Input Neurons Research Articles

Optimizing stroke prediction using gated recurrent unit and feature selection in Sub-Saharan Africa.

Stroke remains a leading cause of death and disability worldwide, with African populations bearing a disproportionately high burden due to limited healthcare infrastructure. Early prediction and intervention are critical to reducing stroke outcomes. This study developed and evaluated a stroke prediction system using Gated Recurrent Units (GRU), a variant of Recurrent Neural Networks (RNN), leveraging the Afrocentric Stroke Investigative Research and Education Network (SIREN) dataset. The study utilized secondary data from the SIREN dataset, comprising 4236 records with 29 phenotypes. Feature selection reduced these to 15 optimal phenotypes based on their significance to stroke occurrence. The GRU model, designed with 128 input neurons and four hidden layers (64, 32, 16, and 8 neurons), was trained and evaluated using 150 epochs, a batch size of 8, and metrics such as accuracy, AUC, and prediction time. Comparisons were made with traditional machine learning algorithms (Logistic Regression, SVM, KNN) and Long Short-Term Memory (LSTM) networks. The GRU-based system achieved a performance accuracy of 77.48 %, an AUC of 0.84, and a prediction time of 0.43 seconds, outperforming all other models. Logistic Regression achieved 73.58 %, while LSTM reached 74.88 % but with a longer prediction time of 2.23 seconds. Feature selection significantly improved the model's performance compared to using all 29 phenotypes. The GRU-based system demonstrated superior performance in stroke prediction, offering an efficient and scalable tool for healthcare. Future research should focus on integrating unstructured data, validating the model on diverse populations, and exploring hybrid architectures to enhance predictive accuracy.

Pearls & Oy-sters: Increased Visibility of Focal Cortical Dysplasia in Cerebral MRI During the First Year of Life.

Early detection of focal cortical dysplasia (FCD) using brain MRI in young children presenting with drug-resistant epilepsy may facilitate prompt surgical treatment, resulting in better control of seizures and decreased associated cognitive difficulties. Characteristics of FCD described in the literature are predominantly based on MRI findings in a fully myelinated brain; therefore, changes occurring during early brain maturation are not well known. In this case report, we describe distinct MRI features of a FCD visualized best before completion of myelination of the cortex and subcortical white matter. Specifically, this lesion exhibited increased visibility because of better delineation of the borders of the pathological area involved in the malformation compared with imaging obtained after completion of myelination. Although the pathophysiology of these imaging features necessitates further investigation, one hypothesis is that excessive neuronal input in young epileptic patients could trigger premature myelination of overstimulated fibers underlying the cortical epileptogenic focus. The aim of this case report was to raise awareness that seizures might induce local changes in brain myelination, which can be detected by MRI during the first year of life. Early identification of FCD may prompt surgical treatment in selected patients with drug-resistant epilepsy.

Long-range inputome of prefrontal GABAergic interneurons in the Alzheimer's disease mouse.

Alzheimer's disease (AD) is the most common neurodegenerative disease, characterized by damage to cortical circuits. However, the mechanisms underlying AD-associated changes in long-range circuits remain poorly understood. In this study, we used viral tracing and fluorescence micro-optical sectioning tomography (fMOST) imaging to investigate whole-brain changes in the input circuit of the frontal cortex of 5×FAD mice. Pathological axonal degeneration was widely observed in upstream regions, including the cortex, hippocampus, and thalamus, across all AD brains examined. The proportion of input neurons projecting to parvalbumin-expressing neurons, compared to those projecting to somatostatin-expressing neurons, decreased in the hippocampus and basal forebrain. This decline was closely related to mouse age and the cell type of the presynaptic input neurons. This study demonstrates the selective vulnerability of long-range circuits in the prelimbic area in AD at the mesoscopic level, thereby enhancing our understanding of circuit architecture degeneration across the brain. We used whole-brain imaging with single-cell resolution to generate brain-wide input maps of the Alzheimer's disease mouse model. The pathological changes in the input proportions showed relevance with the mouse age, distribution, and cell type of the presynaptic input neurons. Compared to the cell body and dendrites of the medial prefrontal cortex input neurons, the pathological changes in the axonal structure are more extensive.

