Fluorescence-mediated Tomography Research Articles

Abstract 1073: Longitudinal Imaging Of NF-kappaB In Murine Pulmonary Embolism

Objective: Pulmonary embolism (PE) is a common condition that urgently needs novel diagnostic and therapeutic strategies. Here, we investigated PE-induced inflammation via longitudinal imaging of transcription factor NF-kappaB (NFkB) activation signals in murine PE. Understanding the temporal activation of inflammatory mediators, such as NFkB, in PE may lead to future medical interventions or better patient stratification. Methods: NFkB activation was assessed using male, 12-16w old, NFkB-GFP-luciferase (NGL) reporter mice (n=20 PE animals). Stasis venous thrombosis (VT) was induced in the femoral veins of donor mice on day 0. On day 1, the thrombus was resected and intravenously injected into recipient mice to create the PE model. PE animals underwent bioluminescence imaging on days -1, 1, 4, and 7. Regions-of-interest (ROI) areas were kept constant and placed over areas of PE as verified in other modalities. Lung and PE morphometrics were imaged by fluorescence-mediated tomography and computed tomography using a fibrin-specific molecular imaging agent, FTP11 (i.v. 150nmol/kg). Cardiac function pre- and post-PE was assessed by ultrasound and pulmonary arterial pressure measurements. Mouse cohorts were sacrificed 1, 4, and 7 days following stasis-VT embolization and underwent histological analysis of the PE. Results: The mean volume of the femoral stasis-VT was 1.21±0.25 mm 3 . The bioluminescence signal plateau was found at 4-8 minutes post-intravenous injection of D-luciferin in PE animals. The six-minute post-injection image was utilized in comparing longitudinal images. Bioluminescence from the activation of NFkB in the PE ROI showed a maximal signal in the day 1 PE with a 55.0% (334 to 151RU) decrease in signal from day 1 to day 7(p<0.05). Histological and protein assays validated the activation and localization of NFkB. In PE animals, immunohistochemistry of infiltrating CD45-positive cells showed increased cell numbers from day 1 to day 7 (p<0.05). Since PE is confided spatially to a lung lobe, contralateral PE-negative lobes also showed increased infiltrating cells compared to sham animals. Conclusions: Quantitative serial imaging pre- and post-PE demonstrates an inflammatory timeline in PE-induced NFkB activation.

Using Fluorescence-Mediated Tomography to Assess Murine Intestinal Inflammation

Using Fluorescence-Mediated Tomography to Assess Murine Intestinal Inflammation

Using Fluorescence-Mediated Tomography to Assess Murine Intestinal Inflammation

Using Fluorescence-Mediated Tomography to Assess Murine Intestinal Inflammation

Lichtwelten in der diagnostischen Medizin

Optical imaging has always played acentral role in the elucidation of biological and physiological mechanisms in modern biology and medicine. Based on the good experiences in light microscopy, sophisticated meso- and macroscopic optical imaging systems have recently been created. All optical imaging methods are characterized by high user-friendliness and sensitivity, they are associated with relatively low costs and do not require any radioactivity. Its clinical performance is seen in the intraoperative imaging of the tissue area to be removed and in the radiotracer-free diagnosis of diseases in body areas with good accessibility to light. From apathomorphological point of view, the focus was particularly on the depiction of tumors and inflammation. Imaging detection of fluorescent dyes with emission characteristics in the near-infrared range of the spectrum is favorable in terms of agood signal-to-background ratio and improved information acquisition from greater tissue depths. Amajor challenge, however, is the diverse photon interactions with the tissue. Previous research and development work has produced various in vivo optical imaging methods, some of which are still in the experimental stage (e.g., fluorescence-mediated tomography, multispectral in vivo imaging, bioluminescence imaging, Raman spectroscopy), while others already have made their way into the clinical setting (e.g., fluorescence reflection imaging, optoacoustic imaging). The most important optical methods are presented in this review article.

