Comput Med Imaging Graph. 2025 May 29;124:102573. doi: 10.1016/j.compmedimag.2025.102573. Online ahead of print.

ABSTRACT

Video laryngoscopes have become increasingly vital in tracheal intubation, providing clear imaging that significantly improves success rates, especially for less experienced clinicians. However, accurate recognition of laryngeal structures remains challenging, which is critical for successful first-attempt intubation in emergency situations. This paper presents MPE-UNet, a deep learning model designed for precise segmentation of laryngeal structures from video laryngoscope images, aiming to assist clinicians in performing tracheal intubation more accurately and efficiently. MPE-UNet follows the classic U-Net architecture, which features an encoder-decoder structure and enhances it with advanced modules and innovative techniques at every stage. In the encoder, we designed an improved multi-scale feature extraction module, which better processes complex throat images. Additionally, a pyramid fusion attention module was incorporated into the skip connections, enhancing the model's ability to capture details by dynamically weighting and merging features from different levels. Moreover, a plug-and-play attention mechanism module was integrated into the decoder, further refining the segmentation process by focusing on important features. The experimental results show that the performance of the proposed method outperforms state-of-the-art methods.

PMID:40472423 | DOI:10.1016/j.compmedimag.2025.102573

J Agric Food Chem. 2025 Jun 5. doi: 10.1021/acs.jafc.5c03646. Online ahead of print.

ABSTRACT

The lack of suitable chemical research methodologies has hindered the discovery of rational daily diet combinations from large-scale dietary-derived compounds. Three deep learning models based on chemical properties for α-glucosidase inhibitors (AGIs), safety, and drug-drug interaction (DDI) were trained. The trained models screened potential AGIs from the FooDB database (approximately 70,000 food-derived compounds) and analyzed the interactions of the selected AGIs. 59 of the 75 selected AGIs from the FooDB database had not been reported before. Betulinic acid in combination with taraxasterol, betulin, and lupeol (all selected from the potential 75 AGIs) was predicted to have a synergistic effect in enhancing the inhibition of α-glucosidase, which was further confirmed by in vitro assays. These collective findings strongly suggest that the potential of deep learning methods based on chemical properties in solving the food chemistry research challenge of developing reasonable daily diet combinations.

PMID:40472393 | DOI:10.1021/acs.jafc.5c03646

Obstet Gynecol. 2025 Jun 5. doi: 10.1097/AOG.0000000000005944. Online ahead of print.

ABSTRACT

Publications on artificial intelligence (AI) applications have dramatically increased for most medical specialties, including obstetrics. Here, we review the most recent pertinent publications on AI programs in obstetrics, describe trends in AI applications for specific obstetric problems, and assess AI's possible effects on obstetric care. Searches were performed in PubMed (MeSH), MEDLINE, Ovid, ClinicalTrials.gov, Google Scholar, and Web of Science using a combination of keywords and text words related to "obstetrics," "pregnancy," "artificial intelligence," "machine learning," "deep learning," and "neural networks," for articles published between June 1, 2019, and May 31, 2024. A total of 1,768 articles met at least one search criterion. After eliminating reviews, duplicates, retractions, inactive research protocols, unspecified AI programs, and non-English-language articles, 207 publications remained for further review. Most studies were conducted outside of the United States, were published in nonobstetric journals, and focused on risk prediction. Study population sizes ranged widely from 10 to 953,909, and model performance abilities also varied widely. Evidence quality was assessed by the description of model construction, predictive accuracy, and whether validation had been performed. Most studies had patient groups differing considerably from U.S. populations, rendering their generalizability to U.S. patients uncertain. Artificial intelligence ultrasound applications focused on imaging issues are those most likely to influence current obstetric care. Other promising AI models include early risk screening for spontaneous preterm birth, preeclampsia, and gestational diabetes mellitus. The rate at which AI studies are being performed virtually guarantees that numerous applications will eventually be introduced into future U.S. obstetric practice. Very few of the models have been deployed in obstetric practice, and more high-quality studies are needed with high predictive accuracy and generalizability. Assuming these conditions are met, there will be an urgent need to educate medical students, postgraduate trainees and practicing physicians to understand how to effectively and safely implement this technology.