Dissection of the long-range circuit of the mouse intermediate retrosplenial cortex

The retrosplenial cortex (RSP) is a complex brain region with multiple interconnected subregions that plays crucial roles in various cognitive functions, including memory, spatial navigation, and emotion. Understanding the afferent and efferent connectivity of the RSP is essential for comprehending the underlying mechanisms of its functions. Here, via viral tracing and fluorescence micro-optical sectioning tomography (fMOST), we systematically investigated the anatomical organisation of the upstream and downstream circuits of glutamatergic and GABAergic neurons in the dorsal and ventral RSP. The cortical connections of the RSP show laminar organisation in which the input neurons are distributed more in the deeper layers of the upstream cortex. Although different types of neurons have similar upstream circuits, GABAergic neurons show bidirectional connections with the hippocampus, whereas glutamatergic neurons only show unidirectional connections. Moreover, GABAergic neurons receive more inputs from the primary sensory cortex than from the prefrontal cortex and association cortex. The dorsal and ventral subregions have preferred circuits such that the dorsal RSP exhibits spatially topological connections with the dorsal visual cortex and lateral thalamus. The systematic study on long-range connections across RSP subregions and cell types may provide useful information for future revealing of RSP working mechanisms.

All Stochastic-Spiking Neural Network (AS-SNN): Noise Induced Spike Pulse Generator for Input and Output Neurons With Resistive Synaptic Array

All Stochastic-Spiking Neural Network (AS-SNN): Noise Induced Spike Pulse Generator for Input and Output Neurons With Resistive Synaptic Array

Evaluation of Evapotranspiration Prediction for Cassava Crop Using Artificial Neural Network Models and Empirical Models over Cross River Basin in Nigeria

The accurate assessment of water availability throughout the cassava cropping season (the initial, developmental, mid-season, and late stages) is crucial for mitigating the impacts of climate change on crop production. Using the Mann–Kendall Test, we investigated the trends in rainfall and cassava crop evapotranspiration (ETc) within the Cross River basin in Nigeria. Reference evapotranspiration (ETo) was based on two approaches, namely Artificial Neural Network (ANN) modelling and three established empirical models—the Penman–Monteith (considered the standard method), Blaney–Morin–Nigeria (BMN), and Hargreaves–Samani (HAG) models. ANN predictions were performed by using inputs from BMN and HAG parameters, denoted as BMN-ANN and HAG-ANN, respectively. The results from the ANN models were compared to those obtained from the Penman–Monteith method. Remotely sensed meteorological data spanning 39 years (1979–2017) were acquired from the Climatic Research Unit (CRU) to estimate ETc, while cassava yield data were acquired from the International Institute of Tropical Agriculture (IITA), Ibadan. The study revealed a significant upward trend in cassava crop ETc over the study period. Additionally, the ANN models outperformed the empirical models in terms of prediction accuracy. The BMN-ANN model with a Tansig activation function and a 3-3-1 architecture (number of input neurons, hidden layers, and output neurons) achieved the highest performance, with a coefficient of determination (R2) of 0.9890, a root mean square error (RMSE) of 0.000056 mm/day, and a Willmott’s index of agreement (d) of 0.9960. There is a decreasing trend in cassava yield in the region and further analysis indicated potential average daily water deficits of approximately −1.1 mm/day during the developmental stage. These deficits could potentially hinder root biomass, yield, and overall cassava yield in the Cross River basin. Our findings highlight the effectiveness of ANN modelling for irrigation planning, especially in the face of a worsening climate change scenario.

Neuromorphic computing for real-time adaptive penetration testing: analysis of human intuition in AI-dominated work space

