Neural Components Research Articles

Abstract B024: Deciphering the neurogenic tumor microenvironment: A novel gene network driving innervation in metastatic melanoma offers prognostic and therapeutic insights

Abstract Melanoma, a highly aggressive and life-threatening cancer, poses a significant therapeutic challenge, particularly for patients with distant metastasis, who face a 5-year survival rate of only approximately 22.5%. This underscores the urgent need for innovative therapeutic approaches. Emerging evidence suggests that sensory and parasympathetic innervation within the tumor microenvironment (TME) plays a crucial role in driving metastasis and immune evasion, partly through the promotion of M2 macrophages and cancer stem cells. To uncover the mechanisms behind tumor-specific innervation in metastatic melanoma, we employed a systems biology approach integrating bulk RNAseq from diverse melanoma cell lines, TCGA-SKCM data analysis, single-cell RNAseq, and spatial transcriptomics. This comprehensive analysis led us to identify a previously uncharacterized gene regulatory network (GRN) orchestrating tumor innervation. This GRN is driven by axon guidance cues (Netrin, Slit, Semaphorin), their receptors (DCC, Robo, Neuropilin, ITGB1), and transcription factors (RAS, RND1, TFAP2, CAMKII). We further discovered that the pioneer factor FOXC2, known for its association with metastasis, plays a critical role in opening the promoters of innervation-specific genes. Additionally, we identified Ephrin and CXCR4-mediated crosstalk between neurons and tumor cells. Machine learning models validated the accuracy of this GRN in predicting melanoma metastasis, and a neural network-based model confirmed the enrichment of nerve cells near metastatic tumor cells within the TME in TCGA immunohistochemistry data. The functional relevance of this network was further supported by CRISPR-mediated knockout studies, which disrupted tumor-specific signaling pathways. Notably, we also observed the enrichment of genes linked to mRNA processing bodies and neuron-specific membrane-less condensates in metastatic melanoma. These findings highlight the critical role of neuronal infiltration in melanoma metastasis. Currently, therapeutic strategies targeting this aspect of the TME remain limited. Our identified gene set offers potential as a metastatic predictor, and targeting the neural component of the TME represents a promising new therapeutic avenue for melanoma. Citation Format: Anjana Goli, Subhajit Dutta. Deciphering the neurogenic tumor microenvironment: A novel gene network driving innervation in metastatic melanoma offers prognostic and therapeutic insights [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Tumor-body Interactions: The Roles of Micro- and Macroenvironment in Cancer; 2024 Nov 17-20; Boston, MA. Philadelphia (PA): AACR; Cancer Res 2024;84(22_Suppl):Abstract nr B024.

The brain atlas of a subsocial bee reflects that of eusocial Hymenoptera.

The evolutionary transition from solitary life to group-living in a society with cooperative brood care, reproductive division of labor and morphological castes is associated with increased cognitive demands for task-specialization. Associated with these demands, the brains of eusocial Hymenoptera divide transcriptomic signatures associated with foraging and reproduction to different populations of cells and also show diverse astrocyte and Kenyon cell types compared with solitary non-hymenopteran insects. The neural architecture of subsocial bees, which represent evolutionary antecedent states to eusocial Hymenoptera, could then show how widely this eusocial brain is conserved across aculeate Hymenoptera. Using single-nucleus transcriptomics, we have created an atlas of neuron and glial cell types from the brain of a subsocial insect, the small carpenter bee (Ceratina calcarata). The proportion of C. calcarata neurons related to the metabolism of classes of neurotransmitters is similar to that of other insects, whereas astrocyte and Kenyon cell types show highly similar gene expression patterns to those of eusocial Hymenoptera. In the winter, the transcriptomic signature across the brain reflected diapause. When the bee was active in the summer, however, genes upregulated in neurons reflected foraging, while the gene expression signature of glia associated with reproductive functions. Like eusocial Hymenoptera, we conclude that neural components for foraging and reproduction in C. calcarata are compartmentalized to different parts of its brain. Cellular examination of the brains of other solitary and subsocial insects can show the extent of neurobiological conservation across levels of social complexity.

