method-for-simultaneous-fmrieeg-data-collection-during-focused

Method for Simultaneous fMRI/EEG Data Collection during a Focused Attention Suggestion for Differential Thermal Sensation

In the present work, we demonstrate a method for concurrent collection of EEG/fMRI data. In our setup, EEG data are collected using a high-density ...

University of California, Los Angeles

Because changes in the 12-lead high-frequency QRS electrocardiogram (HF QRS ECG) more sensitively identify myocardial ischemia than do changes in the ST segments of the conventional ECG, it is important that changes in HF QRS signals that are merely physiological be distinguishable from those that are potentially pathological. We therefore studied the temporal variation of HF QRS measures such as root mean square (RMS) voltage and the presence vs absence of reduced amplitude zones (RAZs) in 107 asymptomatic individuals in the supine position during a brief period of ECG monitoring. In addition, to ascertain the effects of posture on the 12-lead HF QRS ECG, we collected additional seated data from 25 of these individuals and estimated the fifth and 95th percentile of the percent relative change between the supine and seated measurements. In all cases, variation of HF QRS parameters decreased as the number of beats in the signal average increased. For example, in the supine position, the 95th percentile of the percent relative change between consecutive within-lead measurements of RMS voltage for a 50-beat signal average was 12.3% but decreased to 11.7%, 11.2%, and 10.7% for 75, 100, and 150 beat signal averages, respectively (P < .01). After transition from the supine to the seated upright position, changes in some measures of HF QRS were statistically significant, with RMS voltage decreasing significantly in lead V3 and with the number of RAZs lost in the 12-lead HF QRS ECG significantly exceeding the number of RAZs gained. We conclude that most measures of HF QRS ECG are sufficiently stable for routine continuous monitoring.

Temporal and postural variation of 12-lead high-frequency QRS electrocardiographic signals in asymptomatic individuals.

In November of 2005, President George W Bush requested $7.1 billion dollars for a global influenza epidemic preparedness initiative (Brown, 2005). Preparation measures for a biological threat or influenza pandemic focus on rapid quarantine, vaccines, developing antiviral treatments, and economic concerns (Brown, 2005; Ferguson et al., 2006; Reina, 2008). Although these public health measures are vital, they do not consider the acute mental health consequences that could develop during a pandemic and its aftermath. The most recent H1N1 swine flu has now spread to more than 70 countries (CDC, June 2009), and as of June 11, 2009, is considered a Phase 6 pandemic by the World Health Organization, indicative of ongoing community level outbreaks in multiple parts of the globe. Following recent cases of swine flu, global concern of an influenza pandemic has risen, and it is critical that metal health considerations become an integrated part of the pandemic response. Here, potential mental health consequences and high risk populations are identified and reviewed. Mental health professionals, communities, businesses, and organizations can create an infrastructure to help mitigate mental health consequences. These issues, as well as familial stressors and coping methods, are reviewed.

Preparing for pandemic influenza and its aftermath: mental health issues considered.

With the recent approval by the Food and Drug Administration (FDA) of Deep Brain Stimulation (DBS) for Parkinson's Disease, dystonia and obsessive compulsive disorder (OCD), vagus nerve stimulation (VNS) for epilepsy and depression, and repetitive transcranial magnetic stimulation (rTMS) for the treatment of depression, neuromodulation has become increasingly relevant to clinical research. However, these techniques have significant drawbacks (eg, lack of special specificity and depth for the rTMS, and invasiveness and cumbersome maintenance for DBS). This article reviews the background, rationale, and pilot studies to date, using a new brain stimulation method-low-intensity focused ultrasound pulsation (LIFUP). The ability of ultrasound to be focused noninvasively through the skull anywhere within the brain, together with concurrent imaging (ie, functional magnetic resonance imaging [fMRI]) techniques, may create a role for research and clinical use of LIFUP. This technique is still in preclinical testing and needs to be assessed thoroughly before being advanced to clinical trials. In this study, we review over 50 years of research data on the use of focused ultrasound (FUS) in neuronal tissue and live brain, and propose novel applications of this noninvasive neuromodulation method.

A review of low-intensity focused ultrasound pulsation.

Independent component analysis (ICA) is a popular method for the analysis of functional magnetic resonance imaging (fMRI) signals that is capable of revealing connected brain systems of functional significance. To be computationally tractable, estimating the independent components (ICs) inevitably requires one or more dimension reduction steps. Whereas most algorithms perform such reductions in the time domain, the input data are much more extensive in the spatial domain, and there is broad consensus that the brain obeys rules of localization of function into regions that are smaller in number than the number of voxels in a brain image. These functional units apparently reorganize dynamically into networks under different task conditions. Here we develop a new approach to ICA, producing group results by bagging and clustering over hundreds of pooled single-subject ICA results that have been projected to a lower-dimensional subspace. Averages of anatomically based regions are used to compress the single subject-ICA results prior to clustering and resampling via bagging. The computational advantages of this approach make it possible to perform group-level analyses on datasets consisting of hundreds of scan sessions by combining the results of within-subject analysis, while retaining the theoretical advantage of mimicking what is known of the functional organization of the brain. The result is a compact set of spatial activity patterns that are common and stable across scan sessions and across individuals. Such representations may be used in the context of statistical pattern recognition supporting real-time state classification.

Large Sample Group Independent Component Analysis of Functional Magnetic Resonance Imaging Using Anatomical Atlas-Based Reduction and Bootstrapped Clustering.

&#x6CE8;&#x610F;&#x7F3A;&#x9677;&#x591A;&#x52A8;&#x969C;&#x788D; (ADHD) &#x76EE;&#x524D;&#x88AB;&#x8BCA;&#x65AD;&#x513F;&#x7AE5;&#x4E34;&#x5E8A;&#x533B;&#x5E08;&#x901A;&#x8FC7;&#x4E3B;&#x89C2;&#x591A;&#x52A8;&#x75C7;&#x7279;&#x5B9A;&#x884C;&#x4E3A;&#x6587;&#x4E66;&#xFF0C;&#x6765;&#x81EA;&#x7236;&#x6BCD;&#x548C;&#x6559;&#x5E08;&#x7684;&#x62A5;&#x544A;&#x3002;&#x8003;&#x8651;&#x5230;&#x5176;&#x9AD8;&#x6D41;&#x884C;&#x7387;&#x548C;&#x5DE8;&#x5927;&#x7684;&#x7ECF;&#x6D4E;&#x548C;&#x793E;&#x4F1A;&#x4EE3;&#x4EF7;&#xFF0C;&#x4E00;&#x4E2A;&#x5B9A;&#x91CF;&#x7684;&#x5DE5;&#x5177;&#xFF0C;&#x5B83;&#x7684;&#x7279;&#x70B9;&#x57FA;&#x7840;&#x795E;&#x7ECF;&#x751F;&#x7269;&#x5B66;&#x8BCA;&#x65AD;&#x827E;&#x6ECB;&#x75C5;&#x662F;&#x6781;&#x4E3A;&#x5B9D;&#x8D35;&#x7684;&#x3002;&#x8FD9;&#x63D0;&#x4F9B;&#x4E86;&#x591A;&#x52A8;&#x75C7; 200 &#x673A;&#x5668;&#x5B66;&#x4E60; (&#x6BEB;&#x5347;) &#x7ADE;&#x4E89;&#x3001; &#x591A;&#x7AD9;&#x70B9;&#x7684;&#x534F;&#x4F5C;&#x52AA;&#x529B;&#xFF0C;&#x4EE5;&#x8C03;&#x67E5;&#x6709;&#x591A;&#x52A8;&#x75C7;&#x7684;&#x5F71;&#x50CF;&#x5206;&#x7C7B;&#x5668;&#x7684;&#x52A8;&#x673A;&#x3002;&#x5728;&#x8FD9;&#x91CC;&#xFF0C;&#x6211;&#x4EEC;&#x63D0;&#x51FA;&#x6211;&#x4EEC; ML &#x7684;&#x65B9;&#x6CD5;&#xFF0C;&#x7528;&#x4E8E;&#x7ED3;&#x6784;&#x6027;&#x548C;&#x529F;&#x80FD;&#x6027;&#x78C1;&#x5171;&#x632F;&#x6210;&#x50CF;&#x6570;&#x636E;&#xFF0C;&#x52A0;&#x4E0A;&#x4EBA;&#x53E3;&#x4FE1;&#x606F;&#xFF0C;&#x901A;&#x5E38;&#x5728;&#x516B;&#x4E2A;&#x4E0D;&#x540C;&#x7684;&#x7814;&#x7A76;&#x7F51;&#x7AD9;&#x5F00;&#x53D1; (TD) &#x513F;&#x7AE5;&#x4ECE;&#x9884;&#x6D4B;&#x4E2A;&#x4EBA; adhd &#x7684;&#x8BCA;&#x65AD;&#x72B6;&#x6001;&#x3002;&#x7ED3;&#x6784;&#x7279;&#x70B9;&#x5305;&#x62EC;&#x6765;&#x81EA; 113 &#x76AE;&#x5C42;&#x548C;&#x975E;&#x76AE;&#x8D28;&#x533A;&#x57DF;&#x5B9A;&#x91CF;&#x6307;&#x6807;&#x3002;&#x529F;&#x80FD;&#x7279;&#x70B9;&#x5305;&#x62EC;&#x76AE;&#x5C14;&#x900A;&#x76F8;&#x5173;&#x529F;&#x80FD;&#x8FDE;&#x901A;&#x6027;&#x77E9;&#x9635;&#x3001; &#x623F;&#x5BA4;&#x7ED3;&#x53CA;&#x5168;&#x7403;&#x56FE;&#x7406;&#x8BBA;&#x63AA;&#x65BD;&#x3001; &#x8282;&#x70B9;&#x529F;&#x7387;&#x8C31;&#x3001; voxelwise &#x5168;&#x7403;&#x8FDE;&#x901A;&#x6027;&#x548C; voxelwise &#x533A;&#x57DF;&#x7684;&#x540C;&#x8D28;&#x6027;&#x3002;&#x6211;&#x4EEC;&#x4E3A;&#x6BCF;&#x4E2A;&#x7AD9;&#x70B9;&#x6267;&#x884C;&#x529F;&#x80FD;&#x6392;&#x540D;&#x548C;&#x4F7F;&#x7528;&#x591A;&#x4E2A;&#x652F;&#x6301;&#x5411;&#x91CF;&#x673A;&#x9012;&#x5F52;&#x65B9;&#x5F0F;&#x7279;&#x5F81; (SVM RFE) &#x6D88;&#x9664;&#x7B97;&#x6CD5;&#xFF0C;&#x5E76;&#x901A;&#x8FC7;&#x4F18;&#x5316;&#x7684;&#x5F84;&#x5411;&#x57FA;&#x51FD;&#x6570;&#x5185;&#x6838;&#x652F;&#x6301;&#x5411;&#x91CF;&#x673A; (RBF SVM&#xFF09; &#x9884;&#x671F;&#x7684;&#x6CDB;&#x5316;&#x6027;&#x80FD;&#x8DE8;&#x4E00;&#x7CFB;&#x5217;&#x9876;&#x7EA7;&#x529F;&#x80FD;&#x8BAD;&#x7EC3;&#x7684;&#x7279;&#x5F81;&#x5B50;&#x96C6;&#x9009;&#x62E9;&#x3002;&#x7279;&#x5B9A;&#x4E8E;&#x7AD9;&#x70B9;&#x7684; RBF-&#x652F;&#x6301;&#x5411;&#x91CF;&#x673A;&#x4F7F;&#x7528;&#x6BCF;&#x4E2A;&#x5F71;&#x50CF;&#x7684;&#x65B9;&#x5F0F;&#x4ECE;&#x8FD9;&#x4E9B;&#x4F18;&#x5316;&#x529F;&#x80FD;&#x96C6;&#x88AB;&#x7528;&#x6765;&#x9884;&#x6D4B;&#x8FDB;&#x884C;&#x72EC;&#x7ACB;&#x7684;&#x6D4B;&#x8BD5;&#x96C6;&#x7684;&#x7C7B;&#x6807;&#x7B7E;&#x3002;&#x6295;&#x7968;&#x7684;&#x529E;&#x6CD5;&#x7528;&#x6765;&#x7EC4;&#x5408;&#x8FD9;&#x4E9B;&#x591A;&#x4E2A;&#x9884;&#x6D4B;&#x548C;&#x6307;&#x6D3E;&#x6700;&#x7EC8;&#x7C7B;&#x6807;&#x7B7E;&#x3002;&#x7528;&#x8FD9;&#x79CD;&#x65B9;&#x6CD5;&#x6211;&#x4EEC;&#x90FD;&#x80FD;&#x9884;&#x6D4B; 55%&#xFF08;&#x800C;&#x4E0D;&#x662F;&#x5728;&#x6B64;&#x793A;&#x4F8B;&#x4E2D;&#x7684; 39%&#x7684;&#x51E0;&#x7387;&#x7EA7;&#x522B;&#xFF09; &#x7684;&#x51C6;&#x786E;&#x6027;&#x3001; &#x7075;&#x654F;&#x5EA6; 33%&#x548C; 80%&#x7279;&#x5F02;&#x6027;&#x8BCA;&#x65AD;&#x5C0F;&#x513F;&#x591A;&#x52A8;&#x75C7;&#x3002;&#x8FD9;&#x79CD;&#x505A;&#x6CD5;&#x8FD8;&#x4F7F;&#x6211;&#x4EEC;&#x80FD;&#x591F;&#x8BC4;&#x4F30;&#x9884;&#x6D4B;&#x7ED9;&#x4E88;&#x6DF1;&#x5165;&#x4E86;&#x89E3;&#x5F02;&#x5E38;&#x8111;&#x7535;&#x8DEF;&#x4E2D;&#x591A;&#x52A8;&#x75C7;&#x7684;&#x7ED3;&#x6784;&#x548C;&#x529F;&#x80FD;&#x529F;&#x80FD;&#x3002;

