Sci Rep. 2025 May 19;15(1):17372. doi: 10.1038/s41598-025-87529-z.

ABSTRACT

Deep learning models are inherently vulnerable to adversarial examples, particularly in black-box settings where attackers have limited knowledge of the target model. Existing attack algorithms often face challenges in balancing effectiveness and efficiency. Adversarial perturbations generated in such settings can be suboptimal and require large query budgets to achieve high success rates. In this paper, we investigate the transferability of Multiple Universal Adversarial Perturbations (MUAPs), showing that they can affect a large portion of samples across different models. Based on this insight, we propose SMPack, a staged black-box adversarial example generation algorithm that integrates surrogate and query schemes. By combining MUAPs with surrogate models, SMPack effectively overcomes the black-box constraints and improves the efficiency of generating adversarial examples. Additionally, we optimize this process using a Genetic Algorithm (GA), allowing for efficient search of the perturbation space while conserving query budget. We evaluated SMPack against eight popular attack algorithms: OnePixel, SimBA, FNS, GA, SFGSM, SPGD, FGSM, and PGD, using four publicly available datasets: MNIST, SVHN, CIFAR-10, and ImageNet. The experiments involved 500 random correctly classified samples for each dataset. Our results show that SMPack outperforms existing black-box attack methods in both attack success rate (ASR) and query efficiency, while maintaining competitive performance with white-box methods. SMPack provides an efficient and effective solution for generating adversarial examples in black-box settings. The integration of MUAPs, surrogate schemes, and genetic optimization addresses the key limitations of existing methods, offering a robust alternative for generating adversarial perturbations with reduced query budget.

PMID:40389546 | DOI:10.1038/s41598-025-87529-z

Sci Rep. 2025 May 19;15(1):17363. doi: 10.1038/s41598-025-01758-w.

ABSTRACT

Conventional image formats have limited information conveyance, while Hyperspectral Imaging (HSI) offers a broader representation through continuous spectral bands, capturing hundreds of spectral features. However, this abundance leads to redundant information, posing a computational challenge for deep learning models. Thus, models must effectively extract indicative features. HSI's non-linear nature, influenced by environmental factors, necessitates both linear and non-linear modeling techniques for feature extraction. While PCA and ICA, being linear methods, may overlook complex patterns, Autoencoders (AE) can capture and represent non-linear features. Yet, AEs can be biased by unbalanced datasets, emphasizing majority class features and neglecting minority class characteristics, highlighting the need for careful dataset preparation. To address this, the Dual-Path AE (D-Path-AE) model has been proposed, which enhances non-linear feature acquisition through concurrent encoding pathways. This model also employs a down-sampling strategy to reduce bias towards majority classes. The study compared the efficacy of dimensionality reduction using the Naïve Autoencoder (Naïve AE) and D-Path-AE. Classification capabilities were assessed using Decision Tree, Support Vector Machine, and K-Nearest Neighbors (KNN) classifiers on datasets from Pavia Center, Salinas, and Kennedy Space Center. Results demonstrate that the D-Path-AE outperforms both linear dimensionality reduction models and Naïve AE, achieving an Overall Accuracy of up to 98.31% on the Pavia Center dataset using the KNN classifier, indicating superior classification capabilities.

PMID:40389536 | DOI:10.1038/s41598-025-01758-w

Respir Res. 2025 May 19;26(1):192. doi: 10.1186/s12931-025-03270-1.

ABSTRACT

The standard approach to diagnosing idiopathic pulmonary fibrosis (IPF) includes identifying the usual interstitial pneumonia (UIP) pattern via high resolution computed tomography (HRCT) or lung biopsy and excluding known causes of interstitial lung disease (ILD). However, limitations of manual interpretation of lung imaging, along with other reasons such as lack of relevant knowledge and non-specific symptoms have hindered the timely diagnosis of IPF. This review proposes the definition of early IPF, emphasizes the diagnostic urgency of early IPF, and highlights current diagnostic strategies and future prospects for early IPF. The integration of artificial intelligence (AI), specifically machine learning (ML) and deep learning (DL), is revolutionizing the diagnostic procedure of early IPF by standardizing and accelerating the interpretation of thoracic images. Innovative bronchoscopic techniques such as transbronchial lung cryobiopsy (TBLC), genomic classifier, and endobronchial optical coherence tomography (EB-OCT) provide less invasive diagnostic alternatives. In addition, chest auscultation, serum biomarkers, and susceptibility genes are pivotal for the indication of early diagnosis. Ongoing research is essential for refining diagnostic methods and treatment strategies for early IPF.

