Mining the structural knowledge of high-dimensional medical data using isomap

The paper describes an application of a new, non-linear dimensionality reduction method, named Isomap, for mining the structural knowledge from high-dimensional medical data. The algorithm was evaluated on two publicly available medical datasets: the pathological dataset of breast cancer (241 malignant samples) and the gene expression dataset from the lung (186 tumours). It was found by Isomap that the approximate intrinsic dimensionalities of these two datasets were as low as three. The spatial structures of both datasets were presented in low-dimensional space. Isomap, as a general tool for dimensionality reduction analysis, is helpful in revealing the nonlinear structural knowledge of high-dimensional medical data.     

基于薄板样条的MRI图像与脑图谱的配准方法

为了将CT、MRI、PET或SPECT等断层扫描图像用于疾病的辅助诊断、放射治疗、手术计划和引导 ,就必须知道图像中感兴趣区 (ROI)是什么解剖组织 ,即解决医学图像的解剖标识问题。医生通常是从解剖书籍、图谱及自身经验来对ROI做出判断。这些书籍和图谱往往给出的是文字描述和有限数目的 ,固定位置和方向的断层图片 ,很难与患者的实际图像联系起来。对于缺乏临床经验的医生来说尤为困难。数字化 3D人脑解剖图谱使医生对人脑深部组织全方位可视化。因此可以将其通过一定的空间变换 ,与MR体积数据集中的ROI进行比较 ,从而得到ROI的解剖标识。但是 ,没有两个人的大脑是完全一样的。人脑的解剖个体差异较大 ,这就要求利用非线性变形的方法做解剖标识。本研究介绍通过薄板样条变换用Talirach脑图谱对MR图像做解剖标识的方法。     

Extracting the parameters of the double-dispersion Cole bioimpedance model from magnitude response measurements

In the field of bioimpedance measurements, the Cole impedance model is widely used for characterizing biological tissues and biochemical materials. In this work, a nonlinear least squares fitting is applied to extract the double-dispersion Cole impedance parameters from simulated magnitude response datasets without requiring the direct impedance data or phase information. The technique is applied to extract the impedance parameters from MATLAB simulated noisy magnitude datasets showing less than 1.2 % relative error when 60 dB SNR Gaussian white noise is present. This extraction is verified experimentally using apples as the Cole impedances showing less than 3 % relative error between simulated responses (using the extracted impedance parameters) and the experimental results over the entire dataset.     

Correcting frequency and phase offsets in MRS data using robust spectral registration

An algorithm for retrospective correction of frequency and phase offsets in MRS data is presented. The algorithm, termed robust spectral registration (rSR), contains a set of subroutines designed to robustly align individual transients in a given dataset even in cases of significant frequency and phase offsets or unstable lipid contamination and residual water signals. Data acquired by complex multiplexed editing approaches with distinct subspectral profiles are also accurately aligned. Automated removal of unstable lipid contamination and residual water signals is applied first, when needed. Frequency and phase offsets are corrected in the time domain by aligning each transient to a weighted average reference in a statistically optimal order using nonlinear least‐squares optimization. The alignment of subspectra in edited datasets is performed using an approach that specifically targets subtraction artifacts in the frequency domain. Weighted averaging is then used for signal averaging to down‐weight poorer‐quality transients. Algorithm performance was assessed on one simulated and 67 in vivo pediatric GABA‐/GSH‐edited HERMES datasets and compared with the performance of a multistep correction method previously developed for aligning HERMES data. The performance of the novel approach was quantitatively assessed by comparing the estimated frequency/phase offsets against the known values for the simulated dataset or by examining the presence of subtraction artifacts in the in vivo data. Spectral quality was improved following robust alignment, especially in cases of significant spectral distortion. rSR reduced more subtraction artifacts than the multistep method in 64% of the GABA difference spectra and 75% of the GSH difference spectra. rSR overcomes the major challenges of frequency and phase correction.     

