This PDF file contains the front matter associated with SPIE Proceedings Volume 10487 including the Title Page, Copyright information, Table of Contents, Introduction, and Conference Committee listing.

Determining cancer phenotypes early in the development of the disease is fundamental to increase the efficacy of therapy. To this end, genetic profiling of bioptic tissue has become an indispensable tool; however, erroneous sampling of the tissue leads to a considerable number of false negatives. To address this shortcoming, we developed a motorized endoscope (outer diameter 1.35mm) combining near-infrared fluorescence (NIRF) imaging and optical coherence tomography (OCT) at a frame rate of 52 images per second. Here, NIRF imaging is used to guide biopsies by targeting the tumor with fluorescently labeled monoclonal antibodies that attach to receptors on the membrane of cancer cells. Furthermore, the biopsy forceps severely alter the structure of the tissue and OCT can help in reconstructing the morphology thanks to its 10 µm resolution. We demonstrated the device on a mouse model bearing human colon cancer, by administering a sub-therapeutic dose of a monoclonal antibody labeled with both a PET tracer and a near infrared molecule that is already used in several clinical trials (89Zr-Labetuzumab-IRdye800CW). In addition to successfully imaging 1 mm tumors with our endoscope, we developed a NIRF-OCT handheld scanner that allows mapping of the distribution of the antibody with high resolution. Finally, we also performed PET in order to quantify the bio-distribution of the drug in the mouse.

Glioblastoma is a primary malignant brain tumor characterized by highly migratory glioma cells capable of invading into surrounding healthy tissue. The mechanism and the physical paths by which glioma cells are capable of invading healthy brain are not well understood. Using human glioblastoma cell line U251 plated on healthy mouse brain slices, glioma cell migration behavior and dynamics are investigated by multimodality imaging. Polarization-sensitive optical coherence tomography (PS-OCT) is used to delineate nerve fiber tracts. PS-OCT is capable of generating depth-resolved images of reflectivity, phase retardance and optic axis orientation. Because of the birefringence property of myelin sheath, nerve fiber tracts as small as a few tens of micrometers can be resolved from phase retardance images. Swept field confocal imaging system is used to image U251 cells expressing GFP-actin and brain vasculature stained by Isolectin B4. Cell migration is acquired by time-lapse imaging and then correlated with brain vasculature and nerve fiber tracts after fine registration of the two modalities. We found out U251 cells preferentially adhere to and migrate along blood vessels. Our data do not suggest a strong correlation between U251 cell migration and white matter tracts distinguished by tissue birefringence. In addition, U251 cell motility is higher in gray matter compared with white matter. Finally, using higher temporal resolution and high magnification, we are able to observe short time-scale dynamic of U251 cells and the ability of U251 cells to exert forces as they deform blood vessels.

Optical Coherence tomography (OCT) images provide several indicators, e.g., the shape and the thickness of different retinal layers, which can be used for various clinical and non-clinical purposes. We propose an automated classification method to identify different ocular diseases, based on the local binary pattern features. The database consists of normal and diseased human eye SD-OCT images. We use a multiphase approach for building our classifier, including preprocessing, Meta learning, and active learning. Pre-processing is applied to the data to handle missing features from images and replace them with the mean or median of the corresponding feature. All the features are run through a Correlation-based Feature Subset Selection algorithm to detect the most informative features and omit the less informative ones. A Meta learning approach is applied to the data, in which a SVM and random forest are combined to obtain a more robust classifier. Active learning is also applied to strengthen our classifier around the decision boundary. The primary experimental results indicate that our method is able to differentiate between the normal and non-normal retina with an area under the ROC curve (AUC) of 98.6% and also to diagnose the three common retina-related diseases, i.e., Age-related Macular Degeneration, Diabetic Retinopathy, and Macular Hole, with an AUC of 100%, 95% and 83.8% respectively. These results indicate a better performance of the proposed method compared to most of the previous works in the literature.