&#x591A;&#x52A8;&#x75C7;&#x7684;&#x591A;&#x5F0F;&#x8054;&#x8FD0;&#x5F71;&#x50CF;&#x5206;&#x7C7B;&#x7684;&#x89C1;&#x89E3;&#x3002;

Attention deficit hyperactivity disorder (ADHD) currently is diagnosed in children by clinicians via subjective ADHD-specific behavioral instruments and by reports from the parents and teachers. Considering its high prevalence and large economic and societal costs, a quantitative tool that aids in diagnosis by characterizing underlying neurobiology would be extremely valuable. This provided motivation for the ADHD-200 machine learning (ML) competition, a multisite collaborative effort to investigate imaging classifiers for ADHD. Here we present our ML approach, which used structural and functional magnetic resonance imaging data, combined with demographic information, to predict diagnostic status of individuals with ADHD from typically developing (TD) children across eight different research sites. Structural features included quantitative metrics from 113 cortical and non-cortical regions. Functional features included Pearson correlation functional connectivity matrices, nodal and global graph theoretical measures, nodal power spectra, voxelwise global connectivity, and voxelwise regional homogeneity. We performed feature ranking for each site and modality using the multiple support vector machine recursive feature elimination (SVM-RFE) algorithm, and feature subset selection by optimizing the expected generalization performance of a radial basis function kernel SVM (RBF-SVM) trained across a range of the top features. Site-specific RBF-SVMs using these optimal feature sets from each imaging modality were used to predict the class labels of an independent hold-out test set. A voting approach was used to combine these multiple predictions and assign final class labels. With this methodology we were able to predict diagnosis of ADHD with 55% accuracy (versus a 39% chance level in this sample), 33% sensitivity, and 80% specificity. This approach also allowed us to evaluate predictive structural and functional features giving insight into abnormal brain circuitry in ADHD.

Insights into multimodal imaging classification of ADHD.

Attention deficit hyperactivity disorder (ADHD) actuellement est diagnostiqu&eacute;e chez les enfants par les cliniciens via les instruments comportementales sp&eacute;cifiques TDAH subjectives et par les rapports des parents et des enseignants. Compte tenu de sa forte pr&eacute;valence et les co&ucirc;ts &eacute;conomiques et soci&eacute;taux importants, un outil quantitatif qui aide au diagnostic en caract&eacute;risant la neurobiologie sous-jacente serait extr&ecirc;mement utile. Cela a fourni la motivation pour le concours apprentissage (ML) de la machine d&#39;ADHD-200, un effort de collaboration multisite pour &eacute;tudier les classificateurs d&#39;imagerie pour le TDAH. Nous pr&eacute;sentons notre approche ML, servant des donn&eacute;es structurelles et fonctionnelles l&#39;imagerie par r&eacute;sonance magn&eacute;tique, combin&eacute;es avec des donn&eacute;es d&eacute;mographiques, pour pr&eacute;dire le diagnostic &eacute;tat des individus atteints de TDAH des pays en d&eacute;veloppement en g&eacute;n&eacute;ral d&#39;enfants (TD) dans huit sites de recherche diff&eacute;rents. Caract&eacute;ristiques structurales incluent des mesures quantitatives de 113 r&eacute;gions corticales et non-corticales. Caract&eacute;ristiques fonctionnelles incluent matrices de connectivit&eacute; fonctionnelle de corr&eacute;lation Pearson, mesures th&eacute;oriques graphique nodale et global, spectres de puissance nodal, connectivit&eacute; mondiale voxelwise et homog&eacute;n&eacute;it&eacute; r&eacute;gionale voxelwise. Nous avons effectu&eacute; le classement de fonctionnalit&eacute; pour chaque site et modalit&eacute; utilisant la multiples support vector machine r&eacute;cursive disposent d&#39;algorithme d&#39;&eacute;limination (SVM-RFE) et s&eacute;lection de sous-ensemble en vedette en optimisant la performance attendue de g&eacute;n&eacute;ralisation d&#39;un noyau de fonction de base radiale SVM (RBF-SVM) form&eacute;e dans un &eacute;ventail de caract&eacute;ristiques haut de la page. Les RBF-SVMs in sites &agrave; l&#39;aide de ces ensembles de fonctionnalit&eacute;s optimales de chaque modalit&eacute; d&#39;imagerie ont &eacute;t&eacute; utilis&eacute;es pour pr&eacute;dire les &eacute;tiquettes de classe d&#39;un ensemble de tests ind&eacute;pendants de hold-out. Une approche avec droit de vote a &eacute;t&eacute; utilis&eacute;e pour combiner ces pr&eacute;dictions multiples et d&#39;affecter des &eacute;tiquettes de classe finale. Avec cette m&eacute;thode, nous avons pu pr&eacute;dire le diagnostic du TDAH avec pr&eacute;cision de 55 % (par rapport &agrave; un niveau de 39 % de chance dans cet exemple), 33 % de sensibilit&eacute; et sp&eacute;cificit&eacute; de 80 %. Cette approche a &eacute;galement permis d&#39;&eacute;valuer des aspects structurels et fonctionnels pr&eacute;dictifs donner la perspicacit&eacute; dans les circuits c&eacute;r&eacute;braux anormaux dans le TDAH.

Aper&ccedil;u d&#39;une classification d&#39;imagerie multimodale du TDAH.

