Computerized analysis of mammographic microcalcifications in morphological and texture feature spaces

We are developing computerized feature extraction and classification methods to analyze malignant and benign microcalcifications on digitized mammograms. Morphological features that described the size, contrast, and shape of microcalcifications and their variations within a cluster were designed to characterize microcalcifications segmented from the mammographic background. Texture features were derived from the spatial gray-level dependence (SGLD) matrices constructed at multiple distances and directions from tissue regions containing microcalcifications. A genetic algorithm (GA) based feature selection technique was used to select the best feature subset from the multi-dimensional feature spaces. The GA-based method was compared to the commonly used feature selection method based on the stepwise linear discriminant analysis (LDA) procedure. Linear discriminant classifiers using the selected features as input predictor variables were formulated for the classification task. The discriminant scores output from the classifiers were analyzed by receiver operating characteristic (ROC) methodology and the classification accuracy was quantified by the area, Az, under the ROC curve. We analyzed a data set of 145 mammographic microcalcification clusters in this study. It was found that the feature subsets selected by the GA-based method are comparable to or slightly better than those selected by the stepwise LDA method. The texture features (Az = 0.84) were more effective than morphological features (Az = 0.79) in distinguishing malignant and benign microcalcifications. The highest classification accuracy (Az = 0.89) was obtained in the combined texture and morphological feature space. The improvement was statistically significant in comparison to classification in either the morphological (p = 0.002) or the texture (p = 0.04) feature space alone. The classifier using the best feature subset from the combined feature space and an appropriate decision threshold could correctly identify 35% of the benign clusters without missing a malignant cluster. When the average discriminant score from all views of the same cluster was used for classification, the Az value increased to 0.93 and the classifier could identify 50% of the benign clusters at 100% sensitivity for malignancy. Alternatively, if the minimum discriminant score from all views of the same cluster was used, the Az value would be 0.90 and a specificity of 32% would be obtained at 100% sensitivity. The results of this study indicate the potential of using combined morphological and texture features for computer-aided classification of microcalcifications.     

Practical Aspects in Comparing Shock-Capturing Schemes for Dam Break Problems

The results of a survey aimed at comparing the performances of first-order and total variation diminishing (TVD) second-order upwind flux difference splitting schemes, first-order space-centered schemes, and second-order space-centered schemes with the TVD artificial viscosity term are reported here. The schemes were applied to the following dam-break wave cases: in a dry frictionless horizontal channel; in a dry, rough and sloping channel; and in a nonprismatic channel. Among first-order schemes, the diffusive scheme provides only slightly less accurate results than those obtained by the Roe scheme. For TVD second-order schemes, no significant difference between the upwind scheme and central schemes are reported. In the case of a dam break in a dry frictionless horizontal channel, the second-order schemes were two- to five-fold more accurate than the diffusive scheme and Roe’s scheme. These differences in scheme performances drastically reduce when the results obtained for the rough sloping channel test and for the nonprismatic channel test are analyzed. In particular, the accuracy of the diffusive and Roe’s schemes is similar to second-order schemes when such features of dam break wave, relevant from an engineering viewpoint, like wave peak arrival time and maximum water depths, are considered.     

冷轧带钢表面质量自动监测系统的模式识别方法研究

研究了冷轧带钢表面质量自动监测系统的模式识别过程，包括特征的提取和选择，样本的获取，分类器的设计等。提取了缺陷的83个特征量，并且采用基于信息熵的特征选择方法对这些特征量进行了优化选择，构建了基于BP神经网络的分类器模型，通过现场采集的样本对其进行训练作测试，结果表明该分类器能够很好地识别6种常见的冷轧带钢表面缺陷。     

Design of a high-sensitivity classifier based on a genetic algorithm: application to computer-aided diagnosis

