A semi-Markov model for mitosis segmentation in time-lapse phase contrast microscopy image sequences of stem cell populations

We propose a semi-Markov model trained in a max-margin learning framework for mitosis event segmentation in large-scale time-lapse phase contrast microscopy image sequences of stem cell populations. Our method consists of three steps. First, we apply a constrained optimization based microscopy image segmentation method that exploits phase contrast optics to extract candidate subsequences in the input image sequence that contains mitosis events. Then, we apply a max-margin hidden conditional random field (MM-HCRF) classifier learned from human-annotated mitotic and nonmitotic sequences to classify each candidate subsequence as a mitosis or not. Finally, a max-margin semi-Markov model (MM-SMM) trained on manually-segmented mitotic sequences is utilized to reinforce the mitosis classification results, and to further segment each mitosis into four predefined temporal stages. The proposed method outperforms the event-detection CRF model recently reported by Huh as well as several other competing methods in very challenging image sequences of multipolar-shaped C3H10T1/2 mesenchymal stem cells. For mitosis detection, an overall precision of 95.8% and a recall of 88.1% were achieved. For mitosis segmentation, the mean and standard deviation for the localization errors of the start and end points of all mitosis stages were well below 1 and 2 frames, respectively. In particular, an overall temporal location error of 0.73 ± 1.29 frames was achieved for locating daughter cell birth events.     

Image segmentation using hidden Markov Gauss mixture models.

Image segmentation is an important tool in image processing and can serve as an efficient front end to sophisticated algorithms and thereby simplify subsequent processing. We develop a multiclass image segmentation method using hidden Markov Gauss mixture models (HMGMMs) and provide examples of segmentation of aerial images and textures. HMGMMs incorporate supervised learning, fitting the observation probability distribution given each class by a Gauss mixture estimated using vector quantization with a minimum discrimination information (MDI) distortion. We formulate the image segmentation problem using a maximum a posteriori criteria and find the hidden states that maximize the posterior density given the observation. We estimate both the hidden Markov parameter and hidden states using a stochastic expectation-maximization algorithm. Our results demonstrate that HMGMM provides better classification in terms of Bayes risk and spatial homogeneity of the classified objects than do several popular methods, including classification and regression trees, learning vector quantization, causal hidden Markov models (HMMs), and multiresolution HMMs. The computational load of HMGMM is similar to that of the causal HMM.     

基于不同隐马尔科夫模型的图像识别方法

为了更详细地研究隐马尔科夫模型在图像识别中的应用,以指纹识别为例,纵向总结了几种基于隐马尔科夫模型的指纹图像识别算法,包括一维隐马尔科夫模型、伪二维隐马尔科夫模型、二维模型及一维模型组。分别从时间复杂度、识别精确度等方面总结出这四种隐马尔科夫模型在图像识别时的优缺点,得出不同待识别图像适合使用的识别模型的结论。     

Activity modeling using event probability sequences.

Changes in motion properties of trajectories provide useful cues for modeling and recognizing human activities. We associate an event with significant changes that are localized in time and space, and represent activities as a sequence of such events. The localized nature of events allows for detection of subtle changes or anomalies in activities. In this paper, we present a probabilistic approach for representing events using the hidden Markov model (HMM) framework. Using trained HMMs for activities, an event probability sequence is computed for every motion trajectory in the training set. It reflects the probability of an event occurring at every time instant. Though the parameters of the trained HMMs depend on viewing direction, the event probability sequences are robust to changes in viewing direction. We describe sufficient conditions for the existence of view invariance. The usefulness of the proposed event representation is illustrated using activity recognition and anomaly detection. Experiments using the indoor University of Central Florida human action dataset, the Carnegie Mellon University Credo Intelligence, Inc., Motion Capture dataset, and the outdoor Transportation Security Administration airport tarmac surveillance dataset show encouraging results.     

Modeling burst channels using partitioned Fritchman's Markov models

Discrete models based on functions of Markov chains (also referred to as hidden Markov models or finite-state channel (FSC) models) have been used to characterize the error process in communication channels with memory. One important property of these models is that the probability of any observed sequence can be expressed as a linear combination of the probability of a finite set of sequences of finite length, the so-called basis sequences. In this paper, we express the parameters of a class of FSC models as a simple function of the probability of the basis sequences. Based on this approach, we propose a new method for the parameterization of the Fritchman (1967) channel with single-error state as well as the interesting cases of Fritchman channels with more than one error state and the Gilbert-Elliott channel ((GEC) nonrenewal models). To illustrate the method, FSC models for the nonfrequency-selective Rician fading channel are presented. The number of states and the probability of state transitions are estimated for a given set of fading parameters     

Vision-Based Learning Status Monitoring on Color and Depth Live Facial Videos

This study analyzes live facial videos for recognizing nonverbal learning-related facial movements and head poses to discover the learning status of students. First, color and depth facial videos captured by a Kinect are analyzed for face tracking using a three-dimensional (3D) active appearance model (AAM). Second, the facial feature vector sequences are used to train hidden Markov models (HMMs) to recognize seven learning-related facial movements (smile, blink, frown, shake, nod, yawn, and talk). The final stage involves the analysis of the facial movement vector sequence to evaluate three status scores (understanding, interaction, and consciousness), each represents the learning status of a student and is helpful to both teachers and students for improving teaching and learning. Five teaching activities demonstrate that the proposed learning status analysis system promotes the interpersonal communication between teachers and students.     

