Deep inference of seabird dives from GPS-only records: Performance and generalization properties.

At-sea behaviour of seabirds have received significant attention in ecology over the last decades as it is a key process in the ecology and fate of these populations. It is also, through the position of top predator that these species often occupy, a relevant and integrative indicator of the dynamics of the marine ecosystems they rely on. Seabird trajectories are recorded through the deployment of GPS, and a variety of statistical approaches have been tested to infer probable behaviours from these location data. Recently, deep learning tools have shown promising results for the segmentation and classification of animal behaviour from trajectory data. Yet, these approaches have not been widely used and investigation is still needed to identify optimal network architecture and to demonstrate their generalization properties. From a database of about 300 foraging trajectories derived from GPS data deployed simultaneously with pressure sensors for the identification of dives, this work has benchmarked deep neural network architectures trained in a supervised manner for the prediction of dives from trajectory data. It first confirms that deep learning allows better dive prediction than usual methods such as Hidden Markov Models. It also demonstrates the generalization properties of the trained networks for inferring dives distribution for seabirds from other colonies and ecosystems. In particular, convolutional networks trained on Peruvian boobies from a specific colony show great ability to predict dives of boobies from other colonies and from distinct ecosystems. We further investigate accross-species generalization using a transfer learning strategy known as 'fine-tuning'. Starting from a convolutional network pre-trained on Guanay cormorant data reduced by two the size of the dataset needed to accurately predict dives in a tropical booby from Brazil. We believe that the networks trained in this study will provide relevant starting point for future fine-tuning works for seabird trajectory segmentation.

SegSRGAN: Super-resolution and segmentation using generative adversarial networks - Application to neonatal brain MRI.

BACKGROUND AND OBJECTIVE: One of the main issues in the analysis of clinical neonatal brain MRI is the low anisotropic resolution of the data. In most MRI analysis pipelines, data are first re-sampled using interpolation or single image super-resolution techniques and then segmented using (semi-)automated approaches. In other words, image reconstruction and segmentation are then performed separately. In this article, we propose a methodology and a software solution for carrying out simultaneously high-resolution reconstruction and segmentation of brain MRI data. METHODS: Our strategy mainly relies on generative adversarial networks. The network architecture is described in detail. We provide information about its implementation, focusing on the most crucial technical points (whereas complementary details are given in a dedicated GitHub repository). We illustrate the behavior of the proposed method for cortex analysis from neonatal MR images. RESULTS: The results of the method, evaluated quantitatively (Dice, peak signal-to-noise ratio, structural similarity, number of connected components) and qualitatively on a research dataset (dHCP) and a clinical one (Epirmex), emphasize the relevance of the approach, and its ability to take advantage of data-augmentation strategies. CONCLUSIONS: Results emphasize the potential of our proposed method/software with respect to practical medical applications. The method is provided as a freely available software tool, which allows one to carry out his/her own experiments, and involve the method for the super-resolution reconstruction and segmentation of arbitrary cerebral structures from any MR image dataset.

Multiscale brain MRI super-resolution using deep 3D convolutional networks.

The purpose of super-resolution approaches is to overcome the hardware limitations and the clinical requirements of imaging procedures by reconstructing high-resolution images from low-resolution acquisitions using post-processing methods. Super-resolution techniques could have strong impacts on structural magnetic resonance imaging when focusing on cortical surface or fine-scale structure analysis for instance. In this paper, we study deep three-dimensional convolutional neural networks for the super-resolution of brain magnetic resonance imaging data. First, our work delves into the relevance of several factors in the performance of the purely convolutional neural network-based techniques for the monomodal super-resolution: optimization methods, weight initialization, network depth, residual learning, filter size in convolution layers, number of the filters, training patch size and number of training subjects. Second, our study also highlights that one single network can efficiently handle multiple arbitrary scaling factors based on a multiscale training approach. Third, we further extend our super-resolution networks to the multimodal super-resolution using intermodality priors. Fourth, we investigate the impact of transfer learning skills onto super-resolution performance in terms of generalization among different datasets. Lastly, the learnt models are used to enhance real clinical low-resolution images. Results tend to demonstrate the potential of deep neural networks with respect to practical medical image applications.

CNN-Based Target Recognition and Identification for Infrared Imaging in Defense Systems.

