Adaptive relevance matrices in learning vector quantization

Subspace corrected relevance learning with application in neuroimaging

In machine learning, data often comes from different sources, but combining them can introduce extraneous variation that affects both generalization and interpretability. For example, we investigate the classification of neurodegenerative diseases using FDG-PET data collected from multiple neuroimaging centers. However, data collected at different centers introduces unwanted variation due to differences in scanners, scanning protocols, and processing methods. To address this issue, we propose a two-step approach to limit the influence of center-dependent variation on the classification of healthy controls and early vs. late-stage Parkinson's disease patients. First, we train a Generalized Matrix Learning Vector Quantization (GMLVQ) model on healthy control data to identify a "relevance space" that distinguishes between centers. Second, we use this space to construct a correction matrix that restricts a second GMLVQ system's training on the diagnostic problem. We evaluate the effectiveness of this approach on the real-world multi-center datasets and simulated artificial dataset. Our results demonstrate that the approach produces machine learning systems with reduced bias - being more specific due to eliminating information related to center differences during the training process - and more informative relevance profiles that can be interpreted by medical experts. This method can be adapted to similar problems outside the neuroimaging domain, as long as an appropriate "relevance space" can be identified to construct the correction matrix.

Urine steroid metabolomics as a diagnostic tool in primary aldosteronism

Primary aldosteronism (PA) causes 5-10% of hypertension cases, but only a minority of patients are currently diagnosed and treated because of a complex, stepwise, and partly invasive workup. We tested the performance of urine steroid metabolomics, the computational analysis of 24-hour urine steroid metabolome data by machine learning, for the identification and subtyping of PA. Mass spectrometry-based multi-steroid profiling was used to quantify the excretion of 34 steroid metabolites in 24-hour urine samples from 158 adults with PA (88 with unilateral PA [UPA] due to aldosterone-producing adenomas [APAs]; 70 with bilateral PA [BPA]) and 65 sex- and age-matched healthy controls. All APAs were resected and underwent targeted gene sequencing to detect somatic mutations associated with UPA. Patients with PA had increased urinary metabolite excretion of mineralocorticoids, glucocorticoids, and glucocorticoid precursors. Urine steroid metabolomics identified patients with PA with high accuracy, both when applied to all 34 or only the three most discriminative steroid metabolites (average areas under the receiver-operating characteristics curve [AUCs-ROC] 0.95-0.97). Whilst machine learning was suboptimal in differentiating UPA from BPA (average AUCs-ROC 0.65-0.73), it readily identified APA cases harbouring somatic KCNJ5 mutations (average AUCs-ROC 0.79-85). These patients showed a distinctly increased urine excretion of the hybrid steroid 18-hydroxycortisol and its metabolite 18-oxo-tetrahydrocortisol, the latter identified by machine learning as by far the most discriminative steroid. In conclusion, urine steroid metabolomics is a non-invasive candidate test for the accurate identification of PA cases and KCNJ5-mutated APAs.

Predicting incident dementia in cerebral small vessel disease: comparison of machine learning and traditional statistical models

Background

Cerebral small vessel disease (SVD) contributes to 45% of dementia cases worldwide, yet we lack a reliable model for predicting dementia in SVD. Past attempts largely relied on traditional statistical approaches. Here, we investigated whether machine learning (ML) methods improved prediction of incident dementia in SVD from baseline SVD-related features over traditional statistical methods.

Methods

We included three cohorts with varying SVD severity (RUN DMC, n = 503; SCANS, n = 121; HARMONISATION, n = 265). Baseline demographics, vascular risk factors, cognitive scores, and magnetic resonance imaging (MRI) features of SVD were used for prediction. We conducted both survival analysis and classification analysis predicting 3-year dementia risk. For each analysis, several ML methods were evaluated against standard Cox or logistic regression. Finally, we compared the feature importance ranked by different models.

Results

We included 789 participants without missing data in the survival analysis, amongst whom 108 (13.7%) developed dementia during a median follow-up of 5.4 years. Excluding those censored before three years, we included 750 participants in the classification analysis, amongst whom 48 (6.4%) developed dementia by year 3. Comparing statistical and ML models, only regularised Cox/logistic regression outperformed their statistical counterparts overall, but not significantly so in survival analysis. Baseline cognition was highly predictive, and global cognition was the most important feature.

Conclusions

When using baseline SVD-related features to predict dementia in SVD, the ML survival or classification models we evaluated brought little improvement over traditional statistical approaches. The benefits of ML should be evaluated with caution, especially given limited sample size and features.

