搜索结果：architecture

Stem of vascular plants has emerged as hub organ for its connectivity and providing mechanically support. The relationship between stem diameter and reproductive traits has been widely considered. Differences in stem traits of grass and non-grasses influence plant reproductive output and trait correlations as well, which has been neglected in recent studies. Measurements of plant height, stem biomass, volume, basal diameter, total branch length and bulk density were used to describe the different stem size and architectures of 46 species of grasses and non-grasses in the Songnen Grassland, Northeast China. These size and architectural traits were coupled with a measure of leaf and seed traits. Except for bulk density, stem traits were size-dependent which explained much of the variation in leaf and reproductive biomass for both grasses and non-grasses. Fecundity was positively correlated with plant height in grasses, and with total branch length in non-grasses. The effects of stem traits, leaf biomass, and seed mass on plant seed number suggest that the specificity of stem architecture and the ubiquity of trait trade-off relationships combine to explain variation in herbaceous plant seed number, and it is therefore important that phylogenetically determined morphological differences in plant architecture should be incorporated into studies of interspecies competition and coexistence relationships, also the trait-based community dynamic.

Spontaneous coronary artery dissection (SCAD) is a major cause of myocardial infarction in young women without traditional cardiovascular risk factors (Hayes et al., 2018; Adlam et al., 2018 [1, 2]). Despite growing awareness, its biological underpinnings remain incompletely understood, and clinical management is largely based on observational evidence rather than mechanistic insight (Saw et al., 2014; Lettieri et al., 2015; Steg et al., 2024 [3-5]). To systematically integrate genomic, epitranscriptomic, proteomic, and metabolomic data in order to characterize the multi-omic architecture of SCAD and identify potential biomarkers and therapeutic targets. A systematic review was conducted in accordance with the PRISMA 2020 statement (Arbelo et al., 2023 [6]). PubMed/MEDLINE was searched for original studies investigating genomic and multi-omic features of SCAD. Data were extracted on study design, patient characteristics, identified variants, circulating biomarkers, and implicated biological pathways. Functional enrichment analysis was performed using the DAVID bioinformatics resource (Page et al., 2021 [7]). A total of 16 studies were included. Genome-wide association studies consistently identified susceptibility loci related to arterial structure and extracellular matrix integrity, including ADAMTSL4, PHACTR1/EDN1, LRP1, and FBN1 (Huang et al., 2009; Saw et al., 2020; Turley et al., 2020 [8-10]). Rare variant analyses further supported the role of genes involved in extracellular matrix remodeling and vascular smooth muscle cell function, including COL3A1, COL4A1/2, SMAD3, and TLN1 (Adlam et al., 2023; Turley et al., 2021, 2019; Carss et al., 2020; Zekavat et al., 2022; Wang et al., 2022 [11-16]), while ancestry-specific signals such as TSR1 variants were observed in distinct populations (Turley et al., 2023 [17]). Proteogenomic approaches linked genetic susceptibility loci to circulating proteins involved in matrix remodeling and inflammation, including cathepsin B and ECM1 (Maioli et al., 2010 [18]). Epitranscriptomic analyses identified differential microRNA expression profiles associated with vascular injury and repair pathways (Sun et al., 2019 [19]). SCAD is characterized by a complex, multi-layered biological architecture involving genetic susceptibility, extracellular matrix dysregulation, and vascular signaling pathways. Integration of multi-omic data provides novel insights into disease mechanisms and highlights potential biomarkers and targets for precision medicine approaches in SCAD.

Understanding how the spliceosome integrates regulatory cues to generate RNA diversity remains a central question in gene expression control. Emerging evidence reveals a multilayered framework in which splicing is governed by nuclear architecture and the physical state of nuclear speckles. These condensates function as phosphorylation-sensitive hubs that concentrate splicing machinery and couple signaling pathways to RNA processing. Chromatin organization, transcript architecture, and condensate properties are tightly coordinated, adding spatial constraints to spliceosome function. Recent findings further uncover temporal regulation through cell cycle and ultradian dynamics of speckle assembly. In this review, we synthesize these advances and propose a unified model in which charge-dependent phosphorylation of splicing factors drives condensate remodeling, linking nuclear organization to regulated splicing outcomes across space and time.