Convolutional neural networks (CNNs) have rapidly become the state-of-the-art models for image classification applications. They usually require large groundtruthed datasets for training. Here, we address object identification and recognition in the wild for infrared (IR) imaging in defense applications, where no such large-scale dataset is available. With a focus on robustness issues, especially viewpoint invariance, we introduce a compact and fully convolutional CNN architecture with global average pooling. We show that this model trained from realistic simulation datasets reaches a state-of-the-art performance compared with other CNNs with no data augmentation and fine-tuning steps. We also demonstrate a significant improvement in the robustness to viewpoint changes with respect to an operational support vector machine (SVM)-based scheme.

Completing fishing monitoring with spaceborne Vessel Detection System (VDS) and Automatic Identification System (AIS) to assess illegal fishing in Indonesia.

The Indonesian fisheries management system is now equipped with the state-of-the-art technologies to deter and combat Illegal, Unreported and Unregulated (IUU) fishing. Since October 2014, non-cooperative fishing vessels can be detected from spaceborne Vessel Detection System (VDS) based on high resolution radar imagery, which directly benefits to coordinated patrol vessels in operation context. This study attempts to monitor the amount of illegal fishing in the Arafura Sea based on this new source of information. It is analyzed together with Vessel Monitoring System (VMS) and satellite-based Automatic Identification System (Sat-AIS) data, taking into account their own particularities. From October 2014 to March 2015, i.e. just after the establishment of a new moratorium by the Indonesian authorities, the estimated share of fishing vessels not carrying VMS, thus being illegal, ranges from 42 to 47%. One year later in January 2016, this proportion decreases and ranges from 32 to 42%.

Coupling spectral analysis and hidden Markov models for the segmentation of behavioural patterns.

BACKGROUND: Movement pattern variations are reflective of behavioural switches, likely associated with different life history traits in response to the animals' abiotic and biotic environment. Detecting these can provide rich information on the underlying processes driving animal movement patterns. However, extracting these signals from movement time series, requires tools that objectively extract, describe and quantify these behaviours. The inference of behavioural modes from movement patterns has been mainly addressed through hidden Markov models. Until now, the metrics implemented in these models did not allow to characterize cyclic patterns directly from the raw time series. To address these challenges, we developed an approach to i) extract new metrics of cyclic behaviours and activity levels from a time-frequency analysis of movement time series, ii) implement the spectral signatures of these cyclic patterns and activity levels into a HMM framework to identify and classify latent behavioural states. RESULTS: To illustrate our approach, we applied it to 40 high-resolution European sea bass depth time series. Our results showed that the fish had different activity regimes, which were also associated (or not) with the spectral signature of different environmental cycles. Tidal rhythms were observed when animals tended to be less active and dived shallower. Conversely, animals exhibited a diurnal behaviour when more active and deeper in the water column. The different behaviours were well defined and occurred at similar periods throughout the annual cycle amongst individuals, suggesting these behaviours are likely related to seasonal functional behaviours (e.g. feeding, migrating and spawning). CONCLUSIONS: The innovative aspects of our method lie within the combined use of powerful, but generic, mathematical tools (spectral analysis and hidden Markov Models) to extract complex behaviours from 1-D movement time series. It is fully automated which makes it suitable for analyzing large datasets. HMMs also offer the flexibility to include any additional variable in the segmentation process (e.g. environmental features, location coordinates). Thus, our method could be widely applied in the bio-logging community and contribute to prime issues in movement ecology (e.g. habitat requirements and selection, site fidelity and dispersal) that are crucial to inform mitigation, management and conservation strategies.

Validation and Parameter Sensitivity Tests for Reconstructing Swell Field Based on an Ensemble Kalman Filter.

The swell propagation model built on geometric optics is known to work well when simulating radiated swells from a far located storm. Based on this simple approximation, satellites have acquired plenty of large samples on basin-traversing swells induced by fierce storms situated in mid-latitudes. How to routinely reconstruct swell fields with these irregularly sampled observations from space via known swell propagation principle requires more examination. In this study, we apply 3-h interval pseudo SAR observations in the ensemble Kalman filter (EnKF) to reconstruct a swell field in ocean basin, and compare it with buoy swell partitions and polynomial regression results. As validated against in situ measurements, EnKF works well in terms of spatial-temporal consistency in far-field swell propagation scenarios. Using this framework, we further address the influence of EnKF parameters, and perform a sensitivity analysis to evaluate estimations made under different sets of parameters. Such analysis is of key interest with respect to future multiple-source routinely recorded swell field data. Satellite-derived swell data can serve as a valuable complementary dataset to in situ or wave re-analysis datasets.