This research investigates the revolutionary integration of neuromorphic computing with human intuition in the domain of real-time adaptive penetration testing, addressing the critical challenges facing modern cybersecurity in AI-dominated workspaces. The study presents a novel approach that combines the parallel processing capabilities of neuromorphic architectures with the nuanced decision-making abilities of human security experts, resulting in a hybrid system that significantly enhances threat detection and response capabilities. Our research implements a sophisticated neuromorphic computing model featuring 1024 input neurons, 2048 hidden layer neurons, and 512 output neurons, utilizing modified leaky integrate-and-fire algorithms for spike processing. The system incorporates real-time adaptive mechanisms that enable dynamic threat modeling and immediate response generation, while simultaneously learning from human expert intuition through a carefully designed collaborative interface. This architecture demonstrates remarkable improvements in both processing efficiency and threat detection accuracy, achieving a 94.7% true positive rate while maintaining an exceptionally low 2.3% false positive rate. The experimental methodology encompassed extensive testing across a diverse range of attack scenarios, including advanced persistent threats, zero-day vulnerabilities, and sophisticated social engineering attacks. The test environment comprised 500 virtual nodes distributed across multiple security zones, providing a realistic platform for comprehensive system evaluation. Performance metrics revealed significant improvements over traditional approaches, including an 89% success rate in adapting to previously unseen attack patterns and a 76% reduction in decision-making time for complex threats. Quantitative analysis demonstrates the system's superior capabilities in real-time threat detection and response, with average detection latencies of 1.2 milliseconds and consistent performance maintaining up to 100,000 concurrent connections. Qualitative assessment of human-AI synergy, conducted with 50 experienced penetration testers, revealed that 92% reported enhanced decision-making capabilities, while 95% experienced reduced cognitive load during complex scenarios. The research contributes significantly to the field by establishing a new paradigm for adaptive penetration testing that effectively combines neuromorphic computing efficiency with human intuitive expertise. The findings demonstrate that this integration not only enhances detection and response capabilities but also provides a more sustainable approach to cybersecurity in increasingly complex technological environments. Furthermore, the study opens new avenues for research in human-AI collaboration within cybersecurity, suggesting promising directions for future development in adaptive security systems. The implications of this research extend beyond immediate cybersecurity applications, offering insights into the broader field of human-AI collaboration in critical decision-making scenarios. The study also addresses important ethical considerations regarding AI autonomy in security operations, providing guidelines for responsible implementation of AI-driven security solutions while maintaining essential human oversight.

Integral Neuron: A New Concept for Nonlinear Neuron Modeling Using Weight Functions—Creation of XOR Neurons

In the present study, an extension of the idea of dynamic neurons is proposed by replacing the weights with a weight function that is applied simultaneously to all neuron inputs. A new type of artificial neuron called an integral neuron is modeled, in which the total signal is obtained as the integral of the weight function. The integral neuron enhances traditional neurons by allowing the signal shape to be linear and nonlinear. The training of the integral neuron involves finding the parameters of the weight function, where its functional values directly influence the total signal in the neuron’s body. This article presents theoretical and experimental evidence for the applicability and convergence of standard training methods such as gradient descent, Gauss–Newton, and Levenberg–Marquardt in searching for the optimal weight function of an integral neuron. The experimental part of the study demonstrates that a single integral neuron can be trained on the logical XOR function—something that is impossible for single classical neurons due to the linear nature of the summation in their bodies.

Stimulus-repetition effects on macaque V1 and V4 microcircuits explain gamma-synchronization increase.

Under natural conditions, animals repeatedly encounter the same visual scenes, objects or patterns repeatedly. These repetitions constitute statistical regularities, which the brain captures in an internal model through learning. A signature of such learning in primate visual areas V1 and V4 is the gradual strengthening of gamma synchronization. We used a V1-V4 Dynamic Causal Model (DCM) to explain visually induced responses in early and late epochs from a sequence of several hundred grating presentations. The DCM reproduced the empirical increase in local and inter-areal gamma synchronization, revealing specific intrinsic connectivity effects that could explain the phenomenon. In a sensitivity analysis, the isolated modulation of several connection strengths induced increased gamma. Comparison of alternative models showed that empirical gamma increases are better explained by (1) repetition effects in both V1 and V4 intrinsic connectivity (alone or together with extrinsic) than in extrinsic connectivity alone, and (2) repetition effects on V1 and V4 population input rather than output gain. The best input gain model included effects in V1 granular and superficial excitatory populations and in V4 granular and deep excitatory populations. Our findings are consistent with gamma reflecting bottom-up signal precision, which increases with repetition and, therefore, with predictability and learning.

