup.6) may be seeded in a 10-cm ultra-low attachment (ULA) plate to promote spheroid formation. After 24-48 hours, S2 (40-100 .Math.m) spheroid fractions may be pelleted and resuspended in type I rat tail collagen (Corning) at a concentration of 2.5 mg/mL following the addition of 10× PBS with phenol red with pH adjusted using NaOH. The spheroid-collagen mixture may then be injected into the center gel region of the 3D microfluidic culture device (DAX-1 Chip AIM Biotech). After 30 minutes at 37° C., collagen hydrogels containing tumor spheroids may be hydrated with media (10% FBS/DMEM with 1% penicillin-streptomycin) with drugs that induce cell death (e.g. staurosporine STS 1-5 .Math.M) or vehicle control (0.1% DMSO) for 24-48 hours.
[0125] d-.Math.OCT imaging. The entire culture chip may be imaged by d-.Math.OCT in three-dimensions, noting the spatial location of each spheroid for future comparison to confocal or two-photon microscopy on a per-spheroid basis.
[0126] Assay: After d-.Math.OCT imaging, spheroids may be processed for on-chip live/apoptotic/dead staining with Hoechst/PI +/- annexin V-FITC staining. Dual labeling may be performed by loading microfluidic device with Hoechst (Ho) and propidium iodide (PI) from Nexcelom, as previously described.
[0127] Confocal microscopy. Following staining, spheroids may undergo confocal microscopy. Live cells stain with Hoechst-33342, but are negative for surface staining of annexin V, and have intact cell membranes and therefore exclude PI. Early apoptotic cells acquire annexin V positivity but exclude PI. Dead cells (late apoptosis or necrosis) stain for PI with or without annexin V staining.
[0128] Analysis. d-.Math.OCT images may be registered one-to-one with confocal or two-photon microscopy images of each cell. d-.Math.OCT periodogram images may trained and refined via a custom CellProfiler Analyst pipeline, using confocal-determined cell state as ground truth. Agreement of cell state type (live/apoptotic/dead) may be determined prospectively and quantified using Cohen’s kappa statistic.
[0129] Additional approaches - cell type diversity. To confirm accuracy of d-.Math.OCT imaging regardless of growth pattern, multiple human and murine cell types may be evaluated.
[0130] Alternative agents. Initial studies may use STS, a well-established pan-protein kinase C (PKC) inhibitor that induces caspase-dependent and -independent cell death. To evaluate more physiologically or clinically relevant stimuli, spheroids may be treated with exogenous inflammatory cytokines (e.g. TNFa +/- IFNg) to mimic an antitumor immune response and/or small molecules targeting the MAP kinase (MAPK pathway).
[0131] Determine the capacity of d-.Math.OCT to detect tumor response to anti-cancer drugs in tissue.
[0132] Rapidly identifying whether a patient is responding to a systemic therapy is an unmet clinical need in oncology. Preliminary data (
[0133] Approach for in vivo drug response measurements. To expose live tumors to the chemotherapeutic drugs, one may utilize IMDs that are placed directly into the tumor, releasing microdoses of various drugs into confined regions of tumor in well-defined concentration gradients. The drug delivery microdevice is advantageous because it creates highly controllable, confined regions of drug exposure in the tumor. The devices also allow measurement of a large set of drug perturbations per mouse (each device has up to 20 reservoirs). Implant microdevices may be loaded with reservoirs of chemotherapeutic agents such as Doxorubicin, Cisplatin, Olaparib and Topotecan directly into tumors of exemplary sizes of 300-400 mm.sup.3. Intratumoral delivery may be performed at four local doses, corresponding to the IC50, ⅒th IC50, ⅓rd IC50, and 3× IC50. Previous measurements with this system have shown apoptosis induction to be most prevalent between 6 h-48 h of tissue exposure. To capture the various stages of therapeutic activity and account for potentially differential apoptosis induction kinetics among the set of drugs, drug delivery IMDs may remain implanted in the tumors for three time points: for example, 6 h, 24 h and 48 h. For each time point, replicate measurements may be performed for each of the two tumor models. Each of the four drugs may be present on every microdevice at the four concentrations described above (each microdevice may test these 16 conditions in loaded reservoirs, plus we have 4 control/vehicle reservoirs). Using n=8 replicate samples across 3 time points in the two tumor models may require a total of 48 tumors/mice. At the time of tissue harvest, the tumors may be explanted from the animal and immediately transferred to d-.Math.OCT imaging. Unused portions of the tumors may be reserved.
