Biosens Bioelectron. 2025 Jul 3;287:117747. doi: 10.1016/j.bios.2025.117747. Online ahead of print.

ABSTRACT

Surface-enhanced Raman spectroscopy (SERS) has emerged as a powerful technique for bacterial detection, offering high sensitivity and molecular-level specificity. However, conventional label-free SERS methods relie on the spontaneous adsorption of limited chemical components onto the SERS substrate. Here we developed a multidimensional SERS biosensor capable of capturing more comprehensive information through substrate surface modifications. By employing molecular modifiers with distinct chemical characteristics, we modulated the selective adsorption behaviors of bacterial components, enhancing the diversity of physicochemical interactions at the sensing interface. The physicochemical properties of the nanomaterials were characterized using UV-vis spectroscopy, scanning electron microscopy (SEM), dynamic light scattering (DLS), and zeta potential analysis. A database comprising 119,000 SERS profiles from 17 bacterial strains across seven dimensions was constructed. The 1D-convolutional neural network (1D-CNN) model was utilized to analyze 127 dimensional combinations, achieving a maximum accuracy of 99.29 %. The results demonstrate the capability of the multidimensional SERS biosensor to enhance bacterial identification accuracy by leveraging the rich biochemical diversity captured across multiple dimensions. Nevertheless, optimization of the dimensionality is necessary to mitigate problems such as redundancy and overfitting during data processing.

PMID:40617028 | DOI:10.1016/j.bios.2025.117747

Comput Med Imaging Graph. 2025 Jul 3;124:102595. doi: 10.1016/j.compmedimag.2025.102595. Online ahead of print.

ABSTRACT

Low-dose CT (LDCT) significantly reduces the radiation dose received by patients, however, dose reduction introduces additional noise and artifacts. Currently, denoising methods based on convolutional neural networks (CNNs) face limitations in long-range modeling capabilities, while Transformer-based denoising methods, although capable of powerful long-range modeling, suffer from high computational complexity. Furthermore, the denoised images predicted by deep learning-based techniques inevitably exhibit differences in noise distribution compared to normal-dose CT (NDCT) images, which can also impact the final image quality and diagnostic outcomes. This paper proposes CT-Mamba, a hybrid convolutional State Space Model for LDCT image denoising. The model combines the local feature extraction advantages of CNNs with Mamba's strength in capturing long-range dependencies, enabling it to capture both local details and global context. Additionally, we introduce an innovative spatially coherent Z-shaped scanning scheme to ensure spatial continuity between adjacent pixels in the image. We design a Mamba-driven deep noise power spectrum (NPS) loss function to guide model training, ensuring that the noise texture of the denoised LDCT images closely resembles that of NDCT images, thereby enhancing overall image quality and diagnostic value. Experimental results have demonstrated that CT-Mamba performs excellently in reducing noise in LDCT images, enhancing detail preservation, and optimizing noise texture distribution, and exhibits higher statistical similarity with the radiomics features of NDCT images. The proposed CT-Mamba demonstrates outstanding performance in LDCT denoising and holds promise as a representative approach for applying the Mamba framework to LDCT denoising tasks.

PMID:40616952 | DOI:10.1016/j.compmedimag.2025.102595

JACC Adv. 2025 Jul 4;4(8):101901. doi: 10.1016/j.jacadv.2025.101901. Online ahead of print.

ABSTRACT

BACKGROUND: Delays in the diagnosis of transthyretin amyloid cardiomyopathy (ATTR-CM) contribute to the significant morbidity of the condition, especially in the era of disease-modifying therapies. Screening for ATTR-CM with artificial intelligence and other algorithms may improve timely diagnosis, but these algorithms have not been directly compared.

OBJECTIVES: The aim of this study was to compare the performance of 4 algorithms for ATTR-CM detection in a heart failure population and assess the risk for harms due to model bias.

