Data-Efficient Sensor Upgrade Path Using Knowledge Distillation.

Deep neural networks have achieved state-of-the-art performance in image classification. Due to this success, deep learning is now also being applied to other data modalities such as multispectral images, lidar and radar data. However, successfully training a deep neural network requires a large reddataset. Therefore, transitioning to a new sensor modality (e.g., from regular camera images to multispectral camera images) might result in a drop in performance, due to the limited availability of data in the new modality. This might hinder the adoption rate and time to market for new sensor technologies. In this paper, we present an approach to leverage the knowledge of a teacher network, that was trained using the original data modality, to improve the performance of a student network on a new data modality: a technique known in literature as knowledge distillation. By applying knowledge distillation to the problem of sensor transition, we can greatly speed up this process. We validate this approach using a multimodal version of the MNIST dataset. Especially when little data is available in the new modality (i.e., 10 images), training with additional teacher supervision results in increased performance, with the student network scoring a test set accuracy of 0.77, compared to an accuracy of 0.37 for the baseline. We also explore two extensions to the default method of knowledge distillation, which we evaluate on a multimodal version of the CIFAR-10 dataset: an annealing scheme for the hyperparameter α and selective knowledge distillation. Of these two, the first yields the best results. Choosing the optimal annealing scheme results in an increase in test set accuracy of 6%. Finally, we apply our method to the real-world use case of skin lesion classification.

Image Degradation in Microscopic Images: Avoidance, Artifacts, and Solutions.

The goal of modern microscopy is to acquire high-quality image based data sets. A typical microscopy workflow is set up in order to address a specific biological question and involves different steps. The first step is to precisely define the biological question, in order to properly come to an experimental design for sample preparation and image acquisition. A better object representation allows biological users to draw more reliable scientific conclusions. Image restoration can manipulate the acquired data in an effort to reduce the impact of artifacts (spurious results) due to physical and technical limitations, resulting in a better representation of the object of interest. However, precise usage of these algorithms is necessary so as to avoid further artifacts that might influence the data analysis and bias the conclusions. It is essential to understand image acquisition, and how it introduces artifacts and degradations in the acquired data, so that their effects on subsequent analysis can be minimized. This paper provides an overview of the fundamental artifacts and degradations that affect many micrographs. We describe why artifacts appear, in what sense they impact overall image quality, and how to mitigate them by first improving the acquisition parameters and then applying proper image restoration techniques.

MATE: Machine Learning for Adaptive Calibration Template Detection.

The problem of camera calibration is two-fold. On the one hand, the parameters are estimated from known correspondences between the captured image and the real world. On the other, these correspondences themselves-typically in the form of chessboard corners-need to be found. Many distinct approaches for this feature template extraction are available, often of large computational and/or implementational complexity. We exploit the generalized nature of deep learning networks to detect checkerboard corners: our proposed method is a convolutional neural network (CNN) trained on a large set of example chessboard images, which generalizes several existing solutions. The network is trained explicitly against noisy inputs, as well as inputs with large degrees of lens distortion. The trained network that we evaluate is as accurate as existing techniques while offering improved execution time and increased adaptability to specific situations with little effort. The proposed method is not only robust against the types of degradation present in the training set (lens distortions, and large amounts of sensor noise), but also to perspective deformations, e.g., resulting from multi-camera set-ups.

Rosette tracker: an open source image analysis tool for automatic quantification of genotype effects.

Image analysis of Arabidopsis (Arabidopsis thaliana) rosettes is an important nondestructive method for studying plant growth. Some work on automatic rosette measurement using image analysis has been proposed in the past but is generally restricted to be used only in combination with specific high-throughput monitoring systems. We introduce Rosette Tracker, a new open source image analysis tool for evaluation of plant-shoot phenotypes. This tool is not constrained by one specific monitoring system, can be adapted to different low-budget imaging setups, and requires minimal user input. In contrast with previously described monitoring tools, Rosette Tracker allows us to simultaneously quantify plant growth, photosynthesis, and leaf temperature-related parameters through the analysis of visual, chlorophyll fluorescence, and/or thermal infrared time-lapse sequences. Freely available, Rosette Tracker facilitates the rapid understanding of Arabidopsis genotype effects.

2D mapping of strongly deformable cell nuclei-based on contour matching.

The spatiotemporal dynamics of protein complexes and genome loci are functionally linked to cellular health status. To study the inherent motion of subnuclear particles, it is essential to remove any superimposed component stemming from displacement and deformation of the nucleus. In this article, we propose a mapping of the nuclear interior, which is based on the deformation of the nuclear contour and has no shape constraints. This registration procedure enabled an accurate estimation of telomere mobility in living human cells undergoing dramatic nuclear deformations. Given the large variety of pathologies and cellular processes that are associated with strong nuclear shape changes, the contour mapping algorithm has generic value for improving the accuracy of mobility measurements of genome loci and intranuclear macromolecule complexes. © 2011 International Society for Advancement of Cytometry.

Accurate Detection of Dysmorphic Nuclei Using Dynamic Programming and Supervised Classification.