Multi-Class Classification of Human Activity and Gait Events Using Heterogeneous Sensors

The control of active prostheses and orthoses requires the precise classification of instantaneous human activity and the detection of specific events within each activity. Furthermore, such classification helps physiotherapists, orthopedists, and neurologists in kinetic/kinematic analyses of patients’ gaits. To address this need, we propose an innovative deep neural network (DNN)-based approach with a two-step hyperparameter optimization scheme for classifying human activity and gait events, specific for different motor activities, by using the ENABL3S dataset. The proposed architecture sets the baseline accuracy to 93% with a single hidden layer and offers further improvement by adding more layers; however, the corresponding number of input neurons remains a crucial hyperparameter. Our two-step hyperparameter-tuning strategy is employed which first searches for an appropriate number of hidden layers and then carefully modulates the number of neurons within these layers using 10-fold cross-validation. This multi-class classifier significantly outperforms prior machine learning algorithms for both activity and gait event recognition. Notably, our proposed scheme achieves impressive accuracy rates of 98.1% and 99.96% for human activity and gait events per activity, respectively, potentially leading to significant advancements in prosthetic/orthotic controls, patient care, and rehabilitation programs’ definition.

Optimizing Fermentation Parameters for Bioethanol Production from Areca Nut Leaves using Artificial Neural Networks and Response Surface Methodologies

The research covered the entire process, from collecting the Areca nut leaves to purifying the produced bioethanol. Materials and Methods: The Areca nut leaves were pre-treated with sulfuric acid and sodium hydroxide, followed by enzymatic hydrolysis using cellulose enzymes. The hydrolyzed biomass was then fermented by Saccharomyces cerevisiae for 12–72 h to produce bioethanol. The produced bioethanol was purified through distillation using a rotary flask evaporator. To optimize the fermentation process and bioethanol production, the researchers employed two modeling approaches: Artificial neural networks (ANN) and response surface methodology (RSM). Variables such as pH, fermentation time, and disodium hydrogen phosphate (Na2HPO4) concentration, identified from the Plackett- Burman design, were optimized using the central composite design of RSM. Results and Discussion: The R² value for the RSM model was 91.72%, and the adjusted R² was 84.72%. In addition, an ANN algorithm model with 3 input neurons, 10 hidden layer neurons, and 1 output neuron was developed to investigate the relationship between bioethanol production and fermentation parameters. The ANN model achieved an R² of 99.78%, indicating higher accuracy and reliability compared to the RSM approach. The optimal conditions for bioethanol production were identified as pH 5.5, 60 h fermentation time, and 0.45 g of Na2HPO4. Under these conditions, the experimental bioethanol concentration reached 36.54 g/L. Conclusion: This study demonstrates the effective utilization of Areca nut leaves, a readily available agricultural waste, to produce bioethanol. The combination of statistical and machine learning techniques, such as ANN and RSM, allowed for the optimization of the fermentation process and the enhancement of bioethanol yield, showcasing the potential of this approach for sustainable biofuel production.

Analysis of single layer artificial neural network neuromorphic hardware chip

The neuromorphic architectures are hardware network systems designed with neural functions. Neural networks seen in biology serve as an inspiration for network systems. A synapse connects every node or neuron in an artificial neural network (ANN) to every other node. As in biological brains, the amplitude of the linking between nodes referred to as synaptic weights will regulate the connection. In contrast to conventional design, ANN uses many highly organized dealing pieces that work together to solve real-world issues. The design of the neuromorphic hardware chip is discussed in the paper. The target device used is a Virtex-5 Field Programmable Gate Array (FPGA) and the simulation is taken on Xilinx ModelSim software. This chip is designed for 20 neuron inputs, each of the neuron inputs is 8-bit. Each 20-neuron input is multiplied by 20 input weights and each weight is 8-bit so when these 20 input weights are multiplied by 20 neuron inputs in the multiplier it gives 16-bit output. A control logic is used in this neuromorphic hardware chip design which is used to feed multiplier output to each input of the hidden layer. The system-level outcome of the hidden layer is then given to the multiplexer which has 20 inputs and one single output. The multiplexer is used to select any of the 20 outputs of the control logic. Finally, to gain an understanding of the performance of this neuromorphic hardware chip, we have computed the hardware utilization parameters. These parameters include slices, input/output blocks (IOBs), registers used, memory, and the overall propagation delay used by the hardware chip.

Vestibular afferent neurons develop normally in the absence of quantal/glutamatergic input.