[0134] d-.Math.OCT imaging and histological processing. d-.Math.OCT imaging and marking may be conducted on excised tumors as described above. Following d-.Math.OCT, tumor samples may be FFPE processed and sectioned with care to ensure cellular level registration. Slides may undergo multiplexed IHC staining with established markers for apoptosis and cell death (e.g., cleaved caspase-3 and -8, cleaved PARP, ph-g-H2AX, and ph-S6). VHE may be generated.
[0135] Analysis. For each drug, one may score its anti-tumor effect based on the percentage of positively stained cells for each IHC marker (e.g. the apoptotic index = [#cleaved caspase-3+ cells] / [total #cells] in a given ROI). Multiple markers associated with cell death may be used because cleaved caspase-3 is expressed transiently throughout the cellular apoptotic process. d-.Math.OCT periodogram images may be computed as described above. d-.Math.OCT periodograms may be input into software algorithms such as CellProfiler machine learning pipelines, and classification rules generated via software such as CellProfiler Analyst, using IHC indices of apoptosis or necrosis as ground truth.
[0136] Statistical rationale for number of animals. Given the observed variabilities from tumor responses to the drugs being used in this study and the differences in sensitivity between the two cell lines in-vitro, n=8 replicates per measurement may provide statistically significant differences with p<0.05 and a value of α=0.05 in measured drug responses by ground truth IHC. Assuming that d-.Math.OCT has a diagnostic sensitivity/specificity here of 90%, pooled across all time points, 24 animals per tumor model may provide 95% confidence intervals of ± 12%.
[0137] Sex and other relevant biological variables. Female mice may be used, as these are ovarian tumors.
[0138] Additional approaches - alternative tumor models. In addition to SKOV-3 and OVCAR-3, to extend the breadth of cancer types, this study will be conducted with other human tumor models with different responses to commonly used chemotherapeutics, such as A375, BT474 and PC3.
[0139] Determine how the d-.Math.OCT signal changes with cellular activity, proliferation, and metabolism.
[0140] One hallmark of cancer cells is an increased rate of proliferation relative to the host tissue. The proliferative rate of malignant cells has also been shown to be a prognostic marker of tumor aggressiveness and survival as well as a limited predictor of drug response to cytotoxic chemotherapy. A hypothesis, supported by prior coherence-based motility studies, is that proliferation, effectuated by increased metabolism, organelle transport, and intracellular cytoskeletal reorganization may be reflected in the d-.Math.OCT signal. To test this hypothesis, well-characterized pharmacological inhibitors (Table 3) of specific molecular processes related to proliferation may be used to address the contribution of each cellular process on the d-.Math.OCT signature. This has high translational relevance; if d-.Math.OCT could determine a tumor’s proliferative activity, it would enable therapeutic efficacy to be assessed less invasively and much earlier than currently feasible.
[0141] Tumor cells are the ideal model system for this study, as they are highly proliferative, motile, and have significantly elevated metabolism. The SKOV-3 ovarian cancer human tumor model is one example of an appropriate system. This model reliably yields solid, mostly non-necrotic tumors in mice over the course of 3 weeks. It is highly proliferative, with a cellular doubling time of ~24-36 h, rendering it susceptible to the agents in Table 3. Specimens will be from excess tumor tissue generated.