Generalized Pareto for Pattern-Oriented Random Walk Modelling of Organisms' Movements.

How organisms move and disperse is crucial to understand how population dynamics relates to the spatial heterogeneity of the environment. Random walk (RW) models are typical tools to describe movement patterns. Whether Lévy or alternative RW better describes forager movements is keenly debated. We get around this issue using the Generalized Pareto Distribution (GPD). GPD includes as specific cases Normal, exponential and power law distributions, which underlie Brownian, Poisson-like and Lévy walks respectively. Whereas previous studies typically confronted a limited set of candidate models, GPD lets the most likely RW model emerge from the data. We illustrate the wide applicability of the method using GPS-tracked seabird foraging movements and fishing vessel movements tracked by Vessel Monitoring System (VMS), both collected in the Peruvian pelagic ecosystem. The two parameters from the fitted GPD, a scale and a shape parameter, provide a synoptic characterization of the observed movement in terms of characteristic scale and diffusive property. They reveal and quantify the variability, among species and individuals, of the spatial strategies selected by predators foraging on a common prey field. The GPD parameters constitute relevant metrics for (1) providing a synthetic and pattern-oriented description of movement, (2) using top predators as ecosystem indicators and (3) studying the variability of spatial behaviour among species or among individuals with different personalities.

Broad impacts of fine-scale dynamics on seascape structure from zooplankton to seabirds.

In marine ecosystems, like most natural systems, patchiness is the rule. A characteristic of pelagic ecosystems is that their 'substrate' consists of constantly moving water masses, where ocean surface turbulence creates ephemeral oases. Identifying where and when hotspots occur and how predators manage those vagaries in their preyscape is challenging because wide-ranging observations are lacking. Here we use a unique data set, gathering high-resolution and wide-range acoustic and GPS-tracking data. We show that the upper ocean dynamics at scales less than 10 km play the foremost role in shaping the seascape from zooplankton to seabirds. Short internal waves (100 m-1 km) play a major role, while submesoscale (~1-20 km) and mesoscale (~20-100 km) turbulence have a comparatively modest effect. Predicted changes in surface stratification due to global change are expected to have an impact on the number and intensity of physical structures and thus biological interactions from plankton to top predators.

Hidden Markov models: the best models for forager movements?

One major challenge in the emerging field of movement ecology is the inference of behavioural modes from movement patterns. This has been mainly addressed through Hidden Markov models (HMMs). We propose here to evaluate two sets of alternative and state-of-the-art modelling approaches. First, we consider hidden semi-Markov models (HSMMs). They may better represent the behavioural dynamics of foragers since they explicitly model the duration of the behavioural modes. Second, we consider discriminative models which state the inference of behavioural modes as a classification issue, and may take better advantage of multivariate and non linear combinations of movement pattern descriptors. For this work, we use a dataset of >200 trips from human foragers, Peruvian fishermen targeting anchovy. Their movements were recorded through a Vessel Monitoring System (∼1 record per hour), while their behavioural modes (fishing, searching and cruising) were reported by on-board observers. We compare the efficiency of hidden Markov, hidden semi-Markov, and three discriminative models (random forests, artificial neural networks and support vector machines) for inferring the fishermen behavioural modes, using a cross-validation procedure. HSMMs show the highest accuracy (80%), significantly outperforming HMMs and discriminative models. Simulations show that data with higher temporal resolution, HSMMs reach nearly 100% of accuracy. Our results demonstrate to what extent the sequential nature of movement is critical for accurately inferring behavioural modes from a trajectory and we strongly recommend the use of HSMMs for such purpose. In addition, this work opens perspectives on the use of hybrid HSMM-discriminative models, where a discriminative setting for the observation process of HSMMs could greatly improve inference performance.

Geodesics-based image registration: applications to biological and medical images depicting concentric ring patterns.

