Hierarchical grid transformation for image warping in the analysis of two-dimensional electrophoresis gels

Two-dimensional electrophoresis is a widely used method for separating a large number of proteins from complex protein mixtures and for revealing differential patterns of protein expressions. In the computer-assisted proteome research, the comparison of protein separation profiles involves several heuristic steps, ranging from protein spot detection to matching of unknown spots. An important prerequisite for efficient protein spot matching is the image warping step, where the geometric relationship between the gel profiles is modeled on the basis of a given set of known corresponding spots, so-called landmarks, and the locations of unknown spots are predicted using the optimized model. Traditionally, polynomial functions together with least squares optimization has been used, even though this approach is known to be incapable of modeling all the complex distortions inherent in electrophoretic data. To satisfy the need of more flexible gel distortion correction, a hierarchical grid transformation method with stochastic optimization is presented. The method provides an adaptive multiresolution model between the gels, and good correction performance in the practical cross-validation tests suggests that automatic warping of gel images could be based on this approach. We believe that the proposed model also has significance in the ultimate comparison of corresponding protein spots since the matching process should benefit from the closeness of the true spot pairs.     

Distortion Correction in EPI Using an Extended PSF Method with a Reversed Phase Gradient Approach

In echo-planar imaging (EPI), such as commonly used for functional MRI (fMRI) and diffusion-tensor imaging (DTI), compressed distortion is a more difficult challenge than local stretching as spatial information can be lost in strongly compressed areas. In addition, the effects are more severe at ultra-high field (UHF) such as 7T due to increased field inhomogeneity. To resolve this problem, two EPIs with opposite phase-encoding (PE) polarity were acquired and combined after distortion correction. For distortion correction, a point spread function (PSF) mapping method was chosen due to its high correction accuracy and extended to perform distortion correction of both EPIs with opposite PE polarity thus reducing the PSF reference scan time. Because the amount of spatial information differs between the opposite PE datasets, the method was further extended to incorporate a weighted combination of the two distortion-corrected images to maximize the spatial information content of a final corrected image. The correction accuracy of the proposed method was evaluated in distortion-corrected data using both forward and reverse phase-encoded PSF reference data and compared with the reversed gradient approaches suggested previously. Further we demonstrate that the extended PSF method with an improved weighted combination can recover local distortions and spatial information loss and be applied successfully not only to spin-echo EPI, but also to gradient-echo EPIs acquired with both PE directions to perform geometrically accurate image reconstruction.     

Optimal calibration marker mesh for 2D X-ray sensors in 3D reconstruction

Image intensifiers suffer from distortions due to magnetic fields. In order to use this X-ray projections images for computer-assisted medical interventions, image intensifiers need to be calibrated. Opaque markers are often used for the correction of the image distortion and the estimation of the acquisition geometry parameters. Information under the markers is then lost. In this work, we consider the calibration of image intensifiers in the framework of 3D reconstruction from several 2D X-ray projections. In this context, new schemes of marker distributions are proposed for 2D X-ray sensor calibration. They are based on efficient sampling conditions of the parallel-beam X-ray transform when the detector and source trajectory is restricted to a circle around the measured object. Efficient sampling are essentially subset of standard sampling in this situation. The idea is simply to exploit the data redundancy of standard sampling and to replace some holes of efficient schemes by markers. Optimal location of markers in the sparse efficient sampling geometry can thus be found. In this case, the markers can stay on the sensor during the measurement with--theoretically--no loss of information (when the signal-to-noise ratio is large). Even if the theory is based on the parallel-beam X-ray transform, numerical experiments on both simulated and real data are shown in the case of weakly divergent beam geometry. We show that the 3D reconstruction from simulated data with interlaced markers is essentially the same as those obtained from data with no marker. We show that efficient Fourier interpolation formulas based on optimal sparse sampling schemes can be used to recover the information hidden by the markers.     

