Search PubMed⌕ Search

SEARCH · Search PubMed

Results for “Deep Neural Network”

Search indexed PubMed citations on genomics, clinical trials, systematic reviews and public health. Explore titles, authors and supplied subject terms, then open the PubMed record.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 127 records · Page 7Linked to original sources
PubMed2006

Reducing the dimensionality of data with neural networks.

High-dimensional data can be converted to low-dimensional codes by training a multilayer neural network with a small central layer to reconstruct high-dimensional input vectors. Gradient descent can be used for fine-tuning the weights in such "autoencoder" networks, but this works well only if the initial weights are close to a good solution. We describe an effective way of initializing the weights that allows deep autoencoder networks to learn low-dimensional codes that work much better than principal components analysis as a tool to reduce the dimensionality of data.

Journal Article↗
PubMed1999

Adaptable preprocessing units and neural classification for the segmentation of EEG signals.

In this contribution, a methodology for the simultaneous adaptation of preprocessing units (PPUs) for feature extraction and of neural classifiers that can be used for time series classification is presented. The approach is based upon an extension of the backpropagation algorithm for the correction of the preprocessing parameters. In comparison with purely neural systems, the reduced input dimensionality improves the generalization capability and reduces the numerical effort. In comparison with PPUs with fixed parameters, the success of the adaptation is less sensitive to the choice of the parameters. The efficiency of the developed method is demonstrated via the use of quadratic filters with adaptable transmission bands as preprocessing units for the segmentation of two different types of discontinuous EEG: discontinuous neonatal EEG (burst-interburst segmentation) and EEG in deep stages of sedation (burst-suppression segmentation).

Coma↗
PubMed2025

Improving spliced alignment by modeling splice sites with deep learning.

MOTIVATION: Spliced alignment refers to the alignment of messenger RNA (mRNA) or protein sequences to eukaryotic genomes. It plays a critical role in gene annotation and the study of gene functions. Accurate spliced alignment demands sophisticated modeling of splice sites, but current aligners use simple models, which may affect their accuracy given dissimilar sequences. RESULTS: We implemented minisplice to learn splice signals with a one-dimensional convolutional neural network (1D-CNN) and trained a model with 7,026 parameters for vertebrate and insect genomes. It captures conserved splice signals across phyla and reveals GC-rich introns specific to mammals and birds. We used this model to estimate the empirical splicing probability for every GT and AG in genomes, and modified minimap2 and miniprot to leverage pre-computed splicing probability during alignment. Evaluation on human long-read RNA-seq data and cross-species protein datasets showed our method greatly improves the junction accuracy especially for noisy long RNA-seq reads and proteins of distant homology. AVAILABILITY AND IMPLEMENTATION: https://github.com/lh3/minisplice.

Journal Article↗
PubMed1980

[Vessel work development phases in the neural part of the pons in chick embryos].

By means of intravascular India ink injections we have analyzed the early developmental phases of the vessel network in the cranial part of the chick embryo pons. At four and one half days of incubation the first radial endoneural vessels have already entered the ventral portion of the pons and formed with their branches a perigerminative network. Both the external and deep network progressively extend during the subsequent days, and at the 7th day the previously segregated cell groups begin to be encircled by tiny branches from the radial vessels. On the 8th day vessels appear also within the cellular groups and differences in vascular density between the various groups become detectable, apparently depending on their different cell density and differentiation degree. From this stage until the 12th day the vessel denseness undergoes a gradual, general increase, although differences in the vascularization degree between various groups are still appreciable.

Animals↗
PubMed2001

Syntax and the brain: disentangling grammar by selective anomalies.

Many paradigms employed so far with functional imaging in language studies do not allow a clear differentiation of the semantic, morphological, and syntactic components, as traditionally defined within linguistic theory. In fact, many studies simply consider the brain's response to lists of unrelated words, rather than to syntactic structures, or do not neutralize the confounding effect of the semantic component. In the present PET experiment, we isolated the functional correlates of morphological and syntactic processing. The neutralization of the access to the lexical-semantic component was achieved by requiring the detection of anomalies in written sentences consisting of pseudowords. In both syntactic and morphosyntactic processing, the involvement of a selective deep component of Broca's area and of a right inferior frontal region was detected. In addition, within this system, the left caudate nucleus and insula were activated only during syntactic processing, indicating their role in syntactic computation. These findings provide original in vivo evidence that these brain structures, whose individual contribution has been highlighted by clinical studies, constitute a neural network selectively engaged in morphological and syntactic computation.