Prior probability biases perceptual choices by modulating the accumulation rate, rather than the baseline, of decision evidence

Abstract The prior probability of an upcoming stimulus has been shown to influence the formation of perceptual decisions. Computationally these effects have typically been attributed to changes in the starting point (i.e. baseline) of evidence accumulation in sequential sampling models. More recently, it has also been proposed that prior probability might additionally lead to changes in the rate of evidence accumulation. Here, we introduce a neurally-informed behavioural modelling approach to understand whether prior probability influences the starting point, the rate of evidence accumulation or both. To this end, we employ a well-established visual object categorization task for which two neural components underpinning participants’ choices have been characterised using single-trial analysis of the electroencephalogram. These components are reliable measures of trial-by-trial variability in the quality of the relevant decision evidence, which we use to constrain the estimation of a hierarchical drift diffusion model of perceptual choice. We find that, unlike previous computational accounts, constraining the model with the endogenous variability in the relevant decision evidence, results in prior probability effects being explained primarily by changes in the rate of evidence accumulation rather than changes in the starting point or a combination of both. Ultimately, our neurally-informed modelling approach helps disambiguate the mechanistic effect of prior probability on perceptual decision formation, suggesting that prior probability biases primarily the interpretation of sensory evidence towards the most likely stimulus.

HAECNN: An Intensity Analysis and Classification of Natural Disasters Using Hybrid Auto-Encoder - Convolutional Neural Network Model for Unprecedented Accuracy

Natural catastrophes are a serious threat to the environment, human lives, and property. Effective disaster management and response depend on the precise and timely determination of these occurrences' intensity and classification. In this work, we present a unique method for the analysis and categorization of natural disasters: the Hybrid Auto-Encoder - Convolutional Neural Network (HAECNN). By combining the strengths of convolutional neural networks with auto-encoders, HAECNN is able to classify and measure disaster intensity with previously unheard-of accuracy.HAECNN's auto-encoder part is in charge of autonomously extracting key features from unprocessed inputs like pictures, sensor signals, or satellite imagery. By capturing important details about the calamity, these extracted features strengthen and improve the model's ability to handle a variety of data sources. After the features are recovered, the convolutional neural network component analyses the data to identify spatial patterns and relationships.We carried out in-depth tests on a broad dataset comprising many kinds of natural catastrophes, such as hurricanes, earthquakes, floods, and wildfires, in order to validate the performance of HAECNN. With accuracy rate of over 95% in disaster classification and accurate intensity estimation, our model surpassed other available techniques. Because it can successfully adapt and generalise, HAECNN's outstanding accuracy is ascribed to its ability to utilise the intrinsic data structures and relationships present in many disaster types.For the analysis and categorization of natural disasters, the HAECNN model is a viable option that enhances early warning systems, timely reaction plans, and disaster planning. It is an invaluable tool for governments, first responders, and researchers trying to lessen the effects of natural disasters on the environment and society because of its unmatched accuracy and adaptability.

Enhancing the Recognition of Handwritten Arabic Characters through Hybrid Convolutional and Bidirectional Recurrent Neural Network Models

Handwritten Arabic character recognition remains a challenging task in pattern recognition due to the inherent complexities of the cursive script and visual similarities between characters. While deep learning techniques have demonstrated promising results in this domain, further enhancements to the model architecture can drive even greater performance improvements. This study introduces a hybrid deep learning approach that combines Convolutional Neural Networks (CNNs) with Bidirectional Recurrent Neural Networks, specifically Bidirectional Long Short-Term Memory (Bi-LSTM) and Bidirectional Gated Recurrent Units (Bi-GRU). By leveraging the strengths of both convolutional and recurrent neural network components, the proposed models are able to effectively capture spatial features as well as model the temporal dynamics and contextual relationships present in handwritten Arabic text. Experiments conducted on the AHCD and Hijjaa benchmark datasets show that the CNN-Bi-GRU framework achieved state-of-the-art accuracy rates of 97.05% and 91.78% respectively, outperforming previous deep learning-based methods. These results demonstrate the significant performance gains that can be achieved by integrating specialized temporal modeling and contextual representation capabilities into the handwriting recognition pipeline, without the need for explicit segmentation. The findings of this research represent a crucial advancement in the continued development of sophisticated and precise deep learning systems for Arabic handwriting recognition, with broad applications across domains that rely on efficient text extraction from handwritten documents.