Deterministic projection by growing cell structure networks for visualization of high-dimensionality datasets

Recent advances in clinical proteomics data acquisition have led to the generation of datasets of high complexity and dimensionality. We present here a visualization method for high-dimensionality datasets that makes use of neuronal vectors of a trained growing cell structure (GCS) network for the projection of data points onto two dimensions. The use of a GCS network enables the generation of the projection matrix deterministically rather than randomly as in random projection. Three datasets were used to benchmark the performance and to demonstrate the use of this deterministic projection approach in real-life scientific applications. Comparisons are made to an existing self-organizing map projection method and random projection. The results suggest that deterministic projection outperforms existing methods and is suitable for the visualization of datasets of very high dimensionality.     

Nonlinear dimensionality reduction of gene expression data for visualization and clustering analysis of cancer tissue samples

Gene expression data are the representation of nonlinear interactions among genes and environmental factors. Computing analysis of these data is expected to gain knowledge of gene functions and disease mechanisms. Clustering is a classical exploratory technique of discovering similar expression patterns and function modules. However, gene expression data are usually of high dimensions and relatively small samples, which results in the main difficulty for the application of clustering algorithms. Principal component analysis (PCA) is usually used to reduce the data dimensions for further clustering analysis. While PCA estimates the similarity between expression profiles based on the Euclidean distance, which cannot reveal the nonlinear connections between genes. This paper uses nonlinear dimensionality reduction (NDR) as a preprocessing strategy for feature selection and visualization, and then applies clustering algorithms to the reduced feature spaces. In order to estimate the effectiveness of NDR for capturing biologically relevant structures, the comparative analysis between NDR and PCA is exploited to five real cancer expression datasets. Results show that NDR can perform better than PCA in visualization and clustering analysis of complex gene expression data.     

Feature-Motivated Simplified Adaptive PCNN-Based Medical Image Fusion Algorithm in NSST Domain

Multimodality medical image fusion plays a vital role in diagnosis, treatment planning, and follow-up studies of various diseases. It provides a composite image containing critical information of source images required for better localization and definition of different organs and lesions. In the state-of-the-art image fusion methods based on nonsubsampled shearlet transform (NSST) and pulse-coupled neural network (PCNN), authors have used normalized coefficient value to motivate the PCNN-processing both low-frequency (LF) and high-frequency (HF) sub-bands. This makes the fused image blurred and decreases its contrast. The main objective of this work is to design an image fusion method that gives the fused image with better contrast, more detail information, and suitable for clinical use. We propose a novel image fusion method utilizing feature-motivated adaptive PCNN in NSST domain for fusion of anatomical images. The basic PCNN model is simplified, and adaptive-linking strength is used. Different features are used to motivate the PCNN-processing LF and HF sub-bands. The proposed method is extended for fusion of functional image with an anatomical image in improved nonlinear intensity hue and saturation (INIHS) color model. Extensive fusion experiments have been performed on CT-MRI and SPECT-MRI datasets. Visual and quantitative analysis of experimental results proved that the proposed method provides satisfactory fusion outcome compared to other image fusion methods.     

Dose‐response analysis of bromate‐induced DNA damage and mutagenicity is consistent with low‐dose linear,nonthreshold processes

Mutagenic agents have long been inferred to act through low‐dose linear, nonthreshold processes. However, there is debate about this assumption, with various studies interpreting datasets as showing thresholds for DNA damage and mutation. We have applied rigorous statistical analyses to investigate the shape of dose‐response relationships for a series of in vitro and in vivo genotoxicity studies using potassium bromate (KBrO₃), a water ozonation byproduct that is bioactivated to a reactive species causing oxidative damage to DNA. We analyzed studies of KBrO₃ genotoxicity where no‐effect/threshold levels were reported as well as other representative datasets. In all cases, the data were consistent with low‐dose linear models. In the majority of cases, the data were fit either by a linear (straight line) model or a model which was linear at low doses and showed a saturation‐like downward curvature at high doses. Other datasets with apparent upward curvature were still adequately represented by models that were linear at low dose. Sensitivity analysis of datasets showing upward curvature revealed that both low‐dose linear and nonlinear models provide adequate fits. Additionally, a simple biochemical model of selected key processes in bromate‐induced DNA damage was developed and illustrated a situation where response for early primary events suggested an apparent threshold while downstream events were linear. Overall, the statistical analyses of DNA damage and mutations induced by KBrO₃ are consistent with a low‐dose linear response and do not provide convincing evidence for the presence of a threshold. Environ. Mol. Mutagen., 2013. © 2012 Wiley Periodicals, Inc.     

