Pattern Classification Approach Research Articles

Generalized zero-shot learning for classifying unseen wafer map patterns

A data-driven approach for classifying defect patterns in wafer maps is crucial for automating the quality assurance process during semiconductor fabrication. Existing works have proposed machine learning-based methods to accomplish this important problem. However, many of these approaches assume model training from a supervised learning perspective, in which a substantial number of annotated wafer maps, each with its actual defect pattern, need to be available. Due to the cost and time required to generate such an annotated dataset, recently a few works started to explore the problem of identifying previously unseen defect patterns. However, the methods proposed in those works are able to only detect the presence of unseen defect patterns without identifying what those defect patterns are. To address this issue, this paper introduces a generalized zero-shot learning (GZSL)-based framework capable of classifying and identifying both seen and unseen defect patterns. It consists of a visual feature extraction phase based on semi-supervised contrastive learning (CL), a clustering phase using distance-based initialization, and an alignment phase based on the Wasserstein-Distance-based Visual Structure Constraint method. Our framework further reduces the need for expensive annotated datasets previously required to train supervised learning-based models, thereby improving the practicality of the data-driven approach for wafer map pattern classification. Furthermore, our framework introduces an effective way of incorporating additional semantic information in wafer map pattern classification that traditionally relied on training models solely from wafer maps. To the best of the authors’ knowledge, our work is the first work to adopt GZSL in wafer map pattern classification.

A novel approach for wafer defect pattern classification based on topological data analysis

A novel approach for wafer defect pattern classification based on topological data analysis

Diffusion MRI indices of brain microstructure: Evaluating the Ball‐and‐Stick model in distinguishing Alzheimer’s disease

AbstractBackgroundDiffusion MRI is a valuable tool to assess white matter (WM) microstructure biomarkers of brain aging and Alzheimer’s disease (AD). The most common model used to extract diffusion properties is the diffusion tensor model (DTI). However, DTI cannot resolve complex WM microstructure and suffers from partial volume effects. Further, the majority of multi‐site imaging clinical studies, such as ADNI, are limited to single‐shell acquisitions due to time constraints. Therefore, multi‐compartment models are not feasible to overcome these limitations except multi‐tensor models, such as Ball‐and‐Stick (B&S) under some constraints. B&S offers a novel approach to derive biomarkers for early detection of AD because it splits the diffusion‐weighted MR signal into several anisotropic components (Stick) and a single isotropic component (Ball). This model is implemented in Anima software. Here, we propose a classification approach using four DTI and two B&S indices quantified in various anatomical regions of interest (ROIs) to assess and compare their ability to distinguish AD.MethodCross‐sectional data were available for 413 ADNI‐3 participants: 68 AD (age: 76.81±8.3; 50% women) and 345 cognitively normal (CN; age:72.91±7.27; 62% women). We processed T1‐weighted and diffusion sequences following the pipeline in Fig.1. We obtained four DTI maps: fractional anisotropy (FA), axial (AxD), mean (MD), radial (RD) diffusivity; and two B&S maps: free water weight (FWW) and Stick‐axial diffusivity (ST). Diffusion maps were quantified within a set of anatomical ROIs. Then, we performed a high‐dimensional pattern classification approach. For each metric, we consider all brain ROIs jointly and identify a minimal set of regions whose values jointly maximally differentiate between AD and CN.ResultB&S model with regional FWW and ST measures shows the greatest discrimination between AD and CN (FWW ROCAUC = 0.942; ST ROCAUC = 0.938) compared to DTI (Table 1). Left PHG Parahippocampal Gyrus and Right Hippocampus regions have the highest contribution in the classification (Table 2, Fig.2).ConclusionOur results suggest B&S measures are more sensitive to discriminate AD from CN than DTI measures, widely used in clinical research. B&S has the advantage of identifying and separating crossing fibers, and therefore, has the potential to offer greater and novel insights into the underlying pathology of AD.