Multiresolution image classification by hierarchical modeling withtwo-dimensional hidden Markov models

This paper treats a multiresolution hidden Markov model for classifying images. Each image is represented by feature vectors at several resolutions, which are statistically dependent as modeled by the underlying state process, a multiscale Markov mesh. Unknowns in the model are estimated by maximum likelihood, in particular by employing the expectation-maximization algorithm. An image is classified by finding the optimal set of states with maximum a posteriori probability. States are then mapped into classes. The multiresolution model enables multiscale information about context to be incorporated into classification. Suboptimal algorithms based on the model provide progressive classification that is much faster than the algorithm based on single-resolution hidden Markov models     

Infinite hidden Markov models for unusual-event detection in video.

We address the problem of unusual-event detection in a video sequence. Invariant subspace analysis (ISA) is used to extract features from the video, and the time-evolving properties of these features are modeled via an infinite hidden Markov model (iHMM), which is trained using "normal"/"typical" video. The iHMM retains a full posterior density function on all model parameters, including the number of underlying HMM states. Anomalies (unusual events) are detected subsequently if a low likelihood is observed when associated sequential features are submitted to the trained iHMM. A hierarchical Dirichlet process framework is employed in the formulation of the iHMM. The evaluation of posterior distributions for the iHMM is achieved in two ways: via Markov chain Monte Carlo and using a variational Bayes formulation. Comparisons are made to modeling based on conventional maximum-likelihood-based HMMs, as well as to Dirichlet-process-based Gaussian-mixture models.     

Decision‐Tree‐Based Markov Model for Phrase Break Prediction

In this paper, a decision‐tree‐based Markov model for phrase break prediction is proposed. The model takes advantage of the non‐homogeneous‐features‐based classification ability of decision tree and temporal break sequence modeling based on the Markov process. For this experiment, a text corpus tagged with parts‐of‐speech and three break strength levels is prepared and evaluated. The complex feature set, textual conditions, and prior knowledge are utilized; and chunking rules are applied to the search results. The proposed model shows an error reduction rate of about 11.6% compared to the conventional classification model.     

Maximum likelihood hidden Markov modeling using a dominant sequenceof states

Approximate maximum likelihood (ML) hidden Markov modeling using the most likely state sequence (MLSS) is examined and compared with the exact ML approach that considers all possible state sequences. It is shown that for any hidden Markov model (HMM), the difference between the approximate and the exact normalized likelihood functions cannot exceed the logarithm of the number of states divided by the dimension of the output vectors (frame length). Furthermore, for Gaussian HMMs and a given observation sequence, the MLSS is typically the sequence of nearest neighbor states in the Itakura-Saito sense, and the posterior probability of any state sequence which departs from the MLSS in a single time instant, decays exponentially with the frame length. Hence, for a sufficiently large frame length the exact and approximate ML approach provide similar model estimates and likelihood values     

Minimum error rate training for PHMM-based text recognition

Discriminative training is studied to improve the performance of our pseudo two-dimensional (2-D) hidden Markov model (PHMM) based text recognition system. The aim of this discriminative training is to adjust model parameters to directly minimize the classification error rate. Experimental results have shown great reduction in recognition error rate even for PHMMs already well-trained using conventional maximum likelihood (ML) approaches.     

Classification of nonstationary narrowband signals using segmented chirp features and hidden Gauss-Markov models

A method is provided for classifying finite-duration signals with narrow instantaneous bandwidth and dynamic instantaneous frequency (IF). In this method, events are partitioned into nonoverlapping segments, and each segment is modeled as a linear chirp, forming a piecewise-linear IF model. The start frequency, chirp rate, signal energy, and noise energy are estimated in each segment. The resulting sequences of frequency and rate features for each event are classified by evaluating their likelihood under the probability density function (PDF) corresponding to each narrowband class hypothesis. The class-conditional PDFs are approximated using continuous-state hidden Gauss-Markov models (HGMMs), whose parameters are estimated from labeled training data. Previous HGMM algorithms are extended by dynamically weighting the output covariance matrix by the ratio of the estimated signal and noise energies from each segment. This covariance weighting discounts spurious features from segments with low signal-to-noise ratio (SNR), making the algorithm more robust in the presence of dynamic noise levels and fading signals. The classification algorithm is applied in a simulated three-class cross-validation experiment, for which the algorithm exhibits percent correct classification greater than 97% as low as -7 dB SNR.     