基于二叉树支持向量机的蛋白质结构类预测

提出了一种基于二叉树支持向量机(BT-SVM)的蛋白质结构类多类预测新方法.采用26维的向量来表示蛋白质序列的特征.BT-SVM多类分类方法能消除SVM在多分类问题中存在的不可分数据问题.采用两个经典数据集作为测试数据,通过自身一致性和n折叠交叉验证方法测试了新方法的性能.预测结果表明新方法具有良好的预测能力,与使用同一数据集的已有结果相比较,新方法的Jackknife结果和目前最好的方法取得的结果相当,可作为蛋白质结构类预测的一个工具.     

A Novel Method for Automatic Identification of Motion Artifact Beats in ECG Recordings

This paper presents a novel method for automatic identification of motion artifact beats in ECG recordings. The proposed method is based on the ECG complexes clustering, fuzzy logic and multi-parameters decision. Firstly, eight simulated datasets with different signal-to-noise ratio (SNR) were built for identification experiments. Results show that the identification sensitivity of our method is sensitive to SNR levels and acts like a low-pass filter that matches the cardiologists' recognition, while the Norm FP rate and PVB FP rate keep significantly low regardless of SNR. Furthermore, a simulated dataset including random durations of motion activities superimposed segments and two clinical datasets acquired from two different commercial recorders were adopted for the evaluation of accuracy and robustness. The overall identification results on these datasets were: sensitivity >94.69%, Norm FP rate <0.60% and PVB FP rate <2.65%. All the results were obtained without any manual threshold adjustment according to the priori information, thus dissolving the drawbacks of previous published methods. Additionally, the total cost time of our method applied to 24 h recordings is less than 1 s, which is extremely suitable in the situation of magnanimity data in long-term ECG recordings.     

On the Advantages of Multi-Input Single-Output Parallel Cascade Classifiers

Parallel Cascade Identification (PCI) has been successfully applied to build dynamic nonlinear systems that address diverse challenges in the field of bioinformatics. PCI may be used to identify either single-input single-output (SISO) or multi-input single-output (MISO) models. Although SISO PCI models have typically sufficed, it has been suggested that MISO PCI systems could also be used to form bioinformatics classifiers, and indeed they were successfully applied in one study. This paper reports on the first systematic comparison of MISO and SISO PCI classifiers. Motivation for using the MISO structure is given. The construction of MISO parallel cascade models is also briefly reviewed. In order to compare the accuracy of SISO and MISO PCI classifiers, genetic algorithms are applied to optimize the model architecture on a number of equivalent single-input and multi-input biological training datasets. Through evaluation of both model structures on independent test datasets, we establish that MISO PCI is capable of building classifiers of equal accuracy to those resulting from SISO PCI models. Moreover, we discuss and illustrate the benefits of the MISO approach, including significant reduction in training and testing times, and the ability to adjust automatically the weighting of individual inputs according to information content.     

Managing the CT Data Explosion: Initial Experiences of Archiving Volumetric Datasets in a Mini-PACS

Two image datasets (one thick section dataset and another volumetric dataset) were typically reconstructed from each single CT projection data. The volumetric dataset was stored in a mini-PACS with 271-Gb online and 680-Gb nearline storage and routed to radiologists’ workstations, whereas the thick section dataset was stored in the main PACS. Over a 5-month sample period, 278 Gb of CT data (8976 examinations) was stored in the main PACS, and 738 Gb of volumetric datasets (6193 examinations) was stored in the mini-PACS. The volumetric datasets formed 32.8% of total data for all modalities (2.20 Tb) in the main PACS and mini-PACS combined. At the end of this period, the volumetric datasets of 1892 and 5162 examinations were kept online and nearline, respectively. Mini-PACS offers an effective method of archiving every volumetric dataset and delivering it to radiologists.     