PMID:40390073 | DOI:10.1186/s12931-025-03270-1

Ann Am Thorac Soc. 2025 May 19. doi: 10.1513/AnnalsATS.202408-882OC. Online ahead of print.

ABSTRACT

RATIONALE: Combined pulmonary fibrosis and emphysema (CPFE) is a unique phenotype with important prognosis and management implications in patients with idiopathic pulmonary fibrosis (CPFE-IPF) and other forms of fibrotic interstitial lung disease (CPFE-fILD). However, the epidemiology of CPFE is not well characterized, creating a barrier to clinical research needed to advance our understanding and management.

OBJECTIVES: To investigate the incidence, prevalence, and long-term outcomes of CPFE among a regional cohort of Veterans.

METHODS: We retrospectively reviewed records for Veterans in the Veterans Affairs Mid-Atlantic Health Care Network (includes North Carolina and Virginia) with International Classification of Disease (ICD)-9 codes for pulmonary fibrosis between January 1, 2008, and December 31, 2015. We stratified pulmonary fibrosis into IPF and fILD using diagnostic codes and chart review. We reviewed CT reports and classified cases as having CPFE according to documented emphysema; a thoracic radiologist overread a subset of scans for validation. We calculated annual incidence and prevalence of CPFE and compared characteristics between Veterans with CPFE and Veterans with fibrosis without emphysema using Chi-squared testing, Mann Whitney U testing, and paired t-tests. We used Kaplan-Meier and Cox models to determine overall survival from diagnosis.

RESULTS: We identified 2,414 Veterans with fibrotic ILD. Among 1,880 Veterans with IPF, 734 (39.0%) had CPFE-IPF; among 534 Veterans with fILD, 194 (36.3%) had CPFE-fILD. Agreement between CT reports and thoracic radiologist review was high (Kappa = 0.78). Annual CPFE prevalence ranged 71-100 per 100,000 Veterans, and incidence ranged 16-39 per 100,000 Veterans. CPFE was associated with male sex, lower BMI, greater tobacco history, higher FVC, reduced FEV1/FVC ratio, reduced DLCO, and increased oxygen utilization. CPFE was associated with increased mortality in unadjusted models. However, after adjustment for age, sex, and BMI, CPFE was not associated survival for both CPFE-IPF versus IPF without emphysema (HR 1.13, 95% CI 0.96-1.33) as well as CPFE-fILD versus fILD without emphysema (HR 1.16, 95% CI 0.82-1.63).

CONCLUSIONS: CPFE has high incidence and prevalence among Veterans with IPF and fILD and has a distinct phenotype with diagnostic and therapeutic implications. Further studies are merited investigating diagnosis, treatment considerations, and long-term impacts in CPFE.

OBJECTIVES: To investigate the incidence, prevalence, and long-term outcomes of CPFE among a regional cohort of Veterans.

METHODS: We retrospectively reviewed records for Veterans in the Veterans Affairs Mid-Atlantic Health Care Network (includes North Carolina and Virginia) with International Classification of Disease (ICD)-9 codes for pulmonary fibrosis between January 1, 2008, and December 31, 2015. We stratified pulmonary fibrosis into IPF and fILD using these ICD9 codes. We reviewed CT reports and classified cases as having CPFE according to documented emphysema; a thoracic radiologist overread a subset of scans for validation. We compared characteristics between Veterans with CPFE and Veterans with fibrosis without emphysema using Chi-squared testing, Mann Whitney U testing, and paired t-tests as appropriate. We used Kaplan-Meier and Cox models to estimate and compare overall survival from diagnosis.