Choroidal neovascularization (CNV) from age-related macular degeneration (AMD) is the leading cause of permanent vision loss and blindness in adults in the developed world. Early detection of CNV results in improved visual outcomes in patients. The current study investigated the feasibility of low-level laser therapy (LLLT)-assisted vascular epithelial growth factor (VEGF) intravitreal injections to increase angiogenesis. In addition, the developed CNV model was evaluated by photoacoustic microscopy (PAM), optical coherence tomography (OCT), and fluorescein angiography (FA) imaging systems. Dual laser wavelengths of 660 and 780 nm were used to activate VEGF to enhance new blood vessels angiogenesis. Ten New Zealand white rabbits were injected intravitreally with recombinant human VEGF165 at a dose of 10 μg/0.1 mL injection followed by laser at an irradiance of 3 and 4 J/cm2 or control (no laser) New blood vessel formation was monitored by PAM, OCT, and FA. Area of neovascularization was imaged and quantified bi-weekly after treatment for 1 month. Laser-enhanced VEGF increased the area of neovascularization after treatment in comparison with VEGF injection only. The location of CNV was obviously identified by PAM, OCT, and FA. The proposed LLLT-assisted VEGF together with multimodal imaging system can serve as a useful technique for clinical detection and diagnostics of CNV and may provide the functional information in the changes.

Previous works have demonstrated feasibility of combining optical coherence tomography (OCT) and hyper-spectral imaging (HSI) through a single double-clad fiber (DCF). In this proceeding we present the continued development of a system combining both modalities and capable of rapid imaging. We discuss the development of a rapidly scanning, dual-band, polygonal swept-source system which combines NIR (1260-1340 nm) and visible (450-800 nm) wavelengths. The NIR band is used for OCT imaging while visible light allows HSI. Scanning rates up to 24 kHz are reported. Furthermore, we present and discuss the fiber system used for light transport, delivery and collection, and the custom signal acquisition software. Key points include the use of a double-clad fiber coupler as well as important alignments and back-reflection management. Simultaneous and co-registered imaging with both modalities is presented in a bench-top system

Relevant chemotherapy-induced changes to intra-tumor heterogeneity occur at multiple length scales, but few imaging methods are capable of simultaneous multiscale monitoring. We have developed a multiscale imaging technique called Diffuse and Nonlinear Imaging (DNI) that uses Spatial Frequency Domain Imaging (SFDI) for wide-field mapping of tumor optical properties and hemodynamics, and Multiphoton Microscopy (MPM) to image tumor micro-vessel and collagen architecture, and endogenous metabolic molecules with cellular resolution. Importantly, DNI co-registers MPM throughout the SFDI field-of-view. The SFDI system uses LEDs to illuminate a digital micro-mirror device with near-infrared wavelengths (650-850 nm) to project structured intensity patterns onto a sample, with diffuse reflectance collected by a CCD camera. The extended wavelength (680-1300 nm) MPM system consists of a femtosecond tunable laser for two-photon excitation (TPE) of fluorescently labeled microvasculature, Second Harmonic Generation imaging of collagen, and TPE auto-fluorescence of intracellular NADH and FAD. We have conducted in vivo DNI monitoring of a syngeneic murine mammary tumor model (Py230) through a mammary imaging window in 10 untreated C57BL/6 female mice. Our initial results revealed that global hemoglobin concentrations and micro-vessel density were highly correlated (ρ = 0.83). This presentation will report on multiscale relationships among tumor hemodynamic, microenvironment, and metabolic metrics, as well as on Monte Carlo modeling and phantom studies comparing depth penetration between the two imaging regimes. In the future, DNI will be used to evaluate chemo-resistance over a range of spatial scales and contrast mechanisms to obtain a more complete picture of the in vivo tumor state.

In this paper, multimodal optical-resolution frequency-domain photoacoustic and fluorescence scanning microscopy is presented on labeled and unlabeled cells. In many molecules, excited electrons relax radiatively and non-radiatively, leading to fluorescence and photoacoustic signals, respectively. Both signals can then be detected simultaneously. There also exist molecules, e.g. hemoglobin, which do not exhibit fluorescence, but provide photoacoustic signals solely. Other molecules, especially fluorescent dyes, preferentially exhibit fluorescence. The fluorescence quantum yield of a molecule and with it the strength of photoacoustic and fluorescence signals depends on the local environment, e.g. on the pH. Therefore, the local distribution of the simultaneously recorded photoacoustic and fluorescence signals may be used in order to obtain information about the local chemistry.