In many biological or medical applications, images that contain sequences of shapes are common. The existence of high inter-individual variability makes their interpretation complex. In this paper, we address the computer-assisted interpretation of such images and we investigate how we can remove or reduce these image variabilities. The proposed approach relies on the development of an efficient image registration technique. We first show the inadequacy of state-of-the-art intensity-based and feature-based registration techniques for the considered image datasets. Then, we propose a robust variational method which benefits from the geometrical information present in this type of images. In the proposed non-rigid geodesics-based registration, the successive shapes are represented by a level-set representation, which we rely on to carry out the registration. The successive level sets are regarded as elements in a shape space and the corresponding matching is that of the optimal geodesic path. The proposed registration scheme is tested on synthetic and real images. The comparison against results of state-of-the-art methods proves the relevance of the proposed method for this type of images.

Preparation techniques alter the mineral and organic fractions of fish otoliths: insights using Raman micro-spectrometry.

The high spatial resolution analysis of the mineral and organic composition of otoliths using Raman micro-spectrometry involves rigorous protocols for sample preparation previously established for microchemistry and trace elements analyses. These protocols often include otolith embedding in chemically neutral resin (i.e., resins which do not contain, in detectable concentration, elements usually sought in the otoliths). Such embedding may however induce organic contamination. In this paper, Raman micro-spectrometry reveals the presence of organic contamination onto the surface obtained from the use of epoxy resin, specifically Araldite. This contamination level varies depending on otolith structures. Core and checks, known as structural discontinuities, exhibit the most important level of contaminations. Our results suggest that otolith embedding with resin affects the organic matrix of the otolith, probably through an infiltration of the resin in the crystalline structure. The interpretation of chemical otolith signatures, especially Raman otolith signatures, and stable isotope analyses should then be revised in light of these results. In this respect, we propose a method for the correction of Raman otolith signatures for contamination effects.

Protein interaction hotspot identification using sequence-based frequency-derived features.

Finding good descriptors, capable of discriminating hotspot residues from others, is still a challenge in many attempts to understand protein interaction. In this paper, descriptors issued from the analysis of amino acid sequences using digital signal processing (DSP) techniques are shown to be as good as those derived from protein tertiary structure and/or information on the complex. The simulation results show that our descriptors can be used separately to predict hotspots, via a random forest classifier, with an accuracy of 79% and a precision of 75%. They can also be used jointly with features derived from tertiary structures to boost the performance up to an accuracy of 82% and a precision of 80%.

Shedding light on fish otolith biomineralization using a bioenergetic approach.

Otoliths are biocalcified bodies connected to the sensory system in the inner ears of fish. Their layered, biorhythm-following formation provides individual records of the age, the individual history and the natural environment of extinct and living fish species. Such data are critical for ecosystem and fisheries monitoring. They however often lack validation and the poor understanding of biomineralization mechanisms has led to striking examples of misinterpretations and subsequent erroneous conclusions in fish ecology and fisheries management. Here we develop and validate a numerical model of otolith biomineralization. Based on a general bioenergetic theory, it disentangles the complex interplay between metabolic and temperature effects on biomineralization. This model resolves controversial issues and explains poorly understood observations of otolith formation. It represents a unique simulation tool to improve otolith interpretation and applications, and, beyond, to address the effects of both climate change and ocean acidification on other biomineralizing organisms such as corals and bivalves.

The impact of metabolism on stable isotope dynamics: a theoretical framework.

Stable isotope analysis is a powerful tool used for reconstructing individual life histories, identifying food-web structures and tracking flow of elemental matter through ecosystems. The mechanisms determining isotopic incorporation rates and discrimination factors are, however, poorly understood which hinders a reliable interpretation of field data when no experimental data are available. Here, we extend dynamic energy budget (DEB) theory with a limited set of new assumptions and rules in order to study the impact of metabolism on stable isotope dynamics in a mechanistic way. We calculate fluxes of stable isotopes within an organism by following fluxes of molecules involved in a limited number of macrochemical reactions: assimilation, growth but also structure turnover that is here explicitly treated. Two mechanisms are involved in the discrimination of isotopes: (i) selection of molecules occurs at the partitioning of assimilation, growth and turnover into anabolic and catabolic sub-fluxes and (ii) reshuffling of atoms occurs during transformations. Such a framework allows for isotopic routing which is known as a key, but poorly studied, mechanism. As DEB theory specifies the impact of environmental conditions and individual state on molecule fluxes, we discuss how scenario analysis within this framework could help reveal common mechanisms across taxa.