Machine learning classification of conduct disorder with high versus low levels of callous-unemotional traits based on facial emotion recognition abilities

Conduct disorder (CD) with high levels of callous-unemotional traits (CD/HCU) has been theoretically linked to specific difficulties with fear and sadness recognition, in contrast to CD with low levels of callous-unemotional traits (CD/LCU). However, experimental evidence for this distinction is mixed, and it is unclear whether these difficulties are a reliable marker of CD/HCU compared to CD/LCU. In a large sample (N = 1263, 9-18 years), we combined univariate analyses and machine learning classifiers to investigate whether CD/HCU is associated with disproportionate difficulties with fear and sadness recognition over other emotions, and whether such difficulties are a reliable individual-level marker of CD/HCU. We observed similar emotion recognition abilities in CD/HCU and CD/LCU. The CD/HCU group underperformed relative to typically developing (TD) youths, but difficulties were not specific to fear or sadness. Classifiers did not distinguish between youths with CD/HCU versus CD/LCU (52% accuracy), although youths with CD/HCU and CD/LCU were reliably distinguished from TD youths (64% and 60%, respectively). In the subset of classifiers that performed well for youths with CD/HCU, fear and sadness were the most relevant emotions for distinguishing them from youths with CD/LCU and TD youths, respectively. We conclude that non-specific emotion recognition difficulties are common in CD/HCU, but are not reliable individual-level markers of CD/HCU versus CD/LCU. These findings highlight that a reduced ability to recognise facial expressions of distress should not be assumed to be a core feature of CD/HCU.

Quantum Computing Approaches for Vector Quantization-Current Perspectives and Developments

In the field of machine learning, vector quantization is a category of low-complexity approaches that are nonetheless powerful for data representation and clustering or classification tasks. Vector quantization is based on the idea of representing a data or a class distribution using a small set of prototypes, and hence, it belongs to interpretable models in machine learning. Further, the low complexity of vector quantizers makes them interesting for the application of quantum concepts for their implementation. This is especially true for current and upcoming generations of quantum devices, which only allow the execution of simple and restricted algorithms. Motivated by different adaptation and optimization paradigms for vector quantizers, we provide an overview of respective existing quantum algorithms and routines to realize vector quantization concepts, maybe only partially, on quantum devices. Thus, the reader can infer the current state-of-the-art when considering quantum computing approaches for vector quantization.

An Enhanced Hyper-Parameter Optimization of a Convolutional Neural Network Model for Leukemia Cancer Diagnosis in a Smart Healthcare System

Healthcare systems in recent times have witnessed timely diagnoses with a high level of accuracy. Internet of Medical Things (IoMT)-enabled deep learning (DL) models have been used to support medical diagnostics in real time, thus resolving the issue of late-stage diagnosis of various diseases and increasing performance accuracy. The current approach for the diagnosis of leukemia uses traditional procedures, and in most cases, fails in the initial period. Hence, several patients suffering from cancer have died prematurely due to the late discovery of cancerous cells in blood tissue. Therefore, this study proposes an IoMT-enabled convolutional neural network (CNN) model to detect malignant and benign cancer cells in the patient's blood tissue. In particular, the hyper-parameter optimization through radial basis function and dynamic coordinate search (HORD) optimization algorithm was used to search for optimal values of CNN hyper-parameters. Utilizing the HORD algorithm significantly increased the effectiveness of finding the best solution for the CNN model by searching multidimensional hyper-parameters. This implies that the HORD method successfully found the values of hyper-parameters for precise leukemia features. Additionally, the HORD method increased the performance of the model by optimizing and searching for the best set of hyper-parameters for the CNN model. Leukemia datasets were used to evaluate the performance of the proposed model using standard performance indicators. The proposed model revealed significant classification accuracy compared to other state-of-the-art models.

FDG-PET combined with learning vector quantization allows classification of neurodegenerative diseases and reveals the trajectory of idiopathic REM sleep behavior disorder

BACKGROUND AND OBJECTIVES

¹⁸F-fluorodeoxyglucose (FDG) positron emission tomography (PET) combined with principal component analysis (PCA) has been applied to identify disease-related brain patterns in neurodegenerative disorders such as Parkinson's disease (PD), Dementia with Lewy Bodies (DLB) and Alzheimer's disease (AD). These patterns are used to quantify functional brain changes at the single subject level. This is especially relevant in determining disease progression in idiopathic REM sleep behavior disorder (iRBD), a prodromal stage of PD and DLB. However, the PCA method is limited in discriminating between neurodegenerative conditions. More advanced machine learning algorithms may provide a solution. In this study, we apply Generalized Matrix Learning Vector Quantization (GMLVQ) to FDG-PET scans of healthy controls, and patients with AD, PD and DLB. Scans of iRBD patients, scanned twice with an approximate 4 year interval, were projected into GMLVQ space to visualize their trajectory.

METHODS

We applied a combination of SSM/PCA and GMLVQ as a classifier on FDG-PET data of healthy controls, AD, DLB, and PD patients. We determined the diagnostic performance by performing a ten times repeated ten fold cross validation. We analyzed the validity of the classification system by inspecting the GMLVQ space. First by the projection of the patients into this space. Second by representing the axis, that span this decision space, into a voxel map. Furthermore, we projected a cohort of RBD patients, whom have been scanned twice (approximately 4 years apart), into the same decision space and visualized their trajectories.

RESULTS

The GMLVQ prototypes, relevance diagonal, and decision space voxel maps showed metabolic patterns that agree with previously identified disease-related brain patterns. The GMLVQ decision space showed a plausible quantification of FDG-PET data. Distance traveled by iRBD subjects through GMLVQ space per year (i.e. velocity) was correlated with the change in motor symptoms per year (Spearman's rho =0.62, P=0.004).