Optical imaging of whole animals or animal organs is a rapidly growing field in translational research revealing the molecular events underlying disease and disease treatment mechanisms in cardiovascular, cancer and neurological research. Here we present a custom-built imaging system for visualisation of 3D distribution of fluorescent markers with high-resolution tissue structure and vasculature network images of small animals or whole organs. The 3D Fluorescence Imaging Cryomicrotome System (3D-FICS) yields comprehensive structural and functional biological information by combining fluorescence remittance imaging. Moreover, the 3D-FICS is adapted to record large series of high-resolution images (2048 x 2048 pixels, with a selectable resolution of 27, 16 and 8μm, corresponding to a FOV of 53, 31 and 16mm) of the bulk tissue remaining in the cryomicrotome in a fully automated manner. All components are controlled through custom software (Labview) to enable fully automated serial cutting and imaging sessions. To ensure 3D reconstructions with isotropic voxels, the slice thickness of the cryomicrotome is set to match the imaging resolution of the camera. Wavelength-selective illumination of the tissue is carried out using a Supercontinuum laser in conjunction with a tunable bandpass filter (400nm - 830nm) with a tunable bandwidth (5nm - 100nm) Similarly, a tunable emission filter (440nm - 800nm) with a bandwidth ranging from 11nm to 15nm is used for spectrally-resolved imaging of the fluorescent tracers injected in the specimen. The 3D-FICS will be used in experimental studies on small animal organs from which we will present the first obtained images.

Objective: To develop dual-modality Photoacoustic(PA)/Ultrasound(US) system based on clinical US machine and performed imaging study of in vivo human superficial lesions. Methods: A dual-modality PA/US system was developed based on a high-end clinical US machine with a handheld probe for US/PA dual modality imaging, equipped with multi-wavelength laser source. Twenty-three patients were enrolled consecutively from the outpatients and inpatients of Peking Union Medical College Hospital (PUMCH) from Dec. 15th 2016 to Apr.15th 2017. All the patients underwent ultrasound examinations including both gray scale and color Doppler flow imaging ultrasound (CDFI) to evaluate the morphological and vascular information of the lesions. PA/US imaging was performed right after CDFI, the CDFI and the PA/US dual imaging information of each case were compared. Results: The laser source for PAI generates multi-wavelength laser pulses at 10Hz. Therefore, deoxyhemoglobin and hematoglobulin in tumor vasculature, which serve as different optical absorbers at specific wavelengths, can be evaluated and compared through PAI. PA/US dual modality imaging was performed in 23 cases, including 10 thyroid lesions, 10 breast lesions, and 3 soft tissue endometriosis lesions. All of the lesions performed surgery with pathology confirmed diagnosis. Our PA/US dual modality imaging system showed high quality gray scale ultrasonic and dual-modality fusion images. According to the results, significant differences exist between PAI and CDFI. PAI could reveal some blood vessels that were not sensitive for Doppler ultrasound. PAI proved the ability of imaging both the peripheral and intra-nodular vessels of human superficial lesions, as well as detecting the difference of oxygen saturation between benign and malignant tumor. Conclusion: Our study demonstrated that photoacoustic imaging could provide important complementary information for traditional ultrasound in superficial lesion examination, which has a great potential for clinical diagnosis.

Current methods for breast tumor margin detection are invasive, time consuming, and typically result in a reoperative rate of over 25%. This marks a clear clinical need to develop improved tools to intraoperatively differentiate negative versus positive tumor margins. Here, we utilize photoacoustic tomography (PAT), ultrasound (US), and inverted Selective Plane Illumination Microscopy (iSPIM) to assess breast tumor margins in eight human breast biopsies. Our PAT/US system consists of a tunable Nd:YAG laser (NT 300, EKSPLA) coupled with a 40MHz central frequency US probe (Vevo2100, FUJIFILM Visual Sonics). This system allows for the delivery of 10Hz, 5ns pulses with fluence of 40mJ/cm² to the tissue with PAT and US axial resolutions of 125μm and 40μm, respectively. For this study, we used a linear stepper motor to acquire volumetric PAT/US images of the breast biopsies using 1100nm light to identify bloodrich “tumor” regions and 1210nm light to identify lipid-rich “healthy” regions. iSPIM (Applied Scientific Instrumentation) is an advanced microscopy technique with lateral resolution of 1.5μm and axial resolution of 7μm. We used 488nm laser excitation and acridine orange as a general comprehensive histology stain. Our results show that PAT/US can be used to identify lipid-rich regions, dense areas of arterioles and arteries, and other internal structures such as ducts. iSPIM images correlate well with histopathology slides and can verify nuclear features, cell type and density, stromal features, and microcalcifications. Together, this multimodality approach has the potential to improve tumor margin detection with a high degree of sensitivity and specificity.