Kernelized discriminative–collaborative representation-based approach for pattern classification

In representation-based classification methods, Gaussian function is adopted for realizing the nonlinear representation of samples. However, there are some problems. The definition of Gaussian kernel includes exponential operation, which causes the parameters to be sensitive and consumes exponential computational time. In addition, the performance of pattern classification may be dropped with noise (outliers). To solve these problems, a new Euclidean Kernel is designed to replace Gaussian Kernel. Moreover, a new weight and an uncorrelated sparsity constraint are introduced to enhance within-class representation consistency and between-class representation discrepancy. Then, we propose a kernelized discriminative–collaborative representation-based approach for pattern classification by incorporating the cooperative and competitive representation. The proposed model fully excavates the intrinsic factors of the samples and presents promising performance and better interpretability. Finally, extensive experiments are conducted in comparison with other popular methods on different types of databases, and these experiments verify that the designed approach performs better than its competitors.

Stable functional compensation within hippocampal-subregion networks in patients with temporal glioma before and after surgery.

ObjectiveTo identify whether tumor invasion of the temporal lobe induces functional compensation of the hippocampal-subregion (HIPsub) network connectivity before surgery, and to further validate the stability of this functional compensation within the HIPsub network in patients with temporal glioma tumor (TTumor) after surgical resection of the tumor.MethodsIn the first cohort, analysis of HIPsub functional connectivity (FC) was conducted to identify the functional compensation of the altered HIPsub connectivity pattern in TTumor through a pattern classification approach. Then, the second cohort investigated whether functional compensation in TTumor patients changed after surgical resection of the tumor.ResultsIn the first cohort, this study identified altered HIPsub network connectivity patterns and its functional compensation regions (i.e., left parahippocampal gyrus and bilateral cerebellum anterior lobe) in TTumor patients. Second, the altered HIPsub network connectivity patterns had the power to discriminate TTumor patients from healthy controls (CN) on an individual subject basis, with an AUC of 97.0%, sensitivity of 93.5%, and specificity of 90.3%. Finally, in the second cohort, we found that functional connectivities of functional compensation regions within the HIPsub network in TTumor patients did not change between before and after surgery.ConclusionThis study provides novel evidence regarding functional compensation within the HIPsub network in TTumor patients. It has been suggested that the fine hippocampal subregion was more sensitive, which reveals functional compensation induced by tumor invasion of the temporal lobe. Furthermore, this study verified the stability and persistence of this functional compensation in TTumor patients after surgical resection of the tumor.

Altered Structural and Functional Patterns Within Executive Control Network Distinguish Frontal Glioma-Related Epilepsy.

BackgroundThe tumor invasion of the frontal lobe induces changes in the executive control network (ECN). It remains unclear whether epileptic seizures in frontal glioma patients exacerbate the structural and functional alterations within the ECN, and whether these changes can be used to identify glioma-related seizures at an early stage. This study aimed to investigate the altered structural and functional patterns of ECN in frontal gliomas without epilepsy (non-FGep) and frontal gliomas with epilepsy (FGep) and to evaluate whether the patterns can accurately distinguish glioma-related epilepsy.MethodsWe measured gray matter (GM) volume, regional homogeneity (ReHo), and functional connectivity (FC) within the ECN to identify the structural and functional changes in 50 patients with frontal gliomas (29 non-FGep and 21 FGep) and 39 healthy controls (CN). We assessed the relationships between the structural and functional changes and cognitive function using partial correlation analysis. Finally, we applied a pattern classification approach to test whether structural and functional abnormalities within the ECN can distinguish non-FGep and FGep from CN subjects.ResultsWithin the ECN, non-FGep and FGep showed increased local structure (GM) and function (ReHo), and decreased FC between brain regions compared to CN. Also, non-FGep and FGep showed differential patterns of structural and functional abnormalities within the ECN, and these abnormalities are more severe in FGep than in non-FGep. Lastly, FC between the right superior frontal gyrus and right dorsolateral prefrontal cortex was positively correlated with episodic memory scores in non-FGep and FGep. In particular, the support vector machine (SVM) classifier based on structural and functional abnormalities within ECN could accurately distinguish non-FGep and FGep from CN, and FGep from non-FGep on an individual basis with very high accuracy, area under the curve (AUC), sensitivity, and specificity.ConclusionTumor invasion of the frontal lobe induces local structural and functional reorganization within the ECN, exacerbated by the accompanying epileptic seizures. The ECN abnormalities can accurately distinguish the presence or absence of epileptic seizures in frontal glioma patients. These findings suggest that differential ECN patterns can assist in the early identification and intervention of epileptic seizures in frontal glioma patients.