Aufmerksamkeit Aufmerksamkeitsdefizitsyndrom/Hyperaktivit&auml;t (ADHS) derzeit wird von &Auml;rzten &uuml;ber subjektive ADHS-spezifische Verhaltens-Instrumente und durch Berichte aus den Eltern und Lehrern bei Kindern diagnostiziert. In Anbetracht seiner hohen Pr&auml;valenz und gro&szlig;en wirtschaftlichen und gesellschaftlichen Kosten w&auml;re ein quantitativer Tool, das bei der Diagnose hilft, indem Charakterisierung zugrunde liegenden Neurobiologie &auml;u&szlig;erst wertvoll. Dies bot die Motivation f&uuml;r den ADHS-200 Maschine lernen (ML) Wettbewerb, einer multizentrischen Gemeinschaftsarbeit bildgebende Klassifizierungen f&uuml;r ADHS untersuchen. Hier pr&auml;sentieren wir unser ML-Ansatz, die strukturelle und funktionelle Kernspintomographie Daten, kombiniert mit demographischen Informationen verwendet, um diagnostische Status von Personen mit ADHS Vorhersagen von Entwicklungs-in der Regel (TD) Kinder in acht verschiedenen Recherche-Websites. Strukturellen Merkmale enthalten quantitative Metriken aus 113 kortikale und nicht-kortikale Regionen. Funktionale Features enthalten Pearson Korrelation funktionelle Konnektivit&auml;t Matrizen, nodal und global Graph theoretische Ma&szlig;nahmen nodal Leistungsspektren, Voxelwise globale Konnektivit&auml;t und Voxelwise regionale Homogenit&auml;t. Feature Rangliste f&uuml;hrten wir f&uuml;r jeden Standort und Modalit&auml;t mit mehreren Unterst&uuml;tzung Vektor Maschine rekursiven Funktion Beseitigung (SVM-RFE) Algorithmus und Featureauswahl Teilmenge von Leistungsoptimierung erwarteten Verallgemeinerung des radialen Grundlage Funktion Kernes SVM (RBF-SVM) in einer Reihe der wichtigsten Funktionen ausgebildet. Standortspezifische RBF-SVMs &uuml;ber diese optimale Feature-Sets aus jeder bildgebenden Modalit&auml;t wurden verwendet, um die Klasse Etiketten eines unabh&auml;ngigen Hold-Out-Test-Sets vorherzusagen. Ein Abstimmung Ansatz wurde verwendet, um diese mehrere Vorhersagen zu kombinieren und weisen Endklasse Etiketten. Mit dieser Methodik konnten wir die Diagnose von ADHS mit 55 % Genauigkeit (im Gegensatz zu einer 39 % Chance-Ebene in diesem Beispiel), 33 % Sensitivit&auml;t und Spezifit&auml;t von 80 % Vorhersagen. Dieser Ansatz konnten auch vorausschauende strukturelle und funktionelle Merkmale geben Einblick in die anormale Gehirn Schaltungen in ADHS zu bewerten.

Einblicke in multimodale Bildgebung Einstufung von ADHS.

Attenzione deficit e iperattivit&agrave; (ADHD) attualmente &egrave; diagnosticata nei bambini dai medici via soggettivi strumenti comportamentali specifici dell&#39;ADHD e dai rapporti da genitori e insegnanti. Considerando l&#39;elevata prevalenza e grandi costi economici e sociali, uno strumento quantitativo che aiuta nella diagnosi di neurobiologia sottostante la caratterizzazione sarebbe estremamente prezioso. Questo fornito motivazione per il concorso di apprendimento (ML) macchina di ADHD-200, uno sforzo collaborativo multisito per indagare imaging classificatori per l&#39;ADHD. Qui vi presentiamo il nostro approccio ML, che utilizzati dati di risonanza magnetica funzionale e strutturale, combinati con informazioni demografiche, prevedere status diagnostico di individui con ADHD in via di sviluppo in genere i bambini (TD) in otto siti di ricerca differenti. Caratteristiche strutturali incluso metriche quantitative da 113 regioni corticali e non corticale. Caratteristiche funzionali includevano matrici di correlazione connettivit&agrave; funzionale Pearson, misure teoriche nodali e globale grafico, spettri di potenza nodali, voxelwise connettivit&agrave; globale e omogeneit&agrave; regionale voxelwise. Abbiamo eseguito classifica caratteristica per ogni sito e modalit&agrave; utilizzando il pi&ugrave; supporto vettoriale macchina ricorsiva dispongono di algoritmo di eliminazione (SVM-RFE) e dispongono di selezione subset ottimizzando le prestazioni di generalizzazione atteso di un kernel di funzione radiale di base SVM (RBF-SVM) addestrate in una vasta gamma delle caratteristiche top. Site-specific RBF-SVMs utilizzando questi insiemi di funzionalit&agrave; ottimale da ogni modalit&agrave; di imaging sono stati utilizzati per predire le etichette di classe di un set di test di hold-out indipendenti. Un approccio di voto &egrave; stato utilizzato per combinare queste previsioni multiple e assegnare etichette di classe finale. Con questa metodologia, siamo stati in grado di predire la diagnosi di ADHD con precisione del 55% (contro un livello di probabilit&agrave; del 39% in questo esempio), 33% sensibilit&agrave; e specificit&agrave; 80%. Questo approccio ci ha anche permesso a valutare predittive caratteristiche strutturali e funzionali dando insight in circuiti cerebrali anomale nell&#39;ADHD.

Intuizioni nella classificazione imaging multimodale dell&#39;ADHD.