METHODS: We identified patients in an integrated health system from 2010 to 2022 with ATTR-CM and age- and sex-matched them to controls with heart failure to target 5% prevalence. We compared the performance of a claims-based random forest model (Huda et al model), a regression-based score (Mayo ATTR-CM), and 2 deep learning echo models (EchoNet-LVH and EchoGo Amyloidosis). We evaluated for bias using standard fairness metrics.

RESULTS: The analytical cohort included 176 confirmed cases of ATTR-CM and 3,192 control patients with 79.2% self-identified as White and 9.0% as Black. The Huda et al model performed poorly (AUC: 0.49). Both deep learning echo models had a higher AUC when compared to the Mayo ATTR-CM Score (EchoNet-LVH 0.88; EchoGo Amyloidosis 0.92; Mayo ATTR-CM Score 0.79; DeLong P < 0.001 for both). Bias auditing met fairness criteria for equal opportunity among patients who identified as Black.

CONCLUSIONS: Deep learning, echo-based models to detect ATTR-CM demonstrated best overall discrimination when compared to 2 other models in external validation with low risk of harms due to racial bias.

PMID:40616933 | DOI:10.1016/j.jacadv.2025.101901

Commun Biol. 2025 Jul 5;8(1):1007. doi: 10.1038/s42003-025-08422-z.

NO ABSTRACT

PMID:40617886 | DOI:10.1038/s42003-025-08422-z

Nat Commun. 2025 Jul 6;16(1):6226. doi: 10.1038/s41467-025-61575-7.

ABSTRACT

Discovering governing equations of complex network dynamics is a fundamental challenge in contemporary science with rich data, which can uncover the hidden patterns and mechanisms of the formation and evolution of complex phenomena in various fields and assist in decision-making. In this work, we develop a universal computational tool that can automatically, efficiently, and accurately learn the symbolic patterns of changes in complex system states by combining the excellent fitting capability of deep learning with the equation inference ability of pre-trained symbolic regression. We perform extensive and intensive experimental verifications on more than ten representative scenarios from fields such as physics, biochemistry, ecology, and epidemiology. The results demonstrate the remarkable effectiveness and efficiency of our tool compared to state-of-the-art symbolic regression techniques for network dynamics. The application to real-world systems including global epidemic transmission and pedestrian movements has verified its practical applicability. We believe that our tool can serve as a universal solution to dispel the fog of hidden mechanisms of changes in complex phenomena, advance toward interpretability, and inspire further scientific discoveries.

PMID:40617863 | DOI:10.1038/s41467-025-61575-7

Sci Rep. 2025 Jul 5;15(1):24088. doi: 10.1038/s41598-025-09727-z.

ABSTRACT

The rapid advancement of the Internet of Medical Things (IoMT) has revolutionized remote healthcare monitoring, enabling real-time disease detection and patient care. This research introduces a novel AI-driven telemedicine framework that integrates IoMT, cloud computing, and wireless sensor networks for efficient healthcare monitoring. A key innovation of this study is the Transformer-based Self-Attention Model (TL-SAM), which enhances disease classification by replacing conventional convolutional layers with transformer layers. The proposed TL-SAM framework effectively extracts spatial and spectral features from patient health data, optimizing classification accuracy. Furthermore, the model employs an Improved Wild Horse Optimization with Levy Flight Algorithm (IWHOLFA) for hyperparameter tuning, enhancing its predictive performance. Real-time biosensor data is collected and transmitted to an IoMT cloud repository, where AI-driven analytics facilitate early disease diagnosis. Extensive experimentation on the UCI dataset demonstrates the superior accuracy of TL-SAM compared to conventional deep learning models, achieving an accuracy of 98.62%, precision of 97%, recall of 98%, and F1-score of 97%. The study highlights the effectiveness of AI-enhanced IoMT systems in reducing healthcare costs, improving early disease detection, and ensuring timely medical interventions. The proposed approach represents a significant advancement in smart healthcare, offering a scalable and efficient solution for remote patient monitoring and diagnosis.

PMID:40617852 | DOI:10.1038/s41598-025-09727-z

Nat Commun. 2025 Jul 5;16(1):6211. doi: 10.1038/s41467-025-61599-z.