Meiosis is a key stage in the sexual reproduction of eukaryotes. It ensures the continuity of genetic information from generation to generation, while also generating the necessary genetic diversity for the survival and evolution of species. Meiotic progression is often compromised in hybrids between related subspecies, resulting in hybrid sterility and irreversible reproductive isolation. However, most genetic studies to date have not focused on the meiotic phenotypes of hybrid sterility and their molecular mechanisms. This review examines the genetic architecture, as well as the meiotic and molecular phenotypes, of hybrid sterility in the house mouse (Mus musculus) and other mammals. House mice subspecies provide the most widely understood mammalian model of hybrid sterility because of their recent evolutionary divergence, powerful genetic tools and comprehensive cytology of individual meiotic stages. We emphasize the potential impact of meiotic surveillance mechanisms, checkpoint pathways, particularly those leading to the meiotic sex chromosome inactivation and we draw parallels between intraspecific genic and chromosomal sterility and intersubspecific hybrid sterility. Finally, we review the Prdm9-Mir465 incompatibility system, the only vertebrate hybrid sterility model for which the three major genetic components necessary and sufficient to recreate the hybrid sterility genome have been identified. This three-part genetic architecture links Prdm9-dependent meiotic recombination hotspot activation, heterosubspecific homolog pairing, and microRNA-mediated meiotic checkpoint regulation to spermatogenic arrest and male sterility. MiR-465 is apparently the first microRNA which functions as a guardian of the pachytene checkpoint.

The ultra-dense vehicle scenarios envisioned in 6G put high requirements on ultra-low latency, secure cooperation, and efficient task offloading decisions. Existing systems usually optimize latency or energy independently but ignore joint privacy problems and long-term trust sustainability. In this work, a distributed intelligence architecture based on the combination of federated learning (FL) and blockchain based trust management for vehicle-to-vehicle (V2V) edge computing is proposed. The proposed architecture enables collaborative prediction and decentralized incentive enforcement in a privacy-preserving manner without revealing raw vehicle data. In this paper, task allocation is defined as a multi-objective optimization problem, which jointly considers latency, energy consumption, communication stability and privacy exposure. The resultant problem is addressed by a learning-coupled primal-dual optimization, where the federated prediction is used to drive the offloading decisions and the dual update is used to impose the limitations of the system. A light-weight distributed ledger layer ensures secure coordination, automatic incentive allocation and reliable detection of fraudulent nodes. The extensive simulations in the integrated traffic-network-blockchain environments show that the proposed method outperforms the state-of-the-art baselines, achieving up to 30-40% reduction in the service latency, approximately 25% improvement in task completion rate, enhanced privacy preservation by the gradient-based learning, and up to 95% accuracy in detecting the malicious nodes. These results validate the efficacy of the suggested framework for attaining scalable, privacy-aware, and trustworthy distributed intelligence for next-generation 6G vehicular edge networks.

Fatal traffic crashes are a rare yet catastrophically consequential event in real-world crash data, typically constituting less than 1% of total records. This extreme class imbalance poses a fundamental challenge for machine learning-based severity prediction, as standard algorithms tend to ignore the minority class in favor of maximizing overall accuracy. This study investigates whether modern deep tabular learning architectures (TabNet, FT-Transformer) offer consistent advantages over the traditional gradient boosting method XGBoost in predicting fatal crashes under conditions of extreme class imbalance. The analysis is conducted on 5,676 traffic crashes recorded in Batman province of Türkiye between 2013 and 2022, with a fatal crash rate of only 0.8%. Methodologically, a leakage-controlled design was implemented through ex-ante variable selection, structured missing value handling, and SMOTE-based balancing applied exclusively to the training set. Model performance was evaluated not only with decomposition metrics such as ROC-AUC, but also with PR-AUC, Recall@K/Lift, and cost-sensitive analyses, which are more meaningful for imbalanced data. The results show that FT-Transformer achieved the strongest performance with ROC-AUC = 0.820 (vs. XGBoost: 0.752, TabNet: 0.760) and PR-AUC = 0.031 (approximately 3.9× above the random baseline of 0.008). It captured approximately 44% of fatal crashes in the riskiest 10% of cases, providing a ≈ 4.4-fold lift compared to random selection. Calibration analyses revealed that FT-Transformer produced more reliable risk scores: in the predicted probability band of 0.5-0.8, its observed positive rate reached the 8-15% range, representing a 4-7× elevation above the near-zero rates (0-2%) recorded for XGBoost and TabNet across the same probability range. These findings indicate that transformer-based tabular architectures offer consistent statistical, operational, and cost-sensitive advantages under extreme imbalance, supporting their use as decision-support tools in traffic safety management. To examine the generalizability of the framework beyond a single jurisdiction and a single time window, the analysis is complemented by (i) a temporal hold-out within Batman (training on 2013-2020, testing on 2021-2022) and (ii) external benchmarking on an independent publicly available rare-event crash corpus (n = 12,316; fatal rate = 1.28%) [61, 62]; the architectural ranking and rank-based operational gains are reproduced in both regimes.