&#x6CE8;&#x610F;&#x6B20;&#x9665;&#x591A;&#x52D5;&#x6027;&#x969C;&#x5BB3; (ADHD) &#x3092;&#x73FE;&#x5728;&#x5B50;&#x4F9B;&#x306E;&#x4E21;&#x89AA;&#x3084;&#x6559;&#x5E2B;&#x304B;&#x3089;&#x306E;&#x30EC;&#x30DD;&#x30FC;&#x30C8;&#x3067;&#x81E8;&#x5E8A;&#x533B;&#x4E3B;&#x89B3;&#x7684; ADHD &#x56FA;&#x6709;&#x884C;&#x52D5;&#x697D;&#x5668;&#x3092;&#x4ECB;&#x3057;&#x3066;&#x306B;&#x3088;&#x308A;&#x8A3A;&#x65AD;&#x3067;&#x3059;&#x3002;&#x305D;&#x306E;&#x9AD8;&#x6709;&#x75C5;&#x7387;&#x3068;&#x5927;&#x898F;&#x6A21;&#x306A;&#x7D4C;&#x6E08;&#x3068;&#x793E;&#x4F1A;&#x306E;&#x30B3;&#x30B9;&#x30C8;&#x3092;&#x8003;&#x616E;&#x3057;, &#x8A3A;&#x65AD;&#x306B;&#x30A8;&#x30A4;&#x30BA;&#x306E;&#x57FA;&#x306B;&#x306A;&#x308B;&#x795E;&#x7D4C;&#x3092;&#x7279;&#x5FB4;&#x3065;&#x3051;&#x308B;&#x5B9A;&#x91CF;&#x7684;&#x30C4;&#x30FC;&#x30EB;&#x306F;&#x975E;&#x5E38;&#x306B;&#x8CB4;&#x91CD;&#x3067;&#x3057;&#x3087;&#x3046;&#x3002;&#x3053;&#x308C;&#x306F;&#x3001;ADHD 200 &#x6A5F; (ML) &#x5B66;&#x7FD2;&#x7AF6;&#x4E89;&#x610F;&#x8B58;&#x3001;&#x30A4;&#x30E1;&#x30FC;&#x30B8;&#x30F3;&#x30B0;&#x306E;&#x5206;&#x985E;&#x5B50;&#x306E; ADHD &#x8ABF;&#x67FB;&#x3059;&#x308B;&#x30DE;&#x30EB;&#x30C1;&#x30B5;&#x30A4;&#x30C8;&#x306E;&#x5171;&#x540C;&#x306E;&#x52AA;&#x529B;&#x3092;&#x63D0;&#x4F9B;&#x3057;&#x307E;&#x3057;&#x305F;&#x3002;&#x3053;&#x3053;&#x3067;&#x306F;&#x901A;&#x5E38; (TD) &#x5B50;&#x4F9B; 8 &#x3064;&#x306E;&#x7570;&#x306A;&#x308B;&#x7814;&#x7A76;&#x30B5;&#x30A4;&#x30C8;&#x9593;&#x3067;&#x306E;&#x958B;&#x767A;&#x304B;&#x3089;&#x3001;ADHD &#x3092;&#x6301;&#x3064;&#x500B;&#x4EBA;&#x306E;&#x8A3A;&#x65AD;&#x72B6;&#x614B;&#x3092;&#x4E88;&#x6E2C;&#x3059;&#x308B;&#x69CB;&#x9020;&#x7684;&#x30FB;&#x6A5F;&#x80FD;&#x7684;&#x78C1;&#x6C17;&#x5171;&#x9CF4;&#x753B;&#x50CF;&#x30C7;&#x30FC;&#x30BF;&#x3001;&#x4EBA;&#x53E3;&#x7D71;&#x8A08;&#x5B66;&#x7684;&#x60C5;&#x5831;&#x3068;&#x306E;&#x7D44;&#x307F;&#x5408;&#x308F;&#x305B;&#x3092;&#x4F7F;&#x7528;&#x79C1;&#x305F;&#x3061; ML &#x30A2;&#x30D7;&#x30ED;&#x30FC;&#x30C1;&#x3092;&#x63D0;&#x793A;&#x3057;&#x307E;&#x3059;&#x3002;&#x69CB;&#x9020;&#x6A5F;&#x80FD;&#x306B;&#x306F; 113 &#x76AE;&#x8CEA;&#x304A;&#x3088;&#x3073;&#x975E;&#x76AE;&#x8CEA;&#x9818;&#x57DF;&#x304B;&#x3089;&#x5B9A;&#x91CF;&#x7684;&#x306A;&#x6307;&#x6A19;&#x304C;&#x542B;&#x307E;&#x308C;&#x3066;&#x3002;&#x6A5F;&#x80FD;&#x306E;&#x6A5F;&#x80FD;&#x3067;&#x306F;&#x3001;&#x30D4;&#x30A2;&#x30BD;&#x30F3;&#x76F8;&#x95A2;&#x6A5F;&#x80FD;&#x7684;&#x7D50;&#x5408;&#x884C;&#x5217;, &#x7BC0;&#x3068;&#x30B0;&#x30ED;&#x30FC;&#x30D0;&#x30EB; &#x30B0;&#x30E9;&#x30D5;&#x7406;&#x8AD6;&#x7684;&#x63AA;&#x7F6E;&#x3001;&#x30EA;&#x30F3;&#x30D1;&#x7BC0;&#x306E;&#x30D1;&#x30EF;&#x30FC; &#x30B9;&#x30DA;&#x30AF;&#x30C8;&#x30EB;&#x3001;voxelwise &#x30B0;&#x30ED;&#x30FC;&#x30D0;&#x30EB;&#x63A5;&#x7D9A;&#x3001;&#x304A;&#x3088;&#x3073; voxelwise &#x5730;&#x57DF;&#x306E;&#x5747;&#x8CEA;&#x6027;&#x3092;&#x542B;&#x307E;&#x308C;&#x3066;&#x3044;&#x307E;&#x3059;&#x3002;&#x79C1;&#x305F;&#x3061;&#x305D;&#x308C;&#x305E;&#x308C;&#x306E;&#x30B5;&#x30A4;&#x30C8;&#x306E;&#x30E9;&#x30F3;&#x30AD;&#x30F3;&#x30B0;&#x6A5F;&#x80FD;&#x3092;&#x5B9F;&#x884C;&#x3057;&#x3066;&#x30C8;&#x30C3;&#x30D7;&#x306E;&#x6A5F;&#x80FD;&#x306E;&#x7BC4;&#x56F2;&#x306B;&#x308F;&#x305F;&#x3063;&#x3066;&#x8A13;&#x7DF4;&#x52D5;&#x5F84;&#x57FA;&#x5E95;&#x95A2;&#x6570;&#x30AB;&#x30FC;&#x30CD;&#x30EB; SVM (RBF SVM) &#x306E;&#x6C4E;&#x5316;&#x6027;&#x80FD;&#x3092;&#x6700;&#x9069;&#x5316;&#x3059;&#x308B;&#x3053;&#x3068;&#x306B;&#x3088;&#x308A;&#x7279;&#x5FB4;&#x30B5;&#x30D6;&#x30BB;&#x30C3;&#x30C8;&#x9078;&#x629E;&#x8907;&#x6570;&#x306E;&#x30B5;&#x30DD;&#x30FC;&#x30C8; &#x30D9;&#x30AF;&#x30C8;&#x30EB; &#x30DE;&#x30B7;&#x30F3;&#x306E;&#x518D;&#x5E30;&#x3092;&#x4F7F;&#x7528;&#x3057;&#x3066;&#x30E2;&#x30C0;&#x30EA;&#x30C6;&#x30A3;&#x304C;&#x9664;&#x53BB; (SVM RFE) &#x30A2;&#x30EB;&#x30B4;&#x30EA;&#x30BA;&#x30E0;&#x3092;&#x5099;&#x3048;&#x3066;&#x3044;&#x307E;&#x3059;&#x3002;&#x3053;&#x308C;&#x3089;&#x5404;&#x30A4;&#x30E1;&#x30FC;&#x30B8;&#x6295;&#x5C04;&#x69D8;&#x76F8;&#x304B;&#x3089;&#x6700;&#x9069;&#x306A;&#x6A5F;&#x80FD;&#x30BB;&#x30C3;&#x30C8;&#x3092;&#x4F7F;&#x7528;&#x3057;&#x3066;&#x30B5;&#x30A4;&#x30C8;&#x56FA;&#x6709; RBF Svm &#x30AF;&#x30E9;&#x30B9; &#x30E9;&#x30D9;&#x30EB;&#x306E;&#x72EC;&#x7ACB;&#x3057;&#x305F;&#x30DB;&#x30FC;&#x30EB;&#x30C9; &#x30A2;&#x30A6;&#x30C8; &#x30C6;&#x30B9;&#x30C8; &#x30BB;&#x30C3;&#x30C8;&#x3092;&#x4E88;&#x6E2C;&#x3059;&#x308B;&#x305F;&#x3081;&#x306B;&#x4F7F;&#x7528;&#x3055;&#x308C;&#x307E;&#x3057;&#x305F;&#x3002;&#x6295;&#x7968;&#x306E;&#x30A2;&#x30D7;&#x30ED;&#x30FC;&#x30C1;&#x306F;&#x3001;&#x3053;&#x308C;&#x3089;&#x8907;&#x6570;&#x306E;&#x4E88;&#x6E2C;&#x3092;&#x7D44;&#x307F;&#x5408;&#x308F;&#x305B;&#x3066;&#x3001;&#x6700;&#x7D42;&#x7684;&#x306A;&#x30AF;&#x30E9;&#x30B9;&#x306E;&#x30E9;&#x30D9;&#x30EB;&#x3092;&#x5272;&#x308A;&#x5F53;&#x3066;&#x308B;&#x306B;&#x4F7F;&#x7528;&#x3055;&#x308C;&#x307E;&#x3057;&#x305F;&#x3002;&#x3053;&#x306E;&#x624B;&#x6CD5;&#x3067;&#x306F; 55% &#x306E;&#x7CBE;&#x5EA6; &#xFF08;&#x5BFE; 39 &#xFF05; &#x306E;&#x78BA;&#x7387;&#x3067;&#x30EC;&#x30D9;&#x30EB;&#x3053;&#x306E;&#x30B5;&#x30F3;&#x30D7;&#x30EB;&#x3067;&#x306F;)&#x3001;&#x611F;&#x5EA6; 33% &#x3068; 80% &#x7279;&#x7570;&#x6027;&#x3068; ADHD &#x306E;&#x8A3A;&#x65AD;&#x3092;&#x4E88;&#x6E2C;&#x3059;&#x308B;&#x3053;&#x3068;&#x304C;&#x3067;&#x304D;&#x305F;&#x3002;&#x3053;&#x306E;&#x65B9;&#x6CD5;&#x3092; ADHD &#x306E;&#x7570;&#x5E38;&#x306A;&#x8133;&#x56DE;&#x8DEF;&#x3078;&#x306E;&#x6D1E;&#x5BDF;&#x529B;&#x3092;&#x4E0E;&#x3048;&#x3066;&#x3001;&#x4E88;&#x6E2C;&#x306E;&#x69CB;&#x9020;&#x3068;&#x6A5F;&#x80FD;&#x3092;&#x8A55;&#x4FA1;&#x3059;&#x308B;&#x3053;&#x3068;&#x304C;&#x3067;&#x304D;&#x307E;&#x3057;&#x305F;&#x3002;

ADHD &#x306E;&#x30DE;&#x30EB;&#x30C1; &#x30E2;&#x30FC;&#x30C0;&#x30EB; &#x30A4;&#x30E1;&#x30FC;&#x30B8;&#x30F3;&#x30B0;&#x5206;&#x985E;&#x3078;&#x306E;&#x6D1E;&#x5BDF;&#x529B;&#x3002;

