Lossy image compression is essential for efficient transmission and storage. Traditional compression methods mainly rely on discrete cosine transform (DCT) or singular value decomposition (SVD), both of which represent image data in continuous domains and, therefore, necessitate carefully designed quantizers. Notably, these methods consider quantization as a separate step, which prevents quantization errors from being incorporated into the compression process and degrades the reconstruction quality, particularly in SVD-based methods. To address this issue, we introduce a quantization-aware matrix factorization (QMF) to develop a novel lossy image compression method. QMF provides a low-rank representation of the image data as a product of two smaller matrices, with elements constrained to bounded integer values, thereby effectively integrating quantization with low-rank approximation. We propose an efficient, provably convergent iterative algorithm for QMF using a block coordinate descent scheme, with subproblems having closed-form solutions. Our experiments demonstrate that our method consistently outperforms JPEG at low bit rates below 0.25 bits per pixel. We also demonstrated that our method has an improved capability to preserve visual semantics compared to JPEG at low bit rates by evaluating an ImageNet pre-trained classifier on compressed images. The project is available at https://github.com/pashtari/qmf .

Convolutional Neural Networks (CNNs) with U-shaped architectures have dominated medical image segmentation, which is crucial for various clinical purposes. However, the inherent locality of convolution makes CNNs fail to fully exploit global context, essential for better recognition of some structures, e.g., brain lesions. Transformers have recently proven promising performance on vision tasks, including semantic segmentation, mainly due to their capability of modeling long-range dependencies. Nevertheless, the quadratic complexity of attention makes existing Transformer-based models use self-attention layers only after somehow reducing the image resolution, which limits the ability to capture global contexts present at higher resolutions. Therefore, this work introduces a family of models, dubbed Factorizer, which leverages the power of low-rank matrix factorization for constructing an end-to-end segmentation model. Specifically, we propose a linearly scalable approach to context modeling, formulating Nonnegative Matrix Factorization (NMF) as a differentiable layer integrated into a U-shaped architecture. The shifted window technique is also utilized in combination with NMF to effectively aggregate local information. Factorizers compete favorably with CNNs and Transformers in terms of accuracy, scalability, and interpretability, achieving state-of-the-art results on the BraTS dataset for brain tumor segmentation and ISLES'22 dataset for stroke lesion segmentation. Highly meaningful NMF components give an additional interpretability advantage to Factorizers over CNNs and Transformers. Moreover, our ablation studies reveal a distinctive feature of Factorizers that enables a significant speed-up in inference for a trained Factorizer without any extra steps and without sacrificing much accuracy. The code and models are publicly available at https://github.com/pashtari/factorizer.

In machine learning, kernel data analysis represents a new approach to the study of neurological diseases such as Multiple Sclerosis (MS).In this work, a kernelization technique was combined with a tensor factorization method based on Multilinear Singular Value Decomposition (MLSVD) for MS profile classification.Our simple, yet effective, approach generates a meaningful feature embedding of multi-view data, allowing good classification performance.The results presented in this work define an interesting approach, given that only the anatomical T1-weighted image was used, which represents the most important modality in clinical applications.

Nonnegative matrix factorization (NMF) is a powerful tool for parameter estimation applied in numerous robotics applications, such as path planning, motion trajectory prediction, and motion intention detection. In particular, NMF has been successfully used to extract simplified and organized movement primitives from myoelectric signal (MES) for robust control of multi-degree of freedom humanoid robots. However, MES is typically contaminated by complex noise sources. The system performance often degrades due to the simplified Gaussian assumption of the noise distribution in existing NMF methods. Furthermore, most existing NMF models are unable to automatically determine the rank of the latent matrices. To address these issues, this article presents a hybrid variational Bayesian Gaussian mixture and NMF (GMNMF) model with a finite Gaussian mixture model adopted to fit the mixed noise density function of MES. In addition, the automatic relevant determination criterion is applied to automatically infer the number of movement primitives. The coordinate descent update rules for the proposed model are formulated by mean-field variational Bayesian inference. We assess the model performance on five synthetic noise distribution functions and an experimental MES dataset to perform six wrist movements. The results demonstrate that GMNMF yields low error and high robustness in extracting the movement primitives over four competitive methods for robust cybernetic control.