CONCLUSION

In this proof-of-concept study, we show that GMLVQ provides a classification of patients with neurodegenerative disorders, and may be useful in future studies investigating speed of progression in prodromal disease stages.

A multi-birth metric learning framework based on binary constraints

Multi-metric learning plays a significant role in improving the generalization of algorithms related to distance metrics since using a single metric is sometimes insufficient to handle complex data. Metric learning can adjust automatically the distance between samples to make the intra-class samples compact while making the inter-class distance as far as possible. To implement this intention better,in this work, we propose a novel multi-metric learning framework based on the pair constraints instead of triple constraints to reduce computational burden. To solve effectively the problem, we first propose a multi-birth metric learning model (termed MBML), where for each class sample, the global metric and a local metric are jointly trained. Both global and local structural information are adapted to better depict sample information. Then two alternating iterative algorithms are developed to optimize the MBML. The convergence of the proposed algorithm and complexity are analyzed theoretically. Moreover, a fast diagonal multi-metric learning method is proposed based on binary constraints, and problem can be reformulated a linear programming, with fast training speed, low the computational burden and the global optimal solutions. Numerical experiments are carried out on different scales and different types of datasets including an artificial data, benchmark datasets and an image database from binary class and multi-class problems. Experiment results confirm the feasibility and effectiveness of the proposed methods.

A robust and interpretable machine learning approach using multimodal biological data to predict future pathological tau accumulation

The early stages of Alzheimer’s disease (AD) involve interactions between multiple pathophysiological processes. Although these processes are well studied, we still lack robust tools to predict individualised trajectories of disease progression. Here, we employ a robust and interpretable machine learning approach to combine multimodal biological data and predict future pathological tau accumulation. In particular, we use machine learning to quantify interactions between key pathological markers (β-amyloid, medial temporal lobe  atrophy, tau and APOE 4) at mildly impaired and asymptomatic stages of AD. Using baseline non-tau markers we derive a prognostic index that: (a) stratifies patients based on future pathological tau accumulation, (b) predicts individualised regional future rate of tau accumulation, and (c) translates predictions from deep phenotyping patient cohorts to cognitively normal individuals. Our results propose a robust approach for fine scale stratification and prognostication with translation impact for clinical trial design targeting the earliest stages of AD.

The authors present a machine learning approach that combines baseline multimodal data to accurately predict individualised trajectories of future pathological tau accumulation at asymptomatic and mildly impaired stages of Alzheimer’s disease.

Quantum-inspired learning vector quantizers for prototype-based classification

Prototype-based models like the Generalized Learning Vector Quantization (GLVQ) belong to the class of interpretable classifiers. Moreover, quantum-inspired methods get more and more into focus in machine learning due to its potential efficient computing. Further, its interesting mathematical perspectives offer new ideas for alternative learning scenarios. This paper proposes a quantum computing-inspired variant of the prototype-based GLVQ for classification learning. We start considering kernelized GLVQ with real- and complex-valued kernels and their respective feature mapping. Thereafter, we explain how quantum space ideas could be integrated into a GLVQ using quantum bit vector space in the quantum state space $${\mathcal {H}}^{n}$$ H n and show the relations to kernelized GLVQ. In particular, we explain the related feature mapping of data into the quantum state space $${\mathcal {H}}^{n}$$ H n . A key feature for this approach is that $${\mathcal {H}}^{n}$$ H n is an Hilbert space with particular inner product properties, which finally restrict the prototype adaptations to be unitary transformations. The resulting approach is denoted as Qu-GLVQ. We provide the mathematical framework and give exemplary numerical results.

The Resolved Mutual Information Function as a Structural Fingerprint of Biomolecular Sequences for Interpretable Machine Learning Classifiers

In the present article we propose the application of variants of the mutual information function as characteristic fingerprints of biomolecular sequences for classification analysis. In particular, we consider the resolved mutual information functions based on Shannon-, Rényi-, and Tsallis-entropy. In combination with interpretable machine learning classifier models based on generalized learning vector quantization, a powerful methodology for sequence classification is achieved which allows substantial knowledge extraction in addition to the high classification ability due to the model-inherent robustness. Any potential (slightly) inferior performance of the used classifier is compensated by the additional knowledge provided by interpretable models. This knowledge may assist the user in the analysis and understanding of the used data and considered task. After theoretical justification of the concepts, we demonstrate the approach for various example data sets covering different areas in biomolecular sequence analysis.

Next move in movement disorders (NEMO): developing a computer-aided classification tool for hyperkinetic movement disorders

INTRODUCTION

Our aim is to develop a novel approach to hyperkinetic movement disorder classification, that combines clinical information, electromyography, accelerometry and video in a computer-aided classification tool. We see this as the next step towards rapid and accurate phenotype classification, the cornerstone of both the diagnostic and treatment process.