RESULTS: We identified 2,414 Veterans with fibrotic ILD. Among 1,880 Veterans with IPF, 734 (39.0%) had CPFE-IPF; among 534 Veterans with fILD, 194 (36.3%) had CPFE-fILD. Agreement between CT reports and thoracic radiologist review was high (Kappa = 0.78). Overall CPFE prevalence was 107.48 per 100,000, and incidence was 28.53 per 100,000. CPFE was associated with male sex, lower BMI, greater tobacco history, higher FVC, reduced FEV1/FVC ratio, reduced DLCO, and increased oxygen utilization. CPFE was associated with increased mortality in unadjusted models. However, after adjustment for age, sex, and BMI, CPFE was not associated survival for both CPFE-IPF versus IPF without emphysema (HR 1.13, 95% CI 0.96-1.33) as well as CPFE-fILD versus fILD without emphysema (HR 1.16, 95% CI 0.82-1.63).

CONCLUSIONS: CPFE has high incidence and prevalence among Veterans with IPF and fILD and has a distinct phenotype with diagnostic and therapeutic implications. Further studies are merited investigating care utilization, treatment considerations, and long-term impacts in CPFE.

PMID:40388887 | DOI:10.1513/AnnalsATS.202408-882OC

J Hum Hypertens. 2025 May 20. doi: 10.1038/s41371-025-01027-3. Online ahead of print.

ABSTRACT

Hypertension is a prevalent disease that substantially elevates the risk of neurological disorders such as dementia, stroke and Parkinson's disease. MicroRNAs (miRNAs) play a critical role in the regulation of gene expression related to brain function and disorders. Understanding the involvement of miRNAs in these conditions could provide new insights into potential therapeutic targets. The main objective of this study is to target and investigate microRNAs (miRNAs) associated with neurological disorders in patients suffering from hypertension. The genes involved in hypertension were identified from various databases including GeneCard, MalaCard, DisGeNet, OMIM & GEO2R. The key gene for hypertension was identified using a systems biology approach. Also, potent phytochemical for hypertension was determined by computer-aided drug-designing approach. Functional miRNAs were determined for the key target gene using miRNet analytics platform by hypergeometric tests. Further, the gene-miRNA interaction was determined and enrichment analysis was done. RPS27A was identified as a key target gene for hypertension. Naringenin showed effective molecular interaction with RPS27A with a binding energy score (-6.28). Further, a list of miRNAs which were targeting brain disorders was determined from miRNet. A gene-miRNA network was constructed using the PSRR tool for Parkinson's Disease, Autism Spectrum Disorder, Acute Cerebral Infarction, ACTH-Secreting Pituitary Adenoma, & Ependymoma. Further, miRNA 21 & miRNA 16 were found to be associated with four of the neurological disorders. The study identifies specific miRNAs that may serve as potential biomarkers for brain disorders in hypertensive patients. Targeting these miRNAs could open new avenues for therapeutic strategies aimed at mitigating neurological damage in this patient population.

PMID:40389629 | DOI:10.1038/s41371-025-01027-3

Nat Commun. 2025 May 20;16(1):4666. doi: 10.1038/s41467-025-59844-6.

ABSTRACT

Computer-aided design has become common practice in DNA nanotechnology, and many programs are available that make the sophisticated design processes accessible to both the core research community and curious scientists in other fields. However, most of the design tools are committed to the scaffolded DNA origami method. Here we present an automated design pipeline for creating DNA wireframe nanostructures based on a scaffold-free molecular self-assembly approach. Unlike in the DNA origami method, scaffold-free designs are not built around a global backbone strand but are constituted entirely of short, locally intertwined oligonucleotides. This overcomes many limitations inherent in scaffolded nanostructure designs, most notably the size constraints imposed by the length of available scaffold strands, and the topological and algorithmic challenges of finding feasible scaffold-strand routings. In practice, this leads to simpler design flows and opens up new design possibilities. To demonstrate the flexibility and capability of our approach, we generate a variety of complex DNA wireframe designs automatically from 2D and 3D mesh models and successfully realise the respective molecular nanostructures experimentally.

PMID:40389428 | DOI:10.1038/s41467-025-59844-6

EMBO Mol Med. 2025 May 19. doi: 10.1038/s44321-025-00249-9. Online ahead of print.