Phase conjugation model for information transfer in the nervous system

In the nervous system, dispersion in propagation time sometimes brings delay distortion or phase distortion on the information transmission. Also in the memory retrieval processes in the brain, some parts of images may be retrieved more slowly than others. For smooth control of fast movements as well as for keeping exact thinking, these distortions have to be taken out. To understand the distortion cancelling mechanism, new neural network models for compensating the phase distortion are proposed. The models stand on the concept of "phase conjugate mirror" which is used in optical image processing. Simulation studies based on the model resulted in successful cancellation of the delay dispersion involved in the information transmission in the nervous system.     

A local training-pruning approach for recurrent neural networks

The global extended Kalman filtering (EKF) algorithm for recurrent neural networks (RNNs) is plagued by the drawback of high computational cost and storage requirement. In this paper, we present a local EKF training-pruning approach that can solve this problem. In particular, the by-products, obtained along with the local EKF training, can be utilized to measure the importance of the network weights. Comparing with the original global approach, the proposed local approach results in much lower computational cost and storage requirement. Hence, it is more practical in solving real world problems. Simulation showed that our approach is an effective joint-training-pruning method for RNNs under online operation.     

A bayesian recurrent neural network for unsupervised pattern recognition in large incomplete data sets

A recurrent neural network, modified to handle highly incomplete training data is described. Unsupervised pattern recognition is demonstrated in the WHO database of adverse drug reactions. Comparison is made to a well established method, AutoClass, and the performances of both methods is investigated on simulated data. The neural network method performs comparably to AutoClass in simulated data, and better than AutoClass in real world data. With its better scaling properties, the neural network is a promising tool for unsupervised pattern recognition in huge databases of incomplete observations.     

全景绿视率自动识别和计算研究

绿视率是用于绿色空间感知的直观评价标准,传统研究的绿视率多基于平面影像进行计算,不能完全反映三维空间中人对绿量的主观感受。基于全景影像,提出全景绿视率的概念,通过全景相机获取球面全景照片,将等距圆柱投影转换为等积圆柱投影,利用基于语义分割的卷积神经网络模型,自动识别植被区域面积以实现全景绿视率自动化识别和计量。通过比较5项卷积神经网络模型对绿视率的识别效果,显示出Dilated ResNet-105神经网络模型具有最高的识别准确度。以武汉市武昌区紫阳公园为例,对各级园路和广场的全景绿视率进行计算和分析。将卷积神经网络的识别结果同人工判别结果进行对比研究,结果显示：使用Dilated ResNet-105卷积神经网络对绿植范围识别的平均交并比（mIoU）为62.53%,与人工识别的平均差异为9.17%。全景绿视率自动识别和计算可以为相关研究提供新的思路,实现客观准确、快速便捷的绿视率测量评估。     

A Projection and Density Estimation Method for Knowledge Discovery

A key ingredient to modern data analysis is probability density estimation. However, it is well known that the curse of dimensionality prevents a proper estimation of densities in high dimensions. The problem is typically circumvented by using a fixed set of assumptions about the data, e.g., by assuming partial independence of features, data on a manifold or a customized kernel. These fixed assumptions limit the applicability of a method. In this paper we propose a framework that uses a flexible set of assumptions instead. It allows to tailor a model to various problems by means of 1d-decompositions. The approach achieves a fast runtime and is not limited by the curse of dimensionality as all estimations are performed in 1d-space. The wide range of applications is demonstrated at two very different real world examples. The first is a data mining software that allows the fully automatic discovery of patterns. The software is publicly available for evaluation. As a second example an image segmentation method is realized. It achieves state of the art performance on a benchmark dataset although it uses only a fraction of the training data and very simple features.     