&#xC8FC;&#xC758;&#xB825; &#xACB0;&#xD54D; &#xACFC;&#xC789; &#xD589;&#xB3D9; &#xC7A5;&#xC560; (ADHD) &#xD604;&#xC7AC;&#xB294; &#xC8FC;&#xAD00;&#xC801;&#xC778; ADHD &#xAD00;&#xB828; &#xD589;&#xB3D9; &#xC545;&#xAE30;&#xB97C; &#xD1B5;&#xD574; &#xC784;&#xC0C1; &#xD558; &#xACE0; &#xBD80;&#xBAA8;&#xC640; &#xAD50;&#xC0AC; &#xB85C;&#xBD80;&#xD130; &#xBCF4;&#xACE0;&#xC11C; &#xC5EC; &#xC5B4;&#xB9B0;&#xC774; &#xC9C4;&#xB2E8; &#xC774;&#xB2E4;. &#xB192;&#xC740; &#xC815;&#xB3C4;&#xC640; &#xD070; &#xACBD;&#xC81C;&#xC801;, &#xC0AC;&#xD68C;&#xC801; &#xBE44;&#xC6A9;&#xC744; &#xACE0;&#xB824; &#xD558;&#xBA74; &#xC591;&#xC801; &#xB3C4;&#xAD6C; &#xAE30;&#xBCF8; &#xC2E0;&#xACBD; &#xC0DD;&#xBB3C;&#xD559;&#xC744; &#xD2B9;&#xC131;&#xD654; &#xD558; &#xC5EC; &#xC9C4;&#xB2E8;&#xC5D0; &#xB3C4;&#xC6C0;&#xC744; &#xC8FC;&#xB294; &#xB9E4;&#xC6B0; &#xADC0;&#xC911; &#xD55C; &#xAC83;&#xC785;&#xB2C8;&#xB2E4;. ADHD&#xC5D0; &#xB300; &#xD55C; &#xC774;&#xBBF8;&#xC9D5; &#xBD84;&#xB958;&#xC790;&#xB97C; &#xC870;&#xC0AC; &#xD558;&#xAE30; &#xC704;&#xD574; &#xACF5;&#xB3D9; &#xB178;&#xB825;&#xC744; multisite ADHD 200 &#xAE30;&#xACC4; &#xD559;&#xC2B5; (ML) &#xACBD;&#xC7C1;&#xC5D0; &#xB300; &#xD55C; &#xB3D9;&#xAE30; &#xBD80;&#xC5EC; &#xC81C;&#xACF5;. &#xC5EC;&#xAE30;&#xC5D0; &#xC6B0;&#xB9AC;&#xAC00; &#xD604;&#xC7AC; &#xC6B0;&#xB9AC;&#xC758; ML &#xC811;&#xADFC; &#xAD6C;&#xC870; &#xBC0F; &#xAE30;&#xB2A5;&#xC801; &#xC790;&#xAE30; &#xACF5;&#xBA85; &#xC601;&#xC0C1; &#xB370;&#xC774;&#xD130;, &#xC778;&#xAD6C; &#xD1B5;&#xACC4; &#xD559;&#xC801; &#xC815;&#xBCF4;&#xB97C; &#xD568;&#xAED8; &#xC0AC;&#xC6A9; &#xD558;&#xB294; &#xC77C;&#xBC18;&#xC801;&#xC73C;&#xB85C; 8 &#xAC1C;&#xC758; &#xB2E4;&#xB978; &#xC5F0;&#xAD6C; &#xC0AC;&#xC774;&#xD2B8; &#xC804;&#xCCB4; (TD) &#xC5B4;&#xB9B0;&#xC774; &#xAC1C;&#xBC1C;&#xC5D0;&#xC11C; ADHD &#xAC00;&#xC9C4; &#xAC1C;&#xC778;&#xC758; &#xC9C4;&#xB2E8; &#xC0C1;&#xD0DC;&#xB97C; &#xC608;&#xCE21; &#xD558;&#xB294;. &#xAD6C;&#xC870; &#xAE30;&#xB2A5; 113 &#xB300;&#xB1CC; &#xD53C; &#xC9C8;&#xACFC; &#xB300;&#xB1CC; &#xD53C; &#xC9C8;&#xC758; &#xBE44; &#xC601;&#xC5ED;&#xC5D0;&#xC11C; &#xC815;&#xB7C9;&#xC801; &#xD1B5;&#xACC4;&#xB97C; &#xD3EC;&#xD568;. &#xAE30;&#xB2A5; &#xAE30;&#xB2A5; &#xD3EC;&#xD568; &#xD53C;&#xC5B4;&#xC2A8; &#xC0C1;&#xAD00; &#xAD00;&#xACC4; &#xAE30;&#xB2A5; &#xC5F0;&#xACB0; &#xB9E4;&#xD2B8;&#xB9AD;&#xC2A4;, &#xAFB8;&#xBC85;&#xAFB8;&#xBC85; &#xC878; &#xAE30; &#xBC0F; &#xAE00;&#xB85C;&#xBC8C; &#xADF8;&#xB798;&#xD504; &#xC774;&#xB860;&#xC801;&#xC778; &#xCE21;&#xC815;, &#xAFB8;&#xBC85;&#xAFB8;&#xBC85; &#xC878; &#xAE30; &#xD30C;&#xC6CC; &#xC2A4;&#xD399;&#xD2B8;&#xB7FC;, voxelwise &#xAE00;&#xB85C;&#xBC8C; &#xC5F0;&#xACB0; &#xBC0F; voxelwise &#xC9C0;&#xC5ED; &#xB3D9;&#xC9C8;&#xC131;. &#xC6B0;&#xB9AC;&#xB294; &#xAC01; &#xC0AC;&#xC774;&#xD2B8;&#xC5D0; &#xB300; &#xD55C; &#xC21C;&#xC704; &#xAE30;&#xB2A5;&#xC744; &#xC218;&#xD589; &#xD558; &#xACE0; &#xC5EC;&#xB7EC; &#xC9C0;&#xC6D0; &#xBCA1;&#xD130; &#xAE30;&#xACC4; &#xC7AC;&#xADC0;&#xB97C; &#xC0AC;&#xC6A9; &#xD558; &#xC5EC; &#xBAA8;&#xB2EC; &#xAE30;&#xB2A5; &#xC81C;&#xAC70; (SVM RFE) &#xC54C;&#xACE0;&#xB9AC;&#xC998; &#xBC0F; &#xC0C1;&#xC704; &#xAE30;&#xB2A5; &#xBC94;&#xC704;&#xC5D0; &#xAC78;&#xCCD0; &#xD6C8;&#xB828;&#xC744; &#xBC29;&#xC0AC;&#xD615;&#xC73C;&#xB85C; &#xD568;&#xC218; &#xCEE4;&#xB110; SVM (RBF SVM)&#xC758; &#xC608;&#xC0C1;&#xB41C; &#xC77C;&#xBC18;&#xD654; &#xC131;&#xB2A5;&#xC744; &#xCD5C;&#xC801;&#xD654; &#xD558; &#xC5EC; &#xAE30;&#xB2A5; &#xD558;&#xC704; &#xC9D1;&#xD569; &#xC120;&#xD0DD;. &#xC0AC;&#xC774;&#xD2B8;&#xBCC4; RBF SVMs &#xC774;&#xBBF8;&#xC9D5; &#xAC01; &#xC591;&#xC2DD; &#xC801;&#xC784;&#xC5D0;&#xC11C; &#xC774;&#xB7EC;&#xD55C; &#xCD5C;&#xC801;&#xC758; &#xAE30;&#xB2A5; &#xC9D1;&#xD569;&#xC744; &#xC0AC;&#xC6A9; &#xD558; &#xC5EC; &#xC608;&#xCE21; &#xB3C5;&#xB9BD;&#xC801;&#xC778; &#xD640;&#xB4DC; &#xC544;&#xC6C3; &#xD14C;&#xC2A4;&#xD2B8; &#xC9D1;&#xD569;&#xC758; &#xD074;&#xB798;&#xC2A4; &#xB808;&#xC774;&#xBE14;&#xC744; &#xC0AC;&#xC6A9; &#xD588;&#xB2E4;. &#xD22C;&#xD45C; &#xC811;&#xADFC; &#xC774;&#xB7EC;&#xD55C; &#xC5EC;&#xB7EC; &#xC608;&#xCE21;&#xC744; &#xACB0;&#xD569; &#xD558; &#xACE0; &#xCD5C;&#xC885; &#xD074;&#xB798;&#xC2A4; &#xB77C;&#xBCA8;&#xC744; &#xD560;&#xB2F9; &#xD558;&#xB294; &#xB370; &#xC0AC;&#xC6A9;&#xD55C;. &#xC774; &#xBC29;&#xBC95;&#xB860; ADHD&#xC758; &#xC9C4;&#xB2E8; &#xC815;&#xD655;&#xB3C4; (&#xC774; &#xC0D8;&#xD50C;&#xC758; 39% &#xC218;&#xC900;) &#xB300; 55%, 33% &#xAC10;&#xB3C4;&#xC640; 80% &#xD2B9;&#xC774;&#xC131; &#xC608;&#xCE21;&#xD560; &#xC218; &#xC788;&#xC5C8;&#xC2B5;&#xB2C8;&#xB2E4;. &#xC774; &#xBC29;&#xBC95;&#xC740; &#xB610;&#xD55C; ADHD&#xC5D0; &#xBE44;&#xC815;&#xC0C1;&#xC801;&#xC778; &#xB1CC; &#xD68C;&#xB85C;&#xC5D0; &#xB300; &#xD55C; &#xD1B5;&#xCC30;&#xB825;&#xC744; &#xC8FC;&#xB294; &#xC608;&#xCE21; &#xAD6C;&#xC870; &#xBC0F; &#xAE30;&#xB2A5; &#xAE30;&#xB2A5;&#xC744; &#xD3C9;&#xAC00;&#xD560; &#xC218; &#xC788;&#xC5C8;&#xC2B5;&#xB2C8;&#xB2E4;.

ADHD&#xC758; &#xBCF5;&#xD569; &#xC601;&#xC0C1; &#xBD84;&#xB958;&#xC5D0; &#xB300; &#xD55C; &#xD1B5;&#xCC30;&#xB825;&#xC785;&#xB2C8;&#xB2E4;.

Attention deficit hyperactivity disorder (ADHD) actualmente se diagnostica en ni&ntilde;os por los cl&iacute;nicos v&iacute;a subjetivos instrumentos de conducta TDAH espec&iacute;ficas y por los informes de los padres y maestros. Teniendo en cuenta su alta prevalencia y grandes costos econ&oacute;micos y sociales, una herramienta que ayuda en el diagn&oacute;stico por caracterizar la Neurobiolog&iacute;a subyacente ser&iacute;a muy valiosa. Esto proporciona la motivaci&oacute;n para la competici&oacute;n TDAH-200 m&aacute;quina aprendizaje (ML), un esfuerzo de colaboraci&oacute;n multisitio para investigar clasificadores de proyecci&oacute;n de imagen para el TDAH. Aqu&iacute; presentamos nuestro enfoque ML, que utiliza datos de la proyecci&oacute;n de imagen de resonancia magn&eacute;tica estructural y funcional, combinados con informaci&oacute;n demogr&aacute;fica, para predecir estado diagn&oacute;stico de individuos con TDAH suelen desarrollar ni&ntilde;os (TD) en ocho sitios de investigaci&oacute;n. Caracter&iacute;sticas estructurales incluyeron mediciones cuantitativas de 113 regiones corticales y no-corticales. Caracter&iacute;sticas funcionales incluyen matrices de conectividad funcional de correlaci&oacute;n Pearson, medidas te&oacute;rica nodal y global gr&aacute;fico, espectros de energ&iacute;a nodal, conectividad global voxelwise y homogeneidad regional voxelwise. Realizamos la clasificaci&oacute;n caracter&iacute;stica para cada sitio y modalidad utilizando la m&uacute;ltiples apoyo vector m&aacute;quina recursiva cuentan con algoritmo de eliminaci&oacute;n (SVM-RFE) y cuentan con selecci&oacute;n subconjunto optimizando el rendimiento esperado de la generalizaci&oacute;n de un n&uacute;cleo de funci&oacute;n de base radial SVM (RBF-SVM) entrenado en una amplia gama de las caracter&iacute;sticas principales. Site-specific RBF-MVS con estos conjuntos de caracter&iacute;sticas &oacute;ptimas de cada modalidad de proyecci&oacute;n de imagen fueron usados para predecir las etiquetas de clase de un conjunto de pruebas independientes de sujeci&oacute;n. Un acercamiento de la votaci&oacute;n fue utilizado para combinar estas predicciones m&uacute;ltiples y asignar etiquetas de clase final. Con esta metodolog&iacute;a hemos sido capaces de predecir el diagn&oacute;stico de TDAH con precisi&oacute;n de 55% (en comparaci&oacute;n con un nivel de 39% de probabilidades en este ejemplo), 33% sensibilidad y especificidad de 80%. Este enfoque tambi&eacute;n permiti&oacute; evaluar predictivas caracter&iacute;sticas estructurales y funcionales, dando informaci&oacute;n sobre circuitos cerebrales an&oacute;malas en el TDAH.

