Papers by Demetrio Labate
Scientific Reports
Astrocytes, a subtype of glial cells with a complex morphological structure, are active players i... more Astrocytes, a subtype of glial cells with a complex morphological structure, are active players in many aspects of the physiology of the central nervous system (CNS). However, due to their highly involved interaction with other cells in the CNS, made possible by their morphological complexity, the precise mechanisms regulating astrocyte function within the CNS are still poorly understood. This knowledge gap is also due to the current limitations of existing quantitative image analysis tools that are unable to detect and analyze images of astrocyte with sufficient accuracy and efficiency. To address this need, we introduce a new deep learning framework for the automated detection of GFAP-immunolabeled astrocytes in brightfield or fluorescent micrographs. A major novelty of our approach is the applications of YOLOv5, a sophisticated deep learning platform designed for object detection, that we customized to derive optimized classification models for the task of astrocyte detection. Ex...
arXiv (Cornell University), Feb 4, 2015
Region-of-Interest (ROI) tomography aims at reconstructing a region of interest C inside a body u... more Region-of-Interest (ROI) tomography aims at reconstructing a region of interest C inside a body using only x-ray projections intersecting C with the goal to reduce overall radiation exposure when only a small specific region of the body needs to be examined. We consider x-ray acquisition from sources located on a smooth curve Γ in R 3 verifying classical Tuy's condition. In this situation, the non-trucated cone-beam transform Df of smooth densities f admits an explicit inverse Z; however Z cannot directly reconstruct f from ROI-truncated projections. To deal with the ROI tomography problem, we introduce a novel reconstruction approach. For densities f in L ∞ (B) where B is a bounded ball in R 3 , our method iterates an operator U combining ROI-truncated projections, inversion by the operator Z and appropriate regularization operators. Assuming only knowledge of projections corresponding to a spherical ROI C ⊂ B, given > 0, we prove that if C is sufficiently large our iterative reconstruction algorithm converges uniformly to an-accurate approximation of f , where the accuracy depends on the regularity of f quantified in the Sobolev norm W 5 (B). This result shows the existence of a critical ROI radius ensuring the convergence of the ROI reconstruction algorithm to-accurate approximations of f. We numerically verified these theoretical results using simulated acquisition of ROI-truncated cone-beam projection data for multiple acquisition geometries. Numerical experiments indicate that
Journal of Mathematical Imaging and Vision
Blind inpainting algorithms based on deep learning architectures have shown a remarkable performa... more Blind inpainting algorithms based on deep learning architectures have shown a remarkable performance in recent years, typically outperforming model-based methods both in terms of image quality and run time. However, neural network strategies typically lack a theoretical explanation, which contrasts with the well-understood theory underlying model-based methods. In this work, we leverage the advantages of both approaches by integrating theoretically founded concepts from transform domain methods and sparse approximations into a CNN-based approach for blind image inpainting. To this end, we present a novel strategy to learn convolutional kernels that applies a specifically designed filter dictionary whose elements are linearly combined with trainable weights. Numerical experiments demonstrate the competitiveness of this approach. Our results show not only an improved inpainting quality compared to conventional CNNs but also significantly faster network convergence within a lightweight network design.
IEEE Geoscience and Remote Sensing Letters, 2020
Hyperspectral imagery (HSI) has emerged as a highly successful sensing modality for a variety of ... more Hyperspectral imagery (HSI) has emerged as a highly successful sensing modality for a variety of applications ranging from urban mapping to environmental monitoring and precision agriculture. Despite the efforts by the scientific community, developing reliable algorithms of HSI classification remains a challenging problem especially for high-resolution HSI data where there is often larger intraclass variability combined with scarcity of ground truth data and class imbalance. In recent years, deep neural networks have emerged as a promising strategy for problems of HSI classification where they have shown a remarkable potential for learning joint spectral-spatial features efficiently via backpropagation. In this paper, we propose a deep learning strategy for HSI classification that combines different convolutional neural networks especially designed to efficiently learn joint spatial-spectral features over multiple scales. Our method achieves an overall classification accuracy of 66.73% on the 2018 IEEE GRSS hyperspectral dataset-a high-resolution dataset that includes 20 urban land-cover and land-use classes.
