TBD

Surgical resection is the first-line treatment for most malignancies of the brain. However, the intraoperative identification of brain tumor tissue remains a challenge. In previous work, we demonstrated the potential of wide-field Mueller Polarimetric Imaging (MPI) to assess the anisotropy of fresh and fixed specimens of healthy brain, independently. Now, we use the MPI system to acquire polarimetric maps of fresh cadaveric pig cerebral tissue and compare the parameter evolution over time following formaldehyde-fixation. We demonstrated that despite the apparition of tissue morphological changes induced by formaldehyde fixation, this process preserves the polarimetric properties, remaining quantitatively similar to fresh tissue ones.

This report considers the development of several Machine Learning (ML) classifiers for the automatic diagnostic segmentation of Mueller Matrix polarimetric images of the uterine cervix. The analysed dataset includes polarimetric images obtained on 23 conization specimens in an ex vivo study conducted at the Kremlin-Bicêtre University Hospital in France. We demonstrate high level (>98%) accuracy for producing spatial masks of CIN/healthy zones regions when comparing directly with the pathological evaluated ground truth. The detailed results of numerous data tests, ML classifier training processes and avenues for improving the sensitivity and specificity of the developed techniques for “unseen” images are presented.

Examination of breast fine-needle aspiration (FNA) cytology specimens requires subjective assessment of cell morphology. Consequently, low diagnostic accuracy and high error rates (~40%) have been reported. We investigated methylene blue (MB) fluorescence polarization (Fpol) values depending on tumor grades/molecular subtypes. Pathologically diverse FNA specimens were collected from surgically excised breast tissues, stained in MB, and imaged. MB Fpol images provided a quantitative measurement for each cell. Optical imaging results were compared to clinical assessments. Statistical analysis revealed correlation between MB Fpol values and tumor grade/molecular subtype. The findings indicate that this method could provide accurate differentiation of breast cytology specimens.

Preterm birth accounts for 35% of annual infant deaths and is defined as any birth before 37 weeks of gestation. We have developed the Portable PReterm IMaging system capable of a 4x3 MMI, designed for comfort and ease of use to help fill the need for a diagnostic tool to monitor and detect preterm birth risks. Critical design components have been developed to be comfortable for the user and have a robust core structure. Imaging tests and evaluations are performed on volunteer human subjects with clinician assistance at the Florida International University Health Clinic.

Vagus nerve stimulation (VNS) yields promising therapeutic benefits but is often limited by off-target effects. Imaging tools that reveal the fascicular organization within the nerve can be used to better direct stimulation energy to the intended target. Here, we demonstrate polarization-sensitive (PS)-OCT imaging of an intact 7.5 cm ex-vivo porcine vagus nerve. The limited OCT penetration depth was overcome by using a dual-scanning microscope configured to simultaneously acquire B-scans from the upper and lower segments of a nerve sample. Three-dimensional maps of the complex fascicular organization in these nerves were generated.

Delineating the boundary of a tumors from healthy brain tissue is a challenging task in neurosurgery. Mueller polarimetry imaging promises to visualise and segment these borders in real-time, based on optical properties correlated with the directionality of densely packed white-matter fiber-bundles. In prior work, we demonstrated deep-learning methods leveraging Mueller polarimetry outperformed traditional approaches with similar segmentation tasks. However, formalin-fixation vs. fresh sample tissue and differences of human vs. animal brain tissue properties may hinder the direct applicability to neurosurgical scenarios. To overcome this potential limitation, we propose a learning-based strategy by jointly training on augmented multi-domain data together with model fine-tuning to improve tissue segmentation.

The identification of the border between tumor and healthy brain tissue remains a main challenge in glioma surgery. To address this problem we suggest using the Mueller Polarimetric Imaging (MPI) operating in the visible spectral range. In our prior studies, we demonstrated the potential of MPI to assess the anisotropy of healthy brain tissue in fixed and fresh specimens. In this study, we use the MPI system in backscattering geometry in order to evaluate and determine the depth of light penetration through the evaluation of 2D surface polarimetric maps of the formalin-fixed human cerebral corpus callosum sections of different thicknesses.

Particle sizing of biofluids, tissue samples, and therapeutic reagents is highly relevant for biomedical and clinical applications. Traditional dynamic light scattering (DLS) instruments are limited in their capability to measure optically transparent samples and are not conducive for particle sizing in samples that exhibit rich optical scattering and absorption. Here we demonstrate a new approach that utilizes polarization-dependent attributes of laser speckle to measure particle sizes in biosamples that span a range of optical scattering and absorption properties. We further apply this technique for particle sizing in homogenous controls, as well as biofluids such as platelet rich plasma and whole blood.