Simultaneous genes and training samples selection by modified particle swarm optimization for gene expression data classification

Gene expression datasets is a means to classify and predict the diagnostic categories of a patient. Informative genes and representative samples selection are two important aspects for reducing gene expression data. Identifying and pruning redundant genes and samples simultaneously can improve the performance of classification and circumvent the local optima problem. In the present paper, the modified particle swarm optimization was applied to selecting optimal genes and samples simultaneously and support vector machine was used as an objective function to determine the optimum set of genes and samples. To evaluate the performance of the new proposed method, it was applied to three publicly available microarray datasets. It has been demonstrated that the proposed method for gene and sample selection is a useful tool for mining high dimension data.     

Cardiac MRI Segmentation Using Mutual Context Information from Left and Right Ventricle

In this paper, we propose a graphcut method to segment the cardiac right ventricle (RV) and left ventricle (LV) by using context information from each other. Contextual information is very helpful in medical image segmentation because the relative arrangement of different organs is the same. In addition to the conventional log-likelihood penalty, we also include a “context penalty” that captures the geometric relationship between the RV and LV. Contextual information for the RV is obtained by learning its geometrical relationship with respect to the LV. Similarly, RV provides geometrical context information for LV segmentation. The smoothness cost is formulated as a function of the learned context which helps in accurate labeling of pixels. Experimental results on real patient datasets from the STACOM database show the efficacy of our method in accurately segmenting the LV and RV. We also conduct experiments on simulated datasets to investigate our method’s robustness to noise and inaccurate segmentations.     

Multi-observation PET image analysis for patient follow-up quantitation and therapy assessment

In positron emission tomography (PET) imaging, an early therapeutic response is usually characterized by variations of semi-quantitative parameters restricted to maximum SUV measured in PET scans during the treatment. Such measurements do not reflect overall tumor volume and radiotracer uptake variations. The proposed approach is based on multi-observation image analysis for merging several PET acquisitions to assess tumor metabolic volume and uptake variations. The fusion algorithm is based on iterative estimation using a stochastic expectation maximization (SEM) algorithm. The proposed method was applied to simulated and clinical follow-up PET images. We compared the multi-observation fusion performance to threshold-based methods, proposed for the assessment of the therapeutic response based on functional volumes. On simulated datasets the adaptive threshold applied independently on both images led to higher errors than the ASEM fusion and on clinical datasets it failed to provide coherent measurements for four patients out of seven due to aberrant delineations. The ASEM method demonstrated improved and more robust estimation of the evaluation leading to more pertinent measurements. Future work will consist in extending the methodology and applying it to clinical multi-tracer datasets in order to evaluate its potential impact on the biological tumor volume definition for radiotherapy applications.     

Robust prediction of protein subcellular localization combining PCA and WSVMs

Automated prediction of protein subcellular localization is an important tool for genome annotation and drug discovery, and Support Vector Machines (SVMs) can effectively solve this problem in a supervised manner. However, the datasets obtained from real experiments are likely to contain outliers or noises, which can lead to poor generalization ability and classification accuracy. To explore this problem, we adopt strategies to lower the effect of outliers. First we design a method based on Weighted SVMs, different weights are assigned to different data points, so the training algorithm will learn the decision boundary according to the relative importance of the data points. Second we analyse the influence of Principal Component Analysis (PCA) on WSVM classification, propose a hybrid classifier combining merits of both PCA and WSVM. After performing dimension reduction operations on the datasets, kernel-based possibilistic c-means algorithm can generate more suitable weights for the training, as PCA transforms the data into a new coordinate system with largest variances affected greatly by the outliers. Experiments on benchmark datasets show promising results, which confirms the effectiveness of the proposed method in terms of prediction accuracy.     

Gene expression data classification with Kernel principal component analysis

One important feature of the gene expression data is that the number of genes M far exceeds the number of samples N. Standard statistical methods do not work well when N < M. Development of new methodologies or modification of existing methodologies is needed for the analysis of the microarray data. In this paper, we propose a novel analysis procedure for classifying the gene expression data. This procedure involves dimension reduction using kernel principal component analysis (KPCA) and classification with logistic regression (discrimination). KPCA is a generalization and nonlinear version of principal component analysis. The proposed algorithm was applied to five different gene expression datasets involving human tumor samples. Comparison with other popular classification methods such as support vector machines and neural networks shows that our algorithm is very promising in classifying gene expression data.     