ABSTRACT

Mutations in PIK3R1 have recently been identified in patients with overgrowth syndromes and complex vascular malformations. PIK3R1 encodes p85α which acts as the regulatory subunit of the lipid kinase PI3Kα. PIK3R1 mutations result in the excessive activation of the AKT/mTOR pathway. Currently, there are no approved treatments specifically dedicated to patients with PIK3R1 mutations, and medical care primarily focuses on managing symptoms. In this study, we identified three patients, including two children, who had mosaic somatic PIK3R1 mutations affecting the iSH2 domain, along with severe associated symptoms that were unsuccessfully treated with rapamycin. We conducted in vitro experiments to investigate the impact of these mutations, including a double PIK3R1 mutation in cis observed in one patient. Our findings revealed that p85α mutants in the iSH2 domain showed sensitivity to alpelisib, a pharmacological inhibitor of PI3Kα. Based on these findings, we received authorization to administer alpelisib to all three patients. Following drug introduction, patients rapidly demonstrated clinical improvement, pain, fatigue and inflammatory flares were attenuated. Magnetic Resonance Imaging showed a mean decrease of 22.67% in the volume of vascular malformations over twelve months of treatment with alpelisib. No drug-related adverse events were reported during the course of the study. In conclusion, this study provides support for the use of PI3Kα inhibition as a promising therapeutic approach for individuals with PIK3R1-related anomalies.

PMID:40389643 | DOI:10.1038/s44321-025-00249-9

Sci Rep. 2025 May 19;15(1):17301. doi: 10.1038/s41598-025-02359-3.

ABSTRACT

Through an in-depth analysis of ertapenem-associated adverse events (AEs) in the FDA Adverse Event Reporting System (FAERS) database, this study provides a reference for monitoring and safety management of ertapenem. Data from the FAERS database from Q1 2004 to Q1 2024 were analyzed via four nonproportional analysis techniques, including the reporting odds ratio (ROR). Gender, age, and sensitivity analyses were conducted for a more detailed assessment of ertapenem-associated signals. A total of 2,931 reports with ertapenem as the primary suspected drug were collected, covering 27 system organ classes (SOCs). The two SOCs with the strongest signals were nervous system disorders and psychiatric disorders, with overall stronger signals in individuals aged ≥ 65 years. The most frequently reported AEs were confusional state (n = 265) and convulsions (n = 214). Among the strongest signals were oropharyngeal edema (ROR = 191.05, 95% CI: 60.76-601.35) and granulomatous dermatitis (ROR = 150.49, 95% CI: 55.9-405.15). Eleven AEs not listed on the FDA label were identified. The top 20 AEs were predominantly associated with nervous system and psychiatric disorders, with a median time to onset ranging from 3.5 to 8.5 days. This study highlights the neuropsychiatric risks of ertapenem, providing strong evidence for its safety assessment and emphasizing the need for monitoring and individualized management in high-risk patients. Ertapenem, FAERS, Adverse events, Drug safety, Disproportionality analysis.

PMID:40389541 | DOI:10.1038/s41598-025-02359-3

Medicine (Baltimore). 2025 May 16;104(20):e42411. doi: 10.1097/MD.0000000000042411.

