Front Comput Neurosci [JOURNAL]
Sun
200 papers
Estimation of ionic currents and compensation mechanisms from recursive piecewise assimilation of electrophysiological data.
Wells SA, Morris PG, Taylor JD … +1 more , Nogaret A
Front Comput Neurosci · 2025 · PMID 40104428 · Full text
The identification of ion channels expressed in neuronal function and neuronal dynamics is critical to understanding neurological disease. This program calls for advanced parameter estimation methods that infer ion chann...
The identification of ion channels expressed in neuronal function and neuronal dynamics is critical to understanding neurological disease. This program calls for advanced parameter estimation methods that infer ion channel properties from the electrical oscillations they induce across the cell membrane. Characterization of the expressed ion channels would allow detecting channelopathies and help devise more effective therapies for neurological and cardiac disease. Here, we describe Recursive Piecewise Data Assimilation (RPDA), as a computational method that successfully deconvolutes the ionic current waveforms of a hippocampal neuron from the assimilation of current-clamp recordings. The strength of this approach is to simultaneously estimate all ionic currents in the cell from a small but high-quality dataset. RPDA allows us to quantify collateral alterations in non-targeted ion channels that demonstrate the potential of the method as a drug toxicity counter-screen. The method is validated by estimating the selectivity and potency of known ion channel inhibitors in agreement with the standard pharmacological assay of inhibitor potency (IC50).
Editorial: Brain-inspired intelligence: the deep integration of brain science and artificial intelligence.
Yuan Y, Chen X, Liu J
Front Comput Neurosci · 2025 · PMID 40104427 · Full text
Abstract loading — click title to view on PubMed.
A new method for identifying and evaluating depressive disorders in young people based on cognitive neurocomputing: an exploratory study.
Liu J, Li K, Li S … +9 more , Liu S, Wang C, Huang S, Tu Y, Wang B, Zhang P, Luo Y, Sun G, Chen T
Front Comput Neurosci · 2025 · PMID 40070399 · Full text
BACKGROUND: Depressive disorders are one of the most common mental disorders among young people. However, there is still a lack of objective means to identify and evaluate young people with depressive disorders quickly....
BACKGROUND: Depressive disorders are one of the most common mental disorders among young people. However, there is still a lack of objective means to identify and evaluate young people with depressive disorders quickly. Cognitive impairment is one of the core characteristics of depressive disorders, which is of great value in the identification and evaluation of young people with depressive disorders. METHODS: This study proposes a new method for identifying and evaluating depressive disorders in young people based on cognitive neurocomputing. The method evaluates cognitive impairments such as reduced attention, executive dysfunction, and slowed information processing speed that may exist in the youth depressive disorder population through an independently designed digital evaluation paradigm. It also mines digital biomarkers that can effectively identify these cognitive impairments. A total of 50 young patients with depressive disorders and 47 healthy controls were included in this study to validate the method's identification and evaluation capability. RESULTS: The differences analysis results showed that the digital biomarkers of cognitive function on attention, executive function, and information processing speed extracted in this study were significantly different between young depressive disorder patients and healthy controls. Through stepwise regression analysis, four digital biomarkers of cognitive function were finally screened. The area under the curve for them to jointly distinguish patients with depressive disorders from healthy controls was 0.927. CONCLUSION: This new method rapidly characterizes and quantifies cognitive impairment in young people with depressive disorders. It provides a new way for organizations, such as schools, to quickly identify and evaluate the population of young people with depressive disorders based on human-computer interaction.
Time-domain brain: temporal mechanisms for brain functions using time-delay nets, holographic processes, radio communications, and emergent oscillatory sequences.
Baker JM, Cariani P
Front Comput Neurosci · 2025 · PMID 40041740 · Full text
Time is essential for understanding the brain. A temporal theory for realizing major brain functions (e.g., sensation, cognition, motivation, attention, memory, learning, and motor action) is proposed that uses temporal...
