Broad Spectrum Image Deblurring via an Adaptive Super-Network.

In blurry images, the degree of image blurs may vary drastically due to different factors, such as varying speeds of shaking cameras and moving objects, as well as defects of the camera lens. However, current end-to-end models failed to explicitly take into account such diversity of blurs. This unawareness compromises the specialization at each blur level, yielding sub-optimal deblurred images as well as redundant post-processing. Therefore, how to specialize one model simultaneously at different blur levels, while still ensuring coverage and generalization, becomes an emerging challenge. In this work, we propose Ada-Deblur, a super-network that can be applied to a "broad spectrum" of blur levels with no re-training on novel blurs. To balance between individual blur level specialization and wide-range blur levels coverage, the key idea is to dynamically adapt the network architectures from a single well-trained super-network structure, targeting flexible image processing with different deblurring capacities at test time. Extensive experiments demonstrate that our work outperforms strong baselines by demonstrating better reconstruction accuracy while incurring minimal computational overhead. Besides, we show that our method is effective for both synthetic and realistic blurs compared to these baselines. The performance gap between our model and the state-of-the-art becomes more prominent when testing with unseen and strong blur levels. Specifically, our model demonstrates surprising deblurring performance on these images with PSNR improvements of around 1 dB. Our code is publicly available at https://github.com/wuqiuche/Ada-Deblur.

Dependence of the probability density function of laser radiation power on the scintillation index and the size of a receiver aperture.

Power scintillations of a Gaussian laser beam propagated through a 7 km long horizontal atmospheric path for a wide range of turbulence strengths and different sizes of the receiver aperture were studied experimentally. The probability density function (PDF) and its properties were analyzed for a wide range of scintillation conditions. It was shown that the PDF can be described by the fractional exponential distribution in the strong scintillation regime (scintillation index of power measured on aperture σPIB2>1) for apertures with diameter d < a, where a is the size of the isoplanatic region, and by the gamma distribution in the weak scintillation regime (σPIB2<1), as well as by the lognormal distribution for σPIB2≪1. More than one distribution can be considered as a good approximation for experimental data for some ranges of d/a and σPIB2, but the transition from one distribution to another as the best approximation occurs at certain characteristic values of these parameters. In the strong scintillation regime, the aperture averaging effect resulted in the transition from the fractional exponential distribution to the fractional gamma (FG) distribution when the aperture diameter is about the size of the isoplanatic region (d/a∼1). The FG distribution better approximates the experimental PDF because it accounts for the fact that the probability of zero power values becomes zero due to the averaging effect of the aperture. The fractional gamma distribution is the best approximation of the PDF for a finite aperture when σPIB2>σPIB,crit2=[0.7-0.8], while the gamma distribution is the best approximation for σPIB2<σPIB,crit2.

DoGNet: A deep architecture for synapse detection in multiplexed fluorescence images.

Neuronal synapses transmit electrochemical signals between cells through the coordinated action of presynaptic vesicles, ion channels, scaffolding and adapter proteins, and membrane receptors. In situ structural characterization of numerous synaptic proteins simultaneously through multiplexed imaging facilitates a bottom-up approach to synapse classification and phenotypic description. Objective automation of efficient and reliable synapse detection within these datasets is essential for the high-throughput investigation of synaptic features. Convolutional neural networks can solve this generalized problem of synapse detection, however, these architectures require large numbers of training examples to optimize their thousands of parameters. We propose DoGNet, a neural network architecture that closes the gap between classical computer vision blob detectors, such as Difference of Gaussians (DoG) filters, and modern convolutional networks. DoGNet is optimized to analyze highly multiplexed microscopy data. Its small number of training parameters allows DoGNet to be trained with few examples, which facilitates its application to new datasets without overfitting. We evaluate the method on multiplexed fluorescence imaging data from both primary mouse neuronal cultures and mouse cortex tissue slices. We show that DoGNet outperforms convolutional networks with a low-to-moderate number of training examples, and DoGNet is efficiently transferred between datasets collected from separate research groups. DoGNet synapse localizations can then be used to guide the segmentation of individual synaptic protein locations and spatial extents, revealing their spatial organization and relative abundances within individual synapses. The source code is publicly available: https://github.com/kulikovv/dognet.