Translational psychiatry, Jan 10, 2016
Cognitive processing is highly dependent on the functional integrity of gamma-amino-butyric acid ... more Cognitive processing is highly dependent on the functional integrity of gamma-amino-butyric acid (GABA) interneurons in the brain. These cells regulate excitability and synaptic plasticity of principal neurons balancing the excitatory/inhibitory tone of cortical networks. Reduced function of parvalbumin (PV) interneurons and disruption of GABAergic synapses in the cortical circuitry result in desynchronized network activity associated with cognitive impairment across many psychiatric disorders, including schizophrenia. However, the mechanisms underlying these complex phenotypes are still poorly understood. Here we show that in animal models, genetic deletion of fibroblast growth factor 14 (Fgf14), a regulator of neuronal excitability and synaptic transmission, leads to loss of PV interneurons in the CA1 hippocampal region, a critical area for cognitive function. Strikingly, this cellular phenotype associates with decreased expression of glutamic acid decarboxylase 67 (GAD67) and ves...
From Group Representations to Signal Analysis
Applied and Numerical Harmonic Analysis, 2015
In this chapter, we present the point of view that has inspired this book and we explain the pers... more In this chapter, we present the point of view that has inspired this book and we explain the perspective and scope of the four chapters that follow.
Applied and Numerical Harmonic Analysis, 2015
The use of general descriptive names, registered names, trademarks, service marks, etc. in this p... more The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specific statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use.
Mathematical Modelling of Natural Phenomena, 2018
Region of interest (ROI) tomography has gained increasing attention in recent years due to its po... more Region of interest (ROI) tomography has gained increasing attention in recent years due to its potential to reducing radiation exposure and shortening the scanning time. However, tomographic reconstruction from ROI-focused illumination involves truncated projection data and typically results in higher numerical instability even when the reconstruction problem has unique solution. To address this problem, bothad hocanalytic formulas and iterative numerical schemes have been proposed in the literature. In this paper, we introduce a novel approach for ROI tomographic reconstruction, formulated as a convex optimization problem with a regularized term based on shearlets. Our numerical implementation consists of an iterative scheme based on the scaled gradient projection method and it is tested in the context of fan-beam CT. Our results show that our approach is essentially insensitive to the location of the ROI and remains very stable also when the ROI size is rather small.
Journal of Computational and Applied Mathematics
Efficient representations of multivariate functions are critical for the design of state-of-the-a... more Efficient representations of multivariate functions are critical for the design of state-of-the-art methods of data restoration and feature extraction. In this work, we consider the representation of spatio-temporal data such as temporal sequences (videos) of 2-and 3-dimensional images, where conventional separable representations are usually very inefficient, due to their limitations in handling the geometry of the data. To address this challenge, we consider an idealized class EpAq Ă L 2 pR 4 q of functions of 4 variables dominated by hypersurface singularities in the first three coordinates that we apply to model 4-dimensional data corresponding to temporal sequences (videos) of 3-dimensional objects. To provide an efficient representation for this type of data, we introduce a new multiscale directional system of functions based on cylindrical shearlets and prove that this new approach achieves superior approximation properties with respect to conventional multiscale representations. To further illustrate the advantages of our approach, we apply a discrete implementation of the new representation to problems of dynamic tomography using synthetic data.
Wavelets and Sparsity XVIII
Deep neural networks have achieved impressive performance in problems of object detection and obj... more Deep neural networks have achieved impressive performance in problems of object detection and object category classifications. To perform efficiently though, such methods typically require a large number of training samples. Unfortunately, this requirement is highly impractical or impossible in applications such as hyperspectral classification where it is expensive and labor intensive to generate labeled data for training. A few ideas have been proposed in the literature to address this problem such as transfer learning and domain adaptation. In this work, we propose an alternative strategy to reduce the number of network parameters based on Structured Receptive Field Networks (SRFN), a class of convolutional neural networks (CNNs) where each convolutional filter is a linear combination from a predefined dictionary. To better exploit the characteristics of hyperspectral data to be learned, we choose a filter dictionary consisting of directional filters inspired by the theory of shearlets and we train a SRFN by imposing that the convolutional filters form sparse linear combinations in such dictionary. The application of our SRFN to problems of hyperspectral classification shows that this approach achieves very competitive performance as compared to conventional CNNs.