Adult↗
PubMed2006

Effects of GPi stimulation on human thalamic neuronal activity.

OBJECTIVE: Determine the effects of globus pallidus interna (GPi) deep brain stimulation (DBS) on ventral oralis posterior nucleus of the thalamic (Vop) neuronal activity. METHODS: Microelectrode recordings in Vop during high frequency DBS GPi in a patient with dystonia. RESULTS: Twelve (48%) of 25 neurons in five locations neurons decreased their average discharge frequency, 2 (8%) increased and 11(44%) demonstrated no overall change. The patterns of responses were complex with periods of increase and decreased activity. All neurons were inhibited for the time period 3.5-5ms following the DBS pulse. Eighty-eight percent of neurons showed brief but highly consistent increases in the first 1ms following stimulation, 52% showed increased activities from 1.5 to 3ms. Twenty-four percent of neurons increased activity following inhibition. CONCLUSIONS: These findings are consistent with DBS activation of GPi axons to Vop and probable antidromic activation of Vop axons. SIGNIFICANCE: The physiological effects of DBS are far more complicated and will escape any theory that does not address the mechanisms of DBS as stimulation of a complex network of interactions. Further, the findings of post-inhibitory rebound increased raises questions about the role in inhibition in the current concepts of basal ganglia physiology.

Adult↗
PubMed2005

SPE dose prediction using locally weighted regression.

When astronauts are outside earth's protective magnetosphere, they are subject to large radiation doses resulting from solar particle events (SPEs). The total dose received from a major SPE in deep space could cause severe radiation poisoning. The dose is usually received over a 20-40 h time interval but the event's effects may be mitigated with an early warning system. This paper presents a method to predict the total dose early in the event. It uses a locally weighted regression model, which is easier to train and provides predictions as accurate as neural network models previously used.

Algorithms↗
PubMed2005

Architectures for high-performance FPGA implementations of neural models.

As the complexity of neural models continues to increase (larger populations, varied ionic conductances, more detailed morphologies, etc) traditional software-based models have difficulty scaling to reach the performance levels desired. This paper describes the use of FPGAs, or field programmable gate arrays, to easily implement a wide variety of neural models with the performance of custom analogue circuits or computer clusters, the reconfigurability of software, and at a cost rivalling personal computers. FPGAs reach this level of performance by enabling the design of neural models as parallel processed data paths. These architectures provide for a wide range of single-compartment, multi-compartment and population models to be readily converted to FPGA implementations. Generalized architectures are described for the efficient modelling of a first-order, nonlinear differential equation in throughput maximizing or latency minimizing data-path configurations. The homogeneity of population and multicompartment models is exploited to form deep pipelines for improved performance. Limitations of FPGA architectures and future research areas are explored.

Animals↗
PubMed2005

SPE dose prediction using locally weighted regression.

When astronauts are outside Earth's protective magnetosphere, they are subject to large radiation doses resulting from solar particle events. The total dose received from a major solar particle event in deep space could cause severe radiation poisoning. The dose is usually received over a 20-40 h time interval but the event's effects may be reduced with an early warning system. This paper presents a method to predict the total dose early in the event. It uses a locally weighted regression model, which is easier to train, and provides predictions as accurate as the neural network models that were used previously.

Algorithms↗
PubMed2026

Leveraging Interradiomic Feature Relationships for Enhanced Prediction of Distant Metastasis and Characterization of Heterogeneity in Head and Neck Cancer.