Penetraciones en la clasificaci&oacute;n de im&aacute;genes multimodal del TDAH.

&#x53D1;&#x4F5C;&#x95F4;&#x671F;&#x8111;&#x7535;&#x56FE; &#xFF08;EEG&#xFF09; &#x7531;&#x4E8E;&#x5176;&#x4F9D;&#x8D56;&#x7684;&#x75EB;&#x6837;&#x653E;&#x7535;&#x53D1;&#x751F;&#x6709;&#x4E34;&#x5E8A;&#x610F;&#x4E49;&#x9650;&#x5236;&#x5728;&#x5176;&#x654F;&#x611F;&#x6027;&#x548C;&#x7279;&#x5F02;&#x6027;&#x5BF9;&#x766B;&#x75EB;&#x7684;&#x8BCA;&#x65AD;&#x3002;&#x6211;&#x4EEC;&#x5F00;&#x53D1;&#x4E86;&#x4E00;&#x4E2A;&#x8BA1;&#x7B97;&#x673A;&#x8F85;&#x52A9;&#x8BCA;&#x65AD; (CAD) &#x5DE5;&#x5177;&#xFF0C;&#x5BF9;&#x7EDD;&#x5BF9;&#x5149;&#x8C31;&#x80FD;&#x91CF;&#x7684;&#x5E38;&#x89C4;&#x8111;&#x7535;&#x56FE;&#x8FD0;&#x8425;&#xFF0C;&#x5E76;&#x62E5;&#x6709;&#x8F83;&#x9AD8;&#x7684;&#x7075;&#x654F;&#x5EA6;&#x548C;&#x6D88;&#x6781;&#x9884;&#x6D4B;&#x503C;&#x6BD4;&#x53D1;&#x4F5C;&#x95F4;&#x75EB;&#x6837;&#x653E;&#x7535;&#x7684;&#x9274;&#x5B9A;&#x3002;&#x6211;&#x4EEC;&#x7684;&#x65B9;&#x6CD5;&#x7528;&#x4E8E;&#x591A;&#x5C42;&#x611F;&#x77E5;&#x5668;&#x5206;&#x7C7B;&#x89C6;&#x9891;&#x8111;&#x7535;&#x76D1;&#x6D4B; 156 &#x75C5;&#x4EBA;&#x3002;&#x75C5;&#x4EBA;&#x7684;&#x4EBA;&#x53E3;&#x662F;&#x7167;&#x5C04;&#x591A;&#x79CD;&#x591A;&#x6837; &#xFF1B;87 &#x88AB;&#x8BCA;&#x65AD;&#x60A3;&#x6709;&#x5C40;&#x7076;&#x6027;&#x6216;&#x5E7F;&#x4E49;&#x7684;&#x7F09;&#x83B7;&#x91CF;&#x3002;&#x5176;&#x4F59;&#x7684;&#x75C5;&#x4EBA;&#x88AB;&#x8BCA;&#x65AD;&#x4E3A;&#x975E;&#x766B;&#x75EB;&#x53D1;&#x4F5C;&#x3002;&#x7075;&#x654F;&#x5EA6;&#x662F; 92%&#xFF08;95%&#x7F6E;&#x4FE1;&#x533A;&#x95F4; [CI] 85-97%&#xFF09; &#x548C;&#x9634;&#x6027;&#x9884;&#x6D4B;&#x503C;&#x4E3A; 82%&#xFF08;95%CI 67 92%&#xFF09;&#x3002;&#x6211;&#x4EEC;&#x5C06;&#x8BA8;&#x8BBA;&#x5982;&#x4F55;&#x8FD9;&#x4E9B;&#x7814;&#x7A76;&#x7ED3;&#x679C;&#x8868;&#x660E;&#x6B64; CAD &#x53EF;&#x7528;&#x4E8E;&#x8865;&#x5145;&#x57FA;&#x4E8E;&#x4E8B;&#x4EF6;&#x7684;&#x5206;&#x6790;&#x7531;&#x53D7;&#x8FC7;&#x8BAD;&#x7EC3;&#x7684; epileptologists&#x3002;

&#x81EA;&#x52A8;&#x7684;&#x8BCA;&#x65AD;&#x7684;&#x766B;&#x75EB;&#x8111;&#x7535;&#x56FE;&#x529F;&#x7387;&#x8C31;&#x3002;

Interictal electroencephalography (EEG) has clinically meaningful limitations in its sensitivity and specificity in the diagnosis of epilepsy because of its dependence on the occurrence of epileptiform discharges. We have developed a computer-aided diagnostic (CAD) tool that operates on the absolute spectral energy of the routine EEG and has both substantially higher sensitivity and negative predictive value than the identification of interictal epileptiform discharges. Our approach used a multilayer perceptron to classify 156 patients admitted for video-EEG monitoring. The patient population was diagnostically diverse; 87 were diagnosed with either generalized or focal seizures. The remainder of the patients were diagnosed with nonepileptic seizures. The sensitivity was 92% (95% confidence interval [CI] 85-97%) and the negative predictive value was 82% (95% CI 67-92%). We discuss how these findings suggest that this CAD can be used to supplement event-based analysis by trained epileptologists.

Automated diagnosis of epilepsy using EEG power spectrum.

Intercritiques &eacute;lectroenc&eacute;phalographie (EEG) a des limitations cliniquement significatives dans sa sensibilit&eacute; et sp&eacute;cificit&eacute; dans le diagnostic de l&#39;&eacute;pilepsie en raison de sa d&eacute;pendance sur l&#39;occurrence de d&eacute;charges &eacute;pileptiformes. Nous avons d&eacute;velopp&eacute; un outil de diagnostic assist&eacute; par ordinateur (CAD) qui exploite l&#39;&eacute;nergie spectrale absolue de la routine EEG et a la fois sensiblement plus grande sensibilit&eacute; et une valeur pr&eacute;dictive n&eacute;gative que l&#39;identification des d&eacute;charges &eacute;pileptiformes intercritiques. Notre approche a utilis&eacute; un perceptron multicouche pour classer les 156 patients admis pour la surveillance de vid&eacute;o-EEG. La population de patients &eacute;tait diagnostique diverse ; 87 ont &eacute;t&eacute; diagnostiqu&eacute;s avec des convulsions g&eacute;n&eacute;ralis&eacute;es ou focales. Le reste des patients ont &eacute;t&eacute; diagnostiqu&eacute; avec des convulsions &eacute;pileptiques. La sensibilit&eacute; &eacute;tait de 92 % (intervalle de confiance de 95 % [IC] 85-97 %) et la valeur pr&eacute;dictive n&eacute;gative &eacute;tait de 82 % (95 % CI 67-92 %). Nous discutons comment ces r&eacute;sultats sugg&egrave;rent que cette CAD peut &ecirc;tre utilis&eacute; pour compl&eacute;ter l&#39;analyse bas&eacute;e sur des &eacute;v&eacute;nements par les &eacute;pileptologues form&eacute;s.

Automatis&eacute; de diagnostic de l&#39;&eacute;pilepsie en utilisant le spectre de puissance d&#39;EEG.

Interiktale Elektroenzephalographie (EEG) hat klinisch sinnvolle Einschr&auml;nkungen in seiner Sensitivit&auml;t und Spezifit&auml;t bei der Diagnose von Epilepsie aufgrund der Abh&auml;ngigkeit von dem Vorkommen des Epileptiform Entladungen. Wir haben eine computergest&uuml;tzte Diagnose (CAD)-Tool entwickelt, das auf die absolute spektrale Energie der Routine-EEG und hat wesentlich h&ouml;here Sensitivit&auml;t und negativen Vorhersagewert als die Identifizierung von interiktale Epileptiform Entladungen. Unser Ansatz verwendet eine mehrschichtige Perceptron, um 156 Patienten zugelassen f&uuml;r Video-EEG-&Uuml;berwachung zu klassifizieren. Die Patientengruppe war diagnostisch vielf&auml;ltig; 87 wurden generalisierte oder fokale Anf&auml;lle diagnostiziert. Der Rest der Patienten wurden mit nonepileptic Anf&auml;lle diagnostiziert. Die Empfindlichkeit wurde 92 % (95 % Konfidenzintervall [CI] 85-97 %) und der negative Vorhersagewert war 82 % (95 % CI 67-92 %). Wir diskutieren, wie diese Forschungsergebnisse legen nahe, diese CAD kann verwendet werden um ereignisbasierte Analyse zu erg&auml;nzen durch ausgebildete Epileptologists.

Automatisierte Diagnose der Epilepsie mit EEG-Leistungsspektrum.