A genetic algorithm (GA) based feature selection method was developed for the design of high-sensitivity classifiers, which were tailored to yield high sensitivity with high specificity. The fitness function of the GA was based on the receiver operating characteristic (ROC) partial area index, which is defined as the average specificity above a given sensitivity threshold. The designed GA evolved towards the selection of feature combinations which yielded high specificity in the high-sensitivity region of the ROC curve, regardless of the performance at low sensitivity. This is a desirable quality of a classifier used for breast lesion characterization, since the focus in breast lesion characterization is to diagnose correctly as many benign lesions as possible without missing malignancies. The high-sensitivity classifier, formulated as the Fisher's linear discriminant using GA-selected feature variables, was employed to classify 255 biopsy-proven mammographic masses as malignant or benign. The mammograms were digitized at a pixel size of 0.1 mm x 0.1 mm, and regions of interest (ROIs) containing the biopsied masses were extracted by an experienced radiologist. A recently developed image transformation technique, referred to as the rubber-band straightening transform, was applied to the ROIs. Texture features extracted from the spatial grey-level dependence and run-length statistics matrices of the transformed ROIs were used to distinguish malignant and benign masses. The classification accuracy of the high-sensitivity classifier was compared with that of linear discriminant analysis with stepwise feature selection (LDAsfs). With proper GA training, the ROC partial area of the high-sensitivity classifier above a true-positive fraction of 0.95 was significantly larger than that of LDAsfs, although the latter provided a higher total area (Az) under the ROC curve. By setting an appropriate decision threshold, the high-sensitivity classifier and LDAsfs correctly identified 61% and 34% of the benign masses respectively without missing any malignant masses. Our results show that the choice of the feature selection technique is important in computer-aided diagnosis, and that the GA may be a useful tool for designing classifiers for lesion characterization.     

Adaptation of cortical but not hippocampal NMDA receptors after chronic citalopram treatment

Forty three cases of astrocytic tumors and mixed gliomas were studied with the aim of evaluating the reproducibility of the Kernohan grading system vis a vis (a) grading using computer-aided malignancy classifier TESTAST 268 and (b) grading by quantitative morphometric evaluation of the various histological parameters of TESTAST 268. These patients were then followed up for variable periods with a maximum of forty months. High inter and intra-observer variability were observed in the Kernohan grading system. TESTAST 268 was found to be simpler, rapid and more reproducible. However, one drawback observed of this system was that it did not completely eliminate inter-observer variability because there was still some subjectivity in assignment of the categorical values against the histological features. Morphometric evaluation of the semi-quantitative assignment values of the 4 histological variables in the TESTAST 268 classifier using Zeiss Morphomat-30 revealed a statistically significant difference between the clusters of the measured quantitative values. A repeat grading using TESTAST 268 and categorical assignment values of histological features derived from the absolute values obtained by morphometry resulted in complete elimination of inter-observer variability. Thus, this study highlights the importance of objectivisation using TESTAST 268 and histologic morphometry in the grading of gliomas. However, since this is a preliminary study on a small number of cases, no cut off values of these measurements have been proposed.     

Synthesis of algorithms of control over the process of decomposition of nickel tetracarbonyl

Construction of an automated control system over the process of decomposition of nickel tetracarbonyl is considered. Confidence intervals of the obtained first-order and second-order models are analyzed. A model is described that allows one to compensate for the influence of perturbing factors and mutual influence of control effects on neighboring regions. Based on the algorithm presented, it is possible to construct an adaptive system of automated control over the process of decomposition of nickel tetracarbonyl.     