Hemodynamic-based neuroimaging techniques such as functional magnetic resonance imaging (fMRI) and near-infrared spectroscopy (NIRS) sense hemoglobin concentration in cerebral tissue. The local concentration of hemoglobin, which is differentiated into oxy- and deoxy-hemoglobin by NIRS, features spontaneous oscillations over time scales of 10-100 s in response to a number of local and systemic physiological processes. If one of such processes becomes the dominant source of cerebral hemodynamics, there is a high coherence between this process and the associated hemodynamics. In this work, we report a method to identify such conditions of coherent hemodynamics, which may be exploited to study and quantify microvasculature and microcirculation properties. We discuss how a critical value of significant coherence may depend on the specific data collection scheme (for example, the total acquisition time) and the nature of the hemodynamic data (in particular, oxy- and deoxy-hemoglobin concentrations measured with NIRS show an intrinsic level of correlation that must be taken into account). A frequency-resolved study of coherent hemodynamics is the basis for the new technique of coherent hemodynamics spectroscopy (CHS), which aims to provide measures of cerebral blood flow and cerebral autoregulation. While these concepts apply in principle to both fMRI and NIRS data, in this article we focus on NIRS data.

Polarized light imaging and optical spectroscopy can be used to distinguish between healthy and diseased tissue. In this study, the design and testing of a single-pixel hyperspectral imaging system that uses differences in the polarization of light reflected from tissue to differentiate between healthy and thermally damaged tissue is discussed. Thermal lesions were created in porcine skin (n = 8) samples using an IR laser. The damaged regions were clearly visible in the polarized light hyperspectral images. Reflectance hyperspectral and white light imaging was also obtained for all tissue samples. Sizes of the thermally damaged regions as measured via polarized light hyperspectral imaging are compared to sizes of these regions as measured in the reflectance hyperspectral images and white light images. Good agreement between the sizes measured by all three imaging modalities was found. Hyperspectral polarized light imaging can differentiate between healthy and damaged tissue. Possible applications of this imaging system include determination of tumor margins during cancer surgery or pre-surgical biopsy.

Colorectal cancer is the second leading cause of cancer deaths in the United States. Identifying and removing premalignant lesions via colonoscopy can significantly reduce colorectal cancer mortality. Unfortunately, the protective value of screening colonoscopy is limited because more than one quarter of clinically-important lesions are missed on average. Most of these lesions are associated with characteristic 3D topographical shapes that appear subtle to a conventional colonoscope. Photometric stereo endoscopy captures this 3D structure but is inherently qualitative due to the unknown working distances from each point of the object to the endoscope. In this work, we use deep learning to estimate the depth from a monocular endoscope camera. Significant amounts of endoscopy data with known depth maps is required for training a convolutional neural network for deep learning. Moreover, this training problem is challenging because the colon texture is patient-specific and cannot be used to efficiently learn depth. To resolve these issues, we developed a photometric stereo endoscopy simulator and generated data with ground truth depths from a virtual, texture-free colon phantom. These data were used to train a deep convolutional neural field network that can estimate the depth for test data with an accuracy of 84%. We use this depth estimate to implement a smart photometric stereo algorithm that reconstructs absolute depth maps. Applying this technique to an in-vivo human colonoscopy video of a single polyp viewed at varying distance, initial results show a reduction in polyp size measurement variation from 15.5% with conventional to 3.4% with smart photometric reconstruction.