In this paper, we introduce a new variation of the Convolutional Neural Network Inception block, called Sinc, for sleep stage classification in premature newborn babies using electroencephalogram (EEG). In practice, there are many medical centres where only a limited number of EEG channels are recorded. Existing automated algorithms mainly use multi-channel EEGs which perform poorly when fewer numbers of channels are available. The proposed Sinc utilizes multi-scale analysis to place emphasis on the temporal EEG information to be less dependent on the number of EEG channels. In Sinc, we increase the receptive fields through Inception while by additionally sharing the filters that have similar receptive fields, overfitting is controlled and the number of trainable parameters dramatically reduced. To train and test this model, 96 longitudinal EEG recordings from 26 premature infants are used. The Sinc-based model significantly outperforms state-of-the-art neonatal quiet sleep detection algorithms, with mean Kappa 0.77 ± 0.01 (with 8-channel EEG) and 0.75 ± 0.01 (with a single bipolar channel EEG). This is the first study using Inception-based networks for EEG analysis that utilizes filter sharing to improve efficiency and trainability. The suggested network can successfully detect quiet sleep stages with even a single EEG channel making it more practical especially in the hospital setting where cerebral function monitoring is predominantly used.

Extracting low-rank and/or sparse structures using matrix factorization techniques has been extensively studied in the machine learning community. Kernelized matrix factorization (KMF) is a powerful tool to incorporate side information into the low-rank approximation model, which has been applied to solve the problems of data mining, recommender systems, image restoration, and machine vision. However, most existing KMF models rely on specifying the rows and columns of the data matrix through a Gaussian process prior and have to tune manually the rank. There are also computational issues of existing models based on regularization or the Markov chain Monte Carlo. In this article, we develop a hierarchical kernelized sparse Bayesian matrix factorization (KSBMF) model to integrate side information. The KSBMF automatically infers the parameters and latent variables including the reduced rank using the variational Bayesian inference. In addition, the model simultaneously achieves low-rankness through sparse Bayesian learning and columnwise sparsity through an enforced constraint on latent factor matrices. We further connect the KSBMF with the nonlocal image processing framework to develop two algorithms for image denoising and inpainting. Experimental results demonstrate that KSBMF outperforms the state-of-the-art approaches for these image-restoration tasks under various levels of corruption.

Analysis of complex data is still a challenge in medical image analysis. Due to the heterogeneous information that can be extracted from magnetic resonance imaging (MRI) it can be difficult to fuse such data in a proper way. One interesting case is given by the analysis of diffusion imaging (DI) data. DI techniques give an important variety of information about the status of microstructure in the brain. This is interesting information to use especially in longitudinal setting where the temporal evolution of the pathology is an important added value. In this paper, we propose a new tensor-based framework capable to detect longitudinal changes appearing in DI data in multiple sclerosis (MS) patients. We focus our attention to the analysis of longitudinal changes occurring along different white matter (WM) fiber-bundles. Our main goal is to detect which subset of fibers (within a bundle) and which sections of these fibers contain "pathological" longitudinal changes. The framework consists of three main parts: i) preprocessing of longitudinal diffusion acquisitions and WM fiber-bundles extraction, ii) data tensorization and rank selection, iii) application of a parallelized constrained tensor factorization algorithm to detect longitudinal "pathological" changes. The proposed method was applied on simulated longitudinal variations and on real MS data. High level of accuracy and precision were obtained in the detection of small longitudinal changes along the WM fiber-bundles.

Identifying a core set of features is one of the most important steps in the development of an automated seizure detector. In most of the published studies describing features and seizure classifiers, the features were hand-engineered, which may not be optimal. The main goal of the present paper is using deep convolutional neural networks (CNNs) and random forest to automatically optimize feature selection and classification. The input of the proposed classifier is raw multi-channel EEG and the output is the class label: seizure/nonseizure. By training this network, the required features are optimized, while fitting a nonlinear classifier on the features. After training the network with EEG recordings of 26 neonates, five end layers performing the classification were replaced with a random forest classifier in order to improve the performance. This resulted in a false alarm rate of 0.9 per hour and seizure detection rate of 77% using a test set of EEG recordings of 22 neonates that also included dubious seizures. The newly proposed CNN classifier outperformed three data-driven feature-based approaches and performed similar to a previously developed heuristic method.

Singular value thresholding (SVT)- or nuclear norm minimization (NNM)-based nonlocal image denoising methods often rely on the precise estimation of the noise variance. However, most existing methods either assume that the noise variance is known or require an extra step to estimate it. Under the iterative regularization framework, the error in the noise variance estimate propagates and accumulates with each iteration, ultimately degrading the overall denoising performance. In addition, the essence of these methods is still least squares estimation, which can cause a very high mean-squared error (MSE) and is inadequate for handling missing data or outliers. In order to address these deficiencies, we present a hybrid denoising model based on variational Bayesian inference and Stein's unbiased risk estimator (SURE), which consists of two complementary steps. In the first step, the variational Bayesian SVT performs a low-rank approximation of the nonlocal image patch matrix to simultaneously remove the noise and estimate the noise variance. In the second step, we modify the conventional SURE full-rank SVT and its divergence formulas for rank-reduced eigen-triplets to remove the residual artifacts. The proposed hybrid BSSVT method achieves better performance in recovering the true image compared with state-of-the-art methods.