The vestibular nerve is comprised of neuron sub-groups with diverse functions related to their intrinsic biophysical properties. This diversity is partly due to differences in the types and numbers of low-voltage-gated potassium channels found in the neurons' membranes. Expression for some low-voltage gated ion channels like KCNQ4 is upregulated during early post-natal development; suggesting that ion channel composition and neuronal diversity may be shaped by hair cell activity. This idea is consistent with recent work showing that glutamatergic input from hair cells is necessary for the normal diversification auditory neurons. To test if biophysical diversity is similarly dependent on glutamatergic input in vestibular neurons, we examined vestibular function and the maturation of the vestibular epithelium and ganglion neurons by immunohistochemistry and patch-clamp electrophysiology in Vglut3-ko mice whose hair cell synapses lack glutamate. The knockout mice showed no obvious balance deficits and crossed challenging balance beams with little difficulty. Immunolabeling of the Vglut3-ko vestibular epithelia showed normal development as indicated by an identifiable striolar zone with calyceal terminals labeled by molecular marker calretinin, and normal expression of KCNQ4 by the end of the second post-natal week. We found similar numbers of Type I and Type II hair cells in the knockout and wild-type animals, regardless of epithelial zone. Thus, the presumably quiescent Type II hair cells are not cleared from the epithelium. Patch-clamp recordings showed that biophysical diversity of vestibular ganglion neurons in the Vglut3-ko mice is comparable to that found in wild-type controls, with a similar range firing patterns at both immature and juvenile ages. However, our results suggest a subtle biophysical alteration to the largest ganglion cells (putative somata of central zone afferents); those in the knockout had smaller net conductance and were more excitable than those in the wild type. Thus, unlike in the auditory nerve, glutamatergic signaling is unnecessary for producing biophysical diversity in vestibular ganglion neurons. And yet, because the input signals from vestibular hair cells are complex and not solely reliant on quantal release of glutamate, whether diversity of vestibular ganglion neurons is simply hardwired or regulated by a more complex set of input signals remains to be determined.

AgRP neurons mediate activity-dependent development of oxytocin connectivity and autonomic regulation

During postnatal life, leptin specifies neuronal inputs to the paraventricular nucleus of the hypothalamus (PVH) and activates agouti-related peptide (AgRP) neurons in the arcuate nucleus of the hypothalamus. Activity-dependent developmental mechanisms impact refinement of sensory circuits, but whether leptin-mediated postnatal neuronal activity specifies hypothalamic neural projections is largely unexplored. Here, we used chemogenetics to manipulate the activity of AgRP neurons during a discrete postnatal critical period and evaluated the development of AgRP inputs to the PVH and descending efferent outflow to the dorsal vagal complex (DVC). In leptin-deficient mice, targeting of AgRP neuronal outgrowth to PVH oxytocin neurons was reduced, and despite the lack of leptin receptors found on oxytocin neurons in the PVH, oxytocin-containing connections to the DVC were also impaired. Activation of AgRP neurons during early postnatal life not only normalized AgRP inputs to the PVH but also oxytocin outputs to the DVC in leptin-deficient mice. Blocking AgRP neuron activity during the same postnatal period reduced the density of AgRP inputs to the PVH of wild type mice, as well as the density of oxytocin-containing DVC fibers, and these innervation deficits were associated with dysregulated autonomic function. These findings suggest that leptin-mediated AgRP neuronal activity is required for the development of PVH connectivity and represents a unique activity-dependent mechanism for specification of neural pathways involved in the hypothalamic integration of autonomic responses.

INOVASI DALAM ENERGI TERBARUKAN: JARINGAN SYARAF TIRUAN UNTUK MERAMALKAN DAYA FOTOVOLTAIK

Variations in meteorological conditions cause intermittency, voltage spikes, and feedback power flow, impacting the uncertainty of photovoltaic output, which affects the reliability, stability, and scheduling of photovoltaic operations. Optimal prediction of photovoltaic output power is necessary in the planning and operation of power systems. Photovoltaic technology is utilized to generate electrical energy from direct sunlight. Climatic factors such as cloud cover, humidity, and wind speed also contribute to the electrical energy produced by solar modules. This research discusses a model for predicting photovoltaic output power for the next day using the backpropagation neural network method and multi-factor analysis. In this study, there are 2 different input neurons with 10 predetermined network architectures, a learning rate of 0.1, and a minimum target error of 0.001. The best performance prediction results with the smallest Mean Squared Error (MSE) value were obtained using a backpropagation neural network structure from the 5-20-1 model, which is almost close to the actual value. However, further research is needed to improve the prediction results.