ABSTRACT

Plastic waste, particularly polyethylene terephthalate (PET), presents significant environmental challenges, driving extensive research into enzymatic biodegradation. However, existing PET hydrolases (PETases) are limited by narrow sequence diversity and suboptimal performance. This study introduces VenusMine, a protein discovery pipeline that integrates protein language models (PLMs) with a representation tree to identify PETases based on structural similarity using sequence information. Using the crystal structure of IsPETase as a template, VenusMine identifies and clusters target proteins. Candidates are further screened using PLM-based assessments of solubility and thermostability, leading to the selection of 34 proteins for biochemical validation. Results reveal that 14 candidates exhibit PET degradation activity across 30-60 °C. Notably, a PET hydrolase from Kibdelosporangium banguiense (KbPETase) demonstrates a melting temperature (Tm) 32 °C higher than IsPETase and exhibits the highest PET degradation activity within 30 - 65 °C among wild-type PETases. KbPETase also surpasses FastPETase and LCC in catalytic efficiency. X-ray crystallography and molecular dynamics simulations show that KbPETase possesses a conserved catalytic domain and enhanced intramolecular interactions, underpinning its improved functionality and thermostability. This work demonstrates a novel deep learning approach for discovering natural PETases with enhanced properties.

PMID:40617831 | DOI:10.1038/s41467-025-61599-z

Sci Data. 2025 Jul 5;12(1):1151. doi: 10.1038/s41597-025-05480-0.

ABSTRACT

As Earth's climate continues to undergo changes, it is imperative to gain understanding of how high-impact, extreme weather events will change. Researchers are increasingly relying on data-driven, learning-based approaches for the detection and tracking of extreme weather events. While several attempts to generate datasets of hand-labeled weather or climate have been made, a significant challenge has been to gather a sufficient number of expert-annotated samples. To address this challenge, we introduce the largest dataset of expert-guided, hand-labeled segmentation masks of extreme weather events. It contains global annotations for atmospheric rivers, tropical cyclones, and atmospheric blocking events from the European Centre for Medium-Range Weather Forecasting's reanalysis version 5. Every timestep for each event is annotated by two separate annotators to bring the total number of labeled timesteps to 49,184. Professional annotators were trained and guided to identify these features by domain-experts, and event-specific experts were consulted for each of the annotation guides. The resulting annotations are demonstrated to have characteristics similar to other methods and those generated directly by domain experts.

PMID:40617821 | DOI:10.1038/s41597-025-05480-0

Sci Data. 2025 Jul 5;12(1):1149. doi: 10.1038/s41597-025-05442-6.

ABSTRACT

Kenya's road network significantly influences environmental and socio-economic dynamics. High-quality road data is essential for analyzing its impact on various factors, including land-use, biodiversity, migration, livelihoods, and economy. Like many countries, Kenya faces challenges in the availability of accurate and detailed digital historical road datasets. To address this, we used deep learning techniques to extract Kenya's road network from 533 historical topographic maps (1:50,000 and 1:100,000 scale) covering the 1950s-1980s. This involved digitizing, georeferencing, and classifying of 20 different road symbols on all maps, then converting and merging them into a seamless dataset. The statistical evaluation was conducted against manually created roads from seven representative map sheets by calculating precision, recall, and F1 score. Our study provides a detailed historical road dataset for Kenya containing over 56,000 km of historical roads. The statistical validation showed an average F1 score of 0.84, indicating a high classification performance. The methodology offers an applicable approach for national-level historic road network mapping, also transferable to other regions, map types or features.

PMID:40617814 | DOI:10.1038/s41597-025-05442-6

Acad Radiol. 2025 Jul 4:S1076-6332(25)00573-2. doi: 10.1016/j.acra.2025.06.024. Online ahead of print.