PURPOSE: Distant metastasis remains a major cause of treatment failure in head and neck (HN) cancer, highlighting the need for more accurate early risk stratification. This study developed and validated a deep radiomics framework to characterize tumor heterogeneity from pretreatment computed tomography (CT) images and improve prediction of distant metastasis-free survival (DMFS). METHODS AND MATERIALS: This multicenter study included 3421 patients with HN cancer from 4 cohorts across 12 institutions. Radiomics features were extracted from primary tumors and transformed into OmicsMaps, a structured representation that spatially organizes interfeature relationships to facilitate learning of complex prognostic patterns. A convolutional neural network was trained to derive prognostic signatures, which were integrated with key clinical variables to construct an OmicsMap-clinical fusion model for patient risk stratification. Model performance was assessed using the concordance index (C-index) and time-dependent area under the receiver operating characteristic curve (AUC) in the CT Images from Large Head and Neck Cohort (RADCURE), HEAD-NECK-RADIOMICS-HN1 (HN1), and Head-Neck-Positron Emission Tomography-Computed Tomography (HN-PET-CT) cohorts. Radiogenomic analyses using RNA-seq data were conducted in the Cancer Genome Atlas Head-Neck Squamous Cell Carcinoma (TCGA-HNSC) cohort to investigate biological characteristics associated with the imaging-defined risk groups. RESULTS: The OmicsMap achieved C-index values of 0.742, 0.768, and 0.671 in the RADCURE, HN1, and HN-PET-CT cohorts, outperforming the conventional radiomics approach by 5.40%-6.37%. Incorporating clinical variables further improved generalizability, yielding a C-index of 0.864 (HN1) and 0.730 (HN-PET-CT), with time-dependent AUC of 0.727-0.895. The fusion model consistently stratified patients into distinct high- and low-risk groups for both DMFS and overall survival across cohorts (P <.01). Radiogenomic analyses revealed enrichment of immune-related pathways in the low-risk group, whereas the high-risk group exhibited a more aggressive phenotype enriched for proliferation, hypoxia, and epithelial-mesenchymal transition pathways, along with a fibrosis-prone tumor microenvironment characterized by extracellular matrix remodeling. CONCLUSIONS: Modeling interradiomic feature relationships using the OmicsMap representation substantially improves CT-based prediction of DMFS and characterization of tumor heterogeneity in HN cancer, supporting precision risk stratification in clinical oncology.

Journal Article↗
PubMed2002

Dynamics of cortico-subcortical cross-modal operations involved in audio-visual object detection in humans.

Very recently, a number of neuroimaging studies in humans have begun to investigate the question of how the brain integrates information from different sensory modalities to form unified percepts. Already, intermodal neural processing appears to depend on the modalities of inputs or the nature (speech/non-speech) of information to be combined. Yet, the variety of paradigms, stimuli and technics used make it difficult to understand the relationships between the factors operating at the perceptual level and the underlying physiological processes. In a previous experiment, we used event-related potentials to describe the spatio-temporal organization of audio-visual interactions during a bimodal object recognition task. Here we examined the network of cross-modal interactions involved in simple detection of the same objects. The objects were defined either by unimodal auditory or visual features alone, or by the combination of the two features. As expected, subjects detected bimodal stimuli more rapidly than either unimodal stimuli. Combined analysis of potentials, scalp current densities and dipole modeling revealed several interaction patterns within the first 200 micro s post-stimulus: in occipito-parietal visual areas (45-85 micro s), in deep brain structures, possibly the superior colliculus (105-140 micro s), and in right temporo-frontal regions (170-185 micro s). These interactions differed from those found during object identification in sensory-specific areas and possibly in the superior colliculus, indicating that the neural operations governing multisensory integration depend crucially on the nature of the perceptual processes involved.

Acoustic Stimulation↗
PubMed1978

Pharmacological factors influencing permeability of the bronchial microcirculation.

The bronchial microcirculation is organized as a network of interconnected superficial and deep small blood vessels that extend throughout the bronchial tree. The close proximity of the bronchial venules to the origin of the pulmonary lymphatics suggests a role in fluid balance in the lung. The responsiveness of the venous segment of this network to agents influencing permeability clearly indicates a role in the genesis of certain types of noncardiogenic interstitial edema. The responsiveness of the venous segment of the bronchial circulation to neural and humoral stimuli raises the possibility of a significant role in the distribution of blood flow between the mucosa and peribronchial space. Such control may play a significant role in heat and water exchange between blood and air. Thus, these observations indicate that the bronchial microcirculation is involved with aspects of pulmonary function other than the nutrition of the bronchi, nerves, pleura, and large pulmonary blood vessels.