Time is essential for understanding the brain. A temporal theory for realizing major brain functions (e.g., sensation, cognition, motivation, attention, memory, learning, and motor action) is proposed that uses temporal codes, time-domain neural networks, correlation-based binding processes and signal dynamics. It adopts a signal-centric perspective in which neural assemblies produce circulating and propagating characteristic temporally patterned signals for each attribute (feature). Temporal precision is essential for temporal coding and processing. The characteristic spike patterns that constitute the signals enable general-purpose, multimodal, multidimensional vectorial representations of objects, events, situations, and procedures. Signals are broadcast and interact with each other in spreading activation time-delay networks to mutually reinforce, compete, and create new composite patterns. Sequences of events are directly encoded in the relative timings of event onsets. New temporal patterns are created through nonlinear multiplicative and thresholding signal interactions, such as mixing operations found in radio communications systems and wave interference patterns. The newly created patterns then become markers for bindings of specific combinations of signals and attributes (e.g., perceptual symbols, semantic pointers, and tags for cognitive nodes). Correlation operations enable both bottom-up productions of new composite signals and top-down recovery of constituent signals. Memory operates using the same principles: nonlocal, distributed, temporally coded memory traces, signal interactions and amplifications, and content-addressable access and retrieval. A short-term temporary store is based on circulating temporal spike patterns in reverberatory, spike-timing-facilitated circuits. A long-term store is based on synaptic modifications and neural resonances that select specific delay-paths to produce temporally patterned signals. Holographic principles of nonlocal representation, storage, and retrieval can be applied to temporal patterns as well as spatial patterns. These can automatically generate pattern recognition (wavefront reconstruction) capabilities, ranging from objects to concepts, for distributed associative memory applications. The evolution of proposed neural implementations of holograph-like signal processing and associative content-addressable memory mechanisms is discussed. These can be based on temporal correlations, convolutions, simple linear and nonlinear operations, wave interference patterns, and oscillatory interactions. The proposed mechanisms preserve high resolution temporal, phase, and amplitude information. These are essential for establishing high phase coherency and determining phase relationships, for binding/coupling, synchronization, and other operations. Interacting waves can sum constructively for amplification, or destructively, for suppression, or partially. Temporal precision, phase-locking, phase-dependent coding, phase-coherence, synchrony are discussed within the context of wave interference patterns and oscillatory interactions. Sequences of mixed neural oscillations are compared with a cascade of sequential mixing stages in a single-sideband carrier suppressed (SSBCS) radio communications system model. This mechanism suggests a manner by which multiple neural oscillation bands could interact to produce new emergent information-bearing oscillation bands, as well as to abolish previously generated bands. A hypothetical example illustrates how a succession of different oscillation carriers (gamma, beta, alpha, theta, and delta) could communicate and propagate (broadcast) information sequentially through a neural hierarchy of speech and language processing stages. Based on standard signal mixing principles, each stage emergently generates the next. The sequence of oscillatory bands generated in the mixing cascade model is consistent with neurophysiological observations. This sequence corresponds to stages of speech-language processing (sound/speech detection, acoustic-phonetics, phone/clusters, syllables, words/phrases, word sequences/sentences, and concepts/understanding). The oscillatory SSBCS cascade model makes specific predictions for oscillatory band frequencies that can be empirically tested. The principles postulated here may apply broadly for local and global oscillation interactions across the cortex. Sequences of oscillatory interactions can serve many functions, e.g., to regulate the flow and interaction of bottom-up, gamma-mediated and top-down, beta-mediated neural signals, to enable cross-frequency coupling. Some specific guidelines are offered as to how the general time-domain theory might be empirically tested. Neural signals need to be sampled and analyzed with high temporal resolution, without destructive windowing or filtering. Our intent is to suggest what we think is possible, and to widen both the scope of brain theory and experimental inquiry into brain mechanisms, functions, and behaviors.
Artificial intelligence in stroke risk assessment and management via retinal imaging.
Khalafi P, Morsali S, Hamidi S … +4 more , Ashayeri H, Sobhi N, Pedrammehr S, Jafarizadeh A
Front Comput Neurosci · 2025 · PMID 40034651 · Full text
Retinal imaging, used for assessing stroke-related retinal changes, is a non-invasive and cost-effective method that can be enhanced by machine learning and deep learning algorithms, showing promise in early disease dete...
Retinal imaging, used for assessing stroke-related retinal changes, is a non-invasive and cost-effective method that can be enhanced by machine learning and deep learning algorithms, showing promise in early disease detection, severity grading, and prognostic evaluation in stroke patients. This review explores the role of artificial intelligence (AI) in stroke patient care, focusing on retinal imaging integration into clinical workflows. Retinal imaging has revealed several microvascular changes, including a decrease in the central retinal artery diameter and an increase in the central retinal vein diameter, both of which are associated with lacunar stroke and intracranial hemorrhage. Additionally, microvascular changes, such as arteriovenous nicking, increased vessel tortuosity, enhanced arteriolar light reflex, decreased retinal fractals, and thinning of retinal nerve fiber layer are also reported to be associated with higher stroke risk. AI models, such as Xception and EfficientNet, have demonstrated accuracy comparable to traditional stroke risk scoring systems in predicting stroke risk. For stroke diagnosis, models like Inception, ResNet, and VGG, alongside machine learning classifiers, have shown high efficacy in distinguishing stroke patients from healthy individuals using retinal imaging. Moreover, a random forest model effectively distinguished between ischemic and hemorrhagic stroke subtypes based on retinal features, showing superior predictive performance compared to traditional clinical characteristics. Additionally, a support vector machine model has achieved high classification accuracy in assessing pial collateral status. Despite this advancements, challenges such as the lack of standardized protocols for imaging modalities, hesitance in trusting AI-generated predictions, insufficient integration of retinal imaging data with electronic health records, the need for validation across diverse populations, and ethical and regulatory concerns persist. Future efforts must focus on validating AI models across diverse populations, ensuring algorithm transparency, and addressing ethical and regulatory issues to enable broader implementation. Overcoming these barriers will be essential for translating this technology into personalized stroke care and improving patient outcomes.
Editorial: Computer vision and image synthesis for neurological applications.
Zou C, Roychowdhury S, Rasheed S … +1 more , Ali R
Front Comput Neurosci · 2025 · PMID 39995891 · Full text
Abstract loading — click title to view on PubMed.
EEG electrode setup optimization using feature extraction techniques for neonatal sleep state classification.
Siddiqa HA, Qureshi MF, Khurshid A … +5 more , Xu Y, Wang L, Abbasi SF, Chen C, Chen W
Front Comput Neurosci · 2025 · PMID 39959920 · Full text
An optimal arrangement of electrodes during data collection is essential for gaining a deeper understanding of neonatal sleep and assessing cognitive health in order to reduce technical complexity and reduce skin irritat...
An optimal arrangement of electrodes during data collection is essential for gaining a deeper understanding of neonatal sleep and assessing cognitive health in order to reduce technical complexity and reduce skin irritation risks. Using electroencephalography (EEG) data, a long-short-term memory (LSTM) classifier categorizes neonatal sleep states. An 16,803 30-second segment was collected from 64 infants between 36 and 43 weeks of age at Fudan University Children's Hospital to train and test the proposed model. To enhance the performance of an LSTM-based classification model, 94 linear and nonlinear features in the time and frequency domains with three novel features (Detrended Fluctuation Analysis (DFA), Lyapunov exponent, and multiscale fluctuation entropy) are extracted. An imbalance between classes is solved using the SMOTE technique. In addition, the most significant features are identified and prioritized using principal component analysis (PCA). In comparison to other single channels, the C3 channel has an accuracy value of 80.75% ± 0.82%, with a kappa value of 0.76. Classification accuracy for four left-side electrodes is higher (82.71% ± 0.88%) than for four right-side electrodes (81.14% ± 0.77%), while kappa values are respectively 0.78 and 0.76. Study results suggest that specific EEG channels play an important role in determining sleep stage classification, as well as suggesting optimal electrode configuration. Moreover, this research can be used to improve neonatal care by monitoring sleep, which can allow early detection of sleep disorders. As a result, this study captures information effectively using a single channel, reducing computing load and maintaining performance at the same time. With the incorporation of time and frequency-domain linear and nonlinear features into sleep staging, newborn sleep dynamics and irregularities can be better understood.
MUNet: a novel framework for accurate brain tumor segmentation combining UNet and mamba networks.
Yang L, Dong Q, Lin D … +2 more , Tian C, Lü X
Front Comput Neurosci · 2025 · PMID 39944950 · Full text
Brain tumors are one of the major health threats to humans, and their complex pathological features and anatomical structures make accurate segmentation and detection crucial. However, existing models based on Transforme...
Brain tumors are one of the major health threats to humans, and their complex pathological features and anatomical structures make accurate segmentation and detection crucial. However, existing models based on Transformers and Convolutional Neural Networks (CNNs) still have limitations in medical image processing. While Transformers are proficient in capturing global features, they suffer from high computational complexity and require large amounts of data for training. On the other hand, CNNs perform well in extracting local features but have limited performance when handling global information. To address these issues, this paper proposes a novel network framework, MUNet, which combines the advantages of UNet and Mamba, specifically designed for brain tumor segmentation. MUNet introduces the SD-SSM module, which effectively captures both global and local features of the image through selective scanning and state-space modeling, significantly improving segmentation accuracy. Additionally, we design the SD-Conv structure, which reduces feature redundancy without increasing model parameters, further enhancing computational efficiency. Finally, we propose a new loss function that combines mIoU loss, Dice loss, and Boundary loss, which improves segmentation overlap, similarity, and boundary accuracy from multiple perspectives. Experimental results show that, on the BraTS2020 dataset, MUNet achieves DSC values of 0.835, 0.915, and 0.823 for enhancing tumor (ET), whole tumor (WT), and tumor core (TC), respectively, and Hausdorff95 scores of 2.421, 3.755, and 6.437. On the BraTS2018 dataset, MUNet achieves DSC values of 0.815, 0.901, and 0.815, with Hausdorff95 scores of 4.389, 6.243, and 6.152, all outperforming existing methods and achieving significant performance improvements. Furthermore, when validated on the independent LGG dataset, MUNet demonstrated excellent generalization ability, proving its effectiveness in various medical imaging scenarios. The code is available at https://github.com/Dalin1977331/MUNet.
Motion feature extraction using magnocellular-inspired spiking neural networks for drone detection.
Zheng J, Wan Y, Yang X … +3 more , Zhong H, Du M, Wang G
Front Comput Neurosci · 2025 · PMID 39917463 · Full text
Traditional object detection methods usually underperform when locating tiny or small drones against complex backgrounds, since the appearance features of the targets and the backgrounds are highly similar. To address th...
Traditional object detection methods usually underperform when locating tiny or small drones against complex backgrounds, since the appearance features of the targets and the backgrounds are highly similar. To address this, inspired by the magnocellular motion processing mechanisms, we proposed to utilize the spatial-temporal characteristics of the flying drones based on spiking neural networks, thereby developing the Magno-Spiking Neural Network (MG-SNN) for drone detection. The MG-SNN can learn to identify potential regions of moving targets through motion saliency estimation and subsequently integrates the information into the popular object detection algorithms to design the retinal-inspired spiking neural network module for drone motion extraction and object detection architecture, which integrates motion and spatial features before object detection to enhance detection accuracy. To design and train the MG-SNN, we propose a new backpropagation method called Dynamic Threshold Multi-frame Spike Time Sequence (DT-MSTS), and establish a dataset for the training and validation of MG-SNN, effectively extracting and updating visual motion features. Experimental results in terms of drone detection performance indicate that the incorporation of MG-SNN significantly improves the accuracy of low-altitude drone detection tasks compared to popular small object detection algorithms, acting as a cheap plug-and-play module in detecting small flying targets against complex backgrounds.
Automated karyogram analysis for early detection of genetic and neurodegenerative disorders: a hybrid machine learning approach.
Tabassum S, Khan MJ, Iqbal J … +2 more , Waris A, Ijaz MA
Front Comput Neurosci · 2024 · PMID 39911161 · Full text
Anomalous chromosomes are the cause of genetic diseases such as cancer, Alzheimer's, Parkinson's, epilepsy, and autism. Karyotype analysis is the standard procedure for diagnosing genetic disorders. Identifying anomalies...
Anomalous chromosomes are the cause of genetic diseases such as cancer, Alzheimer's, Parkinson's, epilepsy, and autism. Karyotype analysis is the standard procedure for diagnosing genetic disorders. Identifying anomalies is often costly, time-consuming, heavily reliant on expert interpretation, and requires considerable manual effort. Efforts are being made to automate karyogram analysis. However, the unavailability of large datasets, particularly those including samples with chromosomal abnormalities, presents a significant challenge. The development of automated models requires extensive labeled and incredibly abnormal data to accurately identify and analyze abnormalities, which are difficult to obtain in sufficient quantities. Although the deep learning-based architecture has yielded state-of-the-art performance in medical image anomaly detection, it cannot be generalized well because of the lack of anomalous datasets. This study introduces a novel hybrid approach that combines unsupervised and supervised learning techniques to overcome the challenges of limited labeled data and scalability in chromosomal analysis. An Autoencoder-based system is initially trained with unlabeled data to identify chromosome patterns. It is fine-tuned on labeled data, followed by a classification step using a Convolutional Neural Network (CNN). A unique dataset of 234,259 chromosome images, including the training, validation, and test sets, was used. Marking a significant achievement in the scale of chromosomal analysis. The proposed hybrid system accurately detects structural anomalies in individual chromosome images, achieving 99.3% accuracy in classifying normal and abnormal chromosomes. We also used a structural similarity index measure and template matching to identify the part of the abnormal chromosome that differed from the normal one. This automated model has the potential to significantly contribute to the early detection and diagnosis of chromosome-related disorders that affect both genetic health and neurological behavior.
Global remapping emerges as the mechanism for renewal of context-dependent behavior in a reinforcement learning model.
Kappel D, Cheng S
Front Comput Neurosci · 2024 · PMID 39881840 · Full text
INTRODUCTION: The hippocampal formation exhibits complex and context-dependent activity patterns and dynamics, e.g., place cell activity during spatial navigation in rodents or remapping of place fields when the animal s...
INTRODUCTION: The hippocampal formation exhibits complex and context-dependent activity patterns and dynamics, e.g., place cell activity during spatial navigation in rodents or remapping of place fields when the animal switches between contexts. Furthermore, rodents show context-dependent renewal of extinguished behavior. However, the link between context-dependent neural codes and context-dependent renewal is not fully understood. METHODS: We use a deep neural network-based reinforcement learning agent to study the learning dynamics that occur during spatial learning and context switching in a simulated ABA extinction and renewal paradigm in a 3D virtual environment. RESULTS: Despite its simplicity, the network exhibits a number of features typically found in the CA1 and CA3 regions of the hippocampus. A significant proportion of neurons in deeper layers of the network are tuned to a specific spatial position of the agent in the environment-similar to place cells in the hippocampus. These complex spatial representations and dynamics occur spontaneously in the hidden layer of a deep network during learning. These spatial representations exhibit global remapping when the agent is exposed to a new context. The spatial maps are restored when the agent returns to the previous context, accompanied by renewal of the conditioned behavior. Remapping is facilitated by memory replay of experiences during training. DISCUSSION: Our results show that integrated codes that jointly represent spatial and task-relevant contextual variables are the mechanism underlying renewal in a simulated DQN agent.
How to be an integrated information theorist without losing your body.
Cea I, Signorelli CM
Front Comput Neurosci · 2024 · PMID 39850269 · Full text
Abstract loading — click title to view on PubMed.
Multimodal consumer choice prediction using EEG signals and eye tracking.
Usman SM, Khalid S, Tanveer A … +2 more , Imran AS, Zubair M
Front Comput Neurosci · 2024 · PMID 39845093 · Full text
Marketing plays a vital role in the success of a business, driving customer engagement, brand recognition, and revenue growth. Neuromarketing adds depth to this by employing insights into consumer behavior through brain...
Marketing plays a vital role in the success of a business, driving customer engagement, brand recognition, and revenue growth. Neuromarketing adds depth to this by employing insights into consumer behavior through brain activity and emotional responses to create more effective marketing strategies. Electroencephalogram (EEG) has typically been utilized by researchers for neuromarketing, whereas Eye Tracking (ET) has remained unexplored. To address this gap, we propose a novel multimodal approach to predict consumer choices by integrating EEG and ET data. Noise from EEG signals is mitigated using a bandpass filter, Artifact Subspace Reconstruction (ASR), and Fast Orthogonal Regression for Classification and Estimation (FORCE). Class imbalance is handled by employing the Synthetic Minority Over-sampling Technique (SMOTE). Handcrafted features, including statistical and wavelet features, and automated features from Convolutional Neural Network and Long Short-Term Memory (CNN-LSTM), have been extracted and concatenated to generate a feature space representation. For ET data, preprocessing involved interpolation, gaze plots, and SMOTE, followed by feature extraction using LeNet-5 and handcrafted features like fixations and saccades. Multimodal feature space representation was generated by performing feature-level fusion for EEG and ET, which was later fed into a meta-learner-based ensemble classifier with three base classifiers, including Random Forest, Extended Gradient Boosting, and Gradient Boosting, and Random Forest as the meta-classifier, to perform classification between buy vs. not buy. The performance of the proposed approach is evaluated using a variety of performance metrics, including accuracy, precision, recall, and F1 score. Our model demonstrated superior performance compared to competitors by achieving 84.01% accuracy in predicting consumer choices and 83% precision in identifying positive consumer preferences.
Editorial: 15 years of frontiers in computational neuroscience - computational perception and cognition.
Iannella N
Front Comput Neurosci · 2024 · PMID 39845092 · Full text
Abstract loading — click title to view on PubMed.
Memory consolidation from a reinforcement learning perspective.
Lee JW, Jung MW
Front Comput Neurosci · 2024 · PMID 39845091 · Full text
Memory consolidation refers to the process of converting temporary memories into long-lasting ones. It is widely accepted that new experiences are initially stored in the hippocampus as rapid associative memories, which...
Memory consolidation refers to the process of converting temporary memories into long-lasting ones. It is widely accepted that new experiences are initially stored in the hippocampus as rapid associative memories, which then undergo a consolidation process to establish more permanent traces in other regions of the brain. Over the past two decades, studies in humans and animals have demonstrated that the hippocampus is crucial not only for memory but also for imagination and future planning, with the CA3 region playing a pivotal role in generating novel activity patterns. Additionally, a growing body of evidence indicates the involvement of the hippocampus, especially the CA1 region, in valuation processes. Based on these findings, we propose that the CA3 region of the hippocampus generates diverse activity patterns, while the CA1 region evaluates and reinforces those patterns most likely to maximize rewards. This framework closely parallels Dyna, a reinforcement learning algorithm introduced by Sutton in 1991. In Dyna, an agent performs offline simulations to supplement trial-and-error value learning, greatly accelerating the learning process. We suggest that memory consolidation might be viewed as a process of deriving optimal strategies based on simulations derived from limited experiences, rather than merely strengthening incidental memories. From this perspective, memory consolidation functions as a form of offline reinforcement learning, aimed at enhancing adaptive decision-making.
Editorial: Advances in computer science and their impact on data acquisition and analysis in neuroscience.
Koulen P, Mehdizadeh A, Shyu ML … +2 more , Zhang C, Chen SC
Front Comput Neurosci · 2024 · PMID 39835175 · Full text
Abstract loading — click title to view on PubMed.
Learning delays through gradients and structure: emergence of spatiotemporal patterns in spiking neural networks.
Mészáros B, Knight JC, Nowotny T
Front Comput Neurosci · 2024 · PMID 39759584 · Full text
We present a Spiking Neural Network (SNN) model that incorporates learnable synaptic delays through two approaches: per-synapse delay learning via Dilated Convolutions with Learnable Spacings (DCLS) and a dynamic pruning...
We present a Spiking Neural Network (SNN) model that incorporates learnable synaptic delays through two approaches: per-synapse delay learning via Dilated Convolutions with Learnable Spacings (DCLS) and a dynamic pruning strategy that also serves as a form of delay learning. In the latter approach, the network dynamically selects and prunes connections, optimizing the delays in sparse connectivity settings. We evaluate both approaches on the Raw Heidelberg Digits keyword spotting benchmark using Backpropagation Through Time with surrogate gradients. Our analysis of the spatio-temporal structure of synaptic interactions reveals that, after training, excitation and inhibition group together in space and time. Notably, the dynamic pruning approach, which employs DEEP R for connection removal and RigL for reconnection, not only preserves these spatio-temporal patterns but outperforms per-synapse delay learning in sparse networks. Our results demonstrate the potential of combining delay learning with dynamic pruning to develop efficient SNN models for temporal data processing. Moreover, the preservation of spatio-temporal dynamics throughout pruning and rewiring highlights the robustness of these features, providing a solid foundation for future neuromorphic computing applications.
Editorial: Deep learning and neuroimage processing in understanding neurological diseases.
Carvalho J, Abdollahzadeh A, Ferrari RJ
Front Comput Neurosci · 2024 · PMID 39749286 · Full text
Abstract loading — click title to view on PubMed.
Alleviating the medical strain: a triage method via cross-domain text classification.
Xiao X, Wang S, Jiang F … +2 more , Qi T, Wang W
Front Comput Neurosci · 2024 · PMID 39744724 · Full text
It is a universal phenomenon for patients who do not know which clinical department to register in large general hospitals. Although triage nurses can help patients, due to the larger number of patients, they have to sta...
It is a universal phenomenon for patients who do not know which clinical department to register in large general hospitals. Although triage nurses can help patients, due to the larger number of patients, they have to stand in a queue for minutes to consult. Recently, there have already been some efforts to devote deep-learning techniques or pre-trained language models (PLMs) to triage recommendations. However, these methods may suffer two main limitations: (1) These methods typically require a certain amount of labeled or unlabeled data for model training, which are not always accessible and costly to acquire. (2) These methods have not taken into account the distortion of semantic feature structure and the loss of category discriminability in the model training. To overcome these limitations, in this study, we propose a cross-domain text classification method based on prompt-tuning, which can classify patients' questions or texts about their symptoms into several given categories to give suggestions on which kind of consulting room patients could choose. Specifically, first, different prompt templates are manually crafted based on various data contents, embedding source domain information into the prompt templates to generate another text with similar semantic feature structures for performing classification tasks. Then, five different strategies are employed to expand the label word space for modifying prompts, and the integration of these strategies is used as the final verbalizer. The extensive experiments on Chinese Triage datasets demonstrate that our method achieved state-of-the-art performance.
Multimodal sleep staging network based on obstructive sleep apnea.
Fan J, Zhao M, Huang L … +4 more , Tang B, Wang L, He Z, Peng X
Front Comput Neurosci · 2024 · PMID 39744723 · Full text
BACKGROUND: Automatic sleep staging is essential for assessing sleep quality and diagnosing sleep disorders. While previous research has achieved high classification performance, most current sleep staging networks have...
BACKGROUND: Automatic sleep staging is essential for assessing sleep quality and diagnosing sleep disorders. While previous research has achieved high classification performance, most current sleep staging networks have only been validated in healthy populations, ignoring the impact of Obstructive Sleep Apnea (OSA) on sleep stage classification. In addition, it remains challenging to effectively improve the fine-grained detection of polysomnography (PSG) and capture multi-scale transitions between sleep stages. Therefore, a more widely applicable network is needed for sleep staging. METHODS: This paper introduces MSDC-SSNet, a novel deep learning network for automatic sleep stage classification. MSDC-SSNet transforms two channels of electroencephalogram (EEG) and one channel of electrooculogram (EOG) signals into time-frequency representations to obtain feature sequences at different temporal and frequency scales. An improved Transformer encoder architecture ensures temporal consistency and effectively captures long-term dependencies in EEG and EOG signals. The Multi-Scale Feature Extraction Module (MFEM) employs convolutional layers with varying dilation rates to capture spatial patterns from fine to coarse granularity. It adaptively fuses the weights of features to enhance the robustness of the model. Finally, multiple channel data are integrated to address the heterogeneity between different modalities effectively and alleviate the impact of OSA on sleep stages. RESULTS: We evaluated MSDC-SSNet on three public datasets and our collection of PSG records of 17 OSA patients. It achieved an accuracy of 80.4% on the OSA dataset. It also outperformed the state-of-the-art methods in terms of accuracy, F1 score, and Cohen's Kappa coefficient on the remaining three datasets. CONCLUSION: The MSDC-SSRNet multi-channel sleep staging architecture proposed in this study enhances widespread system applicability by supplementing inter-channel features. It employs multi-scale attention to extract transition rules between sleep stages and effectively integrates multimodal information. Our method address the limitations of single-channel approaches, enhancing interpretability for clinical applications.
About
- Frequency
- Sun
- Papers found
- 200
- RSS feed
- Subscribe