Per- and polyfluoroalkyl substances (PFAS), the so-called "forever chemicals," are emerging contaminants that severely disrupt soil ecosystems by rewiring microbial networks that sustain biogeochemical processes. This review deciphers the mechanisms underlying PFAS-induced microbial dysbiosis, revealing how these contaminants reconfigure community architecture, metabolic functions, and enzyme-mediated processes critical for biogeochemical cycling. It further integrates multi-omics approaches, spanning genomics to metabolomics, to elucidate molecular signatures and adaptive responses that govern microbial resilience and vulnerability across trophic hierarchies. Furthermore, the review examines PFAS biotransformation pathways, emphasising oxidoreductase-mediated mechanisms, kinetic bottlenecks, and catalytic constraints within complex soil matrices. By bridging microbial ecology with advanced material science, the review introduces a transformative paradigm of hybrid catalytic systems, including nanozyme-enabled transformations, engineered enzymes, and photocatalytic assemblies for targeted PFAS degradation. Thus, by linking microbial dysfunction with engineered catalytic innovation, the review offers a systems-level blueprint for sustainable and efficient strategies to restore PFAS-contaminated soils. Notably, this review highlights the urgent need for integrated multidisciplinary approaches to mitigate PFAS-induced ecological risks and advance sustainable soil restoration technologies.

The endoplasmic reticulum (ER) is a central hub coordinating protein homeostasis and lipid metabolism in eukaryotic cells. In microalgae, which inhabit highly fluctuating environments, ER stress is increasingly recognized as a driver of lipid remodeling rather than a secondary metabolic consequence. This review synthesizes recent advances in ER stress signaling in microalgae, focusing on Chlamydomonas reinhardtii, and places these findings in a comparative eukaryotic context. Microalgae retain a conserved, IRE1-centered unfolded protein response (UPR) while lacking auxiliary branches found in animals and land plants. Activation of ER stress induces extensive reprogramming of membrane lipid composition, fatty acid desaturation, sterol metabolism, and triacylglycerol (TAG) accumulation. Notably, the Chlamydomonas IRE1/bZIP1 pathway functions to restrain excessive TAG accumulation, thereby prioritizing membrane adaptation and ER homeostasis. The graded and dynamic nature of this response likely compensates for the simplified single-sensor architecture by enabling flexible modulation of downstream outputs depending on stress intensity and duration. Importantly, ER stress responses exhibit distinct modes depending on stress severity: moderate stress promotes adaptive membrane stabilization, whereas severe or prolonged stress redirects membrane-derived fatty acids into TAG for sequestration. This lipid-centered adaptation contrasts with land plants, which stabilize membrane composition without substantial TAG accumulation, and with yeast and animals, where membrane biogenesis and neutral lipid storage occur in parallel, with distinct regulatory features. By integrating insights across eukaryotes, this review highlights ER stress as a framework for understanding lipid remodeling in microalgae and discusses how UPR manipulation may enable rational engineering of lipid production platforms.

The recently developed CRISPR-Combo technology enables simultaneous targeted mutagenesis and transcriptional activation in plants. However, its reliance on SpCas9 limits its use at AT-rich genomic loci, such as promoter regions commonly targeted for transcription activation. To overcome this limitation, we explored the usage of Cas12b and iSpyMacCas9 in the CRISPR-Combo architecture for simultaneous genome editing and gene activation. We tested these expanded CRISPR-Combo systems for hormone-free regeneration of rice plants by transcriptional activation of a morphogenic gene, OsBBM1, while knocking out the genes of interest. The Cas12b-Combo system induced mild OsBBM1 upregulation (~3-fold), which did not affect the genome editing efficiency. By contrast, iSpyMacCas9-Combo achieved approximately 12-fold OsBBM1 transcriptional activation, supporting hormone-free regeneration at a high rate (42%). As a result, iSpyMacCas9-Combo conferred higher genome editing efficiency, including improved multiplex editing, than the standard iSpyMacCas9 system, either with or without hormones during rice regeneration. Hence, our data prove iSpyMacCas9-Combo to be a more efficient system for genome editing in rice, especially at low-efficiency target sites, when coupled with OsBBM1 transcriptional activation. These findings establish iSpyMacCas9-Combo as a useful addition to the CRISPR-Combo toolkit, expanding its genomic targeting scope and enabling more efficient genome editing by activation of an appropriate endogenous gene such as OsBBM1 in rice.