arXiv (Cornell University), Nov 13, 2015
When it comes to computed tomography (CT), the possibility to reconstruct a small region-of-inter... more When it comes to computed tomography (CT), the possibility to reconstruct a small region-of-interest (ROI) using truncated projection data is particularly appealing due to its potential to lower radiation exposure and reduce the scanning time. However, ROI reconstruction from truncated projections is an ill-posed inverse problem, with the ill-posedness becoming more severe when the ROI size is getting smaller. To address this problem, both ad hoc analytic formulas and iterative numerical schemes have been proposed in the literature. In this paper, we introduce a novel approach for ROI CT reconstruction, formulated as a convex optimization problem with a regularized term based on shearlets. Our numerical implementation consists of an iterative scheme based on the scaled gradient projection (SGP) method and is tested in the context of fan beam CT. Our results show that this approach is essentially insensitive to the location of the ROI and remains very stable also when the ROI size is rather small.
Automated identification of the primary components of a neuron and extraction of its subcellular ... more Automated identification of the primary components of a neuron and extraction of its subcellular features are essential steps in many quantitative studies of neuronal networks. The focus of this paper is the development of an algorithm for the automated detection of the location and morphology of somas in confocal images of neuronal network cultures. This problem is motivated by applications in high-content screenings (HCS), where the extraction of multiple morphological features of neurons on large data sets is required. Existing algorithms are not very efficient when applied to the analysis of confocal image stacks of neuronal cultures. In addition to the usual difficulties associated with the processing of fluorescent images, these types of stacks contain a small number of images so that only a small number of pixels are available along the z-direction and it is challenging to apply conventional 3D filters. The algorithm we present in this paper applies a number of innovative ideas from the theory of directional multiscale representations and involves the following steps: (i) image segmentation based on support vector machines with specially designed multiscale filters; (ii) soma extraction and separation of contiguous somas, using a combination of level set method and directional multiscale filters. We also present an approach to extract the soma's surface morphology using the 3D shearlet transform. Extensive numerical experiments show that our algorithms are computationally efficient and highly accurate in segmenting the somas and separating contiguous ones. The algorithms presented in this paper will facilitate the development of a high-throughput quantitative platform for the study of neuronal networks for HCS applications.
two-stage shearlet-based approach for the removal of
Centerline tracing in dendritic structures acquired from confocal images of neurons is an essenti... more Centerline tracing in dendritic structures acquired from confocal images of neurons is an essential tool for the construction of geometrical representations of a neuronal network from its coarse scale up to its fine scale structures. In this paper, we propose a novel algorithm for centerline extraction that is both highly accurate and computationally efficient. The main novelty of the proposed method is the use of a small set of Multiscale Isotropic Laplacian filters for a quick and efficient binarization of the dendritic structure combined with the application of a simple 3D finite-length filter which automatically detects the seed points for the centerline extraction. The performance of this algorithm, which is validated on data from the DIADEM set, is shown to be very competitive against other state-of-the-art algorithms.
Landscapes of Time-Frequency Analysis, 2020
Several diseases including diabetes, hypertension, and glaucoma are known to cause alterations in... more Several diseases including diabetes, hypertension, and glaucoma are known to cause alterations in the human retina that can be visualized noninvasively and in vivo using well-established techniques of fundus photography. Since the treatment of these diseases can be significantly improved with early detection, methods for the quantitative analysis of fundus imaging have been the subject of extensive studies. Following major advances in image processing and machine learning during the last decade, a remarkable progress is being made towards developing automated quantitative methods to identify image-based biomarkers of different pathologies. In this paper, we focus especially on the automated analysis of alterations of retinal microvasculature-a class of structural alterations that is particularly important for early detection of cardiovascular and neurological diseases.