Nonconvulsive status epilepticus is a condition where the patient is exposed to abnormally prolonged epileptic seizures without evident physical symptoms. Since these continuous seizures may cause permanent brain damage, it constitutes a medical emergency. This paper proposes a method to detect nonconvulsive seizures for a further nonconvulsive status epilepticus diagnosis. To differentiate between the normal and seizure electroencephalogram (EEG), a K-Nearest Neighbor, a Radial Basis Support Vector Machine, and a Linear Discriminant Analysis classifier are used. The classifier features are obtained from the Canonical Polyadic Decomposition (CPD) and Block Term Decomposition of the EEG data represented as third order tensor. To expand the EEG into a tensor, Wavelet or Hilbert-Huang transform are used. The algorithm is tested on a scalp EEG database of 139 seizures of different duration. The experimental results suggest that a Hilbert-Huang tensor representation and the CPD analysis provide the most suitable framework for nonconvulsive seizure detection. The Radial Basis Support Vector Machine classifier shows the best performance with sensitivity, specificity, and accuracy values over 98%. A rough comparison with other methods proposed in the literature shows the superior performance of the proposed method for nonconvulsive epileptic seizure detection.

Objective: This paper presents a methodology to automatically screen for sleep apnea based on the detection of apnea and hypopnea events in the blood oxygen saturation (SpO2) signal. Methods: It starts by detecting all desaturations in the SpO2signal. From these desaturations, a total of 143 time-domain features are extracted. After feature selection, the six most discriminative features are used to construct classifiers to predict if desaturations are caused by respiratory events. From these, a random forest classifier yielded the best classification performance. The number of desaturations, classified as caused by respiratory events per hour of recording, can then be used as an estimate of the apnea-hypopnea index (AHI), and to predict whether or not a patient suffers from sleep apneahypopnea syndrome (SAHS). All classifiers were developed based on a subset of 500 subjects of the Sleep Heart Health Study (SHHS) and tested on three different datasets, containing 8052 subjects in total. Results: An averaged desaturation classification accuracy of 82.8% was achieved over the different test sets. Subjects having SAHS with an AHI greater than 15 can be detected with an average accuracy of 87.6%. Conclusion: The achieved SAHS screening outperforms SpO2methods from the literature on the SHHS test dataset. Moreover, the robustness of the method was shown when tested on different independent test sets. Significance: These results show that an algorithm based on simple features of SpO2desaturations can outperform more elaborate methods in the detection of apneic events and the screening of SAHS patients.

OBJECTIVE: Fragmented QRS (fQRS) is an accessible biomarker and indication of myocardial scarring that can be detected from the electrocardiogram (ECG). Nowadays, fQRS scoring is done on a visual basis, which is time consuming and leads to subjective results. This study proposes an automated method to detect and quantify fQRS in a continuous way using features extracted from variational mode decomposition (VMD) and phase-rectified signal averaging (PRSA). METHODS: In the proposed framework, QRS complexes in the ECG signals were first segmented using VMD. Then, ten VMD- and PRSA-based features were computed and fed into well-known classifiers such as support vector machine (SVM), K-nearest neighbors (KNN), Naive Bayesian (NB), and TreeBagger (TB) in order to compare their performance. The proposed method was evaluated with 12-lead ECG data of 616 patients from the University Hospitals Leuven. The presence of fQRS in each ECG lead was scored by five raters. Both detection and quantification of fQRS could be achieved in this way. RESULTS: The experimental results indicated that the proposed method achieved AUC values of 0.95, 0.94, 0.90, and 0.89 using SVM, KNN, NB, and TB classifiers, respectively, for detecting QRS fragmentation. Assessment of quantification performance was done by comparing the fQRS score with the total score, obtained by summing the scores from the individual raters. Results showed that the fQRS score clearly correlated with this estimate of fQRS certainty. CONCLUSION: The proposed method obtained good results in both fQRS detection and quantification, and is a novel way of assessing the certainty of QRS fragmentation in the ECG signal.

OBJECTIVE: This paper proposes an approach to better estimate the sympathovagal balance (SB) and the respiratory sinus arrhythmia (RSA) after separating respiratory influences from the heart rate (HR). METHODS: The separation is performed using orthogonal subspace projections and the approach is first tested using simulated HR and respiratory signals with different spectral properties. Then, RSA and SB are estimated during autonomic blockade and stress using the proposed approach and the classical heart rate variability (HRV) analysis. Both real- and ECG-derived respiration (EDR) are used and the reliability of the EDR is evaluated. RESULTS: Mean absolute percentage errors lower than [Formula: see text] were obtained after removing previously known respiratory signals from simulated HR. The proposed indices were able to improve the quantification of SB during autonomic withdrawal. In the stress data, differences ( ) among relaxed and stressful phases were found with the proposed approach, using both the real respiration and the EDR, but they disappeared when using the classical HRV. CONCLUSION: A better assessment of the autonomic nervous system' response to pharmacological blockade and stress can be achieved after removing respiratory influences from HR, and this can be done using either the real respiration or the EDR. SIGNIFICANCE: This work can be used to better identify vagal withdrawal and increased sympathetic activation when the classical HRV analysis fails due to the respiratory influences on HR. Furthermore, it can be computed using only the ECG, which is an advantage when developing wearable systems with limited number of sensors.