PMID:40472381 | DOI:10.1097/AOG.0000000000005944

PLoS One. 2025 Jun 5;20(6):e0324158. doi: 10.1371/journal.pone.0324158. eCollection 2025.

ABSTRACT

In industrial societies, most fishery research institutes collect the phenotypic data of fish manually, which is time-consuming, labor-intensive, error-prone, and results in incomplete data. Considering their stress reaction and the natural body extension to collect the phenotypic data of fish quickly and accurately, channel catfish was used as the research subject and a deep-learning-based method was developed to explore their phenotypic data, i.e., body length, full length, head length, body height, tail handle width, tail handle height, and body thickness. First, this study applied two cameras and another device built into an image acquisition system to obtain images of fish in the water. We then adopted an Hourglass module network to position nine and ten key points on the top and side view images, building two key point fish skeletons. Finally, 3D coordinate transformation and scale parameters were employed to obtain the phenotypic data. Compared with the ground truth of the phenotypic fish data, our study achieved a 3.7% average relative error in terms of the full length, and an average 9.6% relative error for all seven types of phenotypic data applied. Furthermore, the average time required for the image processing measurements was approximately 1s.

PMID:40472062 | DOI:10.1371/journal.pone.0324158

PLoS One. 2025 Jun 5;20(6):e0318795. doi: 10.1371/journal.pone.0318795. eCollection 2025.

ABSTRACT

This study presents an advanced adaptive network steganography paradigm that integrates deep learning methodologies with multimedia video analysis to enhance the universality and security of network steganography practices. The proposed approach utilizes a deep convolutional generative adversarial network-based architecture capable of fine-tuning steganographic parameters in response to the dynamic foreground, stable background, and spatio-temporal complexities of multimedia videos. Empirical evaluations using the MPII and UCF101 video repositories demonstrate that the proposed algorithm outperforms existing methods in terms of steganographic success and resilience. The framework achieves a 95% steganographic success rate and a peak signal-to-noise ratio (PSNR) of 48.3 dB, showing significant improvements in security and steganographic fidelity compared to contemporary techniques. These quantitative results underscore the potential of the approach for practical applications in secure multimedia communication, marking a step forward in the field of network steganography.

PMID:40472042 | DOI:10.1371/journal.pone.0318795

Brief Bioinform. 2025 May 1;26(3):bbaf260. doi: 10.1093/bib/bbaf260.

ABSTRACT

During DNA transcription, the central dogma states that DNA generates corresponding RNA sequences based on the principle of complementary base pairing. However, in the allopolyploid line by goldfish and common carp hybrids, there is a significant level of transcriptional infidelity. To explore deeper into the causes of transcriptional infidelity in this line, we developed a deep learning model to explore its underlying determinants. First, our model can accurately identify transcriptional infidelity sequences at the nucleotide resolution and effectively distinguish transcriptional infidelity regions at the subregional level. Subsequently, we utilized this model to quantitatively assess the importance of position-specific motifs. Furthermore, by integrating the relationship between transcription factors and their recognition motifs, we unveiled the distribution of position-specific transcription factor families and classes that influence transcriptional infidelity in this line. In summary, our study provides new insights into the deeper determinants of transcriptional infidelity in this line.

PMID:40471993 | DOI:10.1093/bib/bbaf260

PLoS One. 2025 Jun 5;20(6):e0310716. doi: 10.1371/journal.pone.0310716. eCollection 2025.