ABSTRACT

RATIONALE AND OBJECTIVES: Hyperpolarized 129Xenon magnetic resonance imaging (MRI) measures the extent of lung ventilation by ventilation defect percent (VDP), but VDP alone cannot distinguish between diseases. Prior studies have reported anecdotal evidence of disease-specific defect patterns such as wedge-shaped defects in asthma and polka-dot defects in lymphangioleiomyomatosis (LAM). Neural network artificial intelligence can evaluate image shapes and textures to classify images, but this has not been attempted in xenon MRI. We hypothesized that an artificial intelligence network trained on ventilation MRI could classify diseases based on spatial patterns in lung MR images alone.

MATERIALS AND METHODS: Xenon MRI data in six pulmonary conditions (control, asthma, bronchiolitis obliterans syndrome, bronchopulmonary dysplasia, cystic fibrosis, LAM) were used to train convolutional neural networks. Network performance was assessed with top-1 and top-2 accuracy, recall, precision, and one-versus-all area under the curve (AUC). Gradient class-activation-mapping (Grad-CAM) was used to visualize what parts of the images were important for classification.

RESULTS: Training/testing data were collected from 262 participants. The top performing network (VGG-16) had top-1 accuracy=56%, top-2 accuracy=78%, recall=.30, precision=.70, and AUC=.85. The network performed better on larger classes (top-1 accuracy: control=62% [n=57], CF=67% [n=85], LAM=69% [n=61]) and outperformed human observers (human top-1 accuracy=40%, network top-1 accuracy=61% on a single training fold).

CONCLUSION: We developed an artificial intelligence tool that could classify disease from xenon ventilation images alone that outperformed human observers. This suggests that xenon images have additional, disease-specific information that could be useful for cases that are clinically challenging or for disease phenotyping.

PMID:40617705 | DOI:10.1016/j.acra.2025.06.024

Semin Hematol. 2025 Jun 16:S0037-1963(25)00027-7. doi: 10.1053/j.seminhematol.2025.06.002. Online ahead of print.

ABSTRACT

The landscape of acute myeloid leukemia (AML) diagnostics is undergoing a pivotal shift towards a transformative era, driven by the integration of artificial intelligence (AI). This review delves into the pivotal role of AI in reshaping AML diagnostics in the 21st century, highlighting advancements, challenges, and future prospects. AML, marked by the immediate need for accurate diagnosis and treatment, requires precise analysis against the complexity of various diagnostic methods such as cytomorphology, immunophenotyping, cytogenetics, and molecular testing. The introduction of AI in this field promises to address the critical need for rapid and standardized diagnostics, thereby enhancing patient care. AI technologies, including deep learning (DL) and machine learning (ML), are revolutionizing the interpretation of complex diagnostic data. With the use of AI-based models such as deep learning (DL) classifiers or automated karyotyping, promising tools do already exist. When it comes to reporting and reasoning, large language models (LLM) show their potential in efficient data processing and better clinical decision-making. This includes the use of large language models (LLMs) for generating comprehensive diagnostic reports that integrate multi-layered diagnostic information. However, there is a critical need for transparency and interpretability in AI-driven diagnostics. Explainable AI (XAI) models address this need building trust among clinicians and patients. Moreover, this review addresses the growing field of synthetic data that are becoming increasingly accessible due to advances in AI and computational technology. While synthetic data present a promising avenue for augmenting clinical research and potentially optimizing clinical trials in fields such as AML, their application requires careful ethical, regulatory, and methodological considerations. There are several limitations and challenges to consider regarding not only synthetic data but also AI models in general. This includes regulatory hurdles due to the dynamic nature of AI, as well as data privacy concerns and interoperability between different systems. In conclusion, AI has the potential to completely change how we diagnose and treat AML by offering faster, more accurate, and more comprehensive diagnostic insights. This potential is especially crucial for preserving knowledge in times of shortages of human experts. However, realizing this potential will require overcoming significant challenges and fostering collaboration between technologists and clinicians. As we move forward, the synergy between AI and human expertise will undoubtedly redefine the landscape of AML diagnostics, leading in a new era of precision medicine in hematology.

PMID:40617702 | DOI:10.1053/j.seminhematol.2025.06.002

Life Sci. 2025 Jul 3:123835. doi: 10.1016/j.lfs.2025.123835. Online ahead of print.