In neonatal intensive care units, there is a need for around the clock monitoring of electroencephalogram (EEG), especially for recognizing seizures. An automated seizure detector with an acceptable performance can partly fill this need. In order to develop a detector, an extensive dataset labeled by experts is needed. However, accurately defining neonatal seizures on EEG is a challenge, especially when seizure discharges do not meet exact definitions of repetitiveness or evolution in amplitude and frequency. When several readers score seizures independently, disagreement can be high. Commonly used metrics such as good detection rate (GDR) and false alarm rate (FAR) derived from data scored by multiple raters have their limitations. Therefore, new metrics are needed to measure the performance with respect to the different labels. In this paper, instead of defining the labels by consensus or majority voting, popular metrics including GDR, FAR, positive predictive value, sensitivity, specificity, and selectivity are modified such that they can take different scores into account. To this end, 353 hours of EEG data containing seizures from 81 neonates were visually scored by a clinical neurophysiologist, and then processed by an automated seizure detector. The scored seizures were mixed with false detections of an automated seizure detector and were relabeled by three independent EEG readers. Then, all labels were used in the proposed performance metrics and the result was compared with the majority voting technique and showed higher accuracy and robustness for the proposed metrics. Results were confirmed using a bootstrapping test.

Sleep Apnea Hypopnea Syndrome (SAHS) is a sleep disorder where patients experience multiple airflow cessations or reductions during the night. It is recognized as a common condition with a population prevalence of 1% to 6.5%. The Apnea Hypopnea Index (AHI) characterizes the severity of SAHS using signals obtained from Polysomnography (PSG); this requires the use of multiple sensors on the patient and an overnight hospital stay. The development of cheaper and more comfortable screening techniques involving wearable devices is, therefore, desirable. This paper presents a method based on wavelet decomposition and phase space reconstruction with embedding dimensions for feature extraction from oxygen saturation measured in SpO2signals. The proposed characteristics are the areas spanned by each wavelet level in the phase space calculated using the convex hull algorithm. These areas are then fed into a classifier that groups the patients in categories of AHI higher or lower than 5. The results show an accuracy of 93% using K-Nearest Neighbors (Knn), and 88.61% using Least Square Support Vector Machines (LS-SVM).

In Genetic Algorithms, the mutation operator is used to maintain genetic diversity in the population throughout the evolutionary process. Various kinds of mutation may occur over time, typically depending on a fixed probability value called mutation rate. In this work we make use of a novel data-science approach in order to adaptively generate mutation rates for each locus to the Neuroevolution of Augmenting Topologies (NEAT) algorithm. The trail of high quality candidate solutions obtained during the search process is represented as a third-order tensor; factorization of such a tensor reveals the latent relationship between solutions, determining the mutation probability which is likely to yield improvement at each locus. The single pole balancing problem is used as case study to analyze the effectiveness of the proposed approach. Results show that the tensor approach improves the performance of the standard NEAT algorithm for the case study.

In many medical problems, clinicians are suffering from information overload. Nowadays, clinical decision support systems (CDSS) are widely available to alleviate this issue. However, they are often not transparent, in contrast to medical scoring systems originating in the clinical world itself. This work presents an extension of Interval Coded Scoring System (ICS), an approach for semi-automatic data-driven extraction of such scoring systems from clinical data. It focuses on two ICS improvements for large or high-dimensional datasets. Firstly, it offers an alternative elastic net implementation. This can be solved efficiently due to equivalence with Support Vector Machines and the existing efficient solver for its primal formulation. Secondly, an informed preselection of the variables allows to lower ICS' computational burden. ICS is applied to problems as diverse as arrhythmia diagnosis, breast cancer prognosis, functional capacity assessment and diagnosis of spinal diseases and obtains good results. In particular, a comparison on the arrhythmia database shows the importance of using the extensions which yield similar performance as the original ICS while shortening the execution time with a factor of ten.