Advances in Computational Mathematics, 2021
We present a new method for the stable reconstruction of a class of binary images from a small nu... more We present a new method for the stable reconstruction of a class of binary images from a small number of measurements. The images we consider are characteristic functions of algebraic domains, that is, domains defined as zero loci of bivariate polynomials, and we assume to know only a finite set of uniform samples for each image. The solution to such a problem can be set up in terms of linear equations associated to a set of image moments. However, the sensitivity of the moments to noise makes the numerical solution highly unstable. To derive a robust image recovery algorithm, we represent algebraic polynomials and the corresponding image moments in terms of bivariate Bernstein polynomials and apply polynomial-generating, refinable sampling kernels. This approach is robust to noise, computationally fast and simple to implement. We illustrate the performance of our reconstruction algorithm from noisy samples through extensive numerical experiments. Our code is released open source an...
In this paper we apply a time frequency approach to the study of pseudodi er ential operators Bot... more In this paper we apply a time frequency approach to the study of pseudodi er ential operators Both the Weyl and the Kohn Nirenberg correspondences are considered In order to quantify the time frequency content of a function or distribution we use certain function spaces called modulation spaces We deduce a time frequency characterization of the twisted product of two symbols and and we show that modulation spaces provide the natural setting to exactly control the time frequency content of from the time frequency content of and As a consequence we discuss some boundedness and spectral properties of the corresponding operator with symbol Introduction A pseudodi erential operator can be de ned through the Weyl or the Kohn Nirenberg correspondence by bijectively assigning to any distributional symbol S R n a linear operator T S R n S R so that the properties of the operator are in an appropriate way re ected in the properties of the symbol One way to construct a pseudodi erential operat...
The geometric separation problem, initially posed by Donoho and Kutyniok [7], aims to separate a ... more The geometric separation problem, initially posed by Donoho and Kutyniok [7], aims to separate a distribution containing a non-trivial superposition of point and curvilinear singularities into its distinct geometric constituents. The solution proposed in [7] considers expansions with respect to a combined wavelet-curvelet dictionary and applies an `-norm minimization over the expansion coefficients to achieve separation asymptotically at fine scales. However, the original proof of this result uses a heavy machinery relying on sparse representations of Fourier integral operators which does not extend directly to the 3D setting. In this paper, we extend the geometric separation result to the 3D setting using a novel and simpler argument which relies in part on techniques developed by the authors for the shearlet-based analysis of curvilinear edges. Our new result also yields a significantly simpler proof of the original 2D geometric separation problem and extends a prior result by the...
Applied and Computational Harmonic Analysis, 2020
Several strategies have been applied for the recovery of the missing parts in an image, with reco... more Several strategies have been applied for the recovery of the missing parts in an image, with recovery performance depending significantly on the image type and the geometry of missing data. To provide a deeper understanding of such image restoration problem, King and al. recently introduced a rigorous multiscale analysis framework and proved that a shearlet based inpainting approach outperforms methods based on more conventional multiscale representations when missing data are line singularities. In this paper, we extend and improve the analysis of the inpainting problem to the more realistic and more challenging setting of images containing curvilinear singularities. We derive inpainting performance guarantees showing that exact image recovery is achieved if the size of the missing singularity is smaller than the size of the structure elements of appropriate functional representations of the image. Our proof relies critically on the microlocal and sparsity properties of the shearlet representation.
The Journal of Geometric Analysis, 2020
Sparse representations of multidimensional data have received a significant attention in the lite... more Sparse representations of multidimensional data have received a significant attention in the literature due to their applications in problems of data restoration and feature extraction. In this paper, we consider an idealized class C 2 (Z) ⊂ L 2 (R 3) of 3-dimensional data dominated by surface singularities that are orthogonal to the xy plane. To deal with this type of data, we introduce a new multiscale directional representation called cylindrical shearlets and prove that this new approach achieves superior approximation properties not only with respect to conventional multiscale representations but also with respect to 3-dimensional shearlets and curvelets. Specifically, the N-term approximation f S N obtained by selecting the N largest coefficients of the cylindrical shearlet expansion of a function f ∈ C(Z) satisfies the asymptotic estimate f − f S N 2 2 ≤ c N −2 (ln N) 3 , as N → ∞. This is the optimal decay rate, up the logarithmic factor, outperforming 3d wavelet and 3d shearlet approximations which only yield approximation rates of order N −1/2 and N −1 (ignoring logarithmic factors), respectively, on the same type of data.
Uploads
Papers by Demetrio Labate