TABLE-US-00003 Overview of targeted drug treatments MECHANISM DRUG NAME TARGETED EFFECT CYTOSKELETAL EFFECTS Cylochalasin D Prevents actin polymerization Jasplakinolide Enhances actin polymerization Phailoidin Stabilizes actin lilaments Colchicine Prevents microtubule polymerization Demecolcine Microtuble deploymerization Paclitaxel Stabilizes microtubules Vinblastine Prevents microtuble polymerization METABOLIC EFFECTS GNE-140 Glycolysis inhibitor Elomoxir Fatty acid metabolism Sulfasalazine Amino acid metabolism P7C3 NAD biosynthesis inhibitor Oligomycin ATP synthase inhibitor PROLIFERATION AND GROWTH Rapamycin mTOR inhibitor (reduces cell growth) Cisplatin Antiproliferative chemotherapeutic
[0142] Approach for tumor proliferation modulation:
[0143] Longitudinal imaging with an established baseline signal and monitoring at multiple time points over a 24 h period following drug perturbation may provide a comprehensive assessment of d-.Math.OCT signal changes due to drug activity. Therefore, a tissue slice model of SKOV-3 tumors may be employed. Briefly, freshly excised tumors may be sectioned on a tumor matrix or vibratome to a thickness of 500 .Math.m while in culture media. Sections may subsequently be transferred to an organotypic support on a plate capable of holding multiple sections and cross-sectional imaging, filled with culture media. The plate containing the slices may then be kept in a controlled culture chamber at 37° C. and 5% CO.sub.2 in a humidified atmosphere for the imaging duration. For each of the agents listed in Table 3, a given tumor section may be treated with drug at three concentrations: IC50, ⅓rd IC50, and 3× IC50. Additional wells may contain tumor sections treated with vehicle only. For each drug/concentration, n=8 tissue slices may be treated in individual wells under identical conditions.
[0144] d-.Math.OCT imaging and histology: d-.Math.OCT imaging may take place as described above. One may obtain baseline images immediately prior to the addition of the pharmacological compound, and then 30 minutes, 1 h, 2 h, 6 h, 12 h, and 24 h thereafter. After imaging, tissue may be FFPE and sectioned to create 3D serial H&E slides. Additional IHC staining may be used to confirm specific pathway and anti-proliferative effects (e.g. ki67).
[0145] Analysis. d-.Math.OCT data from drug-treated wells may be normalized against data from vehicle treated specimens to account for potential sample degradation. Then, d-.Math.OCT periodogram images may be computed as described above. The change in d-.Math.OCT periodograms may be assessed on a per pixel, cell, and tissue basis as a function of drug, concentration, and time point.
[0146] Additional approaches - tissue degradation. If significant tissue degradation in the untreated vehicle samples is observed, one may exchange the growth media at regular intervals between d-.Math.OCT measurements. Tissue slice thickness can additionally be reduced to improve nutrient diffusion in the center of the specimen. One may also integrate an available media perfusion system into the controlled culture chamber.
Development of New d-.Math.OCT Technology for Improving Intracellular Motion Characterization: Increasing Spatial and Frequency Resolution of d-.Math.OCT
[0147] Optical spatial resolution. Table 4 shows the spatial resolution performance parameters of the current and new d-.Math.OCT systems. Optical spatial resolution may be improved by one or more of: 1) decreasing the center wavelength of the .Math.OCT light source from 800 nm to 700 nm; 2) increasing its bandwidth from 300 nm to 400 nm; 3) increasing the numerical aperture (NA) of the sample arm lens from NA = 0.12 to NA=0.2; and 4) implementing extended depth of focus (EDOF) technology which has been shown to extend the DOF by more than 20-fold, while maintaining cross-sectional imaging over a range of 300 .Math.m. Although one may attempt to exceed this spatial resolution performance, the parameters in Table 4 were conservatively chosen to provide a doubling of axial and lateral resolutions without significantly compromising cross-sectional imaging performance or penetration depth. The spatial resolution of the new d-.Math.OCT system may be validated using knife-edge scanning, z-scanning, and imaging of resolution standards and excised tissue. Success may be defined as demonstration of a factor of two improved resolution along x, y, and/or z dimensions.
TABLE-US-00004 Parameter Current New Center wavelenght (.Math.m) 0.80 0.70 Banewidth (.Math.m, FWHM) 0.30 0.40 Delta z (.Math.m, air) 0.87 0.50 Numerical aperture (NA) 0.12 0.20 Delta x (.Math.m, FWHM) 2.49 1.31 Standard DOF (.Math.m) 35.30 11.16 DOFextension factor 8 27 EDOF(.Math.m) 300 300
[0148] Table 4. Resolution parameters for d-.Math.OCT systems.