Hippocampal-Subregion Mechanisms of Repetitive Transcranial Magnetic Stimulation Causally Associated with Amelioration of Episodic Memory in Amnestic Mild Cognitive Impairment.

Altered hippocampal subregions (HIPsub) and their network connectivity relate to episodic memory decline in amnestic mild cognitive impairment (aMCI), which is significantly limited by over-dependence on correlational associations. To identify whether restoration of HIPsub and its network connectivity using repetitive transcranial magnetic stimulation (rTMS) is causally linked to amelioration of episodic memory in aMCI. In the first cohort, analysis of HIPsub grey matter (GM) and its functional connectivity was performed to identify an episodic memory-related circuit in aMCI by using a pattern classification approach. In the second cohort, this circuit was experimentally modulated with rTMS. Structural equation modeling was employed to investigate rTMS regulatory mechanism in amelioration of episodic memory. First, in the first cohort, this study identified HIPsub circuit pathology of episodic memory decline in aMCI patients. Second, in the second cohort, restoration of HIPc GM and its connectivity with left middle temporal gyrus (MTG.L) are causally associated with amelioration of episodic memory in aMCI after 4 weeks of rTMS. Especially important, the effects of HIPc GM changes on the improvement of episodic memory were significantly mediated by HIPc connectivity with MTG.L changes in aMCI. This study provides novel experimental evidence about a biological substrate for the treatment of the disabling episodic memory in aMCI patients. Correction of breakdown in HIPc structure and its connectivity with MTG can causally ameliorate episodic memory in aMCI.

Neural Correlates of Smooth Pursuit Eye Movements in Schizotypy and Recent Onset Psychosis: A Multivariate Pattern Classification Approach.

Schizotypy refers to a set of personality traits that bear resemblance, at subclinical level, to psychosis. Despite evidence of similarity at multiple levels of analysis, direct comparisons of schizotypy and clinical psychotic disorders are rare. Therefore, we used functional magnetic resonance imaging (fMRI) to examine the neural correlates and task-based functional connectivity (psychophysiological interactions; PPI) of smooth pursuit eye movements (SPEM) in patients with recent onset psychosis (ROP; n = 34), participants with high levels of negative (HNS; n = 46) or positive (HPS; n = 41) schizotypal traits, and low-schizotypy control participants (LS; n = 61) using machine-learning. Despite strong previous evidence that SPEM is a highly reliable marker of psychosis, patients and controls could not be significantly distinguished based on SPEM performance or blood oxygen level dependent (BOLD) signal during SPEM. Classification was, however, significant for the right frontal eye field (FEF) seed region in the PPI analyses but not for seed regions in other key areas of the SPEM network. Applying the right FEF classifier to the schizotypal samples yielded decision scores between the LS and ROP groups, suggesting similarities and dissimilarities of the HNS and HPS samples with the LS and ROP groups. The very small difference between groups is inconsistent with previous studies that showed significant differences between patients with ROP and controls in both SPEM performance and underlying neural mechanisms with large effect sizes. As the current study had sufficient power to detect such differences, other reasons are discussed.