Hyperspectral imaging (HSI) is a technology used in remote sensing, food processing and documentation recovery. Recently, this approach has been applied in the medical field to spectrally interrogate regions of interest within respective substrates. In spectral imaging, a two (spatial) dimensional image is collected, at many different (spectral) wavelengths, to sample spectral signatures from different regions and/or components within a sample. Here, we report on the use of hyperspectral imaging for endoscopic applications. Colorectal cancer is the 3rd leading cancer for incidences and deaths in the US. One factor of severity is the miss rate of precancerous/flat lesions (~65% accuracy). Integrating HSI into colonoscopy procedures could minimize misdiagnosis and unnecessary resections. We have previously reported a working prototype light source with 16 high-powered light emitting diodes (LEDs) capable of high speed cycling and imaging. In recent testing, we have found our current prototype is limited by transmission loss (~99%) through the multi-furcated solid light guide (lightpipe) and the desired framerate (20-30 fps) could not be achieved. Here, we report on a series of experimental and modeling studies to better optimize the lightpipe and the spectral endoscopy system as a whole. The lightpipe was experimentally evaluated using an integrating sphere and spectrometer (Ocean Optics). Modeling the lightpipe was performed using Monte Carlo optical ray tracing in TracePro (Lambda Research Corp.). Results of these optimization studies will aid in manufacturing a revised prototype with the newly designed light guide and increased sensitivity. Once the desired optical output (5-10 mW) is achieved then the HIS endoscope system will be able to be implemented without adding onto the procedure time.

Preclinical imaging using near-infrared (NIR) fluorescent markers is a non-invasive approach to molecular imaging with many applications ranging from drug delivery monitoring to immune cell tracking and image-guided surgery. It is however limited by signal absorption and scattering within tissues, which make it difficult to localize and quantify signal sources. Moreover, broad absorption and emission spectra make it difficult to use multiple NIR dyes for multiplexed imaging. Fluorescence lifetime imaging (FLI) helps solve some of these problems by providing a contrast mechanism that is sensitive to the intracellular environment but not to signal intensity and can separate dyes with similar spectra but different lifetimes. So far, however, in vivo FLI has found few applications due to the complexity of standard FLI data acquisition and analysis. Here we show that in vivo wide-field time-gated macroscopic FLI (MFLI) imaging, combined with phasor analysis of NIR fluorescence lifetime data, addresses most of these issues. We illustrate these capabilities with a study of in vivo transferrin ligand-receptor engagement dynamic in anesthetized mice using Förster resonant energy transfer (FRET), and validate our results by comparing them with standard lifetime fitting. Our approach speeds up data analysis by several orders of magnitude and requires less data, while providing easy-to-use and interpret data visualization, opening up the possibility of real-time in vivo MFLI of fast dynamic processes.

X-ray luminescence computed tomography (XLCT) is an emerging hybrid molecular imaging modality. In XLCT, high energy x-ray photons excite phosphors emitting optical photons for tomographic image reconstruction. During XLCT, the optical signal obtained is thought to only originate from the embedded phosphor particles. However, numerous studies have reported other sources of optical photons such as in air, water, and tissue that are generated from ionization. These sources of optical photons will provide background noise and will limit the molecular sensitivity of XLCT imaging. In this study, using a water-cooled electron multiplying charge-coupled device (EMCCD) camera, we performed luminescence imaging of water, air, and several tissue mimicking phantoms including one embedded with a target containing 0.01 mg/mL of europium-doped gadolinium oxysulfide (GOS:Eu3+) particles during x-ray irradiation using a focused x-ray beam with energy less than the Cerenkov radiation threshold. In addition, a spectrograph was used to measure the x-ray luminescence spectrum. The phantom embedded with the GOS:Eu³⁺ target displayed the greatest luminescence intensity, followed by the tissue phantom, and finally the water phantom. Our results indicate that the x-ray luminescence intensity from a background phantom is equivalent to a GOS:Eu³⁺ concentration of 0.8 μg/mL. We also found a 3-fold difference in the radioluminescence intensity between liquid water and air. From the measurements of the emission spectra, we found that water produced a broad spectrum and that a tissue-mimicking phantom made from Intralipid had a different x-ray emission spectrum than one made with TiO₂ and India ink. The measured spectra suggest that it is better to use Intralipid instead if TiO₂ as optical scatterer for future XLCT imaging.