ABSTRACT

Currently, there remains a significant gap in effective pharmacologic interventions for neonatal respiratory distress syndrome (NRDS). To address this critical unmet medical need, we aimed to systematically identify novel therapeutic targets and preventive strategies through comprehensive integration and analysis of multiple publicly accessible datasets. In this study, we employed an integrative approach combining druggable genome data, cis-expression quantitative trait loci (cis-eQTL) from human blood and lung tissues, and genome-wide association study summary statistics for neonatal respiratory distress. We performed two-sample Mendelian randomization (TSMR) analysis to investigate potential causal relationships between druggable genes and neonatal respiratory distress. To strengthen causal inference, we performed Bayesian co-localization analyses. Furthermore, we conducted phenome-wide Mendelian randomization (Phe-MR) to systematically evaluate potential side effects and alternative therapeutic indications associated with the identified candidate drug targets. Finally, we interrogated existing drug databases to identify actionable pharmacological agents targeting the identified genes. All 3 genes (LTBR, NAAA, CSNK1G2) were analyzed by Bayesian co-localization (PH4 > 75%). CSNK1G2 (lung eQTL, odds ratio [OR]: 0.419, 95% CI: 0.185-0.948, P = .037; blood eQTL, OR: 4.255, 95% CI: 1.346-13.455, P = .014; Gtex whole blood eQTL, OR: 4.966, 95% CI: 1.104-22.332, P = .037). LTBR (lung eQTL, OR: 0.550, 95% CI: 0.354-0.856, P = .008; blood eQTL, OR: 0.347, 95% CI: 0.179-0.671, P = .002; Gtex whole blood eQTL, OR: 0.059, 95% CI: 0.0.007-0.478, P = .008). NAAA (lung eQTL, OR: 0.717, 95% CI: 0.555-0.925, P = .011; Gtex whole blood eQTL, OR: 0.660, 95% CI: 0.476-0.913, P = .012). Drug repurposing analyses support the possibility that etanercept and asciminib hydrochloride may treat neonatal respiratory distress by activating LTBR. This study demonstrated that LTBR, NAAA, and CSNK1G2 may serve as promising biomarkers and therapeutic targets for NRDS.

PMID:40388790 | DOI:10.1097/MD.0000000000042411

Circulation. 2025 May 20;151(20):1449-1450. doi: 10.1161/CIRCULATIONAHA.125.074556. Epub 2025 May 19.

NO ABSTRACT

PMID:40388510 | DOI:10.1161/CIRCULATIONAHA.125.074556

J Vis Exp. 2025 May 2;(219). doi: 10.3791/67831.

ABSTRACT

RASopathies are genetic syndromes caused by ERK hyperactivation and resulting in multisystemic diseases that can also lead to cancer predisposition. Despite a broad genetic heterogeneity, germline gain-of-function mutations in key regulators of the RAS-MAPK pathway underlie the majority of the cases, and, thanks to advanced sequencing techniques, potentially pathogenic variants affecting the RAS-MAPK pathway continue to be identified. Functional validation of the pathogenicity of these variants, essential for accurate diagnosis, requires fast and reliable protocols, preferably in vivo. Given the scarcity of effective treatments in early childhood, such protocols, especially if scalable in cost-effective animal models, can be instrumental in offering a preclinical ground for drug repositioning/repurposing. Here we describe step-by-step the protocol for rapid generation of transient RASopathy models in zebrafish embryos and direct inspection of live disease-associated ERK activity changes occurring already during gastrulation through real-time multispectral Förster resonance energy transfer (FRET) imaging. The protocol uses a transgenic ERK reporter recently established and integrated with the hardware of commercial microscopes. We provide an example application for Noonan syndrome (NS) zebrafish models obtained by expression of the Shp2D61G. We describe a straightforward method that enables registration of ERK signal change in the NS fish model before and after pharmacological signal modulation by available low-dose MEK inhibitors. We detail how to generate, retrieve, and assess ratiometric FRET signals from multispectral acquisitions before and after treatment and how to cross-validate the results via classical immunofluorescence on whole embryos at early stages. We then describe how, via examining standard morphometric parameters, to query late changes in embryo shape, indicative of a resulting impairment of gastrulation, in the same embryos whose ERK activity is assessed by live FRET at 6 h post fertilization.

PMID:40388378 | DOI:10.3791/67831

Pain. 2025 May 20. doi: 10.1097/j.pain.0000000000003656. Online ahead of print.

ABSTRACT

The peripheral nervous system (PNS) plays a critical role in pathological conditions, including chronic pain disorders, that manifest differently in men and women. To investigate this sexual dimorphism at the molecular level, we integrated quantitative proteomic profiling of human dorsal root ganglia (hDRG) and peripheral nerve tissue into the expanding omics framework of the PNS. Using data-independent acquisition (DIA) mass spectrometry, we characterized a comprehensive proteomic profile, validating tissue-specific differences between the hDRG and peripheral nerve. Through multi-omic analyses and in vitro functional assays, we identified sex-specific molecular differences, with TNFα signalling emerging as a key sexually dimorphic pathway with higher prominence in men. Genetic evidence from genome-wide association studies further supports the functional relevance of TNFα signalling in the periphery, while clinical trial data and meta-analyses indicate a sex-dependent response to TNFα inhibitors. Collectively, these findings underscore a functionally sexual dimorphism in the PNS, with direct implications for sensory and pain-related clinical translation.