Automated seizure detection in a home environment has been of increased interest the last couple of decades. The electrocardiogram is one of the signals that is suited for this application. In this paper, a new method is described that classifies different heart rate characteristics in order to detect seizures from temporal lobe epilepsy patients. The used support vector machine classifier is trained on data from other patients, so that the algorithm can be used directly from the start of each new recording. The algorithm was tested on a dataset of more than 918[Formula: see text]h of data coming from 17 patients containing 127 complex partial and generalized partial seizures. The algorithm was able to detect 81.89% of the seizures, with on average 1.97 false alarms per hour. These results show a strong drop in the number of false alarms of more than 50% compared to other heart rate-based patient-independent algorithms from the literature, at the expense of a slightly higher detection delay of 17.8s on average.

Sleep state development in preterm neonates can provide crucial information regarding functional brain maturation and give insight into neurological well being. However, visual labeling of sleep stages from EEG requires expertise and is very time consuming, prompting the need for an automated procedure. We present a robust method for automated detection of preterm sleep from EEG, over a wide postmenstrual age ([Formula: see text] age) range, focusing first on Quiet Sleep (QS) as an initial marker for sleep assessment. Our algorithm, CLuster-based Adaptive Sleep Staging (CLASS), detects QS if it remains relatively more discontinuous than non-QS over PMA. CLASS was optimized on a training set of 34 recordings aged 27-42 weeks PMA, and performance then assessed on a distinct test set of 55 recordings of the same age range. Results were compared to visual QS labeling from two independent raters (with inter-rater agreement [Formula: see text]), using Sensitivity, Specificity, Detection Factor ([Formula: see text] of visual QS periods correctly detected by CLASS) and Misclassification Factor ([Formula: see text] of CLASS-detected QS periods that are misclassified). CLASS performance proved optimal across recordings at 31-38 weeks (median [Formula: see text], median MF 0-0.25, median Sensitivity 0.93-1.0, and median Specificity 0.80-0.91 across this age range), with minimal misclassifications at 35-36 weeks (median [Formula: see text]). To illustrate the potential of CLASS in facilitating clinical research, normal maturational trends over PMA were derived from CLASS-estimated QS periods, visual QS estimates, and nonstate specific periods (containing QS and non-QS) in the EEG recording. CLASS QS trends agreed with those from visual QS, with both showing stronger correlations than nonstate specific trends. This highlights the benefit of automated QS detection for exploring brain maturation.

Magnetic resonance spectroscopic imaging (MRSI) reveals chemical information that characterizes different tissue types in brain tumors. Blind source separation techniques are used to extract the tissue-specific profiles and their corresponding distribution from the MRSI data. We focus on automatic detection of the tumor, necrotic and normal brain tissue types by constructing a 3D MRSI tensor from in vivo 2D-MRSI data of individual glioma patients. Nonnegative canonical polyadic decomposition (NCPD) is applied to the MRSI tensor to differentiate various tissue types. An in vivo study shows that NCPD has better performance in identifying tumor and necrotic tissue type in glioma patients compared to previous matrix-based decompositions, such as nonnegative matrix factorization and hierarchical nonnegative matrix factorization.

Processing of longitudinal diffusion tensor imaging (DTI) data is a crucial challenge to better understand pathological mechanisms of complex brain diseases such as multiple sclerosis (MS) where white-matter (WM) fiber bundles are variably altered by inflammatory events. In this study, we propose a new fully automated method to detect longitudinal changes in diffusivity metrics along WM fiber bundles. The proposed method is divided in three main parts: 1) preprocessing of longitudinal diffusion acquisitions, 2) WM fiber-bundle extraction, and 3) application of nonnegative matrix factorization and density-based local outliers algorithms to detect and delineate longitudinal variations appearing in the cross section of the WM fiber bundle. In order to validate our method, we introduce a new model to simulate real longitudinal changes based on a generalized Gaussian probability density function. Moreover, we applied our method on longitudinal data. High level of performances were obtained for the detection of small longitudinal changes along the WM fiber bundles in MS patients.

© Copyright 2016 by SCITEPRESS - Science and Technology Publications, Lda. All rights reserved. Scoring systems have been used since long in medical practice, but often they are based on experience rather than a structural approach. In literature, the interval coded scoring index (ICS) has been introduced as an alternative. It derives a scoring system from data using optimization techniques. This work discusses an extension, ICS∗, that takes variable interactions into account. Furthermore, a study is performed to give insight into the new model's sensitivity to noise, the size of the data set and the number of non-informative variables. The study shows interactions can mostly be discovered robustly, even in the presence of noise and spurious variables. A final validation on two UCI data sets further indicates the quality of the approach.