Schwann cell-secreted frizzled-related protein 1 dictates neuroinflammation and peripheral nerve degeneration after neurotrauma.

Neurotrauma in limbs can induce sustained neuroinflammation, resulting in persistent disruption of nerve tissue architecture and retardation of axon regrowth. Despite macrophage-mediated inflammation promoting the removal of necrotic neural components and stimulating neo-vessel ingrowth, detrimental shifts in macrophage phenotype exacerbate nerve degeneration. Herein, we find that peripheral nerve injuries (PNIs) result in abundant secreted frizzled-related protein 1 (sFRP1) expression, particularly by Schwann cells (SCs). Heat shock protein 90 (HSP90) in macrophages recognizes sFRP1 and triggers a dysregulated secretion of inflammatory mediators. Single-cell atlas of human injured peripheral nerves reveals the appearance of sFRP1-expressing SCs with mesenchymal traits and macrophages with a proinflammatory genetic profile. Deletion of either SC-specific sFRP1 or macrophage-specific HSP90 alleviates neuroinflammation and prevents the progression of nerve degeneration. Together, our findings implicate the response of macrophages to SC-derived sFRP1 in exacerbating nerve damage following PNIs.

Electric Pulp Test Threshold Responses in Healthy Mature Permanent Teeth

Electric pulp testers assess only the neural component of pulp sensibility in dental practice, while vascular status is evaluated with laser Doppler flowmeters and pulp oximeters, which are more reliable for determining vitality in traumatic teeth. Objective: To determine the accuracy of the electric pulp test in healthy mature permanent teeth. Methods: This cross-sectional study was conducted at the Liaquat University of Medical and Health Sciences Jamshoro by consecutive sampling on 220 participants aged 18-35, who had healthy, mature permanent central incisors, canines, first premolars, and first molars, with no history of cardiac pacemakers and metallic restorations. Teeth were isolated, dried, and tested with a COXO C-Pulse electric pulp tester using toothpaste as a conducting medium. Threshold responses were recorded at increasing currents, with each tooth tested twice to determine the mean value. Accuracy was assessed using additional metrics post-therapy. Chi-square tests were employed to compare accuracy across gender and age groups, with significance set at p<0.05. Results: The mean age of the patients was 25.55 ± 5.41 years. 132 (60%) subjects were female, 88 (40%) subjects were male in this study. Accuracy of the electric pulp test in healthy mature permanent teeth was detected in 181 (82.3%) subjects in this study. Conclusion: It was concluded that the electric pulp tester method seems to be a reliable way to evaluate how sensitive the live nerve tissue inside a tooth is for healthy permanent teeth that have fully formed.

The Association Between Femoral Nerve Tension and Hip Flexor Length

Background: Femoral nerve tension and hip flexor muscle length are critical factors in musculoskeletal health. Their relationship is particularly important in diagnosing and managing conditions like low back pain, often associated with tight hip flexors.Objective: This study aimed to explore the association between femoral nerve tension and hip flexor muscle length in individuals with acute low back pain compared to healthy controls.Methods: A correlational study was conducted with 68 participants (34 healthy controls and 34 acute low back pain patients). Data were collected using self-administered questionnaires and clinical tests, including the Prone Knee Bend Test and Modified Thomas Test. Femoral nerve tension and hip flexor length were measured using a goniometer. Data were analyzed using Pearson correlation and linear regression in SPSS version 25.Results: The Thomas Test showed a statistically significant negative correlation between femoral nerve tension and hip flexor length in Group B (-0.096, p=0.042). In the Prone Knee Bend Test, a minimal positive correlation was observed (0.007, p>0.05). Cross-tabulation highlighted a significant association between test results in the control group (p=0.025).Conclusion: A significant relationship exists between femoral nerve tension and hip flexor length in individuals with acute low back pain. These findings suggest the need for targeted interventions focusing on neural and muscular components.