ABSTRACT

This study presents an initial model for bark beetle identification, serving as a foundational step toward developing a fully functional and practical identification tool. Bark beetles are known for extensive damage to forests globally, as well as for uniform and homoplastic morphology which poses identification challenges. Utilizing a MaxViT-based deep learning backbone which utilizes local and global attention to classify bark beetles down to the genus level from images containing multiple beetles. The methodology involves a process of image collection, preparation, and model training, leveraging pre-classified beetle species to ensure accuracy and reliability. The model's F1 score estimates of 0.99 and 1.0 indicates a strong ability to accurately classify genera in the collected data, including those previously unknown to the model. This makes it a valuable first step towards building a tool for applications in forest management and ecological research. While the current model distinguishes among 12 genera, further refinement and additional data will be necessary to achieve reliable species-level identification, which is particularly important for detecting new invasive species. Despite the controlled conditions of image collection and potential challenges in real-world application, this study provides the first model capable of identifying the bark beetle genera, and by far the largest training set of images for any comparable insect group. We also designed a function that reports if a species appears to be unknown. Further research is suggested to enhance the model's generalization capabilities and scalability, emphasizing the integration of advanced machine learning techniques for improved species classification and the detection of invasive or undescribed species.

PMID:40471899 | DOI:10.1371/journal.pone.0310716

JACC Clin Electrophysiol. 2025 May 6:S2405-500X(25)00253-1. doi: 10.1016/j.jacep.2025.04.003. Online ahead of print.

NO ABSTRACT

PMID:40471767 | DOI:10.1016/j.jacep.2025.04.003

IEEE J Biomed Health Inform. 2025 Jun;29(6):4362-4373. doi: 10.1109/JBHI.2025.3541849.

ABSTRACT

Fully-supervised deep neural networks have achieved remarkable progress in medical image segmentation, yet they heavily rely on extensive manually labeled data and exhibit inflexibility for unseen tasks. Few-shot segmentation (FSS) addresses these issues by predicting unseen classes from a few labeled support examples. However, most existing FSS models struggle to generalize to diverse target tasks distinct from training domains. Furthermore, designing promising network architectures for such tasks is expertise-intensive and laborious. In this paper, we introduce MOE-FewSeg, a novel automatic design method for FSS architectures. Specifically, we construct a U-shaped encoder-decoder search space that incorporates capabilities for information interaction and feature selection, thereby enabling architectures to leverage prior knowledge from publicly available datasets across diverse domains for improved prediction of various target tasks. Given the potential conflicts among disparate target tasks, we formulate the multi-task problem as a multi-objective optimization problem. We employ a multi-objective genetic algorithm to identify the Pareto-optimal architectures for these target tasks within this search space. Furthermore, to mitigate the impact of noisy labels due to dataset quality variations, we propose a noise-robust loss function named NRL, which encourages the model to de-emphasize larger loss values. Empirical results demonstrate that MOE-FewSeg outperforms manually designed architectures and other related approaches.

PMID:40471744 | DOI:10.1109/JBHI.2025.3541849

IEEE Trans Biomed Eng. 2025 Jun 5;PP. doi: 10.1109/TBME.2025.3577084. Online ahead of print.

ABSTRACT

OBJECTIVE: Accurate, non-invasive methods for estimating joint and muscle physiological states have the potential to greatly enhance control of wearable devices during real-world ambulation. Traditional modeling approaches and current estimation methods used to predict muscle dynamics often rely on complex equipment or computationally intensive simulations and have difficulty estimating across a broad spectrum of tasks or subjects.

METHODS: Our approach used deep learning (DL) models trained on kinematic inputs to estimate internal physiological states at the knee, including moment, power, velocity, and force. We assessed each model's performance against ground truth labels from both a commonly used, standard OpenSim musculoskeletal model without EMG (static optimization) and an EMG-informed method (CEINMS), across 28 different cyclic and noncyclic tasks.

RESULTS: EMG provided no benefit for joint moment/power estimation (e.g., biological moment), but was critical for estimating muscle states. Models trained with EMG-informed labels but without EMG as an input to the DL system significantly outperformed models trained without EMG (e.g., 33.7% improvement for muscle moment estimation) (p < 0.05). Models that included EMG-informed labels and EMG as a model input demonstrated even higher performance (49.7% improvement for muscle moment estimation) (p < 0.05), but require the availability of EMG during model deployment, which may be impractical.