基于探测过程建模的探地雷达多目标识别

在探地雷达探测过程中,天线相对目标的远近变化反映在面向深度的一维时域信号(A-scan)所组成的序列的变化过程中,由此提出一种针对变化过程建模的目标识别方法。在特征提取环节,提出将时频分析与图像纹理分析相结合,首先计算A-scan信号的二维时频联合分布图像,再利用特定的图像纹理描述算子构造特征向量。识别过程根据目标与天线间距离的变化,采用无跨越单向连续隐马尔可夫模型(HMM)对序列的变化过程建模。实验表明这种基于变化过程的HMM方法比无序地利用单条A-scan特征的支持向量机方法具有更好的效果。     

Applications of similarity mapping in dynamic MRI

Dynamic images are temporal sequences of images, where the intensities of certain regions of interest (ROI's) change with time, whereas anatomical structures remain stationary. Here, new applications of dynamic image analysis, called similarity mapping, are reviewed. Similarity mapping identifies regions in a dynamic image sequence according to their temporal similarity or dissimilarity with respect to a reference ROI. Pixels in the resulting similarity map whose temporal sequence is similar to the reference ROI have high correlation values and are bright, while those with low correlation values are dark. Therefore, similarity mapping segments structures in a dynamic image sequence based on their temporal responses rather than spatial properties. The authors describe the abilities of similarity mapping to identify different image structures present in several dynamic MRI datasets with potential clinical value. They demonstrate that similarity mapping technique has been successful in identifying the following structures: 1) renal cortex and medulla, 2) activated areas of the brain during photic stimulation, 3) ischemia in the left coronary artery territory, 4) lung tumor, 5) tentorial meningioma, and 6) a region of focal ischemia in brain.     

Shape-based tracking of left ventricular wall motion

An approach for tracking and quantifying the nonrigid, nonuniform motion of the left ventricular (LV) endocardial wall from two-dimensional (2-D) cardiac image sequences, on a point-by-point basis over the entire cardiac cycle, is presented. Given a set of boundaries, motion computation involves first matching local segments on one contour to segments on the next contour in the sequence using a shape-based strategy. Results from the match process are incorporated with a smoothness term into an optimization functional. The global minimum of this functional is found, resulting in a smooth flow field that is consistent with the match data. The computation is performed for all pairs of frames in the temporal sequence and equally sampled points on one contour are tracked throughout the sequence, resulting in a composite flow field over the entire sequence. Two perspectives on characterizing the optimization functional are presented which result in a tradeoff resolved by the confidence in the initial boundary segmentation. Experimental results for contours derived from diagnostic image sequences of three different imaging modalities are presented. A comparison of trajectory estimates with trajectories of gold-standard markers implanted in the LV wall are presented for validation. The results of this comparison confirm that although cardiac motion is a three-dimensional (3-D) problem, two-dimensional (2-D) analysis provides a rich testing ground for algorithm development     

Iterative maximum likelihood displacement field estimation inquantum-limited image sequences

We develop an algorithm for obtaining the maximum likelihood (ML) estimate of the displacement vector field (DVP) from two consecutive image frames of an image sequence acquired under quantum-limited conditions. The estimation of the DVF has applications in temporal filtering, object tracking, stereo matching, and frame registration in low-light level image sequences as well as low-dose clinical X-ray image sequences. In the latter case, a controlled X-ray dosage reduction may be utilized to lower the radiation exposure to the patient and the medical staff. The quantum-limited effect is modeled as an undesirable, Poisson-distributed, signal-dependent noise artifact. A Fisher-Bayesian formulation is used to estimate the DVF and a block component search algorithm is employed in obtaining the solution. Several experiments involving a phantom sequence and a teleconferencing image sequence with realistic motion demonstrate the effectiveness of this estimator in obtaining the DVF under severe quantum noise conditions (20-25 events/pixel).     

Rotation and gray-scale transform-invariant texture classificationusing spiral resampling, subband decomposition, and hidden Markov model

This paper proposes a new texture classification algorithm that is invariant to rotation and gray-scale transformation. First, we convert two-dimensional (2-D) texture images to one-dimensional (1-D) signals by spiral resampling. Then, we use a quadrature mirror filter (QMF) bank to decompose sampled signals into subbands. In each band, we take high-order autocorrelation functions as features. Features in different bands, which form a vector sequence, are then modeled as a hidden Markov model (BMM). During classification, the unknown texture is matched against all the models and the best match is taken as the classification result. Simulations showed that the highest correct classification rate for 16 kinds of texture was 95.14%