ABSTRACT

AIMS: Human Immunodeficiency Virus (HIV) remains a critical global health concern due to its impact on the immune system and its progression to Acquired Immunodeficiency Syndrome (AIDS) if untreated. While antiretroviral therapy has advanced significantly, challenges such as drug resistance, adverse effects, and viral mutation necessitate the development of novel therapeutic strategies. This study aims to improve HIV bioactivity prediction and provide interpretable insights into molecular determinants influencing bioactivity.

MATERIALS AND METHODS: We propose MPNN-CWExplainer, a novel graph-based deep learning framework for molecular property prediction. The model integrates a Message Passing Neural Network (MPNN) with a class-weighted loss function to effectively address class imbalance in HIV datasets. Furthermore, GNNExplainer is incorporated to provide post-hoc interpretability by identifying key atom- and bond-level substructures contributing to model predictions. Model robustness is ensured through Bayesian hyperparameter optimization and multiple independent runs.

KEY FINDINGS: MPNN-CWExplainer achieved state-of-the-art predictive performance on the HIV dataset, with an AUC-ROC of 87.631 % and AUC-PRC of 86.02 %, surpassing existing baseline models. The class-weighted approach enhanced minority class representation, and GNNExplainer successfully highlighted chemically meaningful substructures correlating with bioactivity.

SIGNIFICANCE: The proposed framework not only improves prediction accuracy for HIV bioactivity but also enhances transparency and interpretability, crucial for medicinal chemists in understanding model behaviour. MPNN-CWExplainer serves as a robust and interpretable tool for computational drug discovery, supporting informed decision-making in lead optimization and molecular design.

PMID:40617525 | DOI:10.1016/j.lfs.2025.123835

Neurol Res. 2025 Jul 5:1-16. doi: 10.1080/01616412.2025.2527899. Online ahead of print.

ABSTRACT

BACKGROUND: Multiple Sclerosis (MS) is a chronic neurological disorder affecting millions worldwide. Early detection is vital to prevent long-term disability. Magnetic Resonance Imaging (MRI) plays a crucial role in MS diagnosis, yet differentiating MS lesions from other brain anomalies remains a complex challenge.

OBJECTIVE: To develop and evaluate a novel deep learning framework-2DRK-MSCAN-for the early and accurate detection of MS lesions using MRI data.

METHODS: The proposed approach is validated using three publicly available MRI-based brain tumor datasets and comprises three main stages. First, Gradient Domain Guided Filtering (GDGF) is applied during pre-processing to enhance image quality. Next, an EfficientNetV2L backbone embedded within a U-shaped encoder-decoder architecture facilitates precise segmentation and rich feature extraction. Finally, classification of MS lesions is performed using the 2DRK-MSCAN model, which incorporates deep diffusion residual kernels and multiscale snake convolutional attention mechanisms to improve detection accuracy and robustness.

RESULTS: The proposed framework achieved 99.9% accuracy in cross-validation experiments, demonstrating its capability to distinguish MS lesions from other anomalies with high precision.

CONCLUSION: The 2DRK-MSCAN framework offers a reliable and effective solution for early MS detection using MRI. While clinical validation is ongoing, the method shows promising potential for aiding timely intervention and improving patient care.

PMID:40616778 | DOI:10.1080/01616412.2025.2527899

Med Biol Eng Comput. 2025 Jul 5. doi: 10.1007/s11517-025-03405-y. Online ahead of print.

