Receive a weekly summary and discussion of the top papers of the week by leading researchers in the field.

General General

Machine Learning Improves Cardiovascular Risk Definition for Young, Asymptomatic Individuals.

In Journal of the American College of Cardiology ; h5-index 167.0

BACKGROUND : Clinical practice guidelines recommend assessment of subclinical atherosclerosis using imaging techniques in individuals with intermediate atherosclerotic cardiovascular risk according to standard risk prediction tools.

OBJECTIVES : The purpose of this study was to develop a machine-learning model based on routine, quantitative, and easily measured variables to predict the presence and extent of subclinical atherosclerosis (SA) in young, asymptomatic individuals. The risk of having SA estimated by this model could be used to refine risk estimation and optimize the use of imaging for risk assessment.

METHODS : The Elastic Net (EN) model was built to predict SA extent, defined by a combined metric of the coronary artery calcification score and 2-dimensional vascular ultrasound. The performance of the model for the prediction of SA extension and progression was compared with traditional risk scores of cardiovascular disease (CVD). An external independent cohort was used for validation.

RESULTS : EN-PESA (Progression of Early Subclinical Atherosclerosis) yielded a c-statistic of 0.88 for the prediction of generalized subclinical atherosclerosis. Moreover, EN-PESA was found to be a predictor of 3-year progression independent of the baseline extension of SA. EN-PESA assigned an intermediate to high cardiovascular risk to 40.1% (n = 1,411) of the PESA individuals, a significantly larger number than atherosclerotic CVD (n = 267) and SCORE (Systematic Coronary Risk Evaluation) (n = 507) risk scores. In total, 86.8% of the individuals with an increased risk based on EN-PESA presented signs of SA at baseline or a significant progression of SA over 3 years.

CONCLUSIONS : The EN-PESA model uses age, systolic blood pressure, and 10 commonly used blood/urine tests and dietary intake values to identify young, asymptomatic individuals with an increased risk of CVD based on their extension and progression of SA. These individuals are likely to benefit from imaging tests or pharmacological treatment. (Progression of Early Subclinical Atherosclerosis [PESA]; NCT01410318).

Sánchez-Cabo Fátima, Rossello Xavier, Fuster Valentín, Benito Fernando, Manzano Jose Pedro, Silla Juan Carlos, Fernández-Alvira Juan Miguel, Oliva Belén, Fernández-Friera Leticia, López-Melgar Beatriz, Mendiguren José María, Sanz Javier, Ordovás Jose María, Andrés Vicente, Fernández-Ortiz Antonio, Bueno Héctor, Ibáñez Borja, García-Ruiz José Manuel, Lara-Pezzi Enrique

2020-Oct-06

ASCVD, atherosclerosis, cardiovascular risk scores, machine-learning, subclinical

Cardiology Cardiology

Computerized Analysis of the Ventricular Fibrillation Waveform Allows Identification of Myocardial Infarction: A Proof-of-Concept Study for Smart Defibrillator Applications in Cardiac Arrest.

In Journal of the American Heart Association ; h5-index 70.0

Background In cardiac arrest, computerized analysis of the ventricular fibrillation (VF) waveform provides prognostic information, while its diagnostic potential is subject of study. Animal studies suggest that VF morphology is affected by prior myocardial infarction (MI), and even more by acute MI. This experimental in-human study reports on the discriminative value of VF waveform analysis to identify a prior MI. Outcomes may provide support for in-field studies on acute MI. Methods and Results We conducted a prospective registry of implantable cardioverter defibrillator recipients with defibrillation testing (2010-2014). From 12-lead surface ECG VF recordings, we calculated 10 VF waveform characteristics. First, we studied detection of prior MI with lead II, using one key VF characteristic (amplitude spectrum area [AMSA]). Subsequently, we constructed diagnostic machine learning models: model A, lead II, all VF characteristics; model B, 12-lead, AMSA only; and model C, 12-lead, all VF characteristics. Prior MI was present in 58% (119/206) of patients. The approach using the AMSA of lead II demonstrated a C-statistic of 0.61 (95% CI, 0.54-0.68). Model A performance was not significantly better: 0.66 (95% CI, 0.59-0.73), P=0.09 versus AMSA lead II. Model B yielded a higher C-statistic: 0.75 (95% CI, 0.68-0.81), P<0.001 versus AMSA lead II. Model C did not improve this further: 0.74 (95% CI, 0.67-0.80), P=0.66 versus model B. Conclusions This proof-of-concept study provides the first in-human evidence that MI detection seems feasible using VF waveform analysis. Information from multiple ECG leads rather than from multiple VF characteristics may improve diagnostic accuracy. These results require additional experimental studies and may serve as pilot data for in-field smart defibrillator studies, to try and identify acute MI in the earliest stages of cardiac arrest.

Thannhauser Jos, Nas Joris, Rebergen Dennis J, Westra Sjoerd W, Smeets Joep L R M, Van Royen Niels, Bonnes Judith L, Brouwer Marc A

2020-Oct-02

amplitude spectrum area, cardiac arrest, machine learning, myocardial infarction, ventricular fibrillation

General General

An urban commuters' OD hybrid prediction method based on big GPS data.

In Chaos (Woodbury, N.Y.)

With the quick development of mobile Internet and communication technology, the use of Global Position System (GPS)-enabled devices is rapidly increasing, which facilitates the collection of huge volumes of movement data in the form of trajectories. Trajectory data contain a lot of commuters' travel information, which offer convenience for researchers to study traffic problems and to mine urban commuters' travel information. In this paper, we represent an urban commuters' origin-destination (OD) hybrid prediction method based on big GPS data, which considers the temporal and spatial dependencies of OD volume data simultaneously. The regional division was performed based on a simple grid map, and the data for each grid can be obtained. Based on the grids, the OD pairs can be constructed and the network topology of OD pairs can be established. A graph convolutional network and a long short-term memory deep learning method were introduced to capture the temporal and spatial dependencies, respectively. In addition, an attention mechanism was used to learn the weights of input data. The numerical experiment was performed based on the GPS data in Chengdu, China, and some comparisons were made. The results demonstrated that the proposed hybrid OD prediction method was significant and the accuracy was reasonable.

Wang Yongdong, Xu Dongwei, Peng Peng, Xuan Qi, Zhang Guijun

2020-Sep

General General

Rise of nations: Why do empires expand and fall?

In Chaos (Woodbury, N.Y.)

We consider centralized networks composed of multiple satellites arranged around a few dominating super-egoistic centers. These so-called empires are organized using a divide and rule framework enforcing strong center-satellite interactions while keeping the pairwise interactions between the satellites sufficiently weak. We present a stochastic stability analysis, in which we consider these dynamical systems as stable if the centers have sufficient resources while the satellites have no value. Our model is based on a Hopfield type network that proved its significance in the field of artificial intelligence. Using this model, it is shown that the divide and rule framework provides important advantages: it allows for completely controlling the dynamics in a straight-forward way by adjusting center-satellite interactions. Moreover, it is shown that such empires should only have a single ruling center to provide sufficient stability. To survive, empires should have switching mechanisms implementing adequate behavior models by choosing appropriate local attractors in order to correctly respond to internal and external challenges. By an analogy with Bose-Einstein condensation, we show that if the noise correlations are negative for each pair of nodes, then the most stable structure with respect to noise is a globally connected network. For social systems, we show that controllability by their centers is only possible if the centers evolve slowly. Except for short periods when the state approaches a certain stable state, the development of such structures is very slow and negatively correlated with the size of the system's structure. Hence, increasing size eventually ends up in the "control trap."

Vakulenko S, Lyakhov D A, Weber A G, Lukichev D, Michels D L

2020-Sep

General General

A two-stage deep learning algorithm for talker-independent speaker separation in reverberant conditions.

In The Journal of the Acoustical Society of America

Speaker separation is a special case of speech separation, in which the mixture signal comprises two or more speakers. Many talker-independent speaker separation methods have been introduced in recent years to address this problem in anechoic conditions. To consider more realistic environments, this paper investigates talker-independent speaker separation in reverberant conditions. To effectively deal with speaker separation and speech dereverberation, extending the deep computational auditory scene analysis (CASA) approach to a two-stage system is proposed. In this method, reverberant utterances are first separated and separated utterances are then dereverberated. The proposed two-stage deep CASA system significantly outperforms a baseline one-stage deep CASA method in real reverberant conditions. The proposed system has superior separation performance at the frame level and higher accuracy in assigning separated frames to individual speakers. The proposed system successfully generalizes to an unseen speech corpus and exhibits similar performance to a talker-dependent system.

Delfarah Masood, Liu Yuzhou, Wang DeLiang

2020-Sep

General General

Identification of key discriminating variables between spinner dolphin (Stenella longirostris) whistle types.

In The Journal of the Acoustical Society of America

Descriptions of the six different spinner dolphin (Stenella longirostris) whistle types were developed from a random sample of 600 whistles collected across a 2-yr period from a Fijian spinner dolphin population. An exploratory multivariate visualization suggested an inverse relationship between delta and minimum frequency (58.6%) as well as whistle duration (18.1%) as the most discriminating variables in this dataset. All three of these variables were deemed to be significant when considered jointly in a multivariate analysis of variance (MANOVA): delta frequency (F5594 = 27.167, p < 0.0001), minimum frequency (F5594 = 14.889, p < 0.0001), and duration (F5594 = 24.303, p < 0.0001). Significant differences between at least two of the whistle types were found for all five acoustic parameters in univariate analysis of variation (ANOVA) tests. Constant and sine whistles were found to be the most distinctive whistles, whereas upsweep and downsweep whistles were the most similar. The identification of which parameters differ most markedly between whistle types and the relatively high explanatory power of this study's results provide a logical starting point for objective classification of spinner dolphin whistle types using machine learning techniques.

Simpson Samanunu D, Miller Cara E

2020-Sep