METHODS AND ANALYSIS

The Next Move in Movement Disorders (NEMO) study is a cross-sectional study at Expertise Centre Movement Disorders Groningen, University Medical Centre Groningen. It comprises patients with single and mixed phenotype movement disorders. Single phenotype groups will first include dystonia, myoclonus and tremor, and then chorea, tics, ataxia and spasticity. Mixed phenotypes are myoclonus-dystonia, dystonic tremor, myoclonus ataxia and jerky/tremulous functional movement disorders. Groups will contain 20 patients, or 40 healthy participants. The gold standard for inclusion consists of interobserver agreement on the phenotype among three independent clinical experts. Electromyography, accelerometry and three-dimensional video data will be recorded during performance of a set of movement tasks, chosen by a team of specialists to elicit movement disorders. These data will serve as input for the machine learning algorithm. Labels for supervised learning are provided by the expert-based classification, allowing the algorithm to learn to predict what the output label should be when given new input data. Methods using manually engineered features based on existing clinical knowledge will be used, as well as deep learning methods which can detect relevant and possibly new features. Finally, we will employ visual analytics to visualise how the classification algorithm arrives at its decision.

ETHICS AND DISSEMINATION

Ethical approval has been obtained from the relevant local ethics committee. The NEMO study is designed to pioneer the application of machine learning of movement disorders. We expect to publish articles in multiple related fields of research and patients will be informed of important results via patient associations and press releases.

AI-Based Multi Sensor Fusion for Smart Decision Making: A Bi-Functional System for Single Sensor Evaluation in a Classification Task

Sensor fusion has gained a great deal of attention in recent years. It is used as an application tool in many different fields, especially the semiconductor, automotive, and medical industries. However, this field of research, regardless of the field of application, still presents different challenges concerning the choice of the sensors to be combined and the fusion architecture to be developed. To decrease application costs and engineering efforts, it is very important to analyze the sensors' data beforehand once the application target is defined. This pre-analysis is a basic step to establish a working environment with fewer misclassification cases and high safety. One promising approach to do so is to analyze the system using deep neural networks. The disadvantages of this approach are mainly the required huge storage capacity, the big training effort, and that these networks are difficult to interpret. In this paper, we focus on developing a smart and interpretable bi-functional artificial intelligence (AI) system, which has to discriminate the combined data regarding predefined classes. Furthermore, the system can evaluate the single source signals used in the classification task. The evaluation here covers each sensor contribution and robustness. More precisely, we train a smart and interpretable prototype-based neural network, which learns automatically to weight the influence of the sensors for the classification decision. Moreover, the prototype-based classifier is equipped with a reject option to measure classification certainty. To validate our approach's efficiency, we refer to different industrial sensor fusion applications.

Probabilistic learning vector quantization on manifold of symmetric positive definite matrices

In this paper, we develop a new classification method for manifold-valued data in the framework of probabilistic learning vector quantization. In many classification scenarios, the data can be naturally represented by symmetric positive definite matrices, which are inherently points that live on a curved Riemannian manifold. Due to the non-Euclidean geometry of Riemannian manifolds, traditional Euclidean machine learning algorithms yield poor results on such data. In this paper, we generalize the probabilistic learning vector quantization algorithm for data points living on the manifold of symmetric positive definite matrices equipped with Riemannian natural metric (affine-invariant metric). By exploiting the induced Riemannian distance, we derive the probabilistic learning Riemannian space quantization algorithm, obtaining the learning rule through Riemannian gradient descent. Empirical investigations on synthetic data, image data , and motor imagery electroencephalogram (EEG) data demonstrate the superior performance of the proposed method.

Learning vector quantization as an interpretable classifier for the detection of SARS-CoV-2 types based on their RNA sequences

We present an approach to discriminate SARS-CoV-2 virus types based on their RNA sequence descriptions avoiding a sequence alignment. For that purpose, sequences are preprocessed by feature extraction and the resulting feature vectors are analyzed by prototype-based classification to remain interpretable. In particular, we propose to use variants of learning vector quantization (LVQ) based on dissimilarity measures for RNA sequence data. The respective matrix LVQ provides additional knowledge about the classification decisions like discriminant feature correlations and, additionally, can be equipped with easy to realize reject options for uncertain data. Those options provide self-controlled evidence, i.e., the model refuses to make a classification decision if the model evidence for the presented data is not sufficient. This model is first trained using a GISAID dataset with given virus types detected according to the molecular differences in coronavirus populations by phylogenetic tree clustering. In a second step, we apply the trained model to another but unlabeled SARS-CoV-2 virus dataset. For these data, we can either assign a virus type to the sequences or reject atypical samples. Those rejected sequences allow to speculate about new virus types with respect to nucleotide base mutations in the viral sequences. Moreover, this rejection analysis improves model robustness. Last but not least, the presented approach has lower computational complexity compared to methods based on (multiple) sequence alignment.

SUPPLEMENTARY INFORMATION

The online version contains supplementary material available at 10.1007/s00521-021-06018-2.