Epileptic seizure detection is traditionally done using video/electroencephalography monitoring, which is not applicable for long-term home monitoring. In recent years, attempts have been made to detect the seizures using other modalities. In this study, we investigated the application of four accelerometers (ACM) attached to the limbs and surface electromyography (sEMG) electrodes attached to upper arms for the detection of tonic-clonic seizures. sEMG can identify the tension during the tonic phase of tonic-clonic seizure, while ACM is able to detect rhythmic patterns of the clonic phase of tonic-clonic seizures. Machine learning techniques, including feature selection and least-squares support vector machine classification, were employed for detection of tonic-clonic seizures from ACM and sEMG signals. In addition, the outputs of ACM and sEMG-based classifiers were combined using a late integration approach. The algorithms were evaluated on 1998.3 h of data recorded nocturnally in 56 patients of which seven had 22 tonic-clonic seizures. A multimodal approach resulted in a more robust detection of short and nonstereotypical seizures (91%), while the number of false alarms increased significantly compared with the use of single sEMG modality (0.28-0.5/12h). This study also showed that the choice of the recording system should be made depending on the prevailing pediatric patient-specific seizure characteristics and nonepileptic behavior.

The automated analysis of the EEG pattern of the preterm newborn would be a valuable tool in the neonatal intensive care units for the prognosis of neurological development. The analysis of the (a)symmetry between the two hemispheres can provide useful information about neuronal dysfunction in early stages. Consecutive and subgroup analyses of different brain regions will allow to detect physiologic asymmetry versus pathologic asymmetry. This can improve the assessment of the long-term neurodevelopmental outcome. We show that pathological asymmetry can be measured and detected using the channel symmetry index, which comprises the difference in power spectral density of contralateral EEG signals. To distinguish pathological from physiological normal EEG patterns, we make use of one-class SVM classifiers. Future work will focus on adding relevant features for classification and augmenting the sample size, thus reducing the overall classification error.

Nocturnal home monitoring of epileptic children is often not feasible due to the cumbersome manner of seizure monitoring with the standard method of video/EEG-monitoring. We propose a method for hypermotor seizure detection based on accelerometers attached to the extremities. From the acceleration signals, multiple temporal, frequency, and wavelet-based features are extracted. After determining the features with the highest discriminative power, we classify movement events in epileptic and nonepileptic movements. This classification is only based on a nonparametric estimate of the probability density function of normal movements. Such approach allows us to build patient-specific models to classify movement data without the need for seizure data that are rarely available. If, in the test phase, the probability of a data point (event) is lower than a threshold, this event is considered to be an epileptic seizure; otherwise, it is considered as a normal nocturnal movement event. The mean performance over seven patients gives a sensitivity of 95.24% and a positive predictive value of 60.04%. However, there is a noticeable interpatient difference.

Polysomnography (PSG) is considered the gold standard to assess sleep accurately, but it can be expensive, time-consuming, and uncomfortable, specifically in long-term sleep studies. Actigraphy, on the other hand, is both cheap and userfriendly, but depending on the application lacks detail and accuracy. Our aim was to evaluate cardiorespiratory and movement signals in discriminating between wake, rapid-eye-movement (REM), light (N1N2), and deep (N3) sleep. The dataset comprised 85 nights of PSG from a healthy population. Starting from a total of 750 characteristic variables (features), problem-specific subsets of 40 features were forwardly selected using the combination of a wrapper method (Cohen's kappa statistic on radial basis function (RBF)-kernel support vector machine (SVM) classifier) and filter method (minimum redundancy maximum relevance criterion on mutual information). Final classification was performed using an RBF-kernel SVM. Non-subject-specific wake versus sleep classification resulted in a Cohen’s kappa value of 0.695, while REM versus NREM resulted in 0.558 and N3 versus N1N2 in 0.553. The broad pool of initial features gave insight in which features discriminated best between the different classes. The classification results demonstrate the possibility of making long-term sleep monitoring more widely available.

Epileptic seizure detection is traditionally done using video/electroencephalogram (EEG) monitoring, which is not applicable in a home situation. In recent years, attempts have been made to detect the seizures using other modalities. In this paper we investigate if a combined usage of accelerometers attached to the limbs and video data would increase the performance compared to a single modality approach. Therefore, we used two existing approaches for seizure detection in accelerometers and video and combined them using a linear discriminant analysis (LDA) classifier. The results for a combined detection have a better positive predictive value (PPV) of 95.00% compared to the single modality detection and reached a sensitivity of 83.33%.

MOTIVATION: New application areas of survival analysis as for example based on micro-array expression data call for novel tools able to handle high-dimensional data. While classical (semi-) parametric techniques as based on likelihood or partial likelihood functions are omnipresent in clinical studies, they are often inadequate for modelling in case when there are less observations than features in the data. Support vector machines (svms) and extensions are in general found particularly useful for such cases, both conceptually (non-parametric approach), computationally (boiling down to a convex program which can be solved efficiently), theoretically (for its intrinsic relation with learning theory) as well as empirically. This article discusses such an extension of svms which is tuned towards survival data. A particularly useful feature is that this method can incorporate such additional structure as additive models, positivity constraints of the parameters or regression constraints. RESULTS: Besides discussion of the proposed methods, an empirical case study is conducted on both clinical as well as micro-array gene expression data in the context of cancer studies. Results are expressed based on the logrank statistic, concordance index and the hazard ratio. The reported performances indicate that the present method yields better models for high-dimensional data, while it gives results which are comparable to what classical techniques based on a proportional hazard model give for clinical data.