[0149] Computational spatial super-resolution. With d-.Math.OCT, multiple images or A-lines are acquired from the same location of the sample in the presence of both intracellular motion (the signal) and bulk motion that occurs on a larger scale. In addition, .Math.OCT EDOF technology that focuses annular patterns onto the sample aliases high spatial frequency information into the lens. Bulk motion and aliasing make it possible to computationally increase spatial resolution without modifying the setup or acquisition process.
[0150] Computational frequency super-resolution. Increased frequency resolution could significantly improve the discriminative capacity of d-.Math.OCT, especially at lower frequencies (0-10 Hz), where a considerable portion of intracellular motion takes place. Since the total measurement duration determines the frequency resolution, one may explore whether longer acquisitions provide additional frequency information content. Yet, extending acquisition times is undesirable for many scenarios and is untenable in vivo. One may therefore investigate computing d-.Math.OCT periodograms using high resolution DFT (HRDFT) and extended DFT (EDFT) algorithms that can increase power spectra frequency resolution within a discrete bandwidth without altering the total acquisition time. Frequency resolutions may be tested using magnetically modulated microbeads incorporated in phantoms, cells, and tissues. Success may be defined as demonstration of a two-fold improvement in frequency resolution between 0-10 Hz.
[0151] Develop new algorithms for processing and extracting more information from the d-.Math.OCT signal.
[0152] Time-frequency analysis (TFA). Short time Fourier transform (STFT) analysis of d-.Math.OCT tissue motility signals indicates that the motion within cells is non-stationary. Thus, more information can be potentially extracted from d-.Math.OCT using TFA. While the Stockwell transform (ST) may be used, owing to its super-resolution capabilities at low frequencies, other possible procedures including STFT, Wavelets, Wigner-Ville, the Constant-Q Gabor transform, and other time-frequency representations may be utilized to improve discrimination. TFA may be applied to signals after 3D elastic unwarping. Feature extraction may be conducted using energy concentration hybrid classification schemes and time-frequency cluster analysis. d-.Math.OCT images of magnetomotive phantoms excited via pre-determined, non-stationary waveforms may be obtained. TFA may be compared to the known frequency stimuli, with success defined as less than 10% variation.
[0153] Machine learning. To mine d-.Math.OCT data further, one may train deep neural networks (NN) to jointly model morphological and functional features and learn spatiotemporal representations that may group individual image pixels into biologically relevant categories. The input may be d-.Math.OCT periodograms and matched histology datasets acquired above, and the output may be series of two-dimensional probabilistic segmentation images - one for each biological category of interest. One may employ recurrent convolutional NNs and fully convolutional NNs, which can jointly model dynamic patterns in a spatial tissue context. To train models, one will apply both supervised and semi-supervised learning approaches using graphic image annotation strategies for obtaining pixel-level ground truth labels. To select a final NN model, one will train many models for a wide variety of network architectures and hyperparameters on the training set. One will then chose the model that performs best on a hold-out validation set to identify the optimal model depth, width, and resolution. To evaluate the selected model and estimate the generalization error, one will use a separate test data set compare quantitative measurements derived from semantic segmentations predicted by a model for a given image with the expected histological patterns for the known cell/tissue types and pathology. One will further analyze resulting NN-generated single-cell measurements using multivariate statistics and unsupervised machine learning to cluster cells based on their phenotype, frequency content, viability, and activation states. One may create lower dimensional representations (e.g. t-SNE, UMAP) of these high-dimensional data sets to facilitate visualization and discovery of interesting subpopulations.
[0154] Develop and test methods for conducing d-.Math.OCT in vivo.