How Different Albumin-Binders Drive Probe Distribution of Fluorescent RGD Mimetics

The biodistribution of medical imaging probes depends on the chemical nature of the probe and the preferred metabolization and excretion routes. Especially targeted probes, which have to reach a certain (sub)cellular destination, have to be guided to the tissue of interest. Therefore, small molecular probes need to exhibit a well-balanced polarity and lipophilicity to maintain an advantageous bioavailability. Labelled antibodies circulate for several days due to their size. To alter the biodistribution behavior of probes, different strategies have been pursued, including utilizing serum albumin as an inherent transport mechanism for small molecules. We describe here the modification of an existing fluorescent RGD mimetic probe targeted to integrin αvβ3 with three different albumin binding moieties (ABMs): a diphenylcyclohexyl (DPCH) group, a p-iodophenyl butyric acid (IPBA) and a fatty acid (FA) group with the purpose to identify an optimal ABM for molecular imaging applications. All three modifications result in transient albumin binding and a preservation of the target binding capability. Spectrophotometric measurements applying variable amounts of bovine serum albumin (BSA) reveal considerable differences between the compounds concerning their absorption and emission characteristics and hence their BSA binding mode. In vivo the modified probes were investigated in a murine U87MG glioblastoma xenograft model over the course of 1 wk by fluorescence reflectance imaging (FRI) and fluorescence mediated tomography (FMT). While the unmodified probe was excreted rapidly, the albumin-binding probes were accumulating in tumor tissue for at least 5 days. Considerable differences between the three probes in biodistribution and excretion characteristics were proved, with the DPCH-modified probe showing the highest overall signal intensities, while the FA-modified probe exhibits a low but more specific fluorescent signal. In conclusion, the modification of small molecular RGD mimetics with ABMs can precisely fine-tune probe distribution and offers potential for future clinical applications.

Mixing Matrix-corrected Whole-body Pharmacokinetic Modeling Using Longitudinal Micro-computed Tomography and Fluorescence-mediated Tomography

PurposePharmacokinetic modeling can be applied to quantify the kinetics of fluorescently labeled compounds using longitudinal micro-computed tomography and fluorescence-mediated tomography (μCT-FMT). However, fluorescence blurring from neighboring organs or tissues and the vasculature within tissues impede the accuracy in the estimation of kinetic parameters. Contributions of elimination and retention activities of fluorescent probes inside the kidneys and liver can be hard to distinguish by a kinetic model. This study proposes a deconvolution approach using a mixing matrix to model fluorescence contributions to improve whole-body pharmacokinetic modeling.ProceduresIn the kinetic model, a mixing matrix was applied to unmix the fluorescence blurring from neighboring tissues and blood vessels and unmix the fluorescence contributions of elimination and retention in the kidney and liver compartments. Accordingly, the kinetic parameters of the hepatobiliary and renal elimination routes and five major retention sites (the kidneys, liver, bone, spleen, and lung) were investigated in simulations and in an in vivo study. In the latter, the pharmacokinetics of four fluorescently labeled compounds (indocyanine green (ICG), HITC-iodide-microbubbles (MB), Cy7-nanogels (NG), and OsteoSense 750 EX (OS)) were evaluated in BALB/c nude mice.ResultsIn the simulations, the corrected modeling resulted in lower relative errors and stronger linear relationships (slopes close to 1) between the estimated and simulated parameters, compared to the uncorrected modeling. For the in vivo study, MB and NG showed significantly higher hepatic retention rates (P<0.05 and P<0.05, respectively), while OS had smaller renal and hepatic retention rates (P<0.01 and P<0.01, respectively). Additionally, the bone retention rate of OS was significantly higher (P<0.01).ConclusionsThe mixing matrix correction improves pharmacokinetic modeling and thus enables a more accurate assessment of the biodistribution of fluorescently labeled pharmaceuticals by μCT-FMT.

English

Imaging may be referred to as the &lsquo;eyes of science&rsquo; as it provides scientists with highly informative multi-dimensional and multi-parameter data usually invisible to the naked eye. As instrumentation technologies and genetic engineering advances, it&rsquo;s possible in modern times to observe and image highly dynamic biochemistry processes. This paper reviews positron emission tomography (PET), magnetic resonance imaging (MRI), computed tomography (CT), ultrasound, visible light microscopy, bioluminescence (BLI) and fluorescence mediated tomography (FMT) imaging techniques, highlighting the principles behind the operation of each technique, their major strengths and drawbacks. With the enhancement of the existing techniques and evolution of new ones, the future possibility of refined view of systems invisible to naked human eye is promising. More so is when two or more techniques are combined in biological systems analysis. &nbsp; Key words: Optical, imaging, microscopy, resolution, fluorescence.

