Volume 02 · Issue 02 · 2026

Astrocytes are increasingly recognized as key contributors to the pathophysiology of mood disorders. Dynamic changes in astrocyte subtypes have emerged as critical factors influencing the progression and outcomes of these conditions. This review provides a comprehensive overview of current knowledge regarding astrocyte subtype dynamics in mood disorders, including their underlying mechanisms and potential therapeutic implications. Changes in astrocyte subtypes changes are closely associated with inflammatory responses, metabolic changes, cell signaling pathway, epigenetic regulation, and ion channel functions. Mood disorders are strongly linked to astrocyte dysfunction and phenotypic transformations. This review highlights the central role of astrocyte subtype dynamics in the pathogenesis of mood disorders and underscores the need for further investigation of the underlying mechanisms. Targeting astrocyte subtype dynamics may therefore represent a promising therapeutic strategy for mood disorders.

Alzheimer’s disease (AD), the most prevalent neurodegenerative disorder and the leading cause of dementia worldwide, carries one of the highest disability burdens of brain diseases. Disruptions in GABAergic activity and alterations in gamma oscillations are key pathological features of AD associated with amyloid-beta (Aβ) plaques. Although there are links between GABAergic activity and gamma oscillations in AD neuropathology, an updated and comprehensive synthesis of preclinical and human studies is needed to develop effective interventions. This review provides an updated overview of the literature published in 2005-2025 (n = 127 studies) on GABAergic activity, gamma oscillations, and related therapeutic approaches in AD and mild cognitive impairment (MCI). The review is organized into two major sections covering: (1) GABAergic signaling in AD neuropathology and associated therapeutic approaches, and (2) gamma oscillations alterations in AD neuropathology and corresponding interventions. Each section summarizes findings from basic animal and human studies as well as intervention studies targeting GABAergic signaling or gamma oscillations in both preclinical models and human populations (AD/MCI). The preclinical models consistently characterize GABAergic dysfunction by selective loss and dysregulation of parvalbumin/somatostatin interneurons, imbalanced tonic-phasic inhibition, and network hyperexcitability, that are associated with memory deficits. Human studies with in vivo spectroscopy and post-mortem data similarly report reduced GABA levels, altered receptor and transporter expression, and interneuronal vulnerabilities associated with cognitive and affective symptoms. Gamma oscillations—largely mediated by GABAergic interneurons—show disrupted power and coupling in AD models and human clinical populations, as a hallmark of network dysfunction. In preclinical interventional studies, targeting GABA and interneuron transplantation can restore inhibitory balance, reduce pathology, and improve cognition. Furtheremore, preclinical studies show that 40 Hz stimulation (light/sound, somatosensory, electrical, magnetic) engages microglia, modulates vascular—cerebrospinal fluid dynamics, reduces Aβ/tau burden, enhances long-term potentiation, and improves behavior. In humans, 40 Hz transcranial alternating current stimulation/transcranial magnetic stimulation over prefrontal cortex, temporo-parietal hubs, and precuneus, as well as non-focal multisensory stimulation (audiovisual), increases gamma power and connectivity and shows modest cognitive benefits, although amyloid positron emission tomography outcomes and glymphatic clearance effects are inconclusive. Both GABA and gamma-based interventions appear safe and well-tolerated. Collectively, GABAergic deficits and gamma dysrhythmia form a bidirectional cascade in AD progression, supporting a synergistic intervention targeting both mechanisms. Current research gaps include limited longitudinal and multimodal biomarker datasets across disease stages, underpowered human trials, lacking sham controls and dose-response studies, and limited stratification and heterogeneous cognitive endpoints that hinder comparability and translation.

Traumatic brain injury (TBI) is a major public health challenge worldwide, affecting millions of individuals annually and is a leading cause of mortality and long-term disability. Historically, research has focused on neuronal damage, but accumulating evidence highlights the critical role of glial cells in both pathogenesis and potential recovery from TBI. Microglia, astrocytes, and oligodendrocytes undergo profound phenotypic and functional changes following traumatic injury, orchestrating complex responses that can be both protective and detrimental depending on the timing, injury severity, and cellular context. This review examines the multifaceted roles of different neuroglial cell types in TBI pathophysiology and their involvement in inflammation, tissue repair, blood-brain barrier integrity, and white matter damage. Here, this review discusses emerging therapeutic strategies targeting neuroglial dysfunction, including approaches to modulate microglial polarization, enhance astrocytic neuroprotection, and promote oligodendrocyte survival and remyelination. Understanding the temporal dynamics and context-dependent functions of neuroglial responses is crucial for developing effective TBI therapeutics.