Position displacement, not velocity, is the cue to motion detection of second-order stimuli

Motion detection can be achieved either with mechanisms sensitive to a target's velocity, or sensitive to change in a target's position. Using a procedure to dissociate these two provided by Nakayama and Tyler (Vis Res 1981;21:427-433), we explored detection of first-order (luminance-based) and various second-order (texture-based and stereo-based) motion. In the first experiment, observers viewed annular gratings oscillating in rotational motion at various rates. For each oscillation temporal frequency, we determined the minimum displacement of the pattern for which observers could reliably see motion. For first-order motion, these motion detection thresholds decreased with increasing temporal frequency, and thus were determined by a minimum velocity. In contrast, motion detection thresholds for second-order motion remained roughly constant across temporal frequency, and thus were determined by a minimum displacement. In Experiment 2, luminance-based gratings of different contrasts were tested to show that the velocity-dependence was not an artifact of pattern visibility. In the remaining experiments, results similar to Experiment 1 were obtained with a central presentation of a linear grating, instead of an annular grating (Experiment 3), and with a motion discrimination (phase discrimination) rather than motion detection task (Experiment 4). We conclude that, within the ranges tested here, second-order motion is more readily detected with a mechanism which tracks the change of position of features over time.     

Temporal frequency modulates reaction time responses to first-order and second-order motion.

This study investigated the effect of temporal frequency and modulation depth on reaction times for discriminating the direction of first-order (luminance-defined) and second-order (contrast-defined) motion, equated for visibility using equal multiples of direction-discrimination threshold. Results showed that reaction times were heavily influenced by temporal frequency, especially in the case of second-order motion. At 1 Hz, reaction times were faster for first-order compared with second-order motion. As temporal frequency increased, reaction times for first-order motion decreased slightly, but those for second-order motion decreased more rapidly. At 8 Hz, reaction times for second-order motion were, in many cases, faster than those for first-order motion. Reaction times decreased as stimulus modulation depth increased at approximately the same rate for both motion types. The findings demonstrate that behavioral response latencies to first-order and second-order motion are dependent on specific stimulus parameters and may, in some cases, be shorter in response to second-order compared with first-order motion. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Visual Search Asymmetry With Uncertain Targets.

The underlying mechanism of search asymmetry is still unknown. Many computational models postulate top-down selection of target-defining features as a crucial factor. This feature selection account implies, and other theories implicitly assume, that predefined target identity is necessary for search asymmetry. The authors tested the validity of the feature selection account using a singleton search task without a predefined target. Participants conducted a target-defined and a singleton search task with a circle (O) and a circle with a vertical bar (Q). Search asymmetry was observed in both tasks with almost identical magnitude. The results were not due to trial-by-trial feature selection, because search asymmetry persisted even when the target was completely unpredictable. Asymmetry in the singleton search was also observed with more complex stimuli, Kanji characters. These results suggest that feature selection is not necessary for search asymmetry, and they impose important constraints on current visual search theories. (PsycINFO Database Record (c) 2010 APA, all rights reserved)     

Comparing wall motion quantification methods by numerical classification

Quantitative left ventriculography has been studied for many years as a method for assessing left ventricular function. All currently used quantitative left ventriculography methods evaluate functional status based on different assumptions concerning the time course of contraction, shape of the ventricular silhouette, and/or fixed points of reference. Commonly, analyses are performed using only end-diastolic (ED) and end-systolic (ES) frames. However, myocardial infarction manifests itself not only in spatial parameters, e.g., area of affected myocardium, but also in temporal parameters, e.g., deviation from a synchronous contraction during systole. ED/ES comparisons omit information contained in other frames of the cardiac cycle. Therefore, each single method has a limited reliability or usefulness. The present study used several quantitative methods, fitted curvatures, correlation coefficients of the percentage of radial shortening from end diastole to end systole, and Fourier coefficients of the polar coordinates of the endocardial contour points. Features were derived from a sequence of ventriculograms of a whole cardiac cycle from patients with anterior myocardial infarction (AMI group) and without (NV group) to address three major goals: (1) Determine features derived from LV-grams that discriminate between the two patient groups and do so with the highest accuracy, "recognition rate," or lowest error rates; (2) compare the accuracy of features generated by different quantitative methods; and (3) group features generated by the different quantitative methods to achieve higher accuracy. The discriminant capabilities of several feature sets were tested on a patient sample containing 102 patients (63 patients with AMI, 39 patients with no infarction) using numerical classification. The numerical classification had a Supervised Learning design with two cardiologists assigning the patients to the two patient groups. Each quantification method showed a characteristic pattern for detecting pathologic or normal wall motion. Combining features generated by different extraction methods into one feature set improved the correct classification rates.     