Generalized Learning Riemannian Space Quantization: A Case Study on Riemannian Manifold of SPD Matrices

Learning vector quantization (LVQ) is a simple and efficient classification method, enjoying great popularity. However, in many classification scenarios, such as electroencephalogram (EEG) classification, the input features are represented by symmetric positive-definite (SPD) matrices that live in a curved manifold rather than vectors that live in the flat Euclidean space. In this article, we propose a new classification method for data points that live in the curved Riemannian manifolds in the framework of LVQ. The proposed method alters generalized LVQ (GLVQ) with the Euclidean distance to the one operating under the appropriate Riemannian metric. We instantiate the proposed method for the Riemannian manifold of SPD matrices equipped with the Riemannian natural metric. Empirical investigations on synthetic data and real-world motor imagery EEG data demonstrate that the performance of the proposed generalized learning Riemannian space quantization can significantly outperform the Euclidean GLVQ, generalized relevance LVQ (GRLVQ), and generalized matrix LVQ (GMLVQ). The proposed method also shows competitive performance to the state-of-the-art methods on the EEG classification of motor imagery tasks.

Learning vector quantization and relevances in complex coefficient space

In this contribution, we consider the classification of time series and similar functional data which can be represented in complex Fourier and wavelet coefficient space. We apply versions of learning vector quantization (LVQ) which are suitable for complex-valued data, based on the so-called Wirtinger calculus. It allows for the formulation of gradient-based update rules in the framework of cost-function-based generalized matrix relevance LVQ (GMLVQ). Alternatively, we consider the concatenation of real and imaginary parts of Fourier coefficients in a real-valued feature vector and the classification of time-domain representations by means of conventional GMLVQ. In addition, we consider the application of the method in combination with wavelet-space features to heartbeat classification.

An application of generalized matrix learning vector quantization in neuroimaging

BACKGROUND AND OBJECTIVE

Neurodegenerative diseases like Parkinson's disease often take several years before they can be diagnosed reliably based on clinical grounds. Imaging techniques such as MRI are used to detect anatomical (structural) pathological changes. However, these kinds of changes are usually seen only late in the development. The measurement of functional brain activity by means of [¹⁸F]fluorodeoxyglucose positron emission tomography (FDG-PET) can provide useful information, but its interpretation is more difficult. The scaled sub-profile model principal component analysis (SSM/PCA) was shown to provide more useful information than other statistical techniques. Our objective is to improve the performance further by combining SSM/PCA and prototype-based generalized matrix learning vector quantization (GMLVQ).

METHODS

We apply a combination of SSM/PCA and GMLVQ as a classifier. In order to demonstrate the combination's validity, we analyze FDG-PET data of Parkinson's disease (PD) patients collected at three different neuroimaging centers in Europe. We determine the diagnostic performance by performing a ten times repeated ten fold cross validation. Additionally, discriminant visualizations of the data are included. The prototypes and relevance of GMLVQ are transformed back to the original voxel space by exploiting the linearity of SSM/PCA. The resulting prototypes and relevance profiles have then been assessed by three neurologists.

RESULTS

One important finding is that discriminative visualization can help to identify disease-related properties as well as differences which are due to center-specific factors. Secondly, the neurologist assessed the interpretability of the method and confirmed that prototypes are similar to known activity profiles of PD patients.

CONCLUSION

We have shown that the presented combination of SSM/PCA and GMLVQ can provide useful means to assess and better understand characteristic differences in FDG-PET data from PD patients and HCs. Based on the assessments by medical experts and the results of our computational analysis we conclude that the first steps towards a diagnostic support system have been taken successfully.

Accurate non-invasive diagnosis and staging of non-alcoholic fatty liver disease using the urinary steroid metabolome

BACKGROUND

The development of accurate, non-invasive markers to diagnose and stage non-alcoholic fatty liver disease (NAFLD) is critical to reduce the need for an invasive liver biopsy and to identify patients who are at the highest risk of hepatic and cardio-metabolic complications. Disruption of steroid hormone metabolic pathways has been described in patients with NAFLD.

AIM(S)

To assess the hypothesis that assessment of the urinary steroid metabolome may provide a novel, non-invasive biomarker strategy to stage NAFLD.

METHODS

We analysed the urinary steroid metabolome in 275 subjects (121 with biopsy-proven NAFLD, 48 with alcohol-related cirrhosis and 106 controls), using gas chromatography-mass spectrometry (GC-MS) coupled with machine learning-based Generalised Matrix Learning Vector Quantisation (GMLVQ) analysis.

RESULTS

Generalised Matrix Learning Vector Quantisation analysis achieved excellent separation of early (F0-F2) from advanced (F3-F4) fibrosis (AUC receiver operating characteristics [ROC]: 0.92 [0.91-0.94]). Furthermore, there was near perfect separation of controls from patients with advanced fibrotic NAFLD (AUC ROC = 0.99 [0.98-0.99]) and from those with NAFLD cirrhosis (AUC ROC = 1.0 [1.0-1.0]). This approach was also able to distinguish patients with NAFLD cirrhosis from those with alcohol-related cirrhosis (AUC ROC = 0.83 [0.81-0.85]).

CONCLUSIONS

Unbiased GMLVQ analysis of the urinary steroid metabolome offers excellent potential as a non-invasive biomarker approach to stage NAFLD fibrosis as well as to screen for NAFLD. A highly sensitive and specific urinary biomarker is likely to have clinical utility both in secondary care and in the broader general population within primary care and could significantly decrease the need for liver biopsy.