Interittale elettroencefalografia (EEG) &egrave; clinicamente significative limitazioni nella sua sensibilit&agrave; e specificit&agrave; nella diagnosi di epilessia a causa della sua dipendenza sull&#39;occorrenza degli scarichi epileptiform. Abbiamo sviluppato uno strumento di diagnostica computerizzata (CAD) che opera sull&#39;assoluta energia spettrale dell&#39;EEG di routine e ha sia sostanzialmente pi&ugrave; alta sensibilit&agrave; e valore predittivo negativo rispetto l&#39;identificazione degli scarichi epileptiform interictali. Il nostro approccio utilizzato un percettrone multistrato per classificare 156 pazienti ricoverati per il monitoraggio video-EEG. La popolazione di pazienti era diagnosticamente diversificata; 87 sono stati diagnosticati con convulsioni focali o generalizzate. Il resto dei pazienti sono stato diagnosticato con convulsioni nonepileptic. La sensibilit&agrave; era 92% (95% intervallo di confidenza [CI] 85-97%) e il valore predittivo negativo era 82% (95% CI 67-92%). Discutiamo di come questi risultati suggeriscono che questo CAD pu&ograve; essere utilizzato per integrare l&#39;analisi basata su evento di epileptologists addestrato.

Automatizzato diagnosi di epilessia utilizzando lo spettro di potenza EEG.

&#x767A;&#x4F5C;&#x9593;&#x6B20;&#x671F;&#x8133;&#x6CE2; (EEG) &#x81E8;&#x5E8A;&#x7684;&#x306B;&#x6709;&#x610F;&#x7FA9;&#x306A;&#x5236;&#x9650;&#x305D;&#x306E;&#x611F;&#x5EA6;&#x3068;&#x7279;&#x7570;&#x6027;&#x3066;&#x3093;&#x304B;&#x3093;&#x306E;&#x8A3A;&#x65AD;&#x306B;&#x304A;&#x3051;&#x308B;&#x3066;&#x3093;&#x304B;&#x3093;&#x653E;&#x96FB;&#x306E;&#x767A;&#x751F;&#x306B;&#x4F9D;&#x5B58;&#x306E;&#x305F;&#x3081;&#x304C;&#x3042;&#x308A;&#x307E;&#x3059;&#x3002;&#x6211;&#x3005; &#x306F;&#x3001;&#x30EB;&#x30FC;&#x30C1;&#x30F3;&#x306E;&#x8133;&#x6CE2;&#x306E;&#x7D76;&#x5BFE;&#x30B9;&#x30DA;&#x30AF;&#x30C8;&#x30EB;&#x306E;&#x30A8;&#x30CD;&#x30EB;&#x30AE;&#x30FC;&#x3067;&#x52D5;&#x4F5C;&#x3057;&#x3001;&#x5927;&#x5E45;&#x306B;&#x9AD8;&#x3044;&#x611F;&#x5EA6;&#x3068;&#x767A;&#x4F5C;&#x9593;&#x6B20;&#x671F;&#x3066;&#x3093;&#x304B;&#x3093;&#x653E;&#x96FB;&#x306E;&#x540C;&#x5B9A;&#x3088;&#x308A;&#x306E;&#x8CA0;&#x306E;&#x4E88;&#x6E2C;&#x5024;&#x306E;&#x4E21;&#x65B9;&#x304C;&#x3042;&#x308A;&#x307E;&#x3059;&#x30B3;&#x30F3;&#x30D4;&#x30E5;&#x30FC;&#x30BF;&#x30FC;&#x652F;&#x63F4;&#x8A3A;&#x65AD; (CAD) &#x30C4;&#x30FC;&#x30EB;&#x3092;&#x958B;&#x767A;&#x3057;&#x307E;&#x3057;&#x305F;&#x3002;&#x6211;&#x3005; &#x306E;&#x30A2;&#x30D7;&#x30ED;&#x30FC;&#x30C1;&#x306F;&#x30D3;&#x30C7;&#x30AA;&#x8133;&#x6CE2;&#x30E2;&#x30CB;&#x30BF;&#x30EA;&#x30F3;&#x30B0;&#x306E;&#x305F;&#x3081;&#x5165;&#x9662; 156 &#x60A3;&#x8005;&#x3092;&#x5206;&#x985E;&#x3059;&#x308B;&#x306E;&#x306B;&#x591A;&#x5C64;&#x30D1;&#x30FC;&#x30BB;&#x30D7;&#x30C8;&#x30ED;&#x30F3;&#x3092;&#x4F7F;&#x7528;&#x3057;&#x307E;&#x3057;&#x305F;&#x3002;&#x60A3;&#x8005;&#x96C6;&#x56E3;&#x306F;&#x8A3A;&#x65AD;&#x591A;&#x69D8;&#x3060;&#x3063;&#x305F;;87 &#x4E00;&#x822C;&#x5316;&#x307E;&#x305F;&#x306F;&#x7126;&#x70B9;&#x767A;&#x4F5C;&#x3068;&#x8A3A;&#x65AD;&#x3055;&#x308C;&#x307E;&#x3057;&#x305F;&#x3002;&#x60A3;&#x8005;&#x306E;&#x6B8B;&#x308A;&#x306E;&#x90E8;&#x5206;&#x306F;&#x3001;nonepileptic &#x306E;&#x767A;&#x4F5C;&#x3067;&#x8A3A;&#x65AD;&#x3055;&#x308C;&#x307E;&#x3057;&#x305F;&#x3002;&#x611F;&#x5EA6;&#x306F; 92% (95% &#x4FE1;&#x983C;&#x533A;&#x9593; [CI] 85-97%) &#x3055;&#x308C;&#x8CA0;&#x306E;&#x4E88;&#x6E2C;&#x5024; 82% (95% CI 67 92%) &#x3060;&#x3063;&#x305F;&#x3002;&#x6211;&#x3005; &#x306F;&#x3069;&#x306E;&#x3088;&#x3046;&#x306B;&#x3053;&#x306E; CAD &#x30A4;&#x30D9;&#x30F3;&#x30C8;&#x306B;&#x57FA;&#x3065;&#x3044;&#x3066;&#x5206;&#x6790;&#x3092;&#x88DC;&#x5B8C;&#x3059;&#x308B;&#x305F;&#x3081;&#x306B;&#x8A13;&#x7DF4;&#x3092;&#x53D7;&#x3051;&#x305F; epileptologists &#x306B;&#x3088;&#x3063;&#x3066;&#x4F7F;&#x7528;&#x3053;&#x3068;&#x304C;&#x3067;&#x304D;&#x307E;&#x3059;&#x793A;&#x5506;&#x3092;&#x8B70;&#x8AD6;&#x3057;&#x307E;&#x3059;&#x3002;

&#x8133;&#x6CE2;&#x30D1;&#x30EF;&#x30FC; &#x30B9;&#x30DA;&#x30AF;&#x30C8;&#x30EB;&#x3092;&#x7528;&#x3044;&#x305F;&#x3066;&#x3093;&#x304B;&#x3093;&#x306E;&#x8A3A;&#x65AD;&#x306E;&#x81EA;&#x52D5;&#x5316;&#x3002;

Interictal electroencephalography (&#xB1CC; &#xD30C;) epileptiform &#xBC29;&#xC804;&#xC758; &#xBC1C;&#xC0DD;&#xC5D0; &#xB300; &#xD55C; &#xADF8;&#xAC83;&#xC758; &#xC758;&#xC874; &#xB54C;&#xBB38;&#xC5D0; &#xAC10;&#xB3C4;&#xC5D0; &#xC784;&#xC0C1;&#xC801;&#xC73C;&#xB85C; &#xC758;&#xBBF8; &#xC788;&#xB294; &#xC81C;&#xD55C; &#xBC0F; &#xAC04; &#xC9C8;&#xC758; &#xC9C4;&#xB2E8;&#xC5D0; &#xD2B9;&#xC774;&#xC131;&#xC744;&#xAC00;&#xC9C0;&#xACE0;. &#xC6B0;&#xB9AC;&#xB294; &#xC77C;&#xC0C1;&#xC801;&#xC778; &#xB1CC; &#xD30C;&#xC758; &#xC808;&#xB300; &#xC2A4;&#xD399;&#xD2B8;&#xB7FC; &#xC5D0;&#xB108;&#xC9C0;&#xC5D0;&#xC11C; &#xC791;&#xB3D9; &#xD558; &#xACE0; &#xC2E4;&#xC9C8;&#xC801;&#xC73C;&#xB85C; &#xB354; &#xB192;&#xC740; &#xAC10;&#xB3C4; &#xBD80;&#xC815;&#xC801;&#xC778; &#xC608;&#xCE21; &#xAC12; interictal epileptiform &#xBC29;&#xC804;&#xC758; &#xC2E0;&#xBD84; (CAD) &#xCEF4;&#xD4E8;&#xD130; &#xBCF4;&#xC870; &#xC9C4;&#xB2E8; &#xB3C4;&#xAD6C;&#xB97C; &#xAC1C;&#xBC1C; &#xD588;&#xC2B5;&#xB2C8;&#xB2E4;. &#xC6B0;&#xB9AC;&#xC758; &#xC811;&#xADFC; &#xBC29;&#xBC95; &#xBE44;&#xB514;&#xC624;-&#xB1CC; &#xD30C; &#xBAA8;&#xB2C8;&#xD130;&#xB9C1;&#xC5D0; &#xB300; &#xD55C; &#xC778;&#xC815; 156 &#xD658;&#xC790; &#xBD84;&#xB958;&#xB97C; &#xB2E4;&#xC911; &#xACC4;&#xCE35; &#xD37C;&#xC149;&#xD2B8;&#xB860;&#xC744; &#xC0AC;&#xC6A9;. &#xD658;&#xC790; &#xC778;&#xAD6C;&#xB294; &#xB2E4;&#xC591; &#xD55C; &#xC9C4;&#xB2E8;; 87 &#xC77C;&#xBC18;&#xD654; &#xB610;&#xB294; &#xCD08;&#xC810; &#xBC1C;&#xC791; &#xC9C4;&#xB2E8; &#xD588;&#xB2E4;. &#xD658;&#xC790;&#xC758; &#xB098;&#xBA38;&#xC9C0; nonepileptic &#xBC1C;&#xC791; &#xC9C4;&#xB2E8; &#xD588;&#xB2E4;. &#xAC10;&#xB3C4; 92% (95% &#xC2E0;&#xB8B0; &#xAD6C;&#xAC04; [CI] 85-97%)&#xACFC; &#xBD80;&#xC815;&#xC801;&#xC778; &#xC608;&#xCE21; &#xAC12;&#xC740; 82% (95% CI 67-92%). &#xC774;&#xB7EC;&#xD55C; &#xC5F0;&#xAD6C; &#xACB0;&#xACFC; &#xD6C8;&#xB828; &#xB41C; epileptologists&#xC5D0; &#xC758;&#xD574;&#xC774; CAD &#xC774;&#xBCA4;&#xD2B8; &#xAE30;&#xBC18; &#xBD84;&#xC11D;&#xC744; &#xBCF4;&#xC644; &#xD558;&#xAE30; &#xC704;&#xD574; &#xC0AC;&#xC6A9;&#xD560; &#xC218; &#xC788;&#xC2B5;&#xB2C8;&#xB2E4; &#xC81C;&#xC548; &#xD558;&#xB294; &#xBC29;&#xBC95;&#xC744; &#xC0B4;&#xD3B4;&#xBD05;&#xB2C8;&#xB2E4;.