Prone Single Position Approach to Lateral Lumbar Interbody Fusion: Systematic Review and Meta-Analysis.

Lateral lumbar interbody fusion (LLIF) with posterior screw fusion is a safe and effective treatment for patients suffering from degenerative spine disorders. While LLIF has been shown to restore disc height, decompress neural components, correct sagittal imbalances, and improve pain scores, the approach requires repositioning patients for posterior pedicle fixation, which requires 2 separate surgeries. The evolution of surgical techniques, navigation, and robotics has allowed for a single position approach to LLIF with the patient in the prone position. The purpose of this study was to perform a systematic review and meta-analysis comparing the prone single position (PSP) LLIF approach to the dual position LLIF approach. We hypothesized that PSP LLIF will have a reduced operative time, complication rate, and blood loss compared with the dual position LLIF procedure. A systematic review was performed in accordance with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses 2020 guidelines. PubMed and Embase databases were searched with key terms: (lateral AND [interbody OR "inter body"] AND lumbar AND fusion) AND (prone OR single). Results were extracted and reviewed by 2 authors (MR and RB) per selection criteria. Patient demographics were extracted from the selected studies, along with surgical, patient-reported, and radiographic outcomes. A meta-analysis was performed using an unstandardized mean difference or log odds ratio with a confidence level of 95%. Fifteen studies were included in the systematic review and 5 studies compared PSP LLIF to dual position LLIF for meta-analysis. PSP LLIF had a reduced operative time and length of stay compared with the dual position approach, although there was no significant reduction in estimated blood loss. Additionally, PSP LLIF improved lumbar lordosis more effectively than dual position LLIF. There was no difference in segmental lordosis or pelvic tilt. There was no difference in intraoperative complications, postoperative complications, or reoperations. PSP LLIF reduces operative time and length of stay, with no relative increase in complications or reoperations compared with the dual position approach. Additionally, PSP LLIF improves lumbar lordosis relative to dual position LLIF, which may improve functional outcomes and reduce the risk of developing adjacent segment disease. The associated operative and postoperative benefits of PSP LLIF may improve long-term outcomes of patients undergoing spinal fusion.

Neural Fractional Differential Equations: Optimising the Order of the Fractional Derivative

Neural Fractional Differential Equations (Neural FDEs) represent a neural network architecture specifically designed to fit the solution of a fractional differential equation to given data. This architecture combines an analytical component, represented by a fractional derivative, with a neural network component, forming an initial value problem. During the learning process, both the order of the derivative and the parameters of the neural network must be optimised. In this work, we investigate the non-uniqueness of the optimal order of the derivative and its interaction with the neural network component. Based on our findings, we perform a numerical analysis to examine how different initialisations and values of the order of the derivative (in the optimisation process) impact its final optimal value. Results show that the neural network on the right-hand side of the Neural FDE struggles to adjust its parameters to fit the FDE to the data dynamics for any given order of the fractional derivative. Consequently, Neural FDEs do not require a unique α value; instead, they can use a wide range of α values to fit data. This flexibility is beneficial when fitting to given data is required and the underlying physics is not known.

A multiscale distributed neural computing model database (NCMD) for neuromorphic architecture

Distributed neuromorphic architecture is a promising technique for on-chip processing of multiple tasks. Deploying the constructed model in a distributed neuromorphic system, however, remains time-consuming and challenging due to considerations such as network topology, connection rules, and compatibility with multiple programming languages. We proposed a multiscale distributed neural computing model database (NCMD), which is a framework designed for ARM-based multi-core hardware. Various neural computing components, including ion channels, synapses, and neurons, are encompassed in NCMD. We demonstrated how NCMD constructs and deploys multi-compartmental detailed neuron models as well as spiking neural networks (SNNs) in BrainS, a distributed multi-ARM neuromorphic system. We demonstrated that the electrodiffusive Pinsky–Rinzel (edPR) model developed by NCMD is well-suited for BrainS. All dynamic properties, such as changes in membrane potential and ion concentrations, can be easily explored. In addition, SNNs constructed by NCMD can achieve an accuracy of 86.67% on the test set of the Iris dataset. The proposed NCMD offers an innovative approach to applying BrainS in neuroscience, cognitive decision-making, and artificial intelligence research.