[0155] Multiplane beam scanning. Two-dimensional elastic unwarping can correct in-plane, bulk motion, but is incapable of retrieving data that has moved out of the imaging plane. To mitigate this issue, one may develop new beam scanning techniques that acquire d-.Math.OCT images in multiple, adjacent cross-sectional planes and may conduct registration and elastic unwarping in three dimensions. Our first approach for multiplane scanning may be to raster dither the beam (
[0156] d-.Math.OCT handheld probe with stabilization. It has been possible to mechanically stabilize and acquire d-.Math.OCT images of human skin with a bench top microscope in vivo (
[0157] After the d-.Math.OCT handheld probe is developed, the feasibility of using this technology to obtain skin images in vivo may be tested in a clinical pilot study of patients undergoing elective lesion excision (e.g. n=10: 5 males/5 females). 3D imaging of the lesion may take place in vivo prior to excision. Once excised, the specimen may be placed in media and imaged again in 3D with the bench top d-.Math.OCT system ex vivo. The excision may then be sent to pathology per standard of care. d-.Math.OCT images of human skin obtained in vivo may be compared to those acquired ex vivo using frequency dependent ICM. Assuming a correlation of 0.9 between corresponding in vivo and ex vivo images, statistical significance may be attained with a power of 0.8 and α=0.05.
[0158] Another embodiment of this disclosure is dual-beam phase sensitive d-.Math.OCT (DBPS-d-.Math.OCT) system that can be used as a hand held or endoscopic probe for imaging in vivo or ex vivo. Obtaining motility information in a single scan is critical for imaging living internal organs with small diameter endoscopic probes where precise, micrometer-level control of the beam and thus sequential imaging of the same location in the presence of motion is difficult to achieve. Intracellular motion can potentially be assessed in a single scan by measuring the phase change of light reflected from the same point in the sample over a given time interval (
[0159] The DBPS probe can potentially be used for obtaining endoscopic d-.Math.OCT images in vivo in patients undergoing upper endoscopy. Normal regions and abnormalities (e.g. Barrett’s mucosa) seen by video endoscopy may be identified by a clinician and imaged by d-.Math.OCT in vivo, using the DBPS probe inserted into the scope’s accessory channel, as done in prior studies conducted by the present inventors.
Computer Systems and Process
[0160] Turning to
[0161] In some embodiments, computing device 1810 and/or server 1820 can be any suitable computing device or combination of devices, such as a desktop computer, a laptop computer, a smartphone, a tablet computer, a wearable computer, a server computer, a virtual machine being executed by a physical computing device, etc. As described herein, system for obtaining image data and functional data from biological tissue 1804 can present information about the control signals to a user (e.g., researcher and/or physician). In some embodiments, interferometer 1802 may include an apparatus such as that shown in
[0162] In some embodiments, communication network 1806 can be any suitable communication network or combination of communication networks. For example, communication network 1806 can include a Wi-Fi network (which can include one or more wireless routers, one or more switches, etc.), a peer-to-peer network (e.g., a Bluetooth network), a cellular network (e.g., a 3G network, a 4G network, etc., complying with any suitable standard, such as CDMA, GSM, LTE, LTE Advanced, WiMAX, etc.), a wired network, etc. In some embodiments, communication network 1806 can be a local area network, a wide area network, a public network (e.g., the Internet), a private or semi-private network (e.g., a corporate or university intranet), any other suitable type of network, or any suitable combination of networks. Communications links shown in
[0163]
[0164] In some embodiments, communications systems 1908 can include any suitable hardware, firmware, and/or software for communicating information over communication network 1806 and/or any other suitable communication networks. For example, communications systems 1908 can include one or more transceivers, one or more communication chips and/or chip sets, etc. In a more particular example, communications systems 1908 can include hardware, firmware and/or software that can be used to establish a Wi-Fi connection, a Bluetooth connection, a cellular connection, an Ethernet connection, etc.
[0165] In some embodiments, memory 1910 can include any suitable storage device or devices that can be used to store instructions, values, etc., that can be used, for example, by processor 1902 to present content using display 1904, to communicate with server 1820 via communications system(s) 1908, etc. Memory 1910 can include any suitable volatile memory, non-volatile memory, storage, or any suitable combination thereof. For example, memory 1910 can include RAM, ROM, EEPROM, one or more flash drives, one or more hard disks, one or more solid state drives, one or more optical drives, etc. In some embodiments, memory 1910 can have encoded thereon a computer program for controlling operation of computing device 1810. In such embodiments, processor 1902 can execute at least a portion of the computer program to present content (e.g., images, user interfaces, graphics, tables, etc.), receive content from server 1820, transmit information to server 1820, etc.