Modelling prognostic trajectories of cognitive decline due to Alzheimer's disease

Alzheimer's disease (AD) is characterised by a dynamic process of neurocognitive changes from normal cognition to mild cognitive impairment (MCI) and progression to dementia. However, not all individuals with MCI develop dementia. Predicting whether individuals with MCI will decline (i.e. progressive MCI) or remain stable (i.e. stable MCI) is impeded by patient heterogeneity due to comorbidities that may lead to MCI diagnosis without progression to AD. Despite the importance of early diagnosis of AD for prognosis and personalised interventions, we still lack robust tools for predicting individual progression to dementia. Here, we propose a novel trajectory modelling approach based on metric learning (Generalised Metric Learning Vector Quantization) that mines multimodal data from MCI patients in the Alzheimer's disease Neuroimaging Initiative (ADNI) cohort to derive individualised prognostic scores of cognitive decline due to AD. We develop an integrated biomarker generation- using partial least squares regression- and classification methodology that extends beyond binary patient classification into discrete subgroups (i.e. stable vs. progressive MCI), determines individual profiles from baseline (i.e. cognitive or biological) data and predicts individual cognitive trajectories (i.e. change in memory scores from baseline). We demonstrate that a metric learning model trained on baseline cognitive data (memory, executive function, affective measurements) discriminates stable vs. progressive MCI individuals with high accuracy (81.4%), revealing an interaction between cognitive (memory, executive functions) and affective scores that may relate to MCI comorbidity (e.g. affective disturbance). Training the model to perform the same binary classification on biological data (mean cortical β-amyloid burden, grey matter density, APOE 4) results in similar prediction accuracy (81.9%). Extending beyond binary classifications, we develop and implement a trajectory modelling approach that shows significantly better performance in predicting individualised rate of future cognitive decline (i.e. change in memory scores from baseline), when the metric learning model is trained with biological (r = -0.68) compared to cognitive (r = -0.4) data. Our trajectory modelling approach reveals interpretable and interoperable markers of progression to AD and has strong potential to guide effective stratification of individuals based on prognostic disease trajectories, reducing MCI patient misclassification, that is critical for clinical practice and discovery of personalised interventions.

Deep Adversarial Metric Learning

Learning an effective distance measurement between sample pairs plays an important role in visual analysis, where the training procedure largely relies on hard negative samples. However, hard negative samples usually account for the tiny minority in the training set, which may fail to fully describe the data distribution close to the decision boundary. In this paper, we present a deep adversarial metric learning (DAML) framework to generate synthetic hard negatives from the original negative samples, which is widely applicable to existing supervised deep metric learning algorithms. Different from existing sampling strategies which simply ignore numerous easy negatives, our DAML aim to exploit them by generating synthetic hard negatives adversarial to the learned metric as complements. We simultaneously train the feature embedding and hard negative generator in an adversarial manner, so that adequate and targeted synthetic hard negatives are created to learn more precise distance metrics. As a single transformation may not be powerful enough to describe the global input space under the attack of the hard negative generator, we further propose a deep adversarial multi-metric learning (DAMML) method by learning multiple local transformations for more complete description. We simultaneously exploit the collaborative and competitive relationships among multiple metrics, where the metrics display unity against the generator for effective distance measurement as well as compete for more training data through a metric discriminator to avoid overlapping. Extensive experimental results on five benchmark datasets show that our DAML and DAMML effectively boost the performance of existing deep metric learning approaches through adversarial learning.

Adaptive basis functions for prototype-based classification of functional data

We present a framework for distance-based classification of functional data. We consider the analysis of labeled spectral data and time series by means of generalized matrix relevance learning vector quantization (GMLVQ) as an example. To take advantage of the functional nature, a functional expansion of the input data is considered. Instead of using a predefined set of basis functions for the expansion, a more flexible scheme of an adaptive functional basis is employed. GMLVQ is applied on the resulting functional parameters to solve the classification task. For comparison of the classification, a GMLVQ system is also applied to the raw input data, as well as on data expanded by a different predefined functional basis. Computer experiments show that the methods offer potential to improve classification performance significantly. Furthermore, the analysis of the adapted set of basis functions give further insights into the data structure and yields an option for a drastic reduction of dimensionality.

Counteracting Electrode Shifts in Upper-Limb Prosthesis Control via Transfer Learning

Research on machine learning approaches for upper-limb prosthesis control has shown impressive progress. However, translating these results from the lab to patient's everyday lives remains a challenge because advanced control schemes tend to break down under everyday disturbances, such as electrode shifts. Recently, it has been suggested to apply adaptive transfer learning to counteract electrode shifts using as little newly recorded training data as possible. In this paper, we present a novel, simple version of transfer learning and provide the first user study demonstrating the effectiveness of transfer learning to counteract electrode shifts. For this purpose, we introduce the novel Box and Beans test to evaluate prosthesis proficiency and compare user performance with an initial simple pattern recognition system, the system under electrode shifts, and the system after transfer learning. Our results show that transfer learning could significantly alleviate the impact of electrode shifts on user performance in the Box and Beans test.