Landform pattern recognition and classification for predicting soil types of the Uasin Gishu Plateau, Kenya

Information obtained from landform classification is fundamental for understanding physical, chemical, and biological soil processes. Digital elevation models (DEMs) can be used for landform classification using a geomorphic pattern recognition and classification approach such as geomorphons. In this study, we used geomorphons to predict soil types of the Uasin Gishu Plateau in western Kenya with the aim of improving the existing soil type maps. We ran the geomorphons classification on the 30 m Shuttle Radar Topographic Mission (SRTM) DEM using the module r.geomorphons “add-on” in GRASS GIS with look up distance (L) values of 10, 20, 30, 40, 50, 60, 70, 80, 90 and 100 cells (300 m to 3,000 m) and flatness threshold values of 2, 0.5 and 0.01 degrees. We grouped the resulting geomorphons into three classes: upland summits, upland midslopes, and bottomlands. We then assigned soil types according to the World Reference Base soil classification system to these three landscape positions after dividing the study area into the Lower Plateau and the Upper Plateau based on elevation and geomorphology. The predicted soil types were quantitatively evaluated against 50 soil point observations based on overall accuracy, precision, recall and Cohen’s kappa coefficient (k) metrics. Evaluation results showed that an L value of 20 cells performed best (k = 0.52; overall accuracy = 0.62) followed by an L value of 30 cells (k = 0.50; overall accuracy = 0.58). L values of 60, 70 and 80 cells performed worse (k = 0.35; overall accuracy = 0.48). Although an L value of 20 cells performed slightly better than an L value of 30 cells, the L value of 30 cells better captured the geomorphology and soil-landscape relationships based on our own expert knowledge, legacy data, and the fact that the bottomlands pattern was more continuous for an L value of 30 cells than for an L value of 20 cells. Upland summits occurred over ~ 32% of the plateau and were occupied by Nitisols on the Upper Plateau and Ferralsols on the Lower Plateau. Upland midslopes occurred over ~ 42% of the plateau and were occupied by Acrisols on the Upper Plateau and Acrisols/Ferralsols on the Lower Plateau. Bottomlands occurred over ~ 26% of the plateau and were occupied by Luvisols/Gleysols on both the Upper and Lower Plateau. Geomorphons, as a method of landform classification, correlated to geomorpho-pedological processes and captured soil variations and differences in the study area. The approach is computationally efficient and can be used for large areas, but is limited in that it only classifies landscapes according to shape. This means that the algorithm does not separate or identify soil types with different parent materials occurring within one landform class. Thus, Regosols and Cambisols on very steep slopes and Acrisols and Ferralsols on gentle slopes could not be separated.

A new online approach for classification of pumps vibration patterns based on intelligent IoT system

Machine condition monitoring is a primordial field of study. It allows to avoid downtime in industrial plants, avoiding financial and time losses. In this article, we use an IoT framework to classify the pump’s vibration signal, in order to identify a normal stage of operation, an incipient cavitation stage and a severe cavitation stage. Our approach uses the vibration signal, which is collected with a MEMS sensor, as an image. The feature extractors used in this study: Hu’s Moments, Gray Level Co-occurrence Matrix, Local Binary Patterns, DenseNet169, ResNet50, VGG19 and MobileNet. The classifiers used in this paper were: Gaussian Naive Bayes, Support Vector Machines, Random Forest, Multilayer Perceptron and k-Nearest Neighbors (kNN). The results showed that Hu’s Moments combined with kNN achieved the best accuracy (99.47%) with a score time of 17 ms. Thus, our approach is reliable and efficient to detect cavitation in pumps.