[0166] In some embodiments, server 1820 can include a processor 1912, a display 1914, one or more inputs 1916, one or more communications systems 1918, and/or memory 1920. In some embodiments, processor 1912 can be any suitable hardware processor or combination of processors, such as a central processing unit, a graphics processing unit, etc. In some embodiments, display 1914 can include any suitable display devices, such as a computer monitor, a touchscreen, a television, etc. In some embodiments, inputs 1916 can include any suitable input devices and/or sensors that can be used to receive user input, such as a keyboard, a mouse, a touchscreen, a microphone, etc.
[0167] In some embodiments, communications systems 1918 can include any suitable hardware, firmware, and/or software for communicating information over communication network 1806 and/or any other suitable communication networks. For example, communications systems 1918 can include one or more transceivers, one or more communication chips and/or chip sets, etc. In a more particular example, communications systems 1918 can include hardware, firmware and/or software that can be used to establish a Wi-Fi connection, a Bluetooth connection, a cellular connection, an Ethernet connection, etc.
[0168] In some embodiments, memory 1920 can include any suitable storage device or devices that can be used to store instructions, values, etc., that can be used, for example, by processor 1912 to present content using display 1914, to communicate with one or more computing devices 1810, etc. Memory 1920 can include any suitable volatile memory, non-volatile memory, storage, or any suitable combination thereof. For example, memory 1920 can include RAM, ROM, EEPROM, one or more flash drives, one or more hard disks, one or more solid state drives, one or more optical drives, etc. In some embodiments, memory 1920 can have encoded thereon a server program for controlling operation of server 1820. In such embodiments, processor 1912 can execute at least a portion of the server program to transmit information and/or content (e.g., results of a tissue identification and/or classification, a user interface, etc.) to one or more computing devices 1810, receive information and/or content from one or more computing devices 1810, receive instructions from one or more devices (e.g., a personal computer, a laptop computer, a tablet computer, a smartphone, etc.), etc.
[0169] In some embodiments, any suitable computer readable media can be used for storing instructions for performing the functions and/or processes described herein. For example, in some embodiments, computer readable media can be transitory or non-transitory. For example, non-transitory computer readable media can include media such as magnetic media (such as hard disks, floppy disks, etc.), optical media (such as compact discs, digital video discs, Blu-ray discs, etc.), semiconductor media (such as RAM, Flash memory, electrically programmable read only memory (EPROM), electrically erasable programmable read only memory (EEPROM), etc.), any suitable media that is not fleeting or devoid of any semblance of permanence during transmission, and/or any suitable tangible media. As another example, transitory computer readable media can include signals on networks, in wires, conductors, optical fibers, circuits, or any suitable media that is fleeting and devoid of any semblance of permanence during transmission, and/or any suitable intangible media.
[0170] It should be noted that, as used herein, the term mechanism can encompass hardware, software, firmware, or any suitable combination thereof.
[0171]
[0172] It should be understood that the above described steps of the process of
[0173] In certain embodiments, the steps of any process disclosed herein may be carried out using a processor in communication with a memory having stored thereon instructions which cause the processor to carry out the process. In some embodiments, the memory may include any suitable computer readable media which can be used for storing instructions for performing the functions and/or processes described herein. For example, in some embodiments, computer readable media can be transitory or non-transitory.
[0174] It will be appreciated by those skilled in the art that while the disclosed subject matter has been described above in connection with particular embodiments and examples, the invention is not necessarily so limited, and that numerous other embodiments, examples, uses, modifications and departures from the embodiments, examples and uses are intended to be encompassed by the claims attached hereto. The entire disclosure of each patent and publication cited herein is hereby incorporated by reference, as if each such patent or publication were individually incorporated by reference herein.
[0175] Various features and advantages of the invention are set forth in the following claims.