Background: The cellular signals controlling synapse formation and neuronal network function remain poorly understood. This study investigated the canonical signaling pathways in neurons and their effects on network development using pharmacological assays to inhibit multiple signaling pathways in cultured primary neurons. Methods: Immunofluorescence was applied to detect synapse-related proteins in the neurons. Microelectrode arrays (MEAs) have been used to investigate the developmental characteristics of neuronal networks. Ribonucleic acid (RNA) sequencing was conducted to analyze gene expression profiles related to multiple signaling pathways. Results: We found that the c-Jun N-terminal kinase (JNK) pathway was essential for pre- and postsynaptic formation, while the phosphatidylinositide 3-kinases (PI3K) pathway was crucial for postsynaptic formation and influenced vesicular glutamate transporter 1 (VGlut1) puncta size. Inhibition of the JNK pathway significantly reduced the mean firing rate (MFR), network burst frequency (NBF), and firing regularity after four weeks, but did not influence network synchrony. During network development, the PI3K pathway modulates longer burst durations and decreases network synchrony. The inhibition of these pathways disrupts the signaling related to neurodevelopment and synapses. Additionally, inhibition of the PI3K pathway alters voltage-gated ion channel activity, synaptic transmission, and synaptic plasticity. Conclusions: Overall, this study revealed that the JNK and PI3K pathways play distinct roles in synapse formation and significantly influence neuronal network development. These findings enhance current understanding of the intracellular signaling pathways involved in synapse formation and neuronal network development.

Background: Advances in convolutional neural networks (CNNs) have continued to enhance automated brain tumor diagnosis; however, the trade-off between increasing model complexity and computational efficiency remains insufficiently explored. Methods: This study provides a comprehensive evaluation of ten model variants from the EfficientNetV2, ConvNeXt, and InceptionNeXt families using the Brain tumor Image Segmentation and Classification (BRISC) 2025 benchmark dataset. Results: By examining the efficiency frontier across these architectures, we investigate how predictive performance scales relative to computational cost. All models demonstrated robust predictive performance, with accuracies ranging from 0.972 to 0.992. The InceptionNeXt-Base model achieved the highest accuracy (0.992) and an F1-score of 0.9929. Importantly, our analysis revealed a law of diminishing returns: While computationally intensive models yielded marginal accuracy improvements, efficient architectures such as EfficientNetV2-Medium achieved comparable accuracy (0.990) with substantially lower parameter counts and latency. To validate prediction reliability, gradient-weighted class activation mapping++ (Grad-CAM++) interpretability analysis was employed, confirming that the top-performing models consistently focused on anatomically relevant tumor regions rather than background artifacts. Conclusions: These findings provide practical guidance for selecting resource-efficient architectures that balance high diagnostic accuracy with the computational constraints of clinical environments.

Background: Stroke remains a major cause of morbidity and mortality worldwide. In Indonesia, it is the second most commonly diagnosed disease with a significant public health burden. The Alberta Stroke Program Early Computed Tomography Score (ASPECTS) is a well-established clinical tool for assessing early ischemic changes. Lower ASPECTS (≤7) is associated with poorer clinical outcomes and a higher complication risk, particularly following thrombolytic therapy. The 24-h National Institutes of Health Stroke Scale (NIHSS24) score is a standardized measure of stroke severity and early neurological outcomes and strong predictor of long-term functional recovery. Methods: This retrospective study included 94 patients with acute ischemic stroke who underwent noncontrast brain computed tomography imaging and NIHSS evaluation at Cipto Mangunkusumo Hospital. Deep learning-based ASPECTS calculation was implemented to enhance accuracy. Associations of clinical, laboratory, and imaging parameters with the NIHSS24 score were evaluated. Linear multivariate and nonlinear regression (quadratic, cubic, and exponential) were performed, with Bonferroni and false detection rate (FDR) corrections for multiple comparisons. Results: Left-hemisphere ASPECTS robustly and independently predicted the NIHSS24 score in linear and nonlinear models. The nonlinear analysis demonstrated a ceiling effect: Patients with higher ASPECTS experienced disproportionately greater improvements in the NIHSS24 score (-1.18 points per ASPECTS point lost), whereas those with lower ASPECTS presented only minimal change (-0.06 points). Patient age exhibited borderline significance in the unadjusted model (P = 0.0484) but was non-significant after Bonferroni (P = 0.3872) and FDR corrections (P = 0.1936). No other clinical or laboratory variables independently predicted the outcomes. Although patients with high ASPECTS had better NIHSS24 scores, these differences were not significant after adjustment. Conclusions: Lower ASPECTS, particularly in the left hemisphere, predicted greater neurological impairment 24-h post-stroke. The nonlinear association indicated nonuniform prognostic sensitivity across the ASPECTS spectrum, with patients with near-normal scores benefiting more from prompt intervention, compared with those with low scores.