Blind demixing methods for recovering dense neuronal morphology from barcode imaging data

Cellular barcoding methods offer the exciting possibility of ‘infinite-pseudocolor’ anatomical reconstruction—i.e., assigning each neuron its own random unique barcoded ‘pseudocolor,’ and then using these pseudocolors to trace the microanatomy of each neuron. Here we use simulations, based on densely-reconstructed electron microscopy microanatomy, with signal structure matched to real barcoding data, to quantify the feasibility of this procedure. We develop a new blind demixing approach to recover the barcodes that label each neuron, and validate this method on real data with known barcodes. We also develop a neural network which uses the recovered barcodes to reconstruct the neuronal morphology from the observed fluorescence imaging data, ‘connecting the dots’ between discontiguous barcode amplicon signals. We find that accurate recovery should be feasible, provided that the barcode signal density is sufficiently high. This study suggests the possibility of mapping the morphology and projection pattern of many individual neurons simultaneously, at high resolution and at large scale, via conventional light microscopy.     

A Novel Iterative CT Reconstruction Approach Based on FBP Algorithm

The Filtered Back-Projection (FBP) algorithm and its modified versions are the most important techniques for CT (Computerized tomography) reconstruction, however, it may produce aliasing degradation in the reconstructed images due to projection discretization. The general iterative reconstruction (IR) algorithms suffer from their heavy calculation burden and other drawbacks. In this paper, an iterative FBP approach is proposed to reduce the aliasing degradation. In the approach, the image reconstructed by FBP algorithm is treated as the intermediate image and projected along the original projection directions to produce the reprojection data. The difference between the original and reprojection data is filtered by a special digital filter, and then is reconstructed by FBP to produce a correction term. The correction term is added to the intermediate image to update it. This procedure can be performed iteratively to improve the reconstruction performance gradually until certain stopping criterion is satisfied. Some simulations and tests on real data show the proposed approach is better than FBP algorithm or some IR algorithms in term of some general image criteria. The calculation burden is several times that of FBP, which is much less than that of general IR algorithms and acceptable in the most situations. Therefore, the proposed algorithm has the potential applications in practical CT systems.     

Bayesian Neural Network Models for Censored Data

Neural networks are considered by many to be very promising tools for classification and prediction. The flexibility of the neural network models often result in over-fit. Shrinking the parameters using a penalized likelihood is often used in order to overcome such over-fit. In this paper we extend the approach proposed by FARAGGI and SIMON (1995a) to modeling censored survival data using the input-output relationship associated with a single hidden layer feed-forward neural network. Instead of estimating the neural network parameters using the method of maximum likelihood, we place normal prior distributions on the parameters and make inferences based on derived posterior distributions of the parameters. This Bayesian formulation will result in shrinking the parameters of the neural network model and will reduce the over-fit compared with the maximum likelihood estimators. We illustrate our proposed method on a simulated and a real example.     

Automatic greenhouse insect pest detection and recognition based on a cascaded deep learning classification method

Inspection of insect sticky paper traps is an essential task for an effective integrated pest management (IPM) programme. However, identification and counting of the insect pests stuck on the traps is a very cumbersome task. Therefore, an efficient approach is needed to alleviate the problem and to provide timely information on insect pests. In this research, an automatic method for the multi-class recognition of small-size greenhouse insect pests on sticky paper trap images acquired by wireless imaging devices is proposed. The developed algorithm features a cascaded approach that uses a convolutional neural network (CNN) object detector and CNN image classifiers, separately. The object detector was trained for detecting objects in an image, and a CNN classifier was applied to further filter out non-insect objects from the detected objects in the first stage. The obtained insect objects were then further classified into flies (Diptera: Drosophilidae), gnats (Diptera: Sciaridae), thrips (Thysanoptera: Thripidae) and whiteflies (Hemiptera: Aleyrodidae), using a multi-class CNN classifier in the second stage. Advantages of this approach include flexibility in adding more classes to the multi-class insect classifier and sample control strategies to improve classification performance. The algorithm was developed and tested for images taken by multiple wireless imaging devices installed in several greenhouses under natural and variable lighting environments. Based on the testing results from long-term experiments in greenhouses, it was found that the algorithm could achieve average F₁-scores of 0.92 and 0.90 and mean counting accuracies of 0.91 and 0.90, as tested on a separate 6-month image data set and on an image data set from a different greenhouse, respectively. The proposed method in this research resolves important problems for the automated recognition of insect pests and provides instantaneous information of insect pest occurrences in greenhouses, which offers vast potential for developing more efficient IPM strategies in agriculture.     