Fingerprint pattern identification and classification approach based on convolutional neural networks

Fingerprint pattern recognition and classification can be of assistance in the research on human personality. In some previous studies, fingerprints were classified into four categories to speed up recognition, but the method of that classification is not suitable for researching the diversity of human personalities. Therefore, in this paper, fingerprint patterns were classified into six types and the accuracy of the recognition was improved to facilitate the research on human personality characteristics. Based on this idea, a six-category fingerprint database is annotated manually and a convolutional neural network (CNN) is proposed for identifying real fingerprint patterns. The new CNN consists of four convolutional layers, three max-pooling layers, two norm layers, and three fully connected layers. The best accuracy the model achieved was 94.87% for a six-category fingerprint database and 92.9% accuracy for a four-category fingerprint database. The results of experimental tests show that the proposed model can recognize the pattern features from a large fingerprint database using the automatic learning and feature extraction abilities of the CNN to get a greater accuracy than in previous experiments.

A local binary pattern classification approach for cloud types derived from all-sky imagers

ABSTRACTCloud classification from ground-based observations is a challenging task that attracts increasing attention, favoured by the development of all-sky imaging equipment. In this work, we propose a new method for cloud type classification from all-sky images. Appropriate versions of two descriptors, Regional Local Binary Pattern (R-LBP) and Four Patch-Local Binary Pattern (FP-LBP), are employed on the images in order to extract not only global but also local textural information from the observed cloud type patterns. In the classification stage, a linear Support Vector Machine (SVM) scheme and Linear Discriminant Analysis (LDA) classifiers are adopted for the assignment of the corresponding cloud type label. Our method is evaluated against two state-of-the-art methods and their datasets consisting of 5000 and 2500 images, respectively. According to the results, the proposed method outperforms the previous ones. Due to its promising results and the novelty of local pattern information of clouds, the proposed methodology could be considered as the basis for future studies aiming to overcome the basic disadvantage of all-sky imaging algorithms: to provide regional cloud type information instead of one cloud type for the whole sky.

Corrigendum to "On the integrity of functional brain networks in schizophrenia, Parkinson's disease, and advanced age: Evidence from connectivity-based single-subject classification."

Previous whole-brain functional connectivity studies achieved successful classifications of patients and healthy controls but only offered limited specificity as to affected brain systems. Here, we examined whether the connectivity patterns of functional systems affected in schizophrenia (SCZ), Parkinson's disease (PD), or normal aging equally translate into high classification accuracies for these conditions. We compared classification performance between pre-defined networks for each group and, for any given network, between groups. Separate support vector machine classifications of 86 SCZ patients, 80 PD patients, and 95 older adults relative to their matched healthy/young controls, respectively, were performed on functional connectivity in 12 task-based, meta-analytically defined networks using 25 replications of a nested 10-fold cross-validation scheme. Classification performance of the various networks clearly differed between conditions, as those networks that best classified one disease were usually non-informative for the other. For SCZ, but not PD, emotion-processing, empathy, and cognitive action control networks distinguished patients most accurately from controls. For PD, but not SCZ, networks subserving autobiographical or semantic memory, motor execution, and theory-of-mind cognition yielded the best classifications. In contrast, young-old classification was excellent based on all networks and outperformed both clinical classifications. Our pattern-classification approach captured associations between clinical and developmental conditions and functional network integrity with a higher level of specificity than did previous whole-brain analyses. Taken together, our results support resting-state connectivity as a marker of functional dysregulation in specific networks known to be affected by SCZ and PD, while suggesting that aging affects network integrity in a more global way. Hum Brain Mapp 38:5845-5858, 2017. © 2017 Wiley Periodicals, Inc.

Cumulative and CUB Models for Rating Data: A Comparative Analysis

SummaryOrdinal measurements as ratings, preference and evaluation data are very common in applied disciplines, and their analysis requires a proper modelling approach for interpretation, classification and prediction of response patterns. This work proposes a comparative discussion between two statistical frameworks that serve these goals: the established class of cumulative models and a class of mixtures of discrete random variables, denoted as CUB models, whose peculiar feature is the specification of an uncertainty component to deal with indecision and heterogeneity. After surveying their definition and main features, we compare the performances of the selected paradigms by means of simulation experiments and selected case studies. The paper is tailored to enrich the understanding of the two approaches by running an extensive and comparative analysis of results, relative advantages and limitations, also at graphical level. In conclusion, a summarising review of the key issues of the alternative strategies and some final remarks are given, aimed to support a unifying setting.