While the liver has astonishing regenerative capabilities, its potential wanes in pathological conditions such as fibrosis, cirrhosis and carcinoma, posing significant global health challenges. The regenerative process is critically dependent on the regulated cellular signals, extracellular matrix (ECM), and its mechanical dynamics. Accordingly, hydrogel-based tissue engineering strategies replicate the liver ECM microenvironmental architecture, including bioactivity, biocompatibility, porosity, and stiffness. Studies have demonstrated that hydrogels, whether derived from natural polymers, synthetic materials, or decellularised liver tissue, can be fine-tuned in their properties. However, many fail to recapitulate the dynamic mechanical, immunological, and vascular cues required for regeneration. Studies have reported that the ECM-derived hydrogel helps preserve the phenotype (KRT, AAT, HNF4α) and functions (CYP activity, albumin & urea production) of hepatocytes. Therefore, dECM-based hydrogels are particularly important for cellular transplantation and the delivery of bioactive agents for liver regeneration. However, deviations from the normal stiffness range of 4-6 kPa may trigger pathological responses, such as hepatic stellate cell (HSC) differentiation into myofibroblasts, leading to liver fibrosis. Ineffective decellularisation can cause scaffold rejection during transplantation. Additionally, because vascularisation is critical given the liver's rich blood supply, endothelial cells tend to spread randomly within hydrogel scaffolds. The random spreading of endothelial cells can lead to the deformation of lobular zonation, which profoundly impacts hepatocyte functions and has been documented to affect conditions like hepatocellular carcinoma and alcoholic/non-alcoholic fatty liver disease (AFLD/NAFLD). In discussing hydrogel-based strategies, the article highlights the importance of addressing immunogenic concerns and matrix remodelling during decellularisation. It also argues that current studies lack integration of vascular-based zonation, which is fundamental to accurately mimicking the liver's native structural and functional architecture. This approach will facilitate the development of relevant patient-specific models.

Disordered dendrite growth and corrosion reactions of zinc negative electrodes remain critical challenges in aqueous zinc metal batteries. Regulating water states in shear-thickening non-Newtonian fluid electrolytes has emerged as a promising strategy to simultaneously suppress dendrite growth and corrosion for zinc negative electrodes. Herein, we design a multifunctional shear-thickening non-Newtonian fluid electrolyte, based on carboxymethyl cellulose and sulfonate silicon oxide nanoparticles, which addresses both issues through hierarchical regulation of water molecule states. Carboxymethyl cellulose converts free water into weakly bound water, thereby suppressing water-induced parasitic reactions. Concurrently, sulfonated SiO2 nanoparticles form an integrated shear-thickening network with amylopectin and carboxymethyl cellulose while introducing abundant surface -SO3⁻ groups that disrupt the strongly bound water layer at the zinc interface. This architecture enables localized mechanical stiffening at dendrite tips without compromising ionic conductivity. As a result, Zn | |Zn symmetric cells exhibit stable Zn plating/stripping for 900 h at 50 mA cm⁻2 and 25 mAh cm⁻2, and Zn | |I2 pouch cells with a capacity of 1.5 Ah maintain stability over 200 cycles at 20 mA cm⁻2. These findings offer a alternative pathway toward corrosion-resistant, mechanically adaptive aqueous and practical zinc pouch cells systems.

Accurate prediction of outcomes after cardiac procedures is critical for personalised decision-making and risk stratification. While machine learning (ML) has shown promise in this domain, most prior studies rely on traditional ML methods that require structured data and manual feature engineering, limiting scalability. Many deep learning (DL) architectures offer an alternative by enabling automated feature extraction, particularly from unstructured data such as text, images, and signals. This scoping review summarises recent advances in DL-based prediction of outcomes for four major cardiovascular procedures: percutaneous coronary intervention (PCI), coronary artery bypass grafting (CABG), aortic valve replacement (AVR), and mitral valvuloplasty. Following PRISMA-ScR guidelines, we searched PubMed and IEEE Xplore for studies published between 2020 and March 2025. Finally, 457 studies were retrieved and nine eligible studies were included after screening. DL models demonstrated varying performance across data types, with particularly strong results for text and imaging tasks. Multimodal approaches combining clinical, imaging, and signal data showed added predictive value. Compared with traditional ML, DL models often reduce the need for manual feature engineering, though they still require preprocessing and validation to mitigate overfitting. Overall, these findings suggest that DL has potential to support preoperative risk stratification, although evidence for clinical utility remains preliminary. Moreover, all included studies lacked external validation, and challenges remain regarding generalisability, explainability, and integration into clinical workflows. Future research should prioritise large, diverse cohorts, multimodal data fusion, and interpretable DL models to enable safe and effective clinical implementation.