DMP3: a dynamic multilayer perceptron construction algorithm

This paper presents DMP3 (Dynamic Multilayer Perceptron 3), a multilayer perceptron (MLP) constructive training method that constructs MLPs by incrementally adding network elements of varying complexity to the network. DMP3 differs from other MLP construction techniques in several important ways, and the motivation for these differences are given. Information gain rather than error minimization is used to guide the growth of the network, which increases the utility of newly added network elements and decreases the likelihood that a premature dead end in the growth of the network will occur. The generalization performance of DMP3 is compared with that of several other well-known machine learning and neural network learning algorithms on nine real world data sets. Simulation results show that DMP3 performs better (on average) than any of the other algorithms on the data sets tested. The main reasons for this result are discussed in detail.     

A local training and pruning approach for neural networks

The training of neural networks using the extended Kalman filter (EKF) algorithm is plagued by the drawback of high computational complexity and storage requirement that may become prohibitive even for networks of moderate size. In this paper, we present a local EKF training and pruning approach that can solve this problem. In particular, the by-products obtained along with the local EKF training can be utilized to measure the importance of the network weights. Comparing with the original global approach, the proposed local EKF training and pruning approach results in a much lower computational complexity and storage requirement. Hence, it is more practical in solving real world problems. The performance of the proposed algorithm is demonstrated on one medium- and one large-scale problems, namely, sunspot data prediction and handwritten digit recognition.     

Iterative Beam Hardening Correction for Multi-Material Objects

In this paper, we propose an iterative beam hardening correction method that is applicable for the case with multiple materials. By assuming that the materials composing scanned object are known and that they are distinguishable by their linear attenuation coefficients at some given energy, the beam hardening correction problem is converted into a nonlinear system problem, which is then solved iteratively. The reconstructed image is the distribution of linear attenuation coefficient of the scanned object at a given energy. So there are no beam hardening artifacts in the image theoretically. The proposed iterative scheme combines an accurate polychromatic forward projection with a linearized backprojection. Both forward projection and backprojection have high degree of parallelism, and are suitable for acceleration on parallel systems. Numerical experiments with both simulated data and real data verifies the validity of the proposed method. The beam hardening artifacts are alleviated effectively. In addition, the proposed method has a good tolerance on the error of the estimated x-ray spectrum.     

Neural network approach for image chromatic adaptation for skin color detection

The goal of image chromatic adaptation is to remove the effect of illumination and to obtain color data that reflects precisely the physical contents of the scene. We present in this paper an approach to image chromatic adaptation using Neural Networks (NN) with application for detecting--adapting human skin color. The NN is trained on randomly chosen color images containing human subject under various illuminating conditions, thereby enabling the model to dynamically adapt to the changing illumination conditions. The proposed network predicts directly the illuminant estimate in the image so as to adapt to human skin color. The comparison of our method with Gray World, White Patch and NN on White Patch methods for skin color stabilization is presented. The skin regions in the NN stabilized images are successfully detected using a computationally inexpensive thresholding operation. We also present results on detecting skin regions on a data set of test images. The results are promising and suggest a new approach for adapting human skin color using neural networks.     