Flux-Difference Splitting Schemes for 2D Flood Flows

Two numerical models for 2D flood flows are presented. One model is first-order accurate and another is second-order accurate. Roe's numerical flux is used to develop the first-order accurate model, while second-order accuracy, in space and time, is obtained by using the Lax-Wendroff numerical flux. A simple operator splitting is found to yield the same results as that obtained by using more complicated, and thus, time consuming, operator splitting. Roe's approximate Jacobian is used for conservative properties and Harten and Hyman's procedure is followed for the entropy inequality condition. Flux limiter is used in the second-order accurate model that removes oscillations while maintaining the order of accuracy. The models are verified against available experimental data of a 2D flood wave due to partial dam-break. Numerical experiments are conducted to verify the models' ability to correctly predict behavior of the free surface, in addition to prediction of depth and velocity.     

Modeling of Temperature–Frequency Correlation Using Combined Principal Component Analysis and Support Vector Regression Technique

A good understanding of environmental effects on structural modal properties is essential for reliable performance of vibration-based damage diagnosis methods. In this paper, a method of combining principal component analysis (PCA) and support vector regression (SVR) technique is proposed for modeling temperature-caused variability of modal frequencies for structures instrumented with long-term monitoring systems. PCA is first applied to extract principal components from the measured temperatures for dimensionality reduction. The predominant feature vectors in conjunction with the measured modal frequencies are then fed into a support vector algorithm to formulate regression models that may take into account thermal inertia effect. The research is focused on proper selection of the hyperparameters to obtain SVR models with good generalization performance. A grid search method with cross validation and a heuristic method are utilized for determining the optimal values of SVR hyperparameters. The proposed method is compared with the method directly using measurement data to train SVR models and the multivariate linear regression (MLR) method through the use of long-term measurement data from a cable-stayed bridge. It is shown that PCA-compressed features make the training and validation of SVR models more efficient in both model accuracy and computational costs, and the formulated SVR model performs much better than the MLR model in generalization performance. When continuously measured data is available, the SVR model formulated taking into account thermal inertia effect can achieve more accurate prediction than that without considering thermal inertia effect.     

基于互信息特征筛选偏最小二乘的激光诱导击穿光谱铁矿浆定量分析

目前检测矿浆品位相对准确的方法是传统化学分析,但周期长、有滞后性,无法实现在线检测。实验利用激光诱导击穿光谱(Laser induced breakdown spectroscopy,LIBS)在线、原位、快速等优点,分析了铁矿选矿过程尾矿浆中铁元素的品位值。由于LIBS采集到的光谱数据中存在大量对成分分析无用的冗余信息,进而增加了建模复杂程度,导致建立的模型精确度不够、泛化能力不强。因此,在偏最小二乘(PLS)模型基础上,提出了基于互信息特征筛选的偏最小二乘模型。实验结果表明,与传统的PLS模型相比,基于互信息特征筛选的偏最小二乘模型在分析精度上得到了明显改善,测试样品的决定系数R²从0.52提高到0.90,测试样本的平均绝对误差(MAEP)从2.87%下降到1.38%,总样本的平均绝对误差(MAE)从1.0%下降到0.60%。     

Non-Gaussian Air Gap Response Models for Floating Structures

The air gap response of a specific semisubmersible platform subjected to irregular waves is considered. Statistical analyses are performed on model test data for the absolute near-structure wave elevation, and these measured data are compared with predictions resulting from probabilistic models. Models applied are first-order and second-order diffraction models typical of standard practice, and two new hybrid models that include second-order effects in the incident wave, but not in the diffracted wave. The first of these hybrid models is moment based, while the second relies on narrow-band random process theory. Either of these new models can be implemented in place of the standard linear-only model with little additional computational effort, as only linear diffraction analysis is required. Both new models are found to better predict the air gap demand than standard linear diffraction analysis.     