&#xAC04; &#xC9C8; &#xB1CC; &#xD30C; &#xD30C;&#xC6CC; &#xC2A4;&#xD399;&#xD2B8;&#xB7FC;&#xC744; &#xC0AC;&#xC6A9; &#xD558; &#xC5EC; &#xC790;&#xB3D9;&#xD654; &#xB41C; &#xC9C4;&#xB2E8;&#xC785;&#xB2C8;&#xB2E4;.

La electroencefalograf&iacute;a (EEG) interictal tiene cl&iacute;nicamente significativas limitaciones en su sensibilidad y especificidad en el diagn&oacute;stico de epilepsia por su dependencia de la ocurrencia de descargas epileptiformes. Hemos desarrollado una herramienta de diagn&oacute;stico asistido por ordenador (CAD) que act&uacute;a sobre la energ&iacute;a espectral absoluta de la rutina de EEG con tanto substancialmente m&aacute;s alta sensibilidad y valor predictivo negativo que la identificaci&oacute;n de descargas epileptiformes interictales. Nuestro enfoque utiliza un perceptr&oacute;n multicapa para clasificar 156 pacientes ingresados para la supervisi&oacute;n del v&iacute;deo-EEG. La poblaci&oacute;n de pacientes fue de diagn&oacute;stico diversa; 87 fueron diagnosticados con convulsiones generalizadas o focales. El resto de los pacientes fueron diagnosticado con asimientos nonepileptic. La sensibilidad fue de 92% (intervalo de confianza [IC] del 95% 85-97%) y el valor predictivo negativo fue de 82% (95% CI 67-92%). Discutimos c&oacute;mo estos resultados sugieren que este CAD puede utilizarse para complementar el an&aacute;lisis basado en eventos por epileptologists entrenado.

Diagn&oacute;stico automatizado de epilepsia utilizando el espectro de energ&iacute;a de EEG.

Interictal FDG-PET (iPET) is a core tool for localizing the epileptogenic focus, potentially before structural MRI, that does not require rare and transient epileptiform discharges or seizures on EEG. The visual interpretation of iPET is challenging and requires years of epilepsy-specific expertise. We have developed an automated computer-aided diagnostic (CAD) tool that has the potential to work both independent of and synergistically with expert analysis. Our tool operates on distributed metabolic changes across the whole brain measured by iPET to both diagnose and lateralize temporal lobe epilepsy (TLE). When diagnosing left TLE (LTLE) or right TLE (RTLE) vs. non-epileptic seizures (NES), our accuracy in reproducing the results of the gold standard long term video-EEG monitoring was 82% [95% confidence interval (CI) 69-90%] or 88% (95% CI 76-94%), respectively. The classifier that both diagnosed and lateralized the disease had overall accuracy of 76% (95% CI 66-84%), where 89% (95% CI 77-96%) of patients correctly identified with epilepsy were correctly lateralized. When identifying LTLE, our CAD tool utilized metabolic changes across the entire brain. By contrast, only temporal regions and the right frontal lobe cortex, were needed to identify RTLE accurately, a finding consistent with clinical observations and indicative of a potential pathophysiological difference between RTLE and LTLE. The goal of CADs is to complement - not replace - expert analysis. In our dataset, the accuracy of manual analysis (MA) of iPET (∼80%) was similar to CAD. The square correlation between our CAD tool and MA, however, was only 30%, indicating that our CAD tool does not recreate MA. The addition of clinical information to our CAD, however, did not substantively change performance. These results suggest that automated analysis might provide clinically valuable information to focus treatment more effectively.

Computer-Aided Diagnosis and Localization of Lateralized Temporal Lobe Epilepsy Using Interictal FDG-PET.

The recent Chu et al. (2012) manuscript discusses two key findings regarding feature selection (FS): (1) data driven FS was no better than using whole brain voxel data and (2) a priori biological knowledge was effective to guide FS. Use of FS is highly relevant in neuroimaging-based machine learning, as the number of attributes can greatly exceed the number of exemplars. We strongly endorse their demonstration of both of these findings, and we provide additional important practical and theoretical arguments as to why, in their case, the data-driven FS methods they implemented did not result in improved accuracy. Further, we emphasize that the data-driven FS methods they tested performed approximately as well as the all-voxel case. We discuss why a sparse model may be favored over a complex one with similar performance. We caution readers that the findings in the Chu et al. report should not be generalized to all data-driven FS methods.

The utility of data-driven feature selection: Re: Chu et al. 2012.

The complex task of assessing the veracity of a statement is thought to activate uniquely distributed brain regions based on whether a subject believes or disbelieves a given assertion. In the current work, we present parallel machine learning methods for predicting a subject's decision response to a given propositional statement based on independent component (IC) features derived from EEG and fMRI data. Our results demonstrate that IC features outperformed features derived from event related spectral perturbations derived from any single spectral band, yet were similar to accuracy across all spectral bands combined. We compared our diagnostic IC spatial maps with our conventional general linear model (GLM) results, and found that informative ICs had significant spatial overlap with our GLM results, yet also revealed unique regions like amygdala that were not statistically significant in GLM analyses. Overall, these results suggest that ICs may yield a parsimonious feature set that can be used along with a decision tree structure for interpretation of features used in classifying complex cognitive processes such as belief and disbelief across both fMRI and EEG neuroimaging modalities.

Single trial decoding of belief decision making from EEG and fMRI data using independent components features.

In the multimodal neuroimaging framework, data on a single subject are collected from inherently different sources such as functional MRI, structural MRI, behavioral and/or phenotypic information. The information each source provides is not independent; a subset of features from each modality maps to one or more common latent dimensions, which can be interpreted using generative models. These latent dimensions, or "topics," provide a sparse summary of the generative process behind thefeatures for each individual. Topic modeling, an unsupervised generative model, has been used to map seemingly disparate features to a common domain. We use Non-Negative Matrix Factorization (NMF) to infer the latent structure of multimodal ADHD data containing fMRI, MRI, phenotypic and behavioral measurements. We compare four different NMF algorithms and find that the sparsest decomposition is also the most differentiating between ADHD and healthy patients. We identify dimensions that map to interpretable, recognizable dimensions such as motion, default mode network activity, and other such features of the input data. For example, structural and functional graph theory features related to default mode subnetworks clustered with the ADHD inattentive diagnosis. Structural measurements of the default mode network (DMN) regions such as the posterior cingulate, precuneus, and parahippocampal regions were all related to the ADHD-Inattentive diagnosis. Ventral DMN subnetworks may have more functional connections in ADHD-I, while dorsal DMN may have less. We also find that ADHD topics may be dependent upon diagnostic site, raising the possibility of the diagnostic differences across geographic locations. We assess our findings in light of theADHD-200 classification competition, and contrast our unsupervised, nominated topics with previously published supervised learning methods. Finally, we demonstrate the validity of these latent variables as biomarkers by using them for classification of ADHD in 730 patients. Cumulatively, this manuscript addresses how multimodal data in ADHD can be interpreted by latent dimensions.

Article	Year
Temporal and postural variation of 12-lead high-frequency QRS electrocardiographic signals in asymptomatic individuals. Journal of electrocardiology\| PubMed ID: 16777512	2006
Preparing for pandemic influenza and its aftermath: mental health issues considered. International journal of emergency mental health\| PubMed ID: 20437844	2009
A review of low-intensity focused ultrasound pulsation. Brain stimulation\| PubMed ID: 21777872	2011
Large Sample Group Independent Component Analysis of Functional Magnetic Resonance Imaging Using Anatomical Atlas-Based Reduction and Bootstrapped Clustering. International journal of imaging systems and technology\| PubMed ID: 22049263	2011
Insights into multimodal imaging classification of ADHD. Frontiers in systems neuroscience\| PubMed ID: 22912605	2012
Automated diagnosis of epilepsy using EEG power spectrum. Epilepsia\| PubMed ID: 22967005	2012
Computer-Aided Diagnosis and Localization of Lateralized Temporal Lobe Epilepsy Using Interictal FDG-PET. Frontiers in neurology\| PubMed ID: 23565107	2013
The utility of data-driven feature selection: Re: Chu et al. 2012. NeuroImage\| PubMed ID: 23891886	2014
Single trial decoding of belief decision making from EEG and fMRI data using independent components features. Frontiers in human neuroscience\| PubMed ID: 23914164	2013
Non-negative matrix factorization of multimodal MRI, fMRI and phenotypic data reveals differential changes in default mode subnetworks in ADHD. NeuroImage\| PubMed ID: 24361664	2013

Pamela K. Douglas

Biography

JoVE Journal Publications

Other Publications