Phylogenetic Reconstruction Using an Unsupervised Growing Neural Network That Adopts the Topology of a Phylogenetic Tree

We propose a new type of unsupervised, growing, self-organizing neural network that expands itself by following the taxonomic relationships that exist among the sequences being classified. The binary tree topology of this neutral network, contrary to other more classical neural network topologies, permits an efficient classification of sequences. The growing nature of this procedure allows to stop it at the desired taxonomic level without the necessity of waiting until a complete phylogenetic tree is produced. This novel approach presents a number of other interesting properties, such as a time for convergence which is, approximately, a lineal function of the number of sequences. Computer simulation and a real example show that the algorithm accurately finds the phylogenetic tree that relates the data. All this makes the neural network presented here an excellent tool for phylogenetic analysis of a large number of sequences. Received: 14 May 1996 / Accepted: 6 August 1996     

An automatic matching technique for patient alignment

An automatic system of patient alignment is required in order to monitor changes that occur in the period between magnetic resonance scans. For each scan of the patient a prime requisite is to register the images with respect to each other. The orthogonal relationship between the sagittal and transverse images should, in principle, identify a single common line at the intersection of the two image planes. The basis of the comparison requires spatial registration of the two images to correct for the probable translational and rotational tilts as well as for the geometrical and intensity distortions. This paper describes a number of automatic techniques which compare, pixel-by-pixel, first two synthetic images, and then their application to real images obtained separately from the same head and neck object field. The robustness, computational cost and effectiveness of the techniques presented are discussed, and computed results on real data for the most promising technique based on the Ratio Absolute Difference algorithm are presented.     

A simple and rapid method for electromagnetic field distortion correction when using two Fastrak sensors for biomechanical studies

The article presents a simple and rapid method for the correction of electromagnetic distortions when using electromagnetic Fastrak (Polhemus, USA) sensors. It is based on the minimization of objective functions composed of derivative polynomial functions, hence estimating the distortion of the electromagnetic field. The polynomial functions composing the objective function each contain 35 deformation coefficients. These coefficients are then used to correct the electromagnetic measures in position and orientation. Preliminary results on the efficacy of the method are presented for two subjects who walked on a treadmill, and for whom relative movement of the lower leg with respect to the thigh was recorded using two Fastrak sensors. The corrected Fastrak measurements were compared with optoelectronic measurements (Vicon, USA), which are not affected by distortions as electromagnetic sensors are. Results showed that after 3 min of calibrating a volume of approximately 1m(3), the method proved to be efficient in correcting errors in orientation (56% (2.72-1.12 degrees ), 78% (4.4-0.89 degrees ), and 56% (2.25-0.90 degrees ) of error reduction in the respective flexion/extension, ab/adduction and tibial internal/external rotation) and position (53% (18.9-8.9 mm), 21% (6.6-4.6mm), and 48% (15.9-8.1mm) of error reduction in the respective medial/lateral, anterior/posterior and proximal/distal translations) (values are overall means for two subjects and four calibration procedures). That amount of correction compared favorably with values presented in the literature.     

A hybrid neural learning algorithm using evolutionary learning and derivative free local search method

In this paper we investigate a hybrid model based on the Discrete Gradient method and an evolutionary strategy for determining the weights in a feed forward artificial neural network. Also we discuss different variants for hybrid models using the Discrete Gradient method and an evolutionary strategy for determining the weights in a feed forward artificial neural network. The Discrete Gradient method has the advantage of being able to jump over many local minima and find very deep local minima. However, earlier research has shown that a good starting point for the discrete gradient method can improve the quality of the solution point. Evolutionary algorithms are best suited for global optimisation problems. Nevertheless they are cursed with longer training times and often unsuitable for real world application. For optimisation problems such as weight optimisation for ANNs in real world applications the dimensions are large and time complexity is critical. Hence the idea of a hybrid model can be a suitable option. In this paper we propose different fusion strategies for hybrid models combining the evolutionary strategy with the discrete gradient method to obtain an optimal solution much quicker. Three different fusion strategies are discussed: a linear hybrid model, an iterative hybrid model and a restricted local search hybrid model. Comparative results on a range of standard datasets are provided for different fusion hybrid models.