Variational region-based segmentation using multiple texture statistics.

This paper investigates variational region-level criterion for supervised and unsupervised texture-based image segmentation. The focus is given to the demonstration of the effectiveness and robustness of this region-based formulation compared to most common variational approaches. The main contributions of this global criterion are twofold. First, the proposed methods circumvent a major problem related to classical texture based segmentation approaches. Existing methods, even if they use different and various texture features, are mainly stated as the optimization of a criterion evaluating punctual pixel likelihoods or similarity measure computed within a local neighborhood. These approaches require sufficient dissimilarity between the considered texture features. An additional limitation is the choice of the neighborhood size and shape. These two parameters and especially the neighborhood size significantly influence the classification performances: the neighborhood must be large enough to capture texture structures and small enough to guarantee segmentation accuracy. These parameters are often set experimentally. These limitations are mitigated with the proposed variational methods stated at the region-level. It resorts to an energy criterion defined on image where regions are characterized by nonparametric distributions of their responses to a set of filters. In the supervised case, the segmentation algorithm consists in the minimization of a similarity measure between region-level statistics and texture prototypes and a boundary based functional that imposes smoothness and regularity on region boundaries. In the unsupervised case, the data-driven term involves the maximization of the dissimilarity between regions. The proposed similarity measure is generic and permits optimally fusing various types of texture features. It is defined as a weighted sum of Kullback-Leibler divergences between feature distributions. The optimization of the proposed variational criteria is carried out using a level-set formulation. The effectiveness and the robustness of this formulation at region-level, compared to classical active contour methods, are evaluated for various Brodatz and natural images.

Understanding otolith biomineralization processes: new insights into microscale spatial distribution of organic and mineral fractions from Raman microspectrometry.

It is generally accepted that the formation of otolith microstructures (L- and D-zones) and in particular the organic and mineral fractions vary on a daily basis. Raman microspectrometry provides a nondestructive technique that can be used to provide structural information on organic and mineral compounds. We applied it to thin otolith sections of hake in order to address the following issues: (1) the simultaneous characterization of variations in the organic and mineral fractions both in the core area and along successive otolith microstructures; (2) elucidation of significant differences between these fractions; (3) quantification of the effects of etching and staining protocols on otolith structures. The primordium appeared as a punctual area depicting higher luminescence and greater concentrations in organic compounds containing CH groups. Sulcus side showed similar composition suggesting that the contact of the otolith with the macula and its orientation in otosac occur rapidly (about 10 days). The characterization of L- and D-zones in the opaque zones indicated that both structures contained organic and aragonitic fractions with cyclic and synchronous variations. Contrary to the results obtained after EDTA etching, L-zones depicted greater concentrations in organic compounds containing CH groups, whereas D-zones appear richer in aragonite. This organic fraction seemed to be revealed by Mutvei's staining and was affected by EDTA etching which suggests that it corresponds to the soluble fraction of organic matrix. Such results indicate that L- and D-zones differ in their respective organic constituents. Raman microspectrometry thus appears as a powerful technique to acquire quantitative information that is required for a better understanding of otolith biomineralization.

Nonparametric motion characterization using causal probabilistic models for video indexing and retrieval.

This paper describes an original approach for content-based video indexing and retrieval. We aim at providing a global interpretation of the dynamic content of video shots without any prior motion segmentation and without any use of dense optic flow fields. To this end, we exploit the spatio-temporal distribution, within a shot, of appropriate local motion-related measurements derived from the spatio-temporal derivatives of the intensity function. These distributions are then represented by causal Gibbs models. To be independent of camera movement, the motion-related measurements are computed in the image sequence generated by compensating the estimated dominant image motion in the original sequence. The statistical modeling framework considered makes the exact computation of the conditional likelihood of a video shot belonging to a given motion or more generally to an activity class feasible. This property allows us to develop a general statistical framework for video indexing and retrieval with query-by-example. We build a hierarchical structure of the processed video database according to motion content similarity. This results in a binary tree where each node is associated to an estimated causal Gibbs model. We consider a similarity measure inspired from Kullback-Leibler divergence. Then, retrieval with query-by-example is performed through this binary tree using the maximum a posteriori (MAP) criterion. We have obtained promising results on a set of various real image sequences.