A vast array of pathologies is typified by the presence of nuclei with an abnormal morphology. Dysmorphic nuclear phenotypes feature dramatic size changes or foldings, but also entail much subtler deviations such as nuclear protrusions called blebs. Due to their unpredictable size, shape and intensity, dysmorphic nuclei are often not accurately detected in standard image analysis routines. To enable accurate detection of dysmorphic nuclei in confocal and widefield fluorescence microscopy images, we have developed an automated segmentation algorithm, called Blebbed Nuclei Detector (BleND), which relies on two-pass thresholding for initial nuclear contour detection, and an optimal path finding algorithm, based on dynamic programming, for refining these contours. Using a robust error metric, we show that our method matches manual segmentation in terms of precision and outperforms state-of-the-art nuclear segmentation methods. Its high performance allowed for building and integrating a robust classifier that recognizes dysmorphic nuclei with an accuracy above 95%. The combined segmentation-classification routine is bound to facilitate nucleus-based diagnostics and enable real-time recognition of dysmorphic nuclei in intelligent microscopy workflows.

Bayesian deconvolution of scanning electron microscopy images using point-spread function estimation and non-local regularization.

Microscopy is one of the most essential imaging techniques in life sciences. High-quality images are required in order to solve (potentially life-saving) biomedical research problems. Many microscopy techniques do not achieve sufficient resolution for these purposes, being limited by physical diffraction and hardware deficiencies. Electron microscopy addresses optical diffraction by measuring emitted or transmitted electrons instead of photons, yielding nanometer resolution. Despite pushing back the diffraction limit, blur should still be taken into account because of practical hardware imperfections and remaining electron diffraction. Deconvolution algorithms can remove some of the blur in post-processing but they depend on knowledge of the point-spread function (PSF) and should accurately regularize noise. Any errors in the estimated PSF or noise model will reduce their effectiveness. This paper proposes a new procedure to estimate the lateral component of the point spread function of a 3D scanning electron microscope more accurately. We also propose a Bayesian maximum a posteriori deconvolution algorithm with a non-local image prior which employs this PSF estimate and previously developed noise statistics. We demonstrate visual quality improvements and show that applying our method improves the quality of subsequent segmentation steps.

A novel dictionary based computer vision method for the detection of cell nuclei.

Cell nuclei detection in fluorescent microscopic images is an important and time consuming task in a wide range of biological applications. Blur, clutter, bleed through and partial occlusion of nuclei make individual nuclei detection a challenging task for automated image analysis. This paper proposes a novel and robust detection method based on the active contour framework. Improvement over conventional approaches is achieved by exploiting prior knowledge of the nucleus shape in order to better detect individual nuclei. This prior knowledge is defined using a dictionary based approach which can be formulated as the optimization of a convex energy function. The proposed method shows accurate detection results for dense clusters of nuclei, for example, an F-measure (a measure for detection accuracy) of 0.96 for the detection of cell nuclei in peripheral blood mononuclear cells, compared to an F-measure of 0.90 achieved by state-of-the-art nuclei detection methods.

Computationally efficient locally adaptive demosaicing of color filter array images using the dual-tree complex wavelet packet transform.

Most digital cameras use an array of alternating color filters to capture the varied colors in a scene with a single sensor chip. Reconstruction of a full color image from such a color mosaic is what constitutes demosaicing. In this paper, a technique is proposed that performs this demosaicing in a way that incurs a very low computational cost. This is done through a (dual-tree complex) wavelet interpretation of the demosaicing problem. By using a novel locally adaptive approach for demosaicing (complex) wavelet coefficients, we show that many of the common demosaicing artifacts can be avoided in an efficient way. Results demonstrate that the proposed method is competitive with respect to the current state of the art, but incurs a lower computational cost. The wavelet approach also allows for computationally effective denoising or deblurring approaches.

Brain blood vessel segmentation using line-shaped profiles.

Segmentation of cerebral blood vessels is of great importance in diagnostic and clinical applications, especially for embolization of cerebral aneurysms and arteriovenous malformations (AVMs). In order to perform embolization of the AVM, the structural and geometric information of blood vessels from 3D images is of utmost importance. For this reason, the in-depth segmentation of cerebral blood vessels is usually done as a fusion of different segmentation techniques, often requiring extensive user interaction. In this paper we introduce the idea of line-shaped profiling with an application to brain blood vessel and AVM segmentation, efficient both in terms of resolving details and in terms of computation time. Our method takes into account both local proximate and wider neighbourhood of the processed pixel, which makes it efficient for segmenting large blood vessel tree structures, as well as fine structures of the AVMs. Another advantage of our method is that it requires selection of only one parameter to perform segmentation, yielding very little user interaction.

A novel tracing method for the segmentation of cell wall networks.

Cell wall networks are a common subject of research in biology, which are important for plant growth analysis, organ studies, etc. In order to automate the detection of individual cells in such cell wall networks, we propose a new segmentation algorithm. The proposed method is a network tracing algorithm, exploiting the prior knowledge of the network structure. The method is applicable on multiple microscopy modalities such as fluorescence, but also for images captured using non invasive microscopes such as differential interference contrast (DIC) microscopes.

2D mapping of strongly deformed cell nuclei based on contour warping.

The dynamics of genome regions are associated to the functional or dysfunctional behaviour of the human cell. In order to study these dynamics it is necessary to remove perturbations coming from movement and deformation of the nucleus, i.e. the container holding the genome. In literature models have been proposed to cope with the transformations corresponding to nuclear dynamics of healthy cells. However for pathological cells, the nucleus deforms in an apparently random way, making the use of such models a non trivial task. In this paper we propose a mapping of the cell nucleus which is based on the matching of the nuclear contours. The proposed method does not put constraints on the possible shapes nor on the possible deformations, making this method suited for the analysis of pathological nuclei.