Neural networks in closed-loop systems: Verification using interval arithmetic and formal prover

Machine Learning approaches have been successfully used for the creation of high-performance control components of cyber–physical systems, where the control dynamics result from the combination of many subsystems. However, these approaches may lack the trustworthiness required to guarantee their reliable application in a safety-critical context. In this paper, we propose a combination of interval arithmetic and theorem-proving verification techniques to analyze safety properties in closed-loop systems that embed neural network components. We show the application of the proposed approach to a model-predictive controller for autonomous driving comparing the neural network verification performance with other existing tools. The results show that open-loop neural network verification through interval arithmetic can outperform existing approaches proving properties with a smaller time overhead. Furthermore, we demonstrate the capability of combining the two approaches to construct a formal model of the network in higher-order logic of the controlled system in a closed-loop.

A unified theoretical framework underlying the regulation of motivated behavior.

To orchestrate behaviors for survival, multiple psychological components have evolved. The current theories do not clearly distinguish the distinct components. In this article, we provide a unified theoretical framework. To optimize survival, there should be four components; (1) "need", an alarm based on a predicted deficiency. (2) "motivation", a direct behavior driver. (3) "pleasure", a teacher based on immediate outcomes. (4) "utility", a teacher based on final delayed outcomes. For behavior stability, need should be accumulated into motivation to drive behavior. Based on the immediate outcome of the behavior, the pleasure should teach whether to continue the current behavior. Based on the final delay outcome, the utility should teach whether to increase future behavior by reshaping the other three components. We provide several neural substrate candidates in the food context. The proposed theoretical framework, in combination with appropriate experiments, will unravel the neural components responsible for each theoretical component.

Refining hemodynamic correction in in vivo wide-field fluorescent imaging through linear regression analysis

Accurate interpretation of in vivo wide-field fluorescent imaging (WFFI) data requires precise separation of raw fluorescence signals into neural and hemodynamic components. The classical Beer-Lambert law-based approach, which uses concurrent 530-nm illumination to estimate relative changes in cerebral blood volume (CBV), fails to account for the scattering and reflection of 530-nm photons from non-neuronal components leading to biased estimates of CBV changes and subsequent misrepresentation of neural activity. This study introduces a novel linear regression approach designed to overcome this limitation. This correction provides a more reliable representation of CBV changes and neural activity in fluorescence data. Our method is validated across multiple datasets, demonstrating its superiority over the classical approach.

Technological advances and challenges in constructing complex gut organoid systems.

Recent advancements in organoid technology have heralded a transformative era in biomedical research, characterized by the emergence of gut organoids that replicate the structural and functional complexity of the human intestines. These stem cell-derived structures provide a dynamic platform for investigating intestinal physiology, disease pathogenesis, and therapeutic interventions. This model outperforms traditional two-dimensional cell cultures in replicating cell interactions and tissue dynamics. Gut organoids represent a significant leap towards personalized medicine. They provide a predictive model for human drug responses, thereby minimizing reliance on animal models and paving the path for more ethical and relevant research approaches. However, the transition from basic organoid models to more sophisticated, biomimetic systems that encapsulate the gut's multifaceted environment-including its interactions with microbial communities, immune cells, and neural networks-presents significant scientific challenges. This review concentrates on recent technological strides in overcoming these barriers, emphasizing innovative engineering approaches for integrating diverse cell types to replicate the gut's immune and neural components. It also explores the application of advanced fabrication techniques, such as 3D bioprinting and microfluidics, to construct organoids that more accurately replicate human tissue architecture. They provide insights into the intricate workings of the human gut, fostering the development of targeted, effective treatments. These advancements hold promise in revolutionizing disease modeling and drug discovery. Future research directions aim at refining these models further, making them more accessible and scalable for wider applications in scientific inquiry and clinical practice, thus heralding a new era of personalized and predictive medicine.