The uterine cervix is composed mainly of fibrous connective tissues where collagen and glycosaminoglycans are the main components. Recent work has shown that elastic fibers also contribute to the cervix's change in mechanical function during pregnancy. While cervical collagen can be visualized with non-linear optical techniques, elastic fibers cannot be uniquely identified without an extensive staining protocol. Here we propose the development of a compound Mueller Matrix microscope to image Mice cervices at different gestation points and visualize collagen and elastic fibers using a convolutional neural network (CNN) and K-nearest neighbor (K-NN) classifiers. The study demonstrates a new methodology for classifying collagen and elastic fibers in the uterine cervix that can be applied to any Mueller Matrix polarimeter once initial calibration is conducted.

This presentation reports a snapshot polarimetry system, capable of measuring a Stokes vector distribution within a millisecond timescale. The proposed system measures at a perpendicular backscattering angle and features a polarization state analyzer with no moving parts, comprised of a pair of quarter waveplates and pixel polarization cameras. An additional novel design aspect of the system is its capability to register polarization speckle. Polarization speckle contains both polarization and phase information that is not available from conventional techniques. The device’s acquisition speed and small form factor enable future studies of polarization speckle for biomedical applications.

Mueller matrix imaging has shown promise for sensing tissue alignment, and spatial frequency domain imaging is touted for adding depth sensitivity in turbid media. Combining the two techniques would seem to be straightforward. It is found that the spatial frequency domain Mueller matrix does not have the same behavior as a real-space Mueller matrix: it can be complex-valued, and it is not a convex sum of non-depolarizing Mueller matrices. In this presentation, the theory of polarized spatial frequency imaging will be reviewed and results from measurements will be presented.

Studies on propagation of linearly and circularly polarized vortex beams in tissue-like forward scattering turbid media revealed longer survival of vortex beams with higher topological charge (l). It is shown that an increase in the effective anisotropy parameter of scattering as seen by vortex beams having higher topological charge (l) is responsible for the observed longer survival of vortex beams in tissue and tissue-like turbid media. These findings may open up interesting avenues towards achieving larger depth of imaging in biological tissues by appropriately tailoring orbital angular momentum and spatial polarization states of vector vortex beams.

Mueller matrix coefficients are conventionally derived from averaged measurements of several polarimetric intensity images for each polarisation state. However, averaging large numbers of measurements is not compatible with real-time surgical applications. To overcome this limitation, we introduce a novel learning-based denoising framework aiming at recovering accurate, physically consistent and high signal-to-noise ratio (SNR) polarimetric scans from short-time noisy acquisitions. We formulate a microstructure-aware denoising diffusion network and validate against current state-of-the-art denoising techniques for real images in healthy and diseased brain samples. Ultimately, the performance is analysed for near-real-time applicability and the advantage of the proposed approach is discussed.

Imaging the collagen layer of the human tympanic membrane, on which the human hearing ability depends, is possible in vivo by using endoscopic polarization-sensitive optical coherence tomography (PSOCT), as we have shown recently. Since the resolution of the utilized system was limited, an ex vivo validation of the structural information is presented here. By comparing higher resolution PSOCT images of a temporal bone specimen and polarized light microscopy (PLM) of histological slices cut from the same sample and stained with picrosirius red, the microstructure is investigated. Besides matching PSOCT and PLM images, the methodical differences are discussed.

We developed the theoretical framework for extracting polarimetric parameters relevant to the biomedical diagnosis of tissue (linear retardance, azimuth of optical axis, depolarization) using partial Mueller matrix (MM) data and compare them with corresponding parameters extracted from complete MM. Both parametric datasets demonstrated compelling similarity for both theoretical model and experimental MMs of tape and thin section of mouse uterine cervix measured in reflection. Our findings represent a significant step toward developing new medical imaging partial polarimeters that may provide a good compromise between the simplicity, compactness, speed of acquisition, and the required accuracy of polarimetric measurements for medical diagnosis.

Peripheral nerve fascicle tracing and longitudinal analysis of the damaging effects of electrical overstimulation using novel polarization sensitive optical coherence tomography (PS-OCT) based biomarkers currently require manual image segmentation to determine the boundaries of the nerve tissue within the acquired OCT volume. We developed an algorithm for automated peripheral nerve tissue classification based on image segmentation of the multimodal swept-source PS-OCT volumes. We compare the performance of density-based and neural network-based classifiers, and demonstrate results on fixed nerve tissues as well as on in vivo acquired data from a rat sciatic nerve and mini-pig vagus nerve.