Maintaining Security and Privacy in Health Care System Using Learning Based Deep-Q-Networks

In the recent past, Internet of Things (IoT) plays a significant role in different applications such as health care, industrial sector, defense and research etc.… It provides effective framework in maintaining the security, privacy and reliability of the information in internet environment. Among various applications as mentioned health care place a major role, because security, privacy and reliability of the medical information is maintained in an effective way. Even though, IoT provides the effective protocols for maintaining the information, several intermediate attacks and intruders trying to access the health information which in turn reduce the privacy, security and reliability of the entire health care system in internet environment. As a result and to solve the issues, in this research Learning based Deep-Q-Networks has been introduced for reducing the malware attacks while managing the health information. This method examines the medical information in different layers according to the Q-learning concept which helps to minimize the intermediate attacks with less complexity. The efficiency of the system has been evaluated with the help of experimental results and discussions.

Identifying HIV Associated Neurocognitive Disorder Using Large-Scale Granger Causality Analysis on Resting-State Functional MRI

We investigate the applicability of large-scale Granger Causality (lsGC) for extracting a measure of multivariate information flow between pairs of regional brain activities from resting-state functional MRI (fMRI) and test the effectiveness of these measures for predicting a disease state. Such pairwise multivariate measures of interaction provide high-dimensional representations of connectivity profiles for each subject and are used in a machine learning task to distinguish between healthy controls and individuals presenting with symptoms of HIV Associated Neurocognitive Disorder (HAND). Cognitive impairment in several domains can occur as a result of HIV infection of the central nervous system. The current paradigm for assessing such impairment is through neuropsychological testing. With fMRI data analysis, we aim at non-invasively capturing differences in brain connectivity patterns between healthy subjects and subjects presenting with symptoms of HAND. To classify the extracted interaction patterns among brain regions, we use a prototype-based learning algorithm called Generalized Matrix Learning Vector Quantization (GMLVQ). Our approach to characterize connectivity using lsGC followed by GMLVQ for subsequent classification yields good prediction results with an accuracy of 87% and an area under the ROC curve (AUC) of up to 0.90. We obtain a statistically significant improvement (p<0.01) over a conventional Granger causality approach (accuracy = 0.76, AUC = 0.74). High accuracy and AUC values using our multivariate method to connectivity analysis suggests that our approach is able to better capture changes in interaction patterns between different brain regions when compared to conventional Granger causality analysis known from the literature.

Ordinal regression based on learning vector quantization

Recently, ordinal regression, which predicts categories of ordinal scale, has received considerable attention. In this paper, we propose a new approach to solve ordinal regression problems within the learning vector quantization framework. It extends the previous approach termed ordinal generalized matrix learning vector quantization with a more suitable and natural cost function, leading to more intuitive parameter update rules. Moreover, in our approach the bandwidth of the prototype weights is automatically adapted. Empirical investigation on a number of datasets reveals that overall the proposed approach tends to have superior out-of-sample performance, when compared to alternative ordinal regression methods.

Investigating Changes in Resting-State Connectivity from Functional MRI Data in Patients with HIV Associated Neurocognitive Disorder Using MCA and Machine Learning

Infection of the brain by the Human Immunodeficiency Virus (HIV) causes irreversible damage to the synaptic connections resulting in cognitive impairment. Patients with HIV infection, showing signs of impairment in multiple cognitive domains, as assessed by neuropsychological testing, are said to exhibit symptoms of HIV Associated Neurocognitive Disorder (HAND). In this study, we use resting-state functional MRI (fMRI) data to distinguish between healthy subjects and subjects with symptoms of HAND. To this end, we first establish a measure of interaction between pairs of regional time-series by quantifying their non-linear functional connectivity using Mutual Connectivity Analysis (MCA). Subsequently, we use a classifier to distinguish patterns of interaction between healthy and diseased individuals. Our results, quantified as the mean Area under the ROC curve (AUC) over 75 iterations, indicate that, using fMRI data, we can discriminate between the two cohorts well (AUC > 0.8). Specifically, we find that MCA (mean AUC = 0.89) based connectivity features perform significantly better (p < 0.05) when compared to cross-correlation (mean AUC = 0.82) at the classification task. A higher AUC using our approach suggests that such a nonlinear approach is better able to capture connectivity changes between brain regions and has potential for the development of novel neuro-imaging biomarkers.

Can Learning Vector Quantization be an Alternative to SVM and Deep Learning? - Recent Trends and Advanced Variants of Learning Vector Quantization for Classification Learning

Learning vector quantization (LVQ) is one of the most powerful approaches for prototype based classification of vector data, intuitively introduced by Kohonen. The prototype adaptation scheme relies on its attraction and repulsion during the learning providing an easy geometric interpretability of the learning as well as of the classification decision scheme. Although deep learning architectures and support vector classifiers frequently achieve comparable or even better results, LVQ models are smart alternatives with low complexity and computational costs making them attractive for many industrial applications like intelligent sensor systems or advanced driver assistance systems.

