On the use of pairwise distance learning for brain signal classification with limited observations

Abstract

The increasing access to brain signal data using electroencephalography creates new opportunities to study electrophysiological brain activity and perform ambulatory diagnoses of neurological disorders. This work proposes a pairwise distance learning approach for Schizophrenia classification relying on the spectral properties of the signal. Given the limited number of observations (i.e. the case and/or control individuals) in clinical trials, we propose a Siamese neural network architecture to learn a discriminative feature space from pairwise combinations of observations per channel. In this way, the multivariate order of the signal is used as a form of data augmentation, further supporting the network generalization ability. Convolutional layers with parameters learned under a cosine contrastive loss are proposed to adequately explore spectral images derived from the brain signal. With the proposed method achieving a best of 0.95 ± 0.05, 0.98 ± 0.02 and 0.92 ± 0.07 in terms of Accuracy, Sensitivity and Specificity respectively. Results on a case-control population show that the features extracted using the proposed neural network are superior than baselines to diagnose Schizophrenia, suggesting the existence of non-trivial electrophysiological brain patterns able to capture discriminative neuroplasticity profiles among individuals.

Computer Multimedia-Oriented Chinese Language and Literature Distance Teaching System

“…In formula (2), Q represents the number of itemsets, H represents candidate nodes, and α and β have the same meaning as formula (1). When the sum of the overall utility values is less than the length of the newly added itemset, formula (2) becomes the following form:…”

“…With the development of real-time communication technology, especially the rapid development of WebRTC technology, this one-way education mode will be gradually broken, so that students can learn knowledge from anyone at anytime and anywhere. Due to the rapid increase of social demand for education, the traditional education model can no longer fully meet this need, and online distance education is bound to become a very important way of knowledge dissemination [1,2]. Computer multimedia technology is playing an increasingly obvious role in the field of teaching, especially in web-based open distance teaching [3].…”

Predicting Postoperative Complications in Cancer Patients: A Survey Bridging Classical and Machine Learning Contributions to Postsurgical Risk Analysis

“…In addition, neural networks generally (1) lack interpretability, (2) are unable to provide statistical guarantees on the adequacy of decisions, and (3) depend on the availability of a considerably large cohort of studies to guarantee learning convergence. With the aim of addressing these challenges, recent contributions in the field offer the possibility to (1) extract visual representations on the underlying network patterning [84], (2) provide a Bayesian frame to neural network learning for statistical assessments [85], and (3) have pairwise learning principles for data augmentation [86].…”

From Sound Perception to Automatic Detection of Schizophrenia: An EEG-Based Deep Learning Approach

“…A limitation of these approaches relates to the process underlying feature extraction and selection, which requires domain expertise. The few studies that probed the potential of deep learning in EEG-based classification of SZ (23)(24)(25)(26)(27) achieved the best performances with CNN-based models applied to resting-state EEG data, which is independent of cognitive or sensory processing. Despite their capacity to discriminate healthy from SZ subjects, these models do not inform about auditory processing, which is affected in SZ (4).…”