Monitoring of epileptic patients is usually done by video/EEG-monitoring which is considered as the golden standard. Due to some disadvantages of this method, this method is not feasible to use in long term home monitoring. Video monitoring provides a solution to this problem as it can monitor the patient in a non-contacting way. An algorithm is developed to detect movement epochs in nocturnal datasets for pediatric epileptic patients. The performance was measured using a threefold crossvaildation, which resulted in a sensitivity of 1 and a positive predictive value above 0.85.

Brain metastases and glioblastoma multiforme are the most aggressive and common brain tumours in adults and they require a different clinical management. Anatomical magnetic resonance imaging (MRI) or clinical history, cannot always clearly distinguish between them. This study describes and verifies the use of magnetic resonance spectroscopy (MRS) and magnetic resonance spectroscopic imaging (MRSI) in combination with MRI for differential diagnosis of glioblastomas and metastases. Feature selection methods are applied to the magnetic resonance (MR) spectra of 121 patients and relevant features are detected. Different classification methods are used to distinguish glioblastoma multiforme and metastasis based on the single-voxel MR spectra, but no reliable differentiation is obtained: the accuracy varies from 50 to 78%. Next, MRSI and MRI data from 10 patients (5 glioblastomas, 5 solitary metastases) are used for differentiation purposes. The combination of multivoxel MR data and MRI data suggests a more clear differentiation between glioblastoma multiforme and brain metastasis. The results are visualized based on nosologic images, which are generated by including spectroscopic information in the segmented MR image. The methodology offers a new way that may support clinicians in decision making.

Newton's method for solving the matrix equation F(X) identical to AX-XX(T) AX = 0 runs up against the fact that its zeros are not isolated. This is due to a symmetry of F by the action of the orthogonal group. We show how differential-geometric techniques can be exploited to remove this symmetry and obtain a "geometric" Newton algorithm that finds the zeros of F. The geometric Newton method does not suffer from the degeneracy issue that stands in the way of the original Newton method.

This paper presents our activities towards the development of a clinical decision support (CDS) system for the early diagnosis of pregnancies of unknown location (PULs). PULs are pregnancies that are not visualized on ultrasound. Generally, they turn out to be failing, normal intra-uterine, or ectopic pregnancies. Using 856 PULs collected at a London hospital, we developed probabilistic multi-class models. At the moment, a large multi-center data set is being collected. Data are entered using stand-alone software, and are sent to the database server by ftp. These data will be used to prospectively test existing models, and to build new, hopefully more robust, models. The ultimate aim is to select one model for implementation in a CDS system that is easy to use in routine clinical practice.

The area under the receiver operating characteristic curve (AUC) is a useful and widely used measure to evaluate the performance of binary and multi-class classification models. However, it does not take into account the exact numerical output of the models, but rather looks at how the output ranks the cases. AUC metrics that incorporate the exact numerical output have been developed for binary classification. In this paper, we try to extend such weighted metrics to the multi-class case. Several metrics are suggested. Using real world data, we investigate intercorrelations between these metrics and demonstrate their use.

We propose a new algorithm to impose independence constraints based on second order statistics in one mode of the Parallel Factor Analysis / Canonical Decomposition, also known as the CP model, and show with simulations that it outperforms in some cases the ordinary CP model.

This paper presents a new subspace-based technique for automatic detection of the number of exponentially damped sinusoids. It consists in studying the shift-invariance of the dominant subspace of the Hankel data matrix. No threshold setting and no penalization terms are necessary. This model-based method, easy to implement, can be plugged into most subspace-based harmonic retrieval algorithms.

This paper investigates variable selection (VS) and classification for biomedical datasets with a small sample size and a very high input dimension. The sequential sparse Bayesian learning methods with linear bases are used as the basic VS algorithm. Selected variables are fed to the kernel-based probabilistic classifiers: Bayesian least squares support vector machines (BayLS-SVMs) and relevance vector machines (RVMs). We employ the bagging techniques for both VS and model building in order to improve the reliability of the selected variables and the predictive performance. This modeling strategy is applied to real-life medical classification problems, including two binary cancer diagnosis problems based on microarray data and a brain tumor multiclass classification problem using spectra acquired via magnetic resonance spectroscopy. The work is experimentally compared to other VS methods. It is shown that the use of bagging can improve the reliability and stability of both VS and model prediction.

