Background and Objectives: Atrial fibrillation (AF) is one of the most common arrhythmias clinically. Aging tends to increase the risk of AF, which also increases the burden of other comorbidities, including coronary artery disease (CAD), and even heart failure (HF). The precise detection of AF is a challenge due to its intermittence and unpredictability. A method for the accurate detection of AF is still needed. Methods: A deep learning model was used to detect atrial fibrillation. Here, a distinction was not made between AF and atrial flutter (AFL), both of which manifest as a similar pattern on an electrocardiogram (ECG). This method not only discriminated AF from normal rhythm of the heart, but also detected its onset and offset. The proposed model involved residual blocks and a Transformer encoder. Results and Conclusions: The data used for training were obtained from the CPSC2021 Challenge, and were collected using dynamic ECG devices. Tests on four public datasets validated the availability of the proposed method. The best performance for AF rhythm testing attained an accuracy of 98.67%, a sensitivity of 87.69%, and a specificity of 98.56%. In onset and offset detection, it obtained a sensitivity of 95.90% and 87.70%, respectively. The algorithm with a low FPR of 0.46% was able to reduce troubling false alarms. The model had a great capability to discriminate AF from normal rhythm and to detect its onset and offset. Noise stress tests were conducted after mixing three types of noise. We visualized the model’s features using a heatmap and illustrated its interpretability. The model focused directly on the crucial ECG waveform where showed obvious characteristics of AF.

Perceptual changes in major depressive disorder (MDD) may extend beyond emotional content and include the processing of basic stimulus features. These alterations may ultimately contribute to perceptual bias and anhedonia. To characterize blood oxygen level-dependent (BOLD) signal of perceptual processing, we investigated temporally independent fMRI signal components related to naturalistic stimulus processing in 39 patients with MDD and 36 healthy subjects. Leveraging the capability of multi-echo data to detect BOLD activity changes, we extracted physiologically meaningful group temporal components. For each component that exhibited a significant correlation with the movie content, we localized its underlying brain network and assessed MDD-associated alterations. Two components exhibited significant group differences; one was associated with auditory features (sound pressure level) and one with visual features (temporal contrast of intensity). Notably, these deficits in MDD localized primarily to higher-order processing areas, such as the dorsal prefrontal cortex and insula, rather than primary sensory cortices. For the visual feature component, additional group differences emerged in non-visual primary sensory cortices (auditory and somatosensory) as well as major hubs of the motor system. Our findings support the hypothesis that basic sensory processing deficits represent an inherent feature of MDD which may contribute to anhedonia and negative perceptual bias. These deficits are primarily confined to higher-order processing units, as well as cross-modal primary sensory cortices indicating predominant dysfunction of top-down control and multisensory integration. Therapeutic effects of interventions targeting the prefrontal cortex may be partially mediated by restoring prefrontal control not only over emotional but also sensory processing hubs.

Multi-echo echo-planar imaging (ME-EPI) acquires images at multiple echo times (TEs), enabling the differentiation of BOLD and non-BOLD fluctuations through TE-dependent analysis of transverse relaxation time and initial intensity. Decomposing ME-EPI in tensor space is a promising approach to characterize the distribution of changes across TEs (TE patterns) directly and aid the classification of components by providing information from an additional domain. In this study, the tensorial extension of independent component analysis (tensor-ICA) is used to characterize the TE patterns of neural and non-neural components in ME-EPI data. With the constraints of independent spatial maps, an ME-EPI dataset was decomposed into spatial, temporal, and TE domains to understand the TE patterns of noise or signal-related independent components. Our analysis revealed three distinct groups of components based on their TE patterns. Motion-related and other non-BOLD origin components followed decreased TE patterns. While the long-TE-peak components showed a large overlay on grey matter and signal patterns, the components that peaked at short TEs reflected noise that may be related to the vascular system, respiration, or cardiac pulsation, amongst others. Accordingly, removing short-TE peak components as part of a denoising strategy significantly improved quality control metrics and revealed clearer, more interpretable activation patterns compared to non-denoised data. To our knowledge, this work is the first application of decomposing ME-EPI in a tensor way. Our findings demonstrate that tensor-ICA is efficient in decomposing ME-EPI and characterizing the neural and non-neural TE patterns aiding in classifying components which is important for denoising fMRI data.

Brain–computer interfaces (BCIs) can be used to monitor and provide real-time feedback on brain signals, directly influencing external systems, such as virtual environments (VE), to support self-regulation. We piloted a novel immersive, first-person shooting BCI-VE during which the avatars’ movement speed was directly influenced by neural activity in the supplementary motor area (SMA). Previous analyses revealed behavioral and localized neural effects for active versus reduced contingency neurofeedback in a randomized controlled trial design. However, the modeling of neural dynamics during such complex tasks challenges traditional event-related approaches. To overcome this limitation, we employed a data-driven framework utilizing group-level independent networks derived from BOLD-specific components of the multi-echo fMRI data obtained during the BCI regulation. Individual responses were estimated through dual regression. The spatial independent components corresponded to established cognitive networks and task-specific networks related to gaming actions. Compared to reduced contingency neurofeedback, active regulation induced significantly elevated fractional amplitude of low-frequency fluctuations (fALFF) in a frontoparietal control network, and spatial reweighting of a salience/ventral attention network, with stronger expression in SMA, prefrontal cortex, inferior parietal lobule, and occipital regions. These findings underscore the distributed network engagement of BCI regulation during a behavioral task in an immersive virtual environment.