Targeting Endothelin Receptors in a Murine Model of Myocardial Infarction Using a Small Molecular Fluorescent Probe.

The endothelin (ET) axis plays a pivotal role in cardiovascular diseases. Enhanced levels of circulating ET-1 have been correlated with an inferior clinical outcome after myocardial infarction (MI) in humans. Thus, the evaluation of endothelin-A receptor (ETAR) expression over time in the course of myocardial injury and healing may offer valuable information toward the understanding of the ET axis involvement in MI. We developed an approach to track the expression of ETAR with a customized molecular imaging probe in a murine model of MI. The small molecular probe based on the ETAR-selective antagonist 3-(1,3-benzodioxol-5-yl)-5-hydroxy-5-(4-methoxyphenyl)-4-[(3,4,5-trimethoxyphenyl)methyl]-2(5H)-furanone (PD156707) was labeled with fluorescent dye, IRDye800cw. Mice undergoing permanent ligation of the left anterior descending artery (LAD) were investigated at day 1, 7, and 21 post surgery after receiving an intravenous injection of the ETAR probe. Cryosections of explanted hearts were analyzed by cryotome-based CCD, and fluorescence reflectance imaging (FRI) and fluorescence signal intensities (SI) were extracted. Fluorescence-mediated tomography (FMT) imaging was performed to visualize probe distribution in the target region in vivo. An enhanced fluorescence signal intensity in the infarct area was detected in cryoCCD images as early as day 1 after surgery and intensified up to 21 days post MI. FRI was capable of detecting significantly enhanced SI in infarcted regions of hearts 7 days after surgery. In vivo imaging by FMT localized enhanced SI in the apex region of infarcted mouse hearts. We verified the localization of the probe and ETAR within the infarct area by immunohistochemistry (IHC). In addition, neovascularized areas were found in the affected myocardium by CD31 staining. Our study demonstrates that the applied fluorescent probe is capable of delineating ETAR expression over time in affected murine myocardium after MI in vivo and ex vivo.

Target-Specific Fluorescence-Mediated Tomography for Non-Invasive and Dynamic Assessment of Early Neutrophil Infiltration in Murine Experimental Colitis.

The role of neutrophils in the pathogenesis of inflammatory bowel disease (IBD) is still only incompletely understood. Here, we evaluated target-specific fluorescence-mediated tomography (FMT) for visualization of neutrophil infiltration in murine experimental DSS-induced colitis. Colitis was assessed using clinical, endoscopic, and histopathological parameters. Intestinal neutrophil infiltration was determined at day 0, 4, and 10 by targeted FMT after injection of a neutrophil-specific fluorescence-labelled monoclonal antibody (Gr-1). Complementary, immunofluorescence tissue sections with Gr-1 and ELISA-based assessment of tissue myeloperoxidase (MPO) served as the gold standard for the quantification of neutrophil infiltration. Colitic animals showed decreasing body weight, presence of fecal occult blood, and endoscopic signs of inflammation. FMT revealed a significantly increased level of fluorescence only four days after colitis induction as compared to pre-experimental conditions (pmol tracer 73.2 ± 18.1 versus 738.6 ± 80.7; p < 0.05), while neither body weight nor endoscopic assessment showed significant changes at this early time. Confirmatory, post-mortem immunofluorescence studies and measurements of tissue MPO confirmed the presence of increased neutrophil infiltration in colitic mice compared to controls. Concluding, Gr-1 targeted FMT can detect early colonic infiltration of neutrophils in experimental colitis even before clinical symptoms or endoscopic alterations occur. Therefore, FMT might be an important tool for repetitive and non-invasive monitoring of inflammatory cell infiltrate in intestinal inflammation.

Non-invasive Imaging and Modeling of Liver Regeneration After Partial Hepatectomy.