Animals↗
PubMed2000

Anatomical organization of the parahippocampal-hippocampal network.

The anatomical organization of the parahippocampal-hippocampal network indicates that it consists of different parallel circuits. Considering the topographical distribution of sensory cortical inputs, the hypothesis is that the major parallel circuits carry functionally different information. These functionally different parallel routes reach different portions of the hippocampal network along the longitudinal axis of all fields as well as along the perpendicularly oriented transverse axis of CA1 and the subiculum. In the remaining fields of the hippocampal formation, that is, the dentate gyrus and CA2/CA3, separation along the transverse axis is not present. By contrast, here the functionally different pathways converge onto the same neuronal population. The entorhinal cortex holds a pivotal position among the cortices that make up the parahippocampal region. By way of the networks of the superficial and deep layers, it mediates, respectively, the input and output streams of the hippocampal formation. Moreover, the intrinsic entorhinal network, particularly the interconnections between the deep and superficial layers, may mediate the comparison of hippocampal input and output signals. As such, the entorhinal cortex may form part of a novelty detection network. In addition, the organization of the entorhinal-hippocampal network may facilitate the holding of information. Finally, the terminal organization of the presubicular input to the medial entorhinal cortex indicates that the interactions between the deep and superficial entorhinal layers may be influenced by this input.

Animals↗
PubMed1994

Ultrastructure of interstitial cells of Cajal in the canine distal esophagus.

The ultrastructure of canine distal esophagus was studied focusing on interstitial cells of Cajal (ICC) and their relationships to nerves and muscle. The distal esophagus consisted of two muscle layers composed of intertwining skeletal and smooth muscle bundles. The ICC formed an interconnecting network and were an integral part of these structures. The ICC communicated with one another and with adjacent smooth muscle cells through numerous gap junctions. The morphology of individual ICC resembled that observed in other gut regions. All interstitial cells were densely innervated. The highest density of ICC, just proximal to the lower esophageal sphincter, coincided with the previously reported highest incidence of occurrence of electrical slow wave type action potentials. Examination of a large number of structural associations of ICC led us to conclude that in the distal esophagus, two networks of ICC and nerves exist, one associated with the inner muscle layer, another associated with the outer muscle layer. These networks are not sheet-like structures, such as the network of ICC in the myenteric plexus or deep muscular plexus of the small intestine, but are three dimensional and are interspersed throughout both muscle layers. The networks do not extend into Auerbach's plexus. The main branches of the networks run along the long axis of the esophagus and seem ideally suited to facilitate communication in this direction. These observations suggest that esophageal interstitial cells are structurally organized in such a manner that they may play a role in pacemaking and neural control of esophageal motility.

Animals↗
PubMed2006

Frontoparietal network involved in successful retrieval from episodic memory. Spatial and temporal analyses using fMRI and ERP.

The neural basis for successful recognition of previously studied items, referred to as "retrieval success," has been investigated using either neuroimaging or brain potentials; however, few studies have used both modalities. Our study combined event-related functional magnetic resonance imaging (fMRI) and event-related potential (ERP) in separate groups of subjects. The neural responses were measured while the subjects performed an old/new recognition task with pictures that had been previously studied in either a deep- or shallow-encoding condition. The fMRI experiment showed that among the frontoparietal regions involved in retrieval success, the inferior frontal gyrus and intraparietal sulcus were crucial to conscious recollection because the activity of these regions was influenced by the depth of memory at encoding. The activity of the right parietal region in response to a repeated item was modulated by the repetition lag, indicating that this area would be critical to familiarity-based judgment. The results of structural equation modeling revealed that the functional connectivity among the regions in the left hemisphere was more significant than that in the right hemisphere. The results of the ERP experiment and independent component analysis paralleled those of the fMRI experiment and demonstrated that the repeated item produced an earlier peak than the hit item by approximately 50 ms.