Optical Molecular Imaging (OMI) has the advantages of high sensitivity, low cost and ease of use. By labeling the regions of interest with fluorescent or bioluminescence probes, OMI can noninvasively obtain the distribution of the probes in vivo, which play the key role in cancer research, pharmacokinetics and other biological studies. In preclinical and clinical application, the image depth, resolution and sensitivity are the key factors for researchers to use OMI. In this paper, we report a high sensitivity optical molecular imaging system developed by our group, which can improve the imaging depth in phantom to nearly 5cm, high resolution at 2cm depth, and high image sensitivity. To validate the performance of the system, special designed phantom experiments and weak light detection experiment were implemented. The results shows that cooperated with high performance electron-multiplying charge coupled device (EMCCD) camera, precision design of light path system and high efficient image techniques, our OMI system can simultaneously collect the light-emitted signals generated by fluorescence molecular imaging, bioluminescence imaging, Cherenkov luminance and other optical imaging modality, and observe the internal distribution of light-emitting agents fast and accurately.

Conical mirror is a preferred choice for fluorescence molecular tomography (FMT) because of its ability to collect fluorescent emission photons from the whole surface of the imaged object such as mice. Conical mirror, however, would lead to a fraction of photons to be reflected back to the mice surface, including excitation photons and emission photons, which result in inaccurate source positions and measurements errors in the FMT forward modeling and reconstruction. Based on Monte Carlo simulations, we have studied systematically the effects of multiple reflections of different conical mirror designs. We first generated a multiple reflected photon map for each design of the conical mirror, and then we applied Monte Carlo simulations to model photon propagation inside tissues. Finally, we evaluated the ratio of the multiple reflected photons to the total photons, and figured out the optimized size of the conical mirror. Our simulations demonstrated that a single conical mirror configuration could eliminate the multiple reflection issue while keep the imaging system setup simple when its small aperture radius is larger than 5 centimeters. We then fabricated a conical mirror with the optimized size according to the Monte Carlo simulation results, and performed phantom experiments with both the optimized conical mirror and the non-optimized one. Phantom experiment results show that noises in the reconstructed images are reduced with the optimized conical mirror, and the reconstruction accuracy is improved as well.

To investigate whether quantitative radiomics features extracted from dynamic contrast-enhanced magnetic resonance imaging (DCE-MRI) are associated with Ki67 expression of breast cancer. In this institutional review board approved retrospective study, we collected 377 cases Chinese women who were diagnosed with invasive breast cancer in 2015. This cohort included 53 low-Ki67 expression (Ki67 proliferation index less than 14%) and 324 cases with high-Ki67 expression (Ki67 proliferation index more than 14%). A binary-classification of low- vs. high- Ki67 expression was performed. A set of 52 quantitative radiomics features, including morphological, gray scale statistic, and texture features, were extracted from the segmented lesion area. Three most common machine learning classification methods, including Naive Bayes, k-Nearest Neighbor and support vector machine with Gaussian kernel, were employed for the classification and the least absolute shrink age and selection operator (LASSO) method was used to select most predictive features set for the classifiers. Classification performance was evaluated by the area under receiver operating characteristic curve (AUC), accuracy, sensitivity and specificity. The model that used Naive Bayes classification method achieved the best performance than the other two methods, yielding 0.773 AUC value, 0.757 accuracy, 0.777 sensitivity and 0.769 specificity. Our study showed that quantitative radiomics imaging features of breast tumor extracted from DCE-MRI are associated with breast cancer Ki67 expression. Future larger studies are needed in order to further evaluate the findings.

The purpose of this work is to introduce and study a novel x-ray beam irradiation pattern for X-ray Luminescence Computed Tomography (XLCT), termed multiple intensity-weighted narrow-beam irradiation. The proposed XLCT imaging method is studied through simulations of x-ray and diffuse lights propagation. The emitted optical photons from X-ray excitable nanophosphors were collected by optical fiber bundles from the right-side surface of the phantom. The implementation of image reconstruction is based on the simulated measurements from 6 or 12 angular projections in terms of 3 or 5 x-ray beams scanning mode. The proposed XLCT imaging method is compared against the constant intensity weighted narrow-beam XLCT. From the reconstructed XLCT images, we found that the Dice similarity and quantitative ratio of targets have a certain degree of improvement. The results demonstrated that the proposed method can offer simultaneously high image quality and fast image acquisition.