Enzymes play a crucial role in metabolism. The rate at which enzyme sequence data is emerging far exceeds the rate at which we can measure the associated catalytic constants, posing a significant problem for the advancement of the fields of metabolic engineering and synthetic biology. Here, we present KcatNeuroCortex, an interpretable and novel deep learning framework for enzyme catalytic efficiency prediction. The proposed architecture combines Bi-directional Gated Recurrent Units (Bi-GRU) with a multi-attention mechanism designed to mirror how enzymes work. It first captures local functional motifs through a segmentation-based strategy, and then integrates these into a global representation of long-range interactions that shape the catalytic landscape, leading to improved interpretability and prediction accuracy. KcatNeuroCortex achieves good results, with R² values of 0.74 and RMSE values of 0.77, representing a 57% improvement compared to DLKcat. We demonstrate that our segmentation-aware, attention-enhanced approach achieves competitive performance compared to conventional sequence-based models, especially on diverse and low-similarity sequences. The framework is robust, scalable, and interpretable, making it a valuable tool for enzyme engineering and large-scale kinetic parameter estimation. Also, the work establishes that deep learning can move beyond prediction to provide a biologically meaningful understanding of enzyme catalysis, positioning KcatNeuroCortex as a very valuable tool for the enzyme engineering community.

Accurate time-series prediction from soft sensor signals is essential for sensor-driven rehabilitation systems, yet remains challenging due to sensor noise, nonlinear dynamics, and complex temporal dependencies. In particular, filament-based tactile sensors exhibit long-term drift, hysteresis, and redundant signal components that degrade the performance of conventional recurrent models. To address these limitations, this paper proposes a novel Parallel Attention-Enhanced Long Short-Term Memory (PA-LSTM) architecture for robust displacement prediction from soft sensor data. The proposed model integrates an LSTM-based temporal encoder with a parallel dense embedding pathway and a Bahdanau-style attention mechanism, enabling adaptive weighting of informative time steps while suppressing noise and irrelevant signal fluctuations. By jointly capturing short-term dynamics and global contextual features, PA-LSTM enhances temporal feature selection and representation learning under noisy sensing conditions. The model is evaluated using pressure-displacement data collected from filament-based tactile sensors in a rehabilitation-oriented experimental setup. Extensive experiments demonstrate that PA-LSTM consistently outperforms standard LSTM, GRU, CNN-LSTM, and attention-only baselines. Specifically, the proposed approach achieves an RMSE of 0.047, an MAE of 0.028, and an R² score of 0.963, indicating substantial improvements in prediction accuracy and robustness. These results confirm that PA-LSTM effectively models complex soft-sensor dynamics and is well-suited for real-time displacement estimation in wearable rehabilitation and soft robotic sensing applications.