The liver has a unique regenerative capability upon injury or partial resection. The regeneration process comprises a complex interplay between parenchymal and non-parenchymal cells and is tightly regulated at different scales. Thus, we investigated liver regeneration using multi-scale methods by combining non-invasive imaging with immunohistochemical analyses. In this context, non-invasive imaging can provide quantitative data of processes involved in liver regeneration at organ and body scale. We quantitatively measured liver volume recovery after 70% partial hepatectomy (PHx) by micro computed tomography (μCT) and investigated changes in the density of CD68+ macrophages by fluorescence-mediated tomography (FMT) combined with μCT using a newly developed near-infrared fluorescent probe. In addition, angiogenesis and tissue-resident macrophages were analyzed by immunohistochemistry. Based on the results, a model describing liver regeneration and the interactions between different cell types was established. In vivo analysis of liver volume regeneration over 21 days after PHx by μCT imaging demonstrated that the liver volume rapidly increased after PHx reaching a maximum at day 14 and normalizing until day 21. An increase in CD68+ macrophage density in the liver was detected from day 4 to day 8 by combined FMT-μCT imaging, followed by a decline towards control levels between day 14 and day 21. Immunohistochemistry revealed the highest angiogenic activity at day 4 after PHx that continuously declined thereafter, whereas the density of tissue-resident CD169+ macrophages was not altered. The simulated time courses for volume recovery, angiogenesis and macrophage density reflect the experimental data describing liver regeneration after PHx at organ and tissue scale. In this context, our study highlights the importance of non-invasive imaging for acquiring quantitative organ scale data that enable modeling of liver regeneration.

Abstract 051: A New Model of Murine Stasis Pulmonary Thromboembolism in vivo With Assessment by Noninvasive Multimodal Molecular-Structural Imaging

Objective: Pulmonary embolism (PE) is a life-threatening cardiovascular disease that urgently requires improved diagnostic, prognostic and therapeutic options. Here we established and investigated a novel murine PE model based on embolization of a stasis-induced venous thrombus (VT). We further utilized in vivo molecular-structural imaging of fibrin accessibility to assess the healing status of the PE. Methods: Stasis VT was induced in the femoral veins of donor C57BL6 mice. VT was resected 24 hours (h) after induction and then intravenously injected into recipient mice to create PE. Pulse-wave Doppler of pulmonary artery flow was utilized to evaluate the pulmonary arterial pressure (PAP) before and 24h after embolization. One day after embolization, the fibrin-specific probe FTP11-CyAm7 (750/767nm ex/em, 150 nmoles/kg i.v.) assessed the FTP binding and spatial localization of PE via integrated fluorescence-mediated tomography and computed tomography (FMT-CT) pulmonary angiography. Mice were sacrificed 1, 4, 8, or 21 days following stasis-VT embolization, and underwent histological analysis of the PE. Results: Femoral vein stasis-VT exhibited similar temporal inflammatory neutrophil and monocyte/macrophage infiltration and FTP11 binding (fibrin accessibility) kinetics compared to inferior vena cava stasis-VT models. The mean VT width and length were 2.4±0.2mm and 0.65±0.02mm, respectively (n=29 thrombi). Noninvasive ultrasound assessment of the mouse PAP, assessed via the pulmonary acceleration time (PAT), showed a 22% decrease (p=0.03) in pre-to-post-PE PAT times, consistent with increased PAP due to PE. FMT-CT imaging using FTP11 demonstrated accessible fibrin in day 1 PE and visualized PE confined to the lungs, primarily (91%) in the right lung. Ex vivo fluorescence reflectance imaging and histological analysis verified the location PE reported by FMT-CT. Conclusions: These data validate the first in vivo model of PE based on stasis VT formed in vivo. The model reproducibly generates a survival PE that harbors accessible fibrin in recent stasis VT lodged in the lung, and acute increases PAP. This new clinically relevant model of PE is anticipated to facilitate research of acute PE and chronic thromboembolic pulmonary hypertension.

Abstract 301: A New Model of Murine Stasis Pulmonary Thromboembolism in vivo With Assessment by Noninvasive Multimodal Molecular-Structural Imaging