Aberrant working memory processing in major depression: evidence from multivoxel pattern classification.

Major depressive disorder (MDD) is often accompanied by severe impairments in working memory (WM). Neuroimaging studies investigating the mechanisms underlying these impairments have produced conflicting results. It remains unclear whether MDD patients show hyper- or hypoactivity in WM-related brain regions and how potential aberrations in WM processing may contribute to the characteristic dysregulation of cognition-emotion interactions implicated in the maintenance of the disorder. In order to shed light on these questions and to overcome limitations of previous studies, we applied a multivoxel pattern classification approach to investigate brain activity in large samples of MDD patients (N = 57) and matched healthy controls (N = 61) during a WM task that incorporated positive, negative, and neutral stimuli. Results showed that patients can be distinguished from healthy controls with good classification accuracy based on functional activation patterns. ROI analyses based on the classification weight maps showed that during WM, patients had higher activity in the left DLPFC and the dorsal ACC. Furthermore, regions of the default-mode network (DMN) were less deactivated in patients. As no performance differences were observed, we conclude that patients required more effort, indexed by more activity in WM-related regions, to successfully perform the task. This increased effort might be related to difficulties in suppressing task-irrelevant information reflected by reduced deactivation of regions within the DMN. Effects were most pronounced for negative and neutral stimuli, thus pointing toward important implications of aberrations in WM processes in cognition-emotion interactions in MDD.

A new constrained maximum margin approach to discriminative learning of Bayesian classifiers

We propose a novel discriminative learning approach for Bayesian pattern classification, called ‘constrained maximum margin (CMM)’. We define the margin between two classes as the difference between the minimum decision value for positive samples and the maximum decision value for negative samples. The learning problem is to maximize the margin under the constraint that each training pattern is classified correctly. This nonlinear programming problem is solved using the sequential unconstrained minimization technique. We applied the proposed CMM approach to learn Bayesian classifiers based on Gaussian mixture models, and conducted the experiments on 10 UCI datasets. The performance of our approach was compared with those of the expectation-maximization algorithm, the support vector machine, and other state-of-the-art approaches. The experimental results demonstrated the effectiveness of our approach.

Detection of illegal consumers using pattern classification approach combined with Levenberg-Marquardt method in smart grid

Detection of illegal consumers is an extremely challenging problem in smart grids as well as traditional environments. In a smart grid environment, electrical energy illegal customers can be divided into two types; (1) if the customer consumes all of its required energy in a portion of day illegally, (2) if the customer consumes a portion of its required energy illegally. Many methods about illegal consumption or electricity theft detection have been proposed but they are able to detect only one type of illegal consumptions. In this paper, a combined method is proposed to detect both two types of illegal consumptions. Customer energy consumption pattern classification method based on probabilistic neural network and mathematical model based on Levenberg-Marquardt method are used to detect the first and second type of illegal consumption, respectively. Moreover, the impact of Distributed Generation (DG) sources on illegal consumption of electricity is analyzed and proposed detection algorithm is modified to compensate it. Experimental results are presented to show the effectiveness of this method in detection of both two types of illegal consumption.

Categorization for Faces and Tools-Two Classes of Objects Shaped by Different Experience-Differs in Processing Timing, Brain Areas Involved, and Repetition Effects.