Nowadays, the mathematical theory developed for LVQ delivers sufficient justification of the algorithm making it an appealing alternative to other approaches like support vector machines and deep learning techniques.

This review article reports current developments and extensions of LVQ starting from the generalized LVQ (GLVQ), which is known as the most powerful cost function based realization of the original LVQ. The cost function minimized in GLVQ is an soft-approximation of the standard classification error allowing gradient descent learning techniques. The GLVQ variants considered in this contribution, cover many aspects like bordersensitive learning, application of non-Euclidean metrics like kernel distances or divergences, relevance learning as well as optimization of advanced statistical classification quality measures beyond the accuracy including sensitivity and specificity or area under the ROC-curve.

According to these topics, the paper highlights the basic motivation for these variants and extensions together with the mathematical prerequisites and treatments for integration into the standard GLVQ scheme and compares them to other machine learning approaches. For detailed description and mathematical theory behind all, the reader is referred to the respective original articles.

Thus, the intention of the paper is to provide a comprehensive overview of the stateof- the-art serving as a starting point to search for an appropriate LVQ variant in case of a given specific classification problem as well as a reference to recently developed variants and improvements of the basic GLVQ scheme.

Classifying Cognitive Profiles Using Machine Learning with Privileged Information in Mild Cognitive Impairment

Early diagnosis of dementia is critical for assessing disease progression and potential treatment. State-or-the-art machine learning techniques have been increasingly employed to take on this diagnostic task. In this study, we employed Generalized Matrix Learning Vector Quantization (GMLVQ) classifiers to discriminate patients with Mild Cognitive Impairment (MCI) from healthy controls based on their cognitive skills. Further, we adopted a “Learning with privileged information” approach to combine cognitive and fMRI data for the classification task. The resulting classifier operates solely on the cognitive data while it incorporates the fMRI data as privileged information (PI) during training. This novel classifier is of practical use as the collection of brain imaging data is not always possible with patients and older participants. MCI patients and healthy age-matched controls were trained to extract structure from temporal sequences. We ask whether machine learning classifiers can be used to discriminate patients from controls and whether differences between these groups relate to individual cognitive profiles. To this end, we tested participants in four cognitive tasks: working memory, cognitive inhibition, divided attention, and selective attention. We also collected fMRI data before and after training on a probabilistic sequence learning task and extracted fMRI responses and connectivity as features for machine learning classifiers. Our results show that the PI guided GMLVQ classifiers outperform the baseline classifier that only used the cognitive data. In addition, we found that for the baseline classifier, divided attention is the only relevant cognitive feature. When PI was incorporated, divided attention remained the most relevant feature while cognitive inhibition became also relevant for the task. Interestingly, this analysis for the fMRI GMLVQ classifier suggests that (1) when overall fMRI signal is used as inputs to the classifier, the post-training session is most relevant; and (2) when the graph feature reflecting underlying spatiotemporal fMRI pattern is used, the pre-training session is most relevant. Taken together these results suggest that brain connectivity before training and overall fMRI signal after training are both diagnostic of cognitive skills in MCI.

Dimensionality-Dependent Generalization Bounds for k-Dimensional Coding Schemes

The k-dimensional coding schemes refer to a collection of methods that attempt to represent data using a set of representative k-dimensional vectors and include nonnegative matrix factorization, dictionary learning, sparse coding, k-means clustering, and vector quantization as special cases. Previous generalization bounds for the reconstruction error of the k-dimensional coding schemes are mainly dimensionality-independent. A major advantage of these bounds is that they can be used to analyze the generalization error when data are mapped into an infinite- or high-dimensional feature space. However, many applications use finite-dimensional data features. Can we obtain dimensionality-dependent generalization bounds for k-dimensional coding schemes that are tighter than dimensionality-independent bounds when data are in a finite-dimensional feature space? Yes. In this letter, we address this problem and derive a dimensionality-dependent generalization bound for k-dimensional coding schemes by bounding the covering number of the loss function class induced by the reconstruction error. The bound is of order [Formula: see text], where m is the dimension of features, k is the number of the columns in the linear implementation of coding schemes, and n is the size of sample, [Formula: see text] when n is finite and [Formula: see text] when n is infinite. We show that our bound can be tighter than previous results because it avoids inducing the worst-case upper bound on k of the loss function. The proposed generalization bound is also applied to some specific coding schemes to demonstrate that the dimensionality-dependent bound is an indispensable complement to the dimensionality-independent generalization bounds.

Prototype-based models in machine learning

An overview is given of prototype-based models in machine learning. In this framework, observations, i.e., data, are stored in terms of typical representatives. Together with a suitable measure of similarity, the systems can be employed in the context of unsupervised and supervised analysis of potentially high-dimensional, complex datasets. We discuss basic schemes of competitive vector quantization as well as the so-called neural gas approach and Kohonen's topology-preserving self-organizing map. Supervised learning in prototype systems is exemplified in terms of learning vector quantization. Most frequently, the familiar Euclidean distance serves as a dissimilarity measure. We present extensions of the framework to nonstandard measures and give an introduction to the use of adaptive distances in relevance learning.