ABSTRACT

Knee osteoarthritis (KOA) is a progressive degenerative joint disease and a leading cause of disability worldwide. Manual diagnosis of KOA from X-ray images is subjective and prone to inter- and intra-observer variability, making early detection challenging. While deep learning (DL)-based models offer automation, they often require large labeled datasets, lack interpretability, and do not provide quantitative feature measurements. Our study presents an automated KOA severity assessment system that integrates a pretrained DL model with image processing techniques to extract and quantify key KOA imaging biomarkers. The pipeline includes contrast limited adaptive histogram equalization (CLAHE) for contrast enhancement, DexiNed-based edge extraction, and thresholding for noise reduction. We design customized algorithms that automatically detect and quantify joint space narrowing (JSN) and osteophytes from the extracted edges. The proposed model quantitatively assesses JSN and finds the number of intercondylar osteophytes, contributing to severity classification. The system achieves accuracies of 88% for JSN detection, 80% for osteophyte identification, and 73% for KOA classification. Its key strength lies in eliminating the need for any expensive training process and, consequently, the dependency on labeled data except for validation. Additionally, it provides quantitative data that can support classification in other OA grading frameworks.

PMID:40616750 | DOI:10.1007/s11517-025-03405-y

J Cancer Res Ther. 2025 Apr 1;21(3):538-546. doi: 10.4103/jcrt.jcrt_2524_24. Epub 2025 Jul 5.

ABSTRACT

Cervical cancer remains a significant global health challenge, with high rates of recurrence and mortality, particularly in low-resource regions. Effective prediction of recurrence and survival is crucial for optimizing treatment and improving patient outcomes. Recently, artificial intelligence (AI) has emerged as a transformative tool in oncology, providing advanced methodologies for analyzing large-scale medical data and offering predictive insights into patient outcomes. This review comprehensively explores the role of AI in predicting cervical cancer recurrence and survival, focusing on techniques such as machine learning, deep learning, and natural language processing. The integration of AI with medical imaging, genomics, and clinical data is discussed, along with the associated challenges and limitations. Future directions and the potential impact of AI on personalized medicine in cervical cancer care are also examined.

PMID:40616534 | DOI:10.4103/jcrt.jcrt_2524_24

J Cancer Res Ther. 2025 Apr 1;21(3):523-537. doi: 10.4103/jcrt.jcrt_2006_24. Epub 2025 Jul 5.

ABSTRACT

Cone-beam computed tomography (CBCT) is pivotal in image-guided radiotherapy (IGRT), yet it faces challenges in accurate organ-at-risk (OAR) segmentation, image registration, and dosimetry. Deep learning, particularly Generative Adversarial Networks (GAN) and Deep Convolutional Neural Networks (DCNN) has shown promise in addressing these challenges. This review explores the latest advancements in deep learning-based methodologies for enhancing CBCT application in radiotherapy. GANs have been employed to generate high-fidelity synthetic CT images, improving the accuracy of OAR segmentation and enabling precise dose calculations. DCNNs, on the other hand, have been instrumental in mitigating artifacts, enhancing image quality, and predicting dose distributions with high precision. Studies demonstrate that these techniques significantly improve the accuracy of OAR delineation and registration, leading to better treatment planning and delivery. Integrating deep learning models with traditional CBCT makes it possible to achieve real-time adaptation to anatomical changes and optimize patient-specific treatment protocols. This review highlights key findings, methodological innovations, and clinical implications, underscoring the transformative potential of deep learning in CBCT-based radiotherapy. The evolution of GANs and DCNNs promises to refine dosimetric accuracy and treatment outcomes further, heralding a new era of precision radiotherapy.

PMID:40616533 | DOI:10.4103/jcrt.jcrt_2006_24

J Biomed Mater Res A. 2025 Jul;113(7):e37961. doi: 10.1002/jbm.a.37961.

ABSTRACT

Bioprinting is a growing area in the field of tissue engineering that offers a potential solution to the global shortage of organ transplants. Ensuring high printability is crucial for bioprinting. To better understand printability, a design of experiment model that examines printing speed and pressure in extrusion-based printing was developed. Two biomaterials, hyaluronic acid and sodium alginate, were selected as surrogate biomaterials to understand how rheological properties play a role in printability. Various rheological aspects such as shear-thinning behavior, viscosity, and recovery were investigated. To further evaluate printability, a new method was used that includes deep learning image similarity. The information obtained with the surrogate bioinks was then applied to another biomaterial, methacrylated hyaluronic acid, in combination with corneal keratocytes to demonstrate the successful implementation of the outcome of this design of experiment. As a result of this study, a better understanding of the rheological properties for bioprinting was achieved, leading to a next step towards improving extrusion-based bioprinting, which can be used for a wide range of applications.