联合多种边缘检测算子的无参考质量评价算法

提出了一种联合多种边缘检测算子的无参考质量评价算法,同时考虑一阶和二阶边缘算子,避免了单一算子的局限性.该方法首先将彩色图像转换为灰度图像,然后计算灰度图像的梯度,相对梯度以及LOG特征.本文所使用的特征分为两部分,一部分提取相对梯度方向的标准差,另一部分利用条件熵来量化不同特征之间的相似性和相互关系,并且考虑到人眼特性进行多尺度计算,最后使用自适应增强（AdaBoost）神经网络进行训练和预测.在公共数据库LIVE和TID2008上进行实验,结果表明新方法对失真图像的预测评分与主观评分有较高的一致性,能很好地反映图像质量的视觉感知效果,仅使用10维特征,性能优于现有的主流无参考质量评价算法.      

Macro-EMG and motor unit recruitment threshold: differences between the young and the aged

The relationship between macro-EMG (electromyography) and motor unit recruitment threshold was studied in the first dorsal interosseous (FDI) muscle of normal young and aged subjects. During voluntary isometric contraction, motor unit action potentials (MUAP) were collected by a special quadrifilar electrode and decomposed to each MUAP train (MUAPT) using an EMG signal decomposition technique. Macro-EMG was obtained from the electrode shaft, then triggered and averaged for each MUAPT. A positive linear correlation was observed in both the young and aged subjects. However, the correlation coefficients were significantly lower in the aged individuals than in the young individuals.     

Magnetic transcranial stimulation in healthy humans: influence on the behavior of upper limb motor units

Aim of the study was to analyze the characteristics of motor action potentials recruitment during magnetic trans-cranial stimulation (TCS) of the brain. Coaxial needle recordings from hand and upper limb musculature, as well as surface electrodes were employed in 20 healthy controls during magnetic TCS with regular and figure-of-8 coil in different experimental protocols including: (a) simple reaction time paradigm during which TCS at subthreshold intensity for eliciting MEPs in relaxation was delivered at various intervals between the signal to move and the onset of the voluntary EMG burst; (b) suprathreshold TCS was randomly delivered while the subject was voluntarily firing at a regular rate one 'low' and/or 'high threshold' motor unit action potential (MUAP). The pre- and post-TCS MUAPs recruitment as well as their firing rates were compared; (c) recordings with two separate needles picking up individual MUAPs from the same or from two different muscles were obtained in order to test 'synchrony' of MUAP's discharge before and after TCS; (d) the influence of the time-interval separating the last discharged MUAP from TCS was evaluated. (e) differences between simultaneous surface and depth recordings were examined. The following results were obtained. (a) The same low-amplitude MUAP which is first voluntarily recruited at the onset of the EMG burst is the one initially fired by TCS in the pre-movement period. Latency shortenings and amplitude enlargement of surface MEPs were observed with faster reaction times. Such changes were coupled to the recruitment of high-threshold MUAPs being larger in amplitude and briefer in latency than the initial one. (b) When using suprathreshold TCS, MEPs followed by silent periods were found. The SP was followed by a rebound acceleration of the MUAPs firing rate compared with pre-TCS levels. Besides rebound acceleration, new MUAPs of larger amplitude than the original (= pre-stimulus) ones were recruited beyond the voluntary control. This phenomenon-together with longer SPs- was progressively more pronounced with stronger stimuli. (c) TCS was affecting the 'synchrony' of MUAPs. (d) If the latency difference between the last pre-stimulus spike and the TCS was exceeding the half-cycle of the MUAP 'natural' firing, the SP was longer in duration. (e) SPs not preceded by MEPs were clearly present in depth recordings. Surface recordings mainly reflected the behavior of high-threshold and large MUAPs.     