Objective: Pulmonary embolism (PE) is a life-threatening cardiovascular disease that urgently requires improved diagnostic, prognostic and therapeutic options. Here we established and investigated a novel murine PE model based on embolization of a stasis-induced venous thrombus (VT). We further utilized in vivo molecular-structural imaging of fibrin accessibility to assess the healing status of the PE. Methods: Stasis VT was induced in the femoral veins of donor C57BL6 mice. VT was resected 24 hours (h) after induction and then intravenously injected into recipient mice to create PE. Pulse-wave Doppler of pulmonary artery flow was utilized to evaluate the pulmonary arterial pressure (PAP) before and 24h after embolization. One day after embolization, the fibrin-specific probe FTP11-CyAm7 (750/767nm ex/em, 150 nmoles/kg i.v.) assessed the FTP binding and spatial localization of PE via integrated fluorescence-mediated tomography and computed tomography (FMT-CT) pulmonary angiography. Mice were sacrificed 1, 4, 8, or 21 days following stasis-VT embolization, and underwent histological analysis of the PE. Results: Femoral vein stasis-VT exhibited similar temporal inflammatory neutrophil and monocyte/macrophage infiltration and FTP11 binding (fibrin accessibility) kinetics compared to inferior vena cava stasis-VT models. The mean VT width and length were 2.4±0.2mm and 0.65±0.02mm, respectively (n=29 thrombi). Noninvasive ultrasound assessment of the mouse PAP, assessed via the pulmonary acceleration time (PAT), showed a 22% decrease (p=0.03) in pre-to-post-PE PAT times, consistent with increased PAP due to PE. FMT-CT imaging using FTP11 demonstrated accessible fibrin in day 1 PE and visualized PE confined to the lungs, primarily (91%) in the right lung. Ex vivo fluorescence reflectance imaging and histological analysis verified the location PE reported by FMT-CT. Conclusions: These data validate the first in vivo model of PE based on stasis VT formed in vivo. The model reproducibly generates a survival PE that harbors accessible fibrin in recent stasis VT lodged in the lung, and acute increases PAP. This new clinically relevant model of PE is anticipated to facilitate research of acute PE and chronic thromboembolic pulmonary hypertension.

Fluorescence-mediated Tomography for the Detection and Quantification of Macrophage-related Murine Intestinal Inflammation.

Murine models of disease are indispensable to scientific research. However, many diagnostic tools such as endoscopy or tomographic imaging are not routinely employed in animal models. Conventional experimental readouts often rely on post mortem and ex vivo analyses, which prevent intra-individual follow-up examinations and increase the number of study animals needed. Fluorescence-mediated tomography enables the non-invasive, repetitive, quantitative, three-dimensional assessment of fluorescent probes. It is highly sensitive and permits the use of molecular makers, which allows for the specific detection and characterization of distinct molecular targets. In particular, targeted probes represent an innovative tool for analyzing gene activation and protein expression in inflammation, autoimmune disease, infection, vascular disease, cell migration, tumorigenesis, etc. In this article, we provide step-by-step instructions on this sophisticated imaging technology for the in vivo detection and characterization of inflammation (i.e., F4/80-positive macrophage infiltration) in a widely used murine model of intestinal inflammation. This technique might also be used in other research areas, such as immune cell or stem cell tracking.

Fluorescence-mediated Tomography for the Detection and Quantification of Macrophage-related Murine Intestinal Inflammation

Fluorescence-mediated Tomography for the Detection and Quantification of Macrophage-related Murine Intestinal Inflammation

Front Cover: Sensitivity and accuracy of hybrid fluorescencemediated tomography in deep tissue regions (J. Biophotonics 9/2017)

Front Cover: Sensitivity and accuracy of hybrid fluorescencemediated tomography in deep tissue regions (J. Biophotonics 9/2017)

Sensitivity and accuracy of hybrid fluorescence-mediated tomography in deep tissue regions.

Fluorescence-mediated tomography (FMT) enables noninvasive assessment of the three-dimensional distribution of near-infrared fluorescence in mice. The combination with micro-computed tomography (µCT) provides anatomical data, enabling improved fluorescence reconstruction and image analysis. The aim of our study was to assess sensitivity and accuracy of µCT-FMT under realistic in vivo conditions in deeply-seated regions. Accordingly, we acquired fluorescence reflectance images (FRI) and µCT-FMT scans of mice which were prepared with rectal insertions with different amounts of fluorescent dye. Default and high-sensitivity scans were acquired and background signal was analyzed for three FMT channels (670nm, 745 nm, and 790 nm). Analysis was performed for the original and an improved FMT reconstruction using the µCT data. While FRI and the original FMT reconstruction could detect 100 pmol, the improved FMT reconstruction could detect 10 pmol and significantly improved signal localization. By using a finer sampling grid and increasing the exposure time, the sensitivity could be further improved to detect 0.5 pmol. Background signal was highest in the 670nm channel and most prominent in the gastro-intestinal tract and in organs with high relative amounts of blood. In conclusion, we show that µCT-FMT allows sensitive and accurate assessment of fluorescence in deep tissue regions.