Positron range (PR) limits spatial resolution and quantitative accuracy in PET imaging, particularly for high-energy positron-emitting radionuclides such as [Formula: see text]Ga. This study proposes a deep learning-based approach using 3D residual encoder-decoder convolutional neural networks (3D RED-CNNs), incorporating tissue-dependent anatomical information through a μ-map-dependent loss function. We propose a model trained from realistic simulations and, using the information in the initial PET and CT images, obtain positron range corrected images. We validated our model in both simulations and real acquisitions. Three different 3D RED-CNN architectures-Single-Channel, Two-Channel, and DualEncoder-were trained using simulated PET datasets and evaluated on both synthetic and real PET acquisitions from [Formula: see text]Ga-FH and [Formula: see text]Ga-PSMA-617 mouse studies. The performance of each model was compared to a standard Richardson-Lucy deconvolution (RL-PRC) approach using metrics such as mean absolute error (MAE), structural similarity index (SSIM), contrast recovery (CR), and contrast-to-noise ratio (CNR). In simulations, CNN-based methods achieved up to 19% improvement in SSIM and 13% reduction in MAE compared to RL-PRC. The Two-Channel model showed the highest CR and CNR values, recovering lung activity with 97% agreement to the ground truth, compared to 77% with RL-PRC. Noise remained stable for CNN models (∼5.9%), whereas RL-PRC increased noise by 5.8%. In preclinical acquisitions, the Two-Channel model achieved the highest CNR in different tissues, while maintaining the lowest noise level (9.6%). Although no ground truth was available for real data, analysis confirmed superior tumor delineation and reduced spillover artifacts with the Two-Channel model. These findings demonstrate the potential of CNN-based PRC for improving quantitative PET imaging, particularly for [Formula: see text]Ga. CNN-based methods, particularly the Two-Channel model, outperformed conventional deconvolution in both simulated and real data. Future work will focus on enhancing model generalization through domain adaptation and hybrid training strategies, as well as extending the method to other high-energy PET radionuclides.

Achieving accurate manipulation over perovskite crystal orientation and phase purity is pivotal for realizing efficient and durable perovskite solar cells. Despite the unique advantages of lead iodide (PbI2) template in the two-step deposition process for obtaining high-quality perovskite films, challenges remain in guiding the preferential growth of perovskite crystals due to complex facets of PbI2. Herein, highly crystalline PbI2 with complete (001)-preferred orientation has been attained by incorporating 2-phenoxyacetamidine hydrochloride (PhOAaCl), which proficiently minimizes the crystal plane energy. This rationally architectured PbI2 template orchestrates coherent crystal plane growth of subsequent perovskite with markedly suppressed defect density and exceptional phase purity. Moreover, we further elucidate underlying growth mechanism prevailing at the solid-liquid interface between PbI2 and organic amine salts from integrated thermodynamic and kinetic perspectives. Consequently, the target photovoltaic device attains a champion efficiency of 26.47%. Notably, the unencapsulated devices exhibit significantly outstanding damp-heat endurance and operational robustness, while attaining 91% of initial efficiency after 2500 h of maximum power point tracking under persistent illumination in ambient conditions.

Spatial ATAC-seq enables simultaneous profiling of cellular locations and chromatin accessibility in intact tissues but faces challenges from high dimensionality, noise, and sparsity. Moreover, existing methods often overlook DNA sequence information, which contains critical regulatory motifs. To address these limitations, we introduce SpaDC, a graph-regularized convolutional neural network that integrates spatial location, chromatin accessibility, and DNA sequence. SpaDC employs a triplet loss function to integrate multiple spatial ATAC-seq datasets and remove batch effects. Benchmark analyses on real datasets demonstrate state-of-the-art performance in spatial domain identification, data denoising, and gene regulatory network (GRN) inference. Applied to mouse embryonic brain spatial ATAC-seq data, SpaDC accurately identified known brain structures and recovered chromatin accessibility signals. On P22 mouse brain spatial multi-omics data, SpaDC revealed spatial domain-specific cis-regulatory elements and GRNs. Collectively, SpaDC provides a powerful, sequence-based solution for spatial ATAC-seq analysis, enabling more accurate and robust investigation of tissue architecture and chromatin organization.

Artificial Intelligence (AI) has become integral to the research of neurological diseases due to the rapid expansion of neuroimaging, clinical, physiological, and wearable data. However, the concise synthesis of recent machine learning (ML) and deep learning (DL) remains limited. This systematic review analyzes studies published between January 2021 and March 2026 on five major conditions- Alzheimer's disease, stroke, Parkinson's disease, brain tumors, and traumatic brain injury (TBI)-following the PRISMA 2020 guidelines and a structured search of PubMed, Scopus, and Web of Science, yielding 206 eligible articles. The results show that convolutional and encoder-decoder architectures dominate imaging tasks, whereas hybrid and multimodal approaches increasingly combine imaging with clinical and sensor data. Emerging paradigms, including federated learning, self-supervised learning, and foundation models, address data scarcity, privacy, and cross-institutional variability. Key advances include high-performing transformer-based models for Alzheimer's diagnosis, real-time stroke detection by CT/MRI, improved Parkinson's detection by multimodal fusion, hybrid models for brain tumor classification, and outcome prediction in TBI. Despite these gains, challenges in generalizability, interpretability, and clinical translation persist, underscoring the need for more robust and clinically reliable AI systems to address these issues.