Neuroscience of cancer: Research progress and emerging of the field.

Cancer cells immediately expand and penetrate adjoining tissues, as opposed to metastasis, that is the spread of cancer cells through the circulatory or lymphatic systems to more distant places via the invasion process. We found that a lack of studies discussed tumor development with the nervous system, by the aspects of cancer-tissue invasion (biological) and chemical modulation of growth that cascades by releasing neural-related factors from the nerve endings via chemical substances known as neurotransmitters. In this review, we aimed to carefully demonstrate and describe the cancer invasion and interaction with the nervous system, as well asreveal the research progress and the emerging neuroscience of cancer. An initial set of 160 references underwent systematic review and summarization. Through a meticulous screening process, these data were refined, ultimately leading to the inclusion of 98 studies that adhered to predetermined criteria. The outcomes show that one formidable challenge in the realm of cancer lies in its intrinsic heterogeneity and remarkable capacity for rapid adaptation. Despite advancements in genomics and precision medicine, there is still a need to identify new molecular targets. Considering cancer within its molecular and cellular environment, including neural components, is crucial for addressing this challenge. In conclusion, this review provides good referential data for direct, indirect, biological, and chemical interaction for nerve tissue-tumor interaction, suggesting the establishment ofnew therapy techniques and mechanisms by controlling and modifying neuron networks that supply signals to tumors.

Intrinsic molecular proton sensitivity underlies GPR4 effects on retrotrapezoid nucleus (RTN) neuronal activation and CO2-stimulated breathing.

An interoceptive homeostatic reflex monitors levels of CO2/H+ to maintain blood gas homeostasis and rapidly regulate tissue acid-base balance by driving lung ventilation and CO2 excretion - this CO2-evoked increase in respiration is the hypercapnic ventilatory reflex (HCVR). Retrotrapezoid nucleus (RTN) neurons provide crucial excitatory drive to downstream respiratory rhythm/pattern-generating circuits, and their activity is directly modulated by changes in CO2/H+ RTN neurons express GPR4 and TASK-2, global deletion of which abrogates CO2/H+ activation of RTN neurons and the HCVR. It has not been determined if the intrinsic pH sensitivity of these proton detectors is required for these effects. We used CRISPR/Cas9 genome editing to generate mice with mutations in either of two pH-sensing histidine residues in GPR4 to determine effects on RTN neuronal CO2/H+ sensitivity and the HCVR. In global GPR4(H81F) and GPR4(H167F) mice, CO2-stimulated breathing and CO2-induced RTN neuronal activation were strongly blunted, with no effect on hypoxia-stimulated breathing. In brainstem slices from GPR4(H81F) mice, peak firing of RTN neurons during bath acidification was significantly reduced compared to GPR4 wild type mice, and a subpopulation of RTN neurons was rendered pH-insensitive, phenocopying previous results from GPR4-deleted mice. These effects were independent of changes in RTN number/distribution, neuronal excitability or transcript levels for GPR4 and TASK-2. CO2-stimulated breathing was reduced to a similar extent in GPR4(H81F) and TASK-2-deleted mice, with combined mutation yielding no additional deficit in the HCVR. Together, these data demonstrate that the intrinsic pH sensitivity of GPR4 is necessary for full elaboration of the HCVR.Significance Statement Among the critical mechanisms for whole-body homeostasis, the hypercapnic ventilatory reflex (HCVR) regulates lung ventilation in order to maintain physiological levels of arterial PCO2 and acid-base balance. GPR4 is a proton-activated receptor and putative molecular proton sensor in retrotrapezoid nucleus (RTN) neurons, which are a crucial neural component of this respiratory reflex. In this work, we developed multiple lines of mice in which the intrinsic pH sensitivity of GPR4 was globally disabled by mutations in key histidine residues; in those mice, CO2/H+-sensitivity of RTN neurons and the HCVR were strongly blunted. These data support a role for GPR4 as a direct molecular sensor for CO2/H+ in support of this important homeostatic reflex.