Amorphous Si-Zn-Sn-O/Si-in-Zn-O Bilayer Transistors with P(VDF-TrFE) for Synaptic Integration in Neuromorphic Systems

Synaptic devices, such as memristors and transistors, are at the forefront of neuromorphic computing, designed to replicate the complex functions of the human brain. Unlike traditional computing systems (von Neumann architecture) that rely on the sequential processing, these synaptic devices excel in performing highly efficient parallel computations. This makes them particularly well-suited for tasks that require vast amounts of data processing, similar to how the brain processes information. Field-effect transistors (FETs), in particular, offer significant advantages in emulating synaptic behaviour. Their versatility in material selection, coupled with the ability to fine-tune parameters, makes them ideal for creating devices that mimic synaptic functions. Moreover, the established working principles of FETs provide a reliable foundation for developing sophisticated neuromorphic systems, pushing the boundaries of what's possible in next-generation computing.Our previous work on bilayer (SZTO/SIZO) thin-film transistors demonstrated remarkable high field-effect mobility exceeding 30 cm²V⁻¹s⁻¹, coupled with excellent stability under thermal stress, negative gate bias, and positive gate bias stress. These enhancements in performance and stability suggest significant potential for achieving high-performance artificial intelligence (AI) synaptic devices. Therefore, in this study, we present an AI synaptic device based on a bilayer thin-film transistor. The device features an amorphous SiZnSnO/SiInZnO (SZTO/SIZO) channel layer combined with a flexible copolymer, poly(vinylidene fluoride-trifluoroethylene) (P(VDF-TrFE)), serving as the gate insulator. The complete device structure includes Al/Ti/P(VDF-TrFE)/ZrO₂/SZTO/SIZO/substrate, with a channel width of 250 µm and a length of 50 µm. The transfer curve of the bilayer-based synaptic devices exhibits counterclockwise hysteresis during gate voltage sweeps of ±30 V, ±20 V, ±10 V, and ±5 V, indicating the presence of a ferroelectric layer as the gate insulator as shown in Figure 1. A memory window greater than 15 V was observed during sweeps from -30 to 30 V, with a saturation current exceeding 10 µA at 30 V.Postsynaptic currents (PSCs) were measured using 50 potentiation and 50 depression input pulses of ±10 V. The resulting maximum conductance exceeded 175 nS, derived from the relationship between the presynaptic gate voltage and the postsynaptic drain current. A neural network simulation was conducted using "MLP NeuroSim+ V3.0" software to model a multilayer perceptron (MLP) with 400 input neurons, 100 hidden neurons, and 10 output neurons, connected by artificial synapses. The model was trained on handwritten digit data from the Modified National Institute of Standards and Technology (MNIST) database and achieved a recognition accuracy of over 75% for the digits. Further enhancements in the ferroelectric layer and a deeper understanding of the gate insulator-channel interface could lead to even more advanced synaptic devices in the future. Figure 1

An Island of Reil excitation: Mapping glutamatergic (vGlut1+ and vGlut2+) connections in the medial insular cortex

The insular cortex is a multifunctional and richly connected region of the cerebral cortex, critical in the neural integration of external stimuli and internal signals. Well-served for this role by a large network of afferent and efferent connections, the mouse insula can be simplified into an anterior, medial and posterior portion. Here we focus on the medial subregion, a once over-looked area that has gained recent attention for its involvement in an array of behaviours. Although the connections of medial insular cortex neurons have been previously identified, their precise glutamatergic phenotype remains undefined (typically defined by the presence of the subtype of vesicular glutamate transporters). Hence, we combined Cre knock-in mouse lines and adeno-associated viral tracing to distinguish between the expression of the two major vesicular glutamate transporters, type 1 (vGlut1) and 2 (vGlut2), in the subregion's neuronal inputs and outputs. Our results determined that the medial insula has extensive glutamatergic efferents expressing both vGlut1 and vGlut2 throughout the neuraxis. In contrast, a more conservative number of glutamatergic inputs were observed, with exclusively vGlut2+ projections received from hypothalamic and thalamic regions. Taken together, we demonstrate that vGlut1- and vGlut2-expressing networks of this insular subdivision have distinct connectivity patterns, including a greater abundance of vGlut1+ fibres innervating hypothalamic regions and the extended amygdala. These findings provide insight into the distinct chemo-architecture of this region, which may facilitate further investigation into the role of the medial insula in complex behaviour.