Targeting distinct myeloid cell populations invivo using polymers, liposomes and microbubbles.

Identifying intended or accidental cellular targets for drug delivery systems is highly relevant for evaluating therapeutic and toxic effects. However, limited knowledge exists on the distribution of nano- and micrometer-sized carrier systems at the cellular level in different organs. We hypothesized that clinically relevant carrier materials, differing in composition and size, are able to target distinct myeloid cell subsets that control inflammatory processes, such as macrophages, neutrophils, monocytes and dendritic cells. Therefore, we analyzed the biodistribution and invivo cellular uptake of intravenously injected poly(N-(2-hydroxypropyl) methacrylamide) polymers, PEGylated liposomes and poly(butyl cyanoacrylate) microbubbles in mice, using whole-body imaging (computed tomography - fluorescence-mediated tomography), intra-organ imaging (intravital multi-photon microscopy) and cellular analysis (flow cytometry of blood, liver, spleen, lung and kidney). While the three carrier materials shared accumulation in tissue macrophages in liver and spleen, they notably differed in uptake by other myeloid subsets. Kupffer cells and splenic red pulp macrophages rapidly take up microbubbles. Liposomes efficiently reach dendritic cells in liver, lung and kidney. Polymers exhibit the longest circulation half-life and target endothelial cells in the liver, neutrophils and alveolar macrophages. The identification of such previously unrecognized target cell populations might open up new avenues for more efficient drug delivery.

The 5A apolipoprotein A-I (apoA-I) mimetic peptide ameliorates experimental colitis by regulating monocyte infiltration.

New therapies for inflammatory bowel disease (IBD) are highly desirable. As apolipoprotein (apo)A-I mimetic peptides are beneficial in several animal models of inflammation, we hypothesized that they might be effective at inhibiting murine colitis. Daily injections of 5A peptide, a synthetic bihelical apoA-I mimetic dissolved in PBS, or PBS alone were administered to C57BL/6 mice fed 3% (w v(-1) ) dextran sodium sulfate (DSS) in drinking water or healthy controls. Daily treatment with 5A peptide potently restricted DSS-induced inflammation, as indicated by improved disease activity indices and colon histology, as well as decreased intestinal tissue myeloperoxidase levels and plasma TNFα and IL-6 concentrations. Additionally, plasma levels of monocyte chemoattractant protein-1 and the monocyte expression of adhesion-mediating molecule CD11b were down-regulated, pro-inflammatory CD11b(+) /Ly6c(high) monocytes were decreased, and the number of intestinal monocytes was reduced in 5A peptide-treated animals as determined by intravital macrophage-related peptide-8/14-directed fluorescence-mediated tomography and post-mortem immunhistochemical F4/80 staining. Intravital fluorescence microscopy of colonic microvasculature demonstrated inhibitory effects of 5A peptide on leukocyte adhesion accompanied by reduced plasma levels of the soluble adhesion molecule sICAM-1. In vitro 5A peptide reduced monocyte adhesion and transmigration in TNFα-stimulated monolayers of human intestinal microvascular endothelial cells. Increased susceptibility to DSS-induced inflammation was noted in apoA-I(-/-) mice. The 5A peptide is effective at ameliorating murine colitis by preventing intestinal monocyte infiltration and activation. These findings point to apoA-I mimetics as a potential treatment approach for IBD.

Diagnostic imaging advances in murine models of colitis.