PMID:40388638 | DOI:10.1097/j.pain.0000000000003656

Syst Biol. 2025 May 19:syaf037. doi: 10.1093/sysbio/syaf037. Online ahead of print.

ABSTRACT

Batesian mimicry is an impressive example of convergent evolution driven by predation. However, the observation that many mimics only superficially resemble their models despite strong selective pressures is an apparent paradox. Here, we tested the 'perfecting hypothesis', that posits that inaccurate mimicry may represent a transitional stage at the macro-evolutionary scale by performing the hereto largest phylogenetic analysis (in terms of the number of taxa and genetic data) of ant-mimicking spiders across two speciose but independent clades, the jumping spider tribe Myrmarachnini (Salticidae) and the sac spider sub-family Castianeirinae (Corinnidae). We found that accurate ant mimicry evolved in a gradual process in both clades, by an integration of compound traits contributing to the ant-like habitus with each trait evolving at different speeds. Accurate states were highly unstable at the macro-evolutionary scale likely because strong expression of some of these traits comes with high fitness costs. Instead, the inferred global optimum of mimicry expression was at an inaccurate state. This result reverses the onus of explanation from inaccurate mimicry to explaining the exceptional evolution and maintenance of accurate mimicry and highlights that the evolution of Batesian mimicry is ruled by multiple conflicting selective pressures.

PMID:40388318 | DOI:10.1093/sysbio/syaf037

Physiol Genomics. 2025 May 19. doi: 10.1152/physiolgenomics.00183.2024. Online ahead of print.

ABSTRACT

Background: Risk factors for cardiac arrhythmias that can cause sudden death and heart failure include genetics, age, lifestyle, and other environmental factors. Objectives: The study assessed electrocardiography (ECG) traits in BXD mice and explored associated quantitative trait loci (QTLs). Methods: Five-minute electrocardiograms were recorded in 44 BXD strains at 4-5 months of age (n≥5 mice/sex/strain). ECG and arrhythmia traits were associated with echocardiography, blood pressure, genome and heart transcriptome data followed by expression QTL mapping. Results: A significant variability in ECG parameters and arrhythmias were recorded among BXDs. Among male BXDs, QRS duration was significantly associated with increased left ventricular internal diameter (LVID) and reduced ejection fraction and fractional shortening, while premature ventricular contractions (PVCs) were correlated with LVID, LV volumes and pulmonary vein peak pressure. In female BXDs, PVCs and premature atrial contractions (PACs) significantly related with right ventricular ID and cardiac output. One significant QTL associated with QTc and JT durations was identified on Chromosome (Chr) 3 in male BXDs, while Chr 9 locus was suggestive for association with QTc and QT intervals in female mice. Gon4l was predicted as a strong candidate gene associated with repolarization abnormalities including short or long QT syndromes in humans. Conclusions: Study results suggested an influence of genetic background on expression of ECG parameters and arrhythmias based on significant variations of those traits between mouse strains of the BXD family. We conclude that murine BXD family can serve as a valuable reference for systems biology and comparative predictions of arrhythmia disorders.

PMID:40388294 | DOI:10.1152/physiolgenomics.00183.2024

Expert Opin Drug Discov. 2025 May 19. doi: 10.1080/17460441.2025.2508866. Online ahead of print.

ABSTRACT

INTRODUCTION: Artificial intelligence (AI) has emerged as a transformative tool in drug discovery, particularly in virtual screening (VS), which is a crucial initial step in identifying potential drug candidates. This article highlights the significance of AI in revolutionizing both ligand-based virtual screening (LBVS) and structure-based virtual screening (SBVS) approaches, streamlining and enhancing the drug discovery process.

AREAS COVERED: The authors provide an overview of AI applications in drug discovery, with a focus on LBVS and SBVS approaches utilized in prospective cases where new bioactive molecules were identified and experimentally validated. Discussion includes the use of AI in quantitative structure-activity relationship (QSAR) modeling for LBVS, as well as its role in enhancing SBVS techniques such as molecular docking and molecular dynamics simulations. The article is based on literature searches on all studies published up to March 2025.