Wavelet denoising in voxel-based parametric estimation of small animal PET images: a systematic evaluation of spatial constraints and noise reduction algorithms

Voxel-based estimation of PET images, generally referred to as parametric imaging, can provide invaluable information about the heterogeneity of an imaging agent in a given tissue. Due to high level of noise in dynamic images, however, the estimated parametric image is often noisy and unreliable. Several approaches have been developed to address this challenge, including spatial noise reduction techniques, cluster analysis and spatial constrained weighted nonlinear least-square (SCWNLS) methods. In this study, we develop and test several noise reduction techniques combined with SCWNLS using simulated dynamic PET images. Both spatial smoothing filters and wavelet-based noise reduction techniques are investigated. In addition, 12 different parametric imaging methods are compared using simulated data. With the combination of noise reduction techniques and SCWNLS methods, more accurate parameter estimation can be achieved than with either of the two techniques alone. A less than 10% relative root-mean-square error is achieved with the combined approach in the simulation study. The wavelet denoising based approach is less sensitive to noise and provides more accurate parameter estimation at higher noise levels. Further evaluation of the proposed methods is performed using actual small animal PET datasets. We expect that the proposed method would be useful for cardiac, neurological and oncologic applications.     

Suppressing motion artefacts in MRI using an Inception-ResNet network with motion simulation augmentation

The suppression of motion artefacts from MR images is a challenging task. The purpose of this paper was to develop a standalone novel technique to suppress motion artefacts in MR images using a data-driven deep learning approach. A simulation framework was developed to generate motion-corrupted images from motion-free images using randomly generated motion profiles. An Inception-ResNet deep learning network architecture was used as the encoder and was augmented with a stack of convolution and upsampling layers to form an encoder-decoder network. The network was trained on simulated motion-corrupted images to identify and suppress those artefacts attributable to motion. The network was validated on unseen simulated datasets and real-world experimental motion-corrupted in vivo brain datasets. The trained network was able to suppress the motion artefacts in the reconstructed images, and the mean structural similarity (SSIM) increased from 0.9058 to 0.9338. The network was also able to suppress the motion artefacts from the real-world experimental dataset, and the mean SSIM increased from 0.8671 to 0.9145. The motion correction of the experimental datasets demonstrated the effectiveness of the motion simulation generation process. The proposed method successfully removed motion artefacts and outperformed an iterative entropy minimization method in terms of the SSIM index and normalized root mean squared error, which were 5–10% better for the proposed method. In conclusion, a novel, data-driven motion correction technique has been developed that can suppress motion artefacts from motion-corrupted MR images. The proposed technique is a standalone, post-processing method that does not interfere with data acquisition or reconstruction parameters, thus making it suitable for routine clinical practice.     

Classification and knowledge discovery in protein databases

We consider the problem of classification in noisy, high-dimensional, and class-imbalanced protein datasets. In order to design a complete classification system, we use a three-stage machine learning framework consisting of a feature selection stage, a method addressing noise and class-imbalance, and a method for combining biologically related tasks through a prior-knowledge based clustering. In the first stage, we employ Fisher's permutation test as a feature selection filter. Comparisons with the alternative criteria show that it may be favorable for typical protein datasets. In the second stage, noise and class imbalance are addressed by using minority class over-sampling, majority class under-sampling, and ensemble learning. The performance of logistic regression models, decision trees, and neural networks is systematically evaluated. The experimental results show that in many cases ensembles of logistic regression classifiers may outperform more expressive models due to their robustness to noise and low sample density in a high-dimensional feature space. However, ensembles of neural networks may be the best solution for large datasets. In the third stage, we use prior knowledge to partition unlabeled data such that the class distributions among non-overlapping clusters significantly differ. In our experiments, training classifiers specialized to the class distributions of each cluster resulted in a further decrease in classification error.