Inflammatory bowel diseases (IBD) such as Crohn's disease and ulcerative colitis are chronic-remittent inflammatory disorders of the gastrointestinal tract still evoking challenging clinical diagnostic and therapeutic situations. Murine models of experimental colitis are a vital component of research into human IBD concerning questions of its complex pathogenesis or the evaluation of potential new drugs. To monitor the course of colitis, to the present day, classical parameters like histological tissue alterations or analysis of mucosal cytokine/chemokine expression often require euthanasia of animals. Recent advances mean revolutionary non-invasive imaging techniques for in vivo murine colitis diagnostics are increasingly available. These novel and emerging imaging techniques not only allow direct visualization of intestinal inflammation, but also enable molecular imaging and targeting of specific alterations of the inflamed murine mucosa. For the first time, in vivo imaging techniques allow for longitudinal examinations and evaluation of intra-individual therapeutic response. This review discusses the latest developments in the different fields of ultrasound, molecularly targeted contrast agent ultrasound, fluorescence endoscopy, confocal laser endomicroscopy as well as tomographic imaging with magnetic resonance imaging, computed tomography and fluorescence-mediated tomography, discussing their individual limitations and potential future diagnostic applications in the management of human patients with IBD.

GPU-accelerated adjoint algorithmic differentiation

Many scientific problems such as classifier training or medical image reconstruction can be expressed as minimization of differentiable real-valued cost functions and solved with iterative gradient-based methods. Adjoint algorithmic differentiation (AAD) enables automated computation of gradients of such cost functions implemented as computer programs. To backpropagate adjoint derivatives, excessive memory is potentially required to store the intermediate partial derivatives on a dedicated data structure, referred to as the “tape”. Parallelization is difficult because threads need to synchronize their accesses during taping and backpropagation. This situation is aggravated for many-core architectures, such as Graphics Processing Units (GPUs), because of the large number of light-weight threads and the limited memory size in general as well as per thread. We show how these limitations can be mediated if the cost function is expressed using GPU-accelerated vector and matrix operations which are recognized as intrinsic functions by our AAD software. We compare this approach with naive and vectorized implementations for CPUs. We use four increasingly complex cost functions to evaluate the performance with respect to memory consumption and gradient computation times. Using vectorization, CPU and GPU memory consumption could be substantially reduced compared to the naive reference implementation, in some cases even by an order of complexity. The vectorization allowed usage of optimized parallel libraries during forward and reverse passes which resulted in high speedups for the vectorized CPU version compared to the naive reference implementation. The GPU version achieved an additional speedup of 7.5±4.4, showing that the processing power of GPUs can be utilized for AAD using this concept. Furthermore, we show how this software can be systematically extended for more complex problems such as nonlinear absorption reconstruction for fluorescence-mediated tomography. Program summaryProgram title: AD-GPUCatalogue identifier: AEYX_v1_0Program summary URL: http://cpc.cs.qub.ac.uk/summaries/AEYX_v1_0.htmlProgram obtainable from: CPC Program Library, Queen’s University, Belfast, N. IrelandLicensing provisions: Standard CPC licence, http://cpc.cs.qub.ac.uk/licence/licence.htmlNo. of lines in distributed program, including test data, etc.: 16715No. of bytes in distributed program, including test data, etc.: 143683Distribution format: tar.gzProgramming language: C++ and CUDA.Computer: Any computer with a compatible C++ compiler and a GPU with CUDA capability 3.0 or higher.Operating system: Windows 7 or Linux.RAM: 16 GbyteClassification: 4.9, 4.12, 6.1, 6.5.External routines: CUDA 6.5, Intel MKL (optional) and routines from BLAS, LAPACK and CUBLASNature of problem: Gradients are required for many optimization problems, e.g. classifier training or nonlinear image reconstruction. Often, the function, of which the gradient is required, can be implemented as a computer program. Then, algorithmic differentiation methods can be used to compute the gradient. Depending on the approach this may result in excessive requirements of computational resources, i.e. memory and arithmetic computations. GPUs provide massive computational resources but require special considerations to distribute the workload onto many light-weight threads.Solution method: Adjoint algorithmic differentiation allows efficient computation of gradients of cost functions given as computer programs. The gradient can be theoretically computed using a similar amount of arithmetic operations as one function evaluation. Optimal usage of parallel processors and limited memory is a major challenge which can be mediated by the use of vectorization.Restrictions: To use the GPU-accelerated adjoint algorithmic differentiation method, the cost function must be implemented using the provided AD-GPU intrinsics for matrix and vector operations. Unusual features:GPU-acceleration.Additional comments: The code uses some features of C++11, e.g. std::shared ptr. Alternatively, the boost library can be used.Running time: The time to run the example program is a few minutes or up to a few hours to reproduce the performance measurements.