EXPERT OPINION: AI is rapidly transforming VS in drug discovery, by leveraging increasing amounts of experimental data and expanding its scalability. These innovations promise to enhance efficiency and precision across both LBVS and SBVS approaches, yet challenges such as data curation, rigorous and prospective validation of new models, and efficient integration with experimental methods remain critical for realizing AI's full potential in drug discovery.

PMID:40388244 | DOI:10.1080/17460441.2025.2508866

Rhinology. 2025 May 19. doi: 10.4193/Rhin25.044. Online ahead of print.

ABSTRACT

BACKGROUND: Sinonasal pathology can be complex and requires a systematic and meticulous approach. Artificial Intelligence (AI) has the potential to improve diagnostic accuracy and efficiency in sinonasal imaging, but its clinical applicability remains an area of ongoing research. This systematic review evaluates the methodologies and clinical relevance of AI in detecting sinonasal pathology through radiological imaging.

METHODOLOGY: Key search terms included "artificial intelligence," "deep learning," "machine learning," "neural network," and "paranasal sinuses,". Abstract and full-text screening was conducted using predefined inclusion and exclusion criteria. Data were extracted on study design, AI architectures used (e.g., Convolutional Neural Networks (CNN), Machine Learning classifiers), and clinical characteristics, such as imaging modality (e.g., Computed Tomography (CT), Magnetic Resonance Imaging (MRI)).

RESULTS: A total of 53 studies were analyzed, with 85% retrospective, 68% single-center, and 92.5% using internal databases. CT was the most common imaging modality (60.4%), and chronic rhinosinusitis without nasal polyposis (CRSsNP) was the most studied condition (34.0%). Forty-one studies employed neural networks, with classification as the most frequent AI task (35.8%). Key performance metrics included Area Under the Curve (AUC), accuracy, sensitivity, specificity, precision, and F1-score. Quality assessment based on CONSORT-AI yielded a mean score of 16.0 ± 2.

CONCLUSIONS: AI shows promise in improving sinonasal imaging interpretation. However, as existing research is predominantly retrospective and single-center, further studies are needed to evaluate AI's generalizability and applicability. More research is also required to explore AI's role in treatment planning and post-treatment prediction for clinical integration.

PMID:40388840 | DOI:10.4193/Rhin25.044

Proc Natl Acad Sci U S A. 2025 May 27;122(21):e2412261122. doi: 10.1073/pnas.2412261122. Epub 2025 May 19.

ABSTRACT

Enhancing the reliability and reproducibility of optical microscopy by reducing specimen irradiance continues to be an important biotechnology target. As irradiance levels are reduced, however, the particle nature of light is heightened, giving rise to Poisson noise, or photon sparsity that restricts only a few (0.5%) image pixels to comprise a photon. Photon sparsity can be addressed by collecting approximately 200 photons per pixel; this, however, requires long acquisitions and, as such, suboptimal imaging rates. Here, we introduce near-zero photon bioimaging, a method that operates at kHz rates and 10,000-fold lower irradiance than standard microscopy. To achieve this level of performance, we uniquely combined a judiciously designed epifluorescence microscope enabling ultralow background levels and AI that learns to reconstruct biological images from as low as 0.01 photons per pixel. We demonstrate that near-zero photon bioimaging captures the structure of multicellular and subcellular features with high fidelity, including features represented by nearly zero photons. Beyond optical microscopy, the near-zero photon bioimaging paradigm can be applied in remote sensing, covert applications, and biomedical imaging that utilize damaging or quantum light.

PMID:40388622 | DOI:10.1073/pnas.2412261122

PLoS One. 2025 May 19;20(5):e0310230. doi: 10.1371/journal.pone.0310230. eCollection 2025.