Beyond-local neural information processing in neuronal networks

While there is much knowledge about local neuronal circuitry, considerably less is known about how neuronal input is integrated and combined across neuronal networks to encode higher order brain functions. One challenge lies in the large number of complex neural interactions. Neural networks use oscillating activity for information exchange between distributed nodes. To better understand building principles underlying the observation of synchronized oscillatory activity in a large-scale network, we developed a reductionistic neuronal network model. Fundamental building principles are laterally and temporally interconnected virtual nodes (microcircuits), wherein each node was modeled as a local oscillator. By this building principle, the neuronal network model can integrate information in time and space. The simulation gives rise to a wave interference pattern that spreads over all simulated columns in form of a travelling wave. The model design stabilizes states of efficient information processing across all participating neuronal equivalents. Model-specific oscillatory patterns, generated by complex input stimuli, were similar to electrophysiological high-frequency signals that we could confirm in the primate visual cortex during a visual perception task. Important oscillatory model pre-runners, limitations and strength of our reductionistic model are discussed. Our simple scalable model shows unique integration properties and successfully reproduces a variety of biological phenomena such as harmonics, coherence patterns, frequency-speed relationships, and oscillatory activities. We suggest that our scalable model simulates aspects of a basic building principle underlying oscillatory, large-scale integration of information in small and large brains.

CNSC-67. ANATOMICAL ORGANIZATION AND ELECTRICAL PROPERTIES OF GLIOMA-INNERVATING NEURONS

Abstract Gliomas are the most common malignant primary brain tumors and are often associated with severe neurological deficits and mortality. Unlike many cancers, gliomas rarely metastasize outside the brain, indicating a possible dependency on the brain’s unique microenvironment. Recent discoveries have highlighted the existence of neuron-glioma synapses, suggesting that glioma cells depend on neuronal inputs and AMPA-receptor-mediated glutamate signaling for their proliferation. Yet, the specific locations and properties of the neurons that innervate gliomas remain elusive. In this study, we utilized transsynaptic tracing with a pseudotyped, glycoprotein-deleted rabies virus to specifically infect TVA and glycoprotein-expressing human glioblastoma cells in an orthotopic xenograft model. This approach allowed us to identify the presynaptic neuronal inputs to the glioma. Comprehensive whole-brain mapping and quantification revealed that these glioma-innervating neurons, predominantly glutamatergic, exhibit a consistent input pattern influenced by the tumor’s locations and engage selective neuromodulatory pathways. Notably, the electrophysiological properties of these neurons mirror those innervating the cortex in vivo. Our study introduces a novel method for investigating glioma-innervating neurons and reveals that these neurons are consistently integrated into the glioma network, displaying distinctive location-dependent neuromodulatory patterns. Our findings offer a promising scaffold for future interventions aimed at disrupting these critical neuronal inputs, potentially redefining therapeutic strategies against brain tumors.

CNSC-57. THE ROLE OF NEURON-TO-GLIOMA SYNAPTIC COMMUNICATION IN THERAPEUTIC RESISTANCE

Abstract Glioblastomas, primary brain tumors known for their extensive brain colonization and high therapeutic resistance, pose significant treatment challenges. Recent research has revealed that glioblastoma cells can form functional multicellular networks both with themselves, surrounding astrocytes and neurons. However, the investigation of the glioblastoma neuronal connectome has been limited by a lack of suitable methodologies. In this study, we utilized an adapted rabies-based retrograde tracing approach to characterize tumor-connected neurons and examine the neuronal connectome of the tumor in the context of therapy resistance. Our initial findings demonstrated that radiotherapy induces hyperexcitability in tumor-connected neurons, as shown by patch-clamp electrophysiology. This hyperexcitability was associated with an increase in neuron-tumor connectivity post-irradiation. To mitigate this increased neuronal input to therapy-resistant glioblastoma cells, we combined radiotherapy with the anti-epileptic drug perampanel. This combination therapy, which inhibits overall neuronal activity while delivering tumor cell-toxic photon irradiation, proved significantly more effective in reducing the number of therapy-resistant tumor cells. In summary, we employed a novel methodology to investigate the role of the neuronal tumor connectome in therapeutic resistance, identifying tumor-connected neurons as a potential target for improving glioblastoma treatment.