PMID:40616386 | DOI:10.1002/jbm.a.37961

Future Sci OA. 2025 Dec;11(1):2527505. doi: 10.1080/20565623.2025.2527505. Epub 2025 Jul 4.

ABSTRACT

Artificial intelligence (AI) has demonstrated remarkable potential in transforming medical diagnostics across various healthcare domains. This paper explores AI applications in cancer detection, dental medicine, brain tumor database management, and personalized treatment planning. AI technologies such as machine learning and deep learning have enhanced diagnostic accuracy, improved data management, and facilitated personalized treatment strategies. In cancer detection, AI-driven imaging analysis aids in early diagnosis and precise treatment decisions. In dental healthcare, AI applications improve oral disease detection, treatment planning, and workflow efficiency. AI-powered brain tumor databases streamline medical data management, enhancing diagnostic precision and research outcomes. Personalized treatment planning benefits from AI algorithms that analyze genetic, clinical, and lifestyle data to recommend tailored interventions. Despite these advancements, AI integration faces challenges related to data privacy, algorithm bias, and regulatory concerns. Addressing these issues requires improved data governance, ethical frameworks, and interdisciplinary collaboration among healthcare professionals, researchers, and policymakers. Through comprehensive validation, educational initiatives, and standardized protocols, AI adoption in healthcare can enhance patient outcomes and optimize clinical decision-making, advancing the future of precision medicine and personalized care.

PMID:40616302 | DOI:10.1080/20565623.2025.2527505

Microsc Res Tech. 2025 Jul 4. doi: 10.1002/jemt.70019. Online ahead of print.

ABSTRACT

The analysis of microscopic characteristics of closed-cell polymeric foams, particularly bead size, is relevant for understanding properties such as thermal insulation, energy absorption, and compressive structural strength of these materials. This study presents an automated method based on Deep Learning models to measure the bead size of Expanded Polystyrene foams in low-resolution micrographs. The results of this approach were compared with manual measurements at two expanded polystyrene foam densities: 8.5 and 24 kg/m3. Hypothesis tests, including Student's t-test, Levene's test, and Mann-Whitney U test, were conducted and showed no significant differences between manual and automatic measurements. Student's t-test and Levene's test indicated that both methods have comparable means and variances, while the Two One-Sided Test confirmed that they were equivalent for bead size measurement. Additionally, the Mann-Whitney U test revealed no differences in medians, and Bland-Altman plot analyses demonstrated no systematic bias between the methods. Taken together, these results suggest that the proposed Deep Learning-based method is a reliable and precise substitute for the manual method in measuring the bead size of expanded polystyrene, making it suitable for practical use in the bead microstructural analysis of expanded polystyrene material.

PMID:40616216 | DOI:10.1002/jemt.70019

Commun Biol. 2025 Jul 4;8(1):1001. doi: 10.1038/s42003-025-08453-6.

ABSTRACT

The blood-brain barrier (BBB) plays a crucial role in maintaining brain homeostasis. During ageing, the BBB undergoes structural alterations. Electron microscopy (EM) is the gold standard for studying the structural alterations of the brain vasculature. However, analysis of EM images is time-intensive and can be prone to selection bias, limiting our understanding of the structural effect of ageing on the BBB. Here, we introduce 3BTRON, a deep learning framework for the automated analysis of electron microscopy images of the BBB. Using age as a readout, we trained and validated our model on a unique dataset (n = 359). We show that the proposed model could confidently identify the BBB of aged mouse brains from young mouse brains across three different brain regions, achieving a sensitivity of 77.8% and specificity of 80.0% post-stratification when predicting on unseen data. Additionally, feature importance methods revealed the spatial features of each image that contributed most to the predictions. These findings demonstrate a new data-driven approach to analysing age-related changes in the architecture of the BBB.

PMID:40615521 | DOI:10.1038/s42003-025-08453-6