ABSTRACT

The rapid growth of Android applications has led to an increase in security threats, while traditional detection methods struggle to combat advanced malware, such as polymorphic and metamorphic variants. To address these challenges, this study introduces a hybrid deep learning model (DBN-GRU) that integrates Deep Belief Networks (DBN) for static analysis and Gated Recurrent Units (GRU) for dynamic behavior modeling to enhance malware detection accuracy and efficiency. The model extracts static features (permissions, API calls, intent filters) and dynamic features (system calls, network activity, inter-process communication) from Android APKs, enabling a comprehensive analysis of application behavior.The proposed model was trained and tested on the Drebin dataset, which includes 129,013 applications (5,560 malware and 123,453 benign).Performance evaluation against NMLA-AMDCEF, MalVulDroid, and LinRegDroid demonstrated that DBN-GRU achieved 98.7% accuracy, 98.5% precision, 98.9% recall, and an AUC of 0.99, outperforming conventional models.In addition, it exhibits faster preprocessing, feature extraction, and malware classification times, making it suitable for real-time deployment.By bridging static and dynamic detection methodologies, the DBN-GRU enhances malware detection capabilities while reducing false positives and computational overhead.These findings confirm the applicability of the proposed model in real-world Android security applications, offering a scalable and high-performance malware detection solution.

PMID:40388500 | DOI:10.1371/journal.pone.0310230

PLoS One. 2025 May 19;20(5):e0313692. doi: 10.1371/journal.pone.0313692. eCollection 2025.

ABSTRACT

Surveillance systems are integral to ensuring public safety by detecting unusual incidents, yet existing methods often struggle with accuracy and robustness. This study introduces an advanced framework for anomaly recognition in surveillance, leveraging deep learning to address these challenges and achieve significant improvements over current techniques. The framework begins with preprocessing input images using histogram equalization to enhance feature visibility. It then employs two DCNNs for feature extraction: a novel 63-layer CNN, "Up-to-the-Minute-Net," and the established Inception-Resnet-v2. The features extracted by both models are fused and optimized through two sophisticated feature selection techniques: Dragonfly and Genetic Algorithm (GA). The optimization process involves rigorous experimentation with 5- and 10-fold cross-validation to evaluate performance across various feature sets. The proposed approach achieves an unprecedented 99.9% accuracy in 5-fold cross-validation using the GA optimizer with 2500 selected features, demonstrating a substantial leap in accuracy compared to existing methods. This study's contribution lies in its innovative combination of deep learning models and advanced feature optimization techniques, setting a new benchmark in the field of anomaly recognition for surveillance systems and showcasing the potential for practical real-world applications.

PMID:40388481 | DOI:10.1371/journal.pone.0313692

PLoS One. 2025 May 19;20(5):e0324047. doi: 10.1371/journal.pone.0324047. eCollection 2025.

ABSTRACT

This study aims to develop a predictive hybrid model for a grid-connected PV system with DC-DC optimizers, designed to operate in extreme altitude conditions at 3800 m above sea level. This approach seeks to address the "curse of dimensionality" by reducing model complexity and improving its accuracy by combining the recursive feature removal (RFE) method with advanced regularization techniques, such as Lasso, Ridge, and Bayesian Ridge. The research used a photovoltaic system composed of monocrystalline modules, DC-DC optimizers and a 3000 W inverter. The data obtained from the system were divided into training and test sets, where RFE identified the most relevant variables, eliminating the reactive power of AC. Subsequently, the three regularization models were trained with these selected variables and evaluated using metrics such as precision, mean absolute error, mean square error and coefficient of determination. The results showed that RFE - Bayesian Ridge obtained the highest accuracy (0.999935), followed by RFE - Ridge, while RFE - Lasso had a slightly lower performance and also obtained an exceptionally low MASE (0.0034 for Bayesian and Ridge, compared to 0.0065 for Lasso). All models complied with the necessary statistical validations, including linearity, error normality, absence of autocorrelation and homoscedasticity, which guaranteed their reliability. This hybrid approach proved effective in optimizing the predictive performance of PV systems under challenging conditions. Future work will explore the integration of these models with energy storage systems and smart control strategies to improve operational stability. In addition, the application of the hybrid model in extreme climates, such as desert or polar areas, will be investigated, as well as its extension through deep learning techniques to capture non-linear relationships and increase adaptability to abrupt climate variations.

PMID:40388424 | DOI:10.1371/journal.pone.0324047