FET-LM: Flow-Enhanced Variational Autoencoder for Topic-Guided Language Modeling.

Variational autoencoder (VAE) is widely used in tasks of unsupervised text generation due to its potential of deriving meaningful latent spaces, which, however, often assumes that the distribution of texts follows a common yet poor-expressed isotropic Gaussian. In real-life scenarios, sentences with different semantics may not follow simple isotropic Gaussian. Instead, they are very likely to follow a more intricate and diverse distribution due to the inconformity of different topics in texts. Considering this, we propose a flow-enhanced VAE for topic-guided language modeling (FET-LM). The proposed FET-LM models topic and sequence latent separately, and it adopts a normalized flow composed of householder transformations for sequence posterior modeling, which can better approximate complex text distributions. FET-LM further leverages a neural latent topic component by considering learned sequence knowledge, which not only eases the burden of learning topic without supervision but also guides the sequence component to coalesce topic information during training. To make the generated texts more correlative to topics, we additionally assign the topic encoder to play the role of a discriminator. Encouraging results on abundant automatic metrics and three generation tasks demonstrate that the FET-LM not only learns interpretable sequence and topic representations but also is fully capable of generating high-quality paragraphs that are semantically consistent.

Neural tracking of speech acoustics in noise is coupled with lexical predictability as estimated by large language models.

Adults heard recordings of two spatially separated speakers reading newspaper and magazine articles. They were asked to listen to one of them and ignore the other, and EEG was recorded to assess their neural processing. Machine learning extracted neural sources that tracked the target and distractor speakers at three levels: the acoustic envelope of speech (delta- and theta-band modulations), lexical frequency for individual words, and the contextual predictability of individual words estimated by GPT-4 and earlier lexical models. To provide a broader view of speech perception, half of the subjects completed a simultaneous visual task, and the listeners included both native and non-native English speakers. Distinct neural components were extracted for these levels of auditory and lexical processing, demonstrating that native English speakers had greater target-distractor separation compared to non-native English speakers on most measures, and that lexical processing was reduced by the visual task. Moreover, there was a novel interaction of lexical predictability and frequency with auditory processing; acoustic tracking was stronger for lexically harder words, suggesting that people listened harder to the acoustics when needed for lexical selection. This demonstrates that speech perception is not simply a feedforward process from acoustic processing to the lexicon. Rather, the adaptable context-sensitive processing long known to occur at a lexical level has broader consequences for perception, coupling with the acoustic tracking of individual speakers in noise.Significance Statement In challenging listening conditions, people use focused attention to help understand individual talkers and ignore others, which changes their neural processing for speech at auditory through lexical levels. However, lexical processing for natural materials (e.g., conversations, audiobooks) has been difficult to measure, because of limitations of tools to estimate the predictability of individual words in longer discourses. The present investigation uses a contemporary large language model, GPT-4, to estimate word predictability, and demonstrates that listeners make online adaptations to their auditory neural processing in accord with these predictions; neural activity more closely tracks the acoustics of the target talker when words are less predictable from the context.

The fMRI global signal and its association with the signal from cranial bone

The nature of the global signal, i.e. the average signal from sequential functional imaging scans of the brain or the cortex, is not well understood, but is thought to include vascular and neural components. Using resting state data, we report on the strong association between the global signal and the average signal from the part of the volume that includes the cranial bone and subdural vessels and venous collectors, separated from each other and the subdural space by multispectral segmentation procedures. While subdural vessels carried a signal with a phase delay relative to the cortex, the association with the cortical signal was strongest in the parts of the scan corresponding to the laminae of the cranial bone, reaching 80% shared variance in some individuals. These findings suggest that in resting state data vascular components may play a prominent role in the genesis of fluctuations of the global signal. Evidence from other studies on the existence of neural sources of the global signal suggests that it may reflect the action of multiple mechanisms (including cerebrovascular reactivity and autonomic control) concurrently acting to regulate global cerebral perfusion.