Application of 3-D imaging sensor for tracking minipigs in the open field test.

BACKGROUND: The minipig is a promising model in neurobiology and psychopharmacology. However, automated tracking of minipig behavior is still unresolved problem. NEW METHOD: The study was carried out on white, agouti and black (or spotted) minipiglets (n=108) bred in the Institute of Cytology and Genetics. New method of automated tracking of minipig behavior is based on Microsoft Kinect 3-D image sensor and the 3-D image reconstruction with EthoStudio software. The algorithms of distance run and time in the center evaluation were adapted for 3-D image data and new algorithm of vertical activity quantification was developed. RESULTS: The 3-D imaging system successfully detects white, black, spotted and agouti pigs in the open field test (OFT). No effect of sex or color on horizontal (distance run), vertical activities and time in the center was shown. Agouti pigs explored the arena more intensive than white or black animals, respectively. The OFT behavioral traits were compared with the fear reaction to experimenter. Time in the center of the OFT was positively correlated with fear reaction rank (ρ=0.21, p<0.05). Black pigs were significantly more fearful compared with white or agouti animals. COMPARISON WITH EXISTING METHOD: The 3-D imaging system has three advantages over existing automated tracking systems: it avoids perspective distortion, distinguishes animals any color from any background and automatically evaluates vertical activity. CONCLUSION: The 3-D imaging system can be successfully applied for automated measurement of minipig behavior in neurobiological and psychopharmacological experiments.

Spatial learning in the Morris water maze in mice genetically different in the predisposition to catalepsy: the effect of intraventricular treatment with brain-derived neurotrophic factor.

Hereditary catalepsy in mice is accompanied with volume reduction of some brain structures and high vulnerability to inflammatory agents. Here an association between hereditary catalepsy and spatial learning deficit in the Morris water maze (MWM) in adult mouse males of catalepsy-resistant AKR, catalepsy-prone CBA and AKR.CBA-D13Mit76 (D13) strains was studied. Recombinant D13 strain was created by means of the transfer of the CBA-derived allele of the major gene of catalepsy to the AKR genome. D13 mice showed a low MWM performance in the acquisition test and high expression of the gene coding proinflammatory interleukin-6 (Il-6) in the hippocampus and cortex compared with mice of the parental AKR and CBA strains. An acute ivc administration of 300 ng of brain derived neurotrophic factor (BDNF) normalized the performance in the MWM, but did not decrease the high Il-6 gene expression in the brain of D13 mice. These results indicated a possible association between the hereditary catalepsy, MWM performance, BDNF and level of Il-6 mRNA in the brain, although the relation between these characteristics seems to be more complex. D13 recombinant mice with deficit of spatial learning is a promising model for study of the genetic and molecular mechanisms of learning disorders as well as for screening potential cognitive enhancers.

Method of estimation of turbulence characteristic scales.

We propose an optical method that uses phase data of a laser beam obtained from a Shack-Hartmann sensor to estimate both the inner and outer scales of turbulence. The method is based on the sequential analysis of normalized correlation functions of Zernike coefficients. It allows the exclusion C(n)(2) from the analysis and reduces the solution of a two-parameter problem to a sequential solution of two single-parameter problems. The method has been applied to estimate the outer and inner scales of turbulence induced in the water cell.

Association between tryptophan hydroxylase-2 genotype and the antidepressant effect of citalopram and paroxetine on immobility time in the forced swim test in mice.

Tryptophan hydroxylase-2 (TPH2) is the rate limiting enzyme of serotonin synthesis in the brain. The 1473G allele of the C1473G polymorphism in mTPH2 gene is associated with reduced enzyme activity and serotonin synthesis rate in the mouse brain. Here, the influence of the 1473G allele on the antidepressant effect of selective serotonin reuptake inhibitors (SSRIs), citalopram (2.5 or 5.0mg/kg) and paroxetine (5.0 or 10.0mg/kg), in the forced swim test was studied using B6-1473G and B6-1473C congenic mouse lines with the 1473G (decreased TPH2 activity) or 1473C (normal TPH2 activity) alleles, respectively, transferred to the genome of C57BL/6 mouse strain. Paroxetine (5.0 or 10.0mg/kg) and citalopram (2.5 or 5.0mg/kg) decreased immobility time in B6-1473C mice, while both doses of paroxetine and 2.5mg/kg of citaloprame did not alter immobility time in B6-1473G mice. However, 5.0mg/kg of citalopram reduced immobility in B6-1473G mice. The results provided genetic evidence of moderate association between 1473G allele and reduced sensitivity to SSRIs in mice.