CONCLUSION/SIGNIFICANCE: While EMG information is not necessary for estimating joint level states, there is a clear benefit during muscle level state estimation. Our results demonstrate excellent tracking of these states with EMG included only during training, highlighting the practicality of real-time deployment of this approach.

PMID:40471740 | DOI:10.1109/TBME.2025.3577084

JAMA Ophthalmol. 2025 Jun 5. doi: 10.1001/jamaophthalmol.2025.1513. Online ahead of print.

ABSTRACT

IMPORTANCE: As on-axis metrics, spherical equivalent refraction (SER) and axial length (AL) are limited in capturing individual-level differences in posterior segment anatomy.

OBJECTIVE: To propose a fundus-level metric-fundus refraction offset (FRO)-and investigate its association with ocular parameters derived from optical coherence tomography (OCT).

DESIGN, SETTING, AND PARTICIPANTS: This cross-sectional, population-based study used data from 45 180 healthy eyes in the UK Biobank (2009-2010). Fundus photographs from a random subset (70%) were used to train a deep learning model to predict SER, with the goal of developing a model that learned to capture the nonpathological variations in fundus appearance from -15.50 D to 9.25 D. The trained model was applied to the remaining subset (internal unseen set) to derive FRO for each eye. FRO was also computed for an external dataset (the Caledonian cohort, 2023-2024) with enhanced depth imaging OCT and AL data for 152 right eyes. Data were analyzed from July to November 2024.

EXPOSURE: FRO, defined as the error in fundus-predicted SER. A more negative FRO indicated a more myopic-looking fundus than typical for an eye with the same SER.

MAIN OUTCOMES AND MEASURES: The association between FRO and macular thickness (MT) was tested using linear mixed-effects regression in the internal unseen set, controlling for SER, age, sex, and race. In the external dataset, the associations of FRO with choroidal area, choroidal vascularity index (CVI), and MT were examined using linear fixed-effects regression, controlling for SER (and subsequently AL) and other aforementioned covariates.

RESULTS: High-quality OCT data were available from 9524 eyes in the internal unseen set and 152 eyes in the external dataset among individuals with a mean (SD) age of 54.5 (8.2) years and 19.3 (3.8) years, respectively. In the internal unseen set, a more negative FRO was independently associated with lower MT (β, 0.64; 95% CI, 0.37-0.90; P < .001). A similar association was observed in the external dataset-whether adjusted for SER (β, 2.45; 95% CI, 0.64-4.26; P = .008) or AL (β, 2.09; 95% CI, 0.28-3.91; P = .02). Additionally, CVI decreased as FRO became more negative-both in the SER-adjusted (β, 0.01; 95% CI, 0.01-0.02; P < .001) and AL-adjusted (β, 0.01, 95% CI, 0.004-0.02; P = .001) analyses.

CONCLUSION AND RELEVANCE: In this study, FRO reflected the individual-level mismatch between SER (or AL) and the anatomical severity of ametropia. This may have prognostic relevance for personalized risk prediction of myopia and its complications.

PMID:40471629 | DOI:10.1001/jamaophthalmol.2025.1513

Interdiscip Sci. 2025 Jun 5. doi: 10.1007/s12539-025-00718-2. Online ahead of print.

ABSTRACT

The medical field uses Magnetic Resonance Imaging (MRI) as an essential diagnostic tool which provides doctors non-invasive images of brain structures and pathological conditions. Brain tumor detection stands as a vital application that needs specific and effective approaches for both medical diagnosis and treatment procedures. The challenges from manual examination of MRI scans stem from inconsistent tumor features including heterogeneity and irregular dimensions which results in inaccurate assessments of tumor size. To address these challenges, this paper proposes an Automated Classification and Grading Diagnosis Model (ACGDM) using MRI images. Unlike conventional methods, ACGDM introduces a Multi-Scale Graph Neural Network (MSGNN), which dynