A comparison of visual and auditory EEG interfaces for robot multi-stage task control.

Shared autonomy holds promise for assistive robotics, whereby physically-impaired people can direct robots to perform various tasks for them. However, a robot that is capable of many tasks also introduces many choices for the user, such as which object or location should be the target of interaction. In the context of non-invasive brain-computer interfaces for shared autonomy-most commonly electroencephalography-based-the two most common choices are to provide either auditory or visual stimuli to the user-each with their respective pros and cons. Using the oddball paradigm, we designed comparable auditory and visual interfaces to speak/display the choices to the user, and had users complete a multi-stage robotic manipulation task involving location and object selection. Users displayed differing competencies-and preferences-for the different interfaces, highlighting the importance of considering modalities outside of vision when constructing human-robot interfaces.

Deep neural network-based classification of cardiotocograms outperformed conventional algorithms.

Cardiotocography records fetal heart rates and their temporal relationship to uterine contractions. To identify high risk fetuses, obstetricians inspect cardiotocograms (CTGs) by eye. Therefore, CTG traces are often interpreted differently among obstetricians, resulting in inappropriate interventions. However, few studies have focused on quantitative and nonbiased algorithms for CTG evaluation. In this study, we propose a newly constructed deep neural network model (CTG-net) to detect compromised fetal status. CTG-net consists of three convolutional layers that extract temporal patterns and interrelationships between fetal heart rate and uterine contraction signals. We aimed to classify the abnormal group (umbilical artery pH < 7.20 or Apgar score at 1 min < 7) and the normal group from CTG data. We evaluated the performance of the CTG-net with the F1 score and compared it with conventional algorithms, namely, support vector machine and k-means clustering, and another deep neural network model, long short-term memory. CTG-net showed the area under the receiver operating characteristic curve of 0.73 ± 0.04, which was significantly higher than that of long short-term memory. CTG-net, a quantitative and automated diagnostic aid system, enables early intervention for putatively abnormal fetuses, resulting in a reduction in the number of cases of hypoxic injury.

Simultaneous monitoring of mouse respiratory and cardiac rates through a single precordial electrode.

Normal respiratory and circulatory functions are crucial for survival. However, conventional methods of monitoring respiration, some of which use sensors inserted into the nasal cavity, may interfere with naïve respiratory rates. In this study, we conducted a single-point measurement of electrocardiograms (ECGs) from the pectoral muscles of anesthetized and waking mice and found low-frequency oscillations in the ECG baseline. Using the fast Fourier transform of simultaneously recorded respiratory signals, we demonstrated that the low-frequency oscillations corresponded to respiratory rhythms. Moreover, the baseline oscillations changed in parallel with the respiratory rhythm when the latter was altered by pharmacological manipulation. We also demonstrated that this method could be combined with in vivo whole-cell patch-clamp recordings from the hippocampus. Thus, we developed a non-invasive form of respirometry in mice. Our recording method using a simple derivation algorithm is applicable to a variety of physiological and pharmacological experiments, providing an experimental platform in studying the mechanisms underlying the interaction of the central nervous system and the peripheral functions.

[Machine Learning-based Prediction of Seizure-inducing Action as an Adverse Drug Effect].

　During the preclinical research period of drug development, animal testing is widely used to help screen out a drug's dangerous side effects. However, it remains difficult to predict side effects within the central nervous system. Here, we introduce a machine learning-based in vitro system designed to detect seizure-inducing side effects before clinical trial. We recorded local field potentials from the CA1 alveus in acute mouse neocortico-hippocampal slices that were bath-perfused with each of 14 different drugs, and at 5 different concentrations of each drug. For each of these experimental conditions, we collected seizure-like neuronal activity and merged their waveforms as one graphic image, which was further converted into a feature vector using Caffe, an open framework for deep learning. In the space of the first two principal components, the support vector machine completely separated the vectors (i.e., doses of individual drugs) that induced seizure-like events, and identified diphenhydramine, enoxacin, strychnine and theophylline as "seizure-inducing" drugs, which have indeed been reported to induce seizures in clinical situations. Thus, this artificial intelligence-based classification may provide a new platform to pre-clinically detect seizure-inducing side effects of drugs.