The brain mechanisms that integrate the separate features of sensory input into a meaningful percept depend upon the prior experience of interaction with the object and differ between categories of objects. Recent studies using representational similarity analysis (RSA) have characterized either the spatial patterns of brain activity for different categories of objects or described how category structure in neuronal representations emerges in time, but never simultaneously. Here we applied a novel, region-based, multivariate pattern classification approach in combination with RSA to magnetoencephalography data to extract activity associated with qualitatively distinct processing stages of visual perception. We asked participants to name what they see whilst viewing bitonal visual stimuli of two categories predominantly shaped by either value-dependent or sensorimotor experience, namely faces and tools, and meaningless images. We aimed to disambiguate the spatiotemporal patterns of brain activity between the meaningful categories and determine which differences in their processing were attributable to either perceptual categorization per se, or later-stage mentalizing-related processes. We have extracted three stages of cortical activity corresponding to low-level processing, category-specific feature binding, and supra-categorical processing. All face-specific spatiotemporal patterns were associated with bilateral activation of ventral occipito-temporal areas during the feature binding stage at 140–170 ms. The tool-specific activity was found both within the categorization stage and in a later period not thought to be associated with binding processes. The tool-specific binding-related activity was detected within a 210–220 ms window and was located to the intraparietal sulcus of the left hemisphere. Brain activity common for both meaningful categories started at 250 ms and included widely distributed assemblies within parietal, temporal, and prefrontal regions. Furthermore, we hypothesized and tested whether activity within face and tool-specific binding-related patterns would demonstrate oppositely acting effects following procedural perceptual learning. We found that activity in the ventral, face-specific network increased following the stimuli repetition. In contrast, tool processing in the dorsal network adapted by reducing its activity over the repetition period. Altogether, we have demonstrated that activity associated with visual processing of faces and tools during the categorization stage differ in processing timing, brain areas involved, and in their dynamics underlying stimuli learning.

Automated VSS-based Burn Scar Assessment using Combined Texture and Color Features of Digital Images in Error-Correcting Output Coding

Assessment of burn scars is an important study in both medical research and clinical settings because it can help determine response to burn treatment and plan optimal surgical procedures. Scar rating has been performed using both subjective observations and objective measuring devices. However, there is still a lack of consensus with respect to the accuracy, reproducibility, and feasibility of the current methods. Computerized scar assessment appears to have potential for meeting such requirements but has been rarely found in literature. In this paper an image analysis and pattern classification approach for automating burn scar rating based on the Vancouver Scar Scale (VSS) was developed. Using the image data of pediatric patients, a rating accuracy of 85% was obtained, while 92% and 98% were achieved for the tolerances of one VSS score and two VSS scores, respectively. The experimental results suggest that the proposed approach is very promising as a tool for clinical burn scar assessment that is reproducible and cost-effective.

On the integrity of functional brain networks in schizophrenia, Parkinson's disease, and advanced age: Evidence from connectivity-based single-subject classification.

Previous whole-brain functional connectivity studies achieved successful classifications of patients and healthy controls but only offered limited specificity as to affected brain systems. Here, we examined whether the connectivity patterns of functional systems affected in schizophrenia (SCZ), Parkinson's disease (PD), or normal aging equally translate into high classification accuracies for these conditions. We compared classification performance between pre-defined networks for each group and, for any given network, between groups. Separate support vector machine classifications of 86 SCZ patients, 80 PD patients, and 95 older adults relative to their matched healthy/young controls, respectively, were performed on functional connectivity in 12 task-based, meta-analytically defined networks using 25 replications of a nested 10-fold cross-validation scheme. Classification performance of the various networks clearly differed between conditions, as those networks that best classified one disease were usually non-informative for the other. For SCZ, but not PD, emotion-processing, empathy, and cognitive action control networks distinguished patients most accurately from controls. For PD, but not SCZ, networks subserving autobiographical or semantic memory, motor execution, and theory-of-mind cognition yielded the best classifications. In contrast, young-old classification was excellent based on all networks and outperformed both clinical classifications. Our pattern-classification approach captured associations between clinical and developmental conditions and functional network integrity with a higher level of specificity than did previous whole-brain analyses. Taken together, our results support resting-state connectivity as a marker of functional dysregulation in specific networks known to be affected by SCZ and PD, while suggesting that aging affects network integrity in a more global way. Hum Brain Mapp 38:5845-5858, 2017. © 2017 Wiley Periodicals, Inc.