Effects of LPS and serotonergic drugs on hygienic behavior in mice.

Hygienic self-grooming is a behavioral adaptation for removing litter particles and pathogenic agents from animal fur and skin. We studied contribution of brain serotonin system into mechanisms regulating hygienic behavior in intact mice and mice with LPS(lipopolysaccharide)-induced sickness. A spot of fluorescent dye was applied on the back of a mouse, and the decrease in its fluorescence served as an index of fur cleaning efficiency estimated using original classifier algorithm. Agonist of 5-HT(1A) receptor (8-OH-DPAT) or 5-HT(2A/2C) receptor (DOI) attenuated fur cleaning at a dose of 1 mg/kg but not of 0.2 mg/kg. MAO-A inhibitor clorgyline decreased hygienic self-grooming at a dose of 10 but not of 5 mg/kg. SSRI paroxetine had no effect while fluoxetine diminished hygienic behavior at the higher dose used (20 mg/kg). Inhibitory effect of LPS treatment (50 µg/kg) on fur cleaning was not altered by administration of p-MPPI (5-HT(1A) receptor antagonist, 1mg/kg) or DOI (1mg/kg) while 8-OH-DPAT (1 mg/kg) produced additive effect. The results suggest the involvement of 5-HT(1A) and 5-HT(2A/2C) brain serotonin receptors and MAO-A in the inhibition of hygienic behavior in mice. However, LPS-induced depression of fur cleaning appeared to be mediated via different mechanisms and enhanced by 5-HT(1A) receptor activation.

Novel approach to the study of fur cleaning in inbred mice: effects of genotype, stress, and lipopolysaccharide.

Body care (grooming) is a behavioral adaptation for removing litter particles, pathogenic microbes, and parasites from animal fur and skin. It also serves as an indicator of animal health. Here a technique of direct measurement of fur cleaning has been developed. A spot of fluorescent dye applied on the back of a mouse was scanned under blue light (450 nm) immediately and rescanned 1, 2, 4, 8, and 24 hours later with a digital camera. The spot fluorescence intensity was measured using ColorScan software, which we developed using a classifier algorithm. The decrease in the spot fluorescence served as an index of fur cleaning, with significant interstrain differences in the dynamics of fur cleaning: mice of C57BL/6, CBA, CC57BR, and DD strains removed the fluorescent spot rapidly (1-2 h) whereas AKR and DBA2 mice did so slowly (24 h). To study the association between fur cleaning and stress or sickness we investigated the effect of restriction stress (for 30 min) and of the bacterial toxin lipopolysaccharide (LPS) on fur cleaning in two fast-cleaning mouse strains, CBA and C57BL/6. Restriction substantially reduced fur cleaning in the CBA mice but had no effect on the C57BL/6 mice. LPS decreased fur cleaning in a dose-dependent manner in both strains. The described technique is fairly simple and sensitive enough to estimate the effects of both stress and LPS treatment. It can be applied to study vulnerability (or resistance) to stressors, pathogenic organisms, and toxic substances.

Automated measurement of spatial preference in the open field test with transmitted lighting.

New modification of the open field was designed to improve automation of the test. The main innovations were: (1) transmitted lighting and (2) estimation of probability to find pixels associated with an animal in the selected region of arena as an objective index of spatial preference. Transmitted (inverted) lighting significantly ameliorated the contrast between an animal and arena and allowed to track white animals with similar efficacy as colored ones. Probability as a measure of preference of selected region was mathematically proved and experimentally verified. A good correlation between probability and classic indices of spatial preference (number of region entries and time spent therein) was shown. The algorithm of calculation of probability to find pixels associated with an animal in the selected region was implemented in the EthoStudio software. Significant interstrain differences in locomotion and the central zone preference (index of anxiety) were shown using the inverted lighting and the EthoStudio software in mice of six inbred strains. The effects of arena shape (circle or square) and a novel object presence in the center of arena on the open field behavior in mice were studied.