The Generalized Mapping Regressor (GMR) neural network is able to solve for inverse problems even when multiple solutions are given. In this case, it does not only identify these solutions (even if infinite, e.g. contours), but also specifies to which branch of the underlying mapping it belongs. It is also able to model mapping with discontinuities. The basic idea is the transformation of the mapping problem in a pattern recognition problem in a higher dimensional space (where the function branches are represented by clusters). Training is given by a multiresolution quantization represented by a pool of neurons whose number is determined by the training set. Then, neurons are linked each other by using some kind of local principal component analysis (LPCA). This phase is the most important and original. Other techniques (e.g. SVM's, mixture-of-experts) could work a priori on the same problems, but are not able to understand automatically when to stop the data quantization. This linking phase can be viewed as a reconstruction phase in which the correct clusters are recovered. The production phase uses a Gaussian kernel interpolation technique. Some examples conclude the paper.

The minor component analysis (MCA) deals with the recovery of the eigenvector associated to the smallest eigenvalue of the autocorrelation matrix of the input data and is a very important tool for signal processing and data analysis. It is almost exclusively solved by linear neurons. This paper presents a linear neuron endowed with a novel learning law, called MCA EXINn and analyzes its features. The neural literature about MCA is very poor, in the sense that both a little theoretical basis is given (almost always focusing on the ODE asymptotic approximation) and only experiments on toy problems (at most four-dimensional problems) are presented, without any numerical analysis. This work addresses these problems and lays sound theoretical foundations for the neural MCA theory. In particular, it classifies the MCA neurons according to the Riemannian metric and justifies, from the analysis of the degeneracy of the error cost; the different behavior in approaching convergence. The cost landscape is studied and used as a basis for the analysis of the asymptotic behavior. All the phases of the dynamics of the MCA algorithms are investigated in detail and, together with the numerical analysis, lead to the identification of three possible kinds of divergence, here called sudden, dynamic, and numerical. The importance of the choice of low initial conditions is also explained. A lot of importance is given to the experimental part, where simulations on high-dimensional problems are,presented and analyzed. The orthogonal regression or total least squares (TLS) technique is also presented, together with a real-world application on the identification of the parameters of an electrical machine. It can be concluded that MCA EXIN is the best MCA neuron in terms of stability (no finite time divergence), speed, and accuracy.

The pre-operative discrimination between malignant and benign masses is a crucial issue in gynaecology. Next to the large amount of background knowledge, there is a growing amount of collected patient data that can be used in inductive techniques. These two sources of information result in two different modelling strategies. Based on the background knowledge, various discrimination models have been constructed by leading experts in the field, tuned and tested by observations. Based on the patient observations, various statistical models have been developed, such as logistic regression models and artificial neural network models. For the efficient combination of prior background knowledge and observations, Bayesian network models are suggested. We summarize the applicability of this technique, report the performance of such models in ovarian cancer diagnosis and outline a possible hybrid usage of this technique.

Linear regression problems deal with the solution of an overdetermined set of linear equations under different assumptions about noise in the experimental data. Some parameterized techniques, designed for a compact treatment of these problems, are briefly reviewed. Then, the GeTLS method is introduced and applied, as learning law, to a novel linear neuron, GeTLS EXIN. Some numerical considerations follow, together with software examples. This neuron yields very good results, especially for large systems of equations.

We present a new speech coding algorithm, based on an all-pole model of the vocal tract. Whereas current autoregressive (AR) based modeling techniques (e.g. CELP, LPC-10) minimize a prediction error, which is considered to be the input to the all-pole model, our approach determines the closest (in L/sub 2/ norm) signal, which exactly satisfies an all-pole model. Each frame is then encoded by storing the parameters of the complex damped exponentials deduced from the all-pole model and its initial conditions. Decoding is performed by adding the complex damped exponentials based on the transmitted parameters. The new algorithm is demonstrated on a speech signal. The quality is compared with that of a standard coding algorithm at comparable compression ratios, by using the segmental signal-to-noise ratio (SNR).

In parameter estimation of closely-spaced sinusoids, prior knowledge of some known signal poles can be incorporated as done in the LCTLS-LP algorithm proposed by Dowling et al. (1992, 1994) and based on forward linear prediction and total least squares. In this paper, two better algorithms, HTLS-PK and HTLN-PK, derived respectively from Kung et al.'s (1983) state-space method using the TLS principle and the newly developed structured total least norm technique are presented that incorporate the same prior knowledge but clearly outperform the LCTLS-LP method in both resolution and parameter accuracy.

A new fast algorithm is presented for quantifying signals modeled by a sum of exponentially damped sinusoids. The new algorithm is applied to the quantitative analysis of Nuclear Magnetic Resonance (NMR) data and is shown to be up to an order of magnitude more efficient than currently used linear prediction and state-space based methods.>