Quantitative image sequence analysis of fundus fluorescein angiography

Interpretation of retinal angiographic studies has heretofore been largely qualitative. The temporal properties associated with fluorescein fluorescence of normal and pathologic fundus features are quantified to potentiate pixel assignment and fundus feature quantitation for clinical studies where precise image metrology is vital. Fluorescein angiography studies were digitized, and temporally sequential images were spatially registered with polynomial warping algorithms, allowing for the construction of a three-dimensional angiogram vector. Temporal profiles through spatially registered, temporally sequential pixels were computed. Fundus feature fluorescence behavior was quantified. Fundus features may be discriminated on the basis of spatio-temporal fluorescein fluorescence properties. Pixel assignment based on spatio-temporal relationships will facilitate fundus feature quantitation.     

Mapping of retinal function in diabetic retinopathy using the multifocal electroretinogram

PURPOSE: To investigate focal abnormalities in the electroretinogram (ERG) signal in diabetic patients, with and without retinopathy, using a multifocal ERG. METHODS: Sixteen patients with diabetes mellitus, 8 of whom had diabetic retinopathy (mean duration of diabetes: 18.5 years) and 19 approximately age-matched healthy volunteers underwent multifocal ERG testing. One hundred three retinal locations within the central 50 degrees were stimulated concurrently, according to a pseudorandom m-sequence. Response components were extracted for each stimulated retinal location. RESULTS: In diabetic patients with retinopathy, the overall amplitudes were reduced (P < 0.01), and peak implicit times were increased (P < 0.01) in the first-order component (mean flash response) and in the first slice of the second-order component (local two flash interaction). In addition, local reductions of amplitude could be seen in the first- and second-order components. In patients without retinopathy, only amplitudes of the second-order component were reduced (P < 0.01). Another salient difference was observed in a special feature of the second-order component that was reduced in diabetic patients, with and without retinopathy (P < 0.05). CONCLUSIONS: Second-order components depend on nonlinear dynamics. Thus our findings indicate changes in the nonlinear dynamics of a fast-gain control in diabetic patients, presumably located in the inner retina. This suggests that early functional changes of the inner retina are evident in diabetic patients before impairment of the outer retina is observed. Multifocal nonlinear analysis permits the detection of subclinical diabetic retinopathy and offers the advantage of topographic mapping of retinal dysfunction.     

Effects of localized pontine lesions on auditory brain-stem evoked potentials and binaural processing in humans

Potassium channels are involved in the control of neuronal excitability by fixing the membrane potential, shaping the action potential, and setting firing rates. Recently, attention has been focused on identifying the factors influencing excitability in second-order auditory and vestibular neurons. Located in the brainstem, second-order auditory and vestibular neurons are sites for convergence of inputs from first-order auditory or vestibular ganglionic cells with other sensory systems and also motor areas. Typically, second-order auditory neurons exhibit two distinct firing patterns in response to depolarization: tonic, with a repetitive firing of action potentials, and phasic, characterized by only one or a few action potentials. In contrast, all mature vestibular second-order neurons fire tonically on depolarization. Already, certain fundamental roles have emerged for potassium currents in these neurons. In mature auditory and vestibular neurons, I(K), the delayed rectifier, is required for the fast repolarization of action potentials. In tonically firing auditory neurons, I(A), the transient outward rectifier, defines the discharge pattern. I(DS), a delayed rectifier-like current distinguished by its low threshold of activation, is found in phasically firing auditory and some developing vestibular neurons where it limits firing to one or a few spikes, and also may contribute to forming short-duration excitatory postsynaptic potential (EPSPs). Also, I(DS) sets the threshold for action potential generation rather high, which may prevent spontaneous discharge in phasically firing cells. During development, there is a gradual acquisition and loss of some potassium conductances, suggesting developmental regulation. As there are similarities in membrane properties of second-order auditory and vestibular neurons, investigations on firing pattern and its underlying mechanisms in one system should help to uncover fundamental properties of the other.