Brain Mapping↗
PubMed2026

IQ-NET: fast and accurate quartet phylogenetic inference using deep learning trained on empirical DNA alignments.

Phylogenetic inference is fundamental to modern biology, with many applications including evolutionary biology, epidemiology, and comparative genomics. While maximum likelihood and Bayesian methods remain the gold standard for phylogenetic analysis, they rely on simplifying assumptions and are computationally intensive. Recent machine learning approaches for phylogenetics offer speed advantages, but have several limitations: exclusive reliance on simulated data for training, inadequate handling of gaps, and sensitivity to input sequence order. Here, we introduce IQ-NET (Intelligent Quartet NETwork), a deep learning framework that solves these limitations to infer four-taxon trees. IQ-NET estimates both tree topology and branch lengths directly from gapped alignments. IQ-NET outperforms existing machine learning methods in terms of accuracy, and obtained a 24-fold speedup compared with the widely used maximum likelihood software, IQ-TREE. We finally introduce a pipeline using IQ-NET and the ASTRAL software to reconstruct a larger species tree, i.e., with more than four taxa.

Empirical data training↗
PubMed1999

[Expression of c-fos prooncogene in rat brain induced by an olfactory stimulus].

OBJECTIVE: In order to investigate how widespread the activated regions induced by an olfactory stimulus throughout the central nervous system (CNS), c-fos expressions in CNS of the rat observed, which is known as a combined morphologically and functionally probe. METHODS: Rats were stimulated by iso-amyl acetate for about 10 times, then one hour later sacrificed with perfusion of fixatives. After ABC immunocytochemi cal staining with anti-c-fos polyclonal antiserum, the c-Fos protein labelling areas on sections of the brain were observed. RESULTS: The results showed that the labelling areas were estremely widespread throughout the brain and cervical segments of the spinal cord in the rat, expecially much more in the areas above the level of the pons. Nearly all neuronal structures in the limbic system were labeled in addition to structures of the olfactory pathways, suggesting there were neuronal activities in multiple neural circuits, which may be associated with emotional, somatic and visceral responses to the olfactory stimulus. Next, many brain areas which were associated with mechanisms of attention and central state regulation were widely labeled. Unexpected heavy labellings in deep layers of the superior colliculus, inferior colliculus and pontine nuclei were presumably relevant to integrating activity between sensations and motions. CONCLUSIONS: The present study indicated that a specific olfactory stimulus did induce widespread activities of so-called pan-brain networks within the central nervous system.

Animals↗
PubMed1999

Central nervous system plasticity and persistent pain.

Nerve signals arising from sites of tissue or nerve injury lead to long-term changes in the central nervous system and contribute to hyperalgesia and the amplification and persistence of pain. These nociceptor activity-induced changes are referred to as central sensitization. Central sensitization involves an increase in the excitability of medullary and spinal dorsal horn neurons brought about by a cascade of events, including neuronal depolarization, removal of the voltage-dependent magnesium block of the N-methyl-D-aspartate (NMDA) receptor, calcium entry into neurons, phosphorylation of the NMDA receptor, a change in the cell's excitability, and an increase in synaptic strength. These changes also include activation of other ionotropic and metabotropic excitatory amino acid receptors, neuropeptides such as substance P, neurotrophins, and kinases involved in the phosphorylation process. Central sensitization occurs in trigeminal nociceptive pathways, and more robust neuronal hyperexcitability occurs following deep tissue stimulation than following cutaneous stimulation. By means of Fos protein immunocytochemistry, researchers have found that 2 distinct regions are activated: the subnucleus interpolaris/caudalis transition zone (Vi/Vc) and the caudal subnucleus caudalis. The latter exhibits changes very similar to those in the spinal dorsal horn, but the Vi/Vc zone likely is involved in autonomic nervous system processing and activation of the pituitary-adrenal axis. Descending systems are also an important component of the central sensitization process and provide the neural networks by which cognitive, attentional, and motivational aspects of the pain experience modulate pain transmission. These findings of nociceptor activity-induced neuronal plasticity have important clinical implications in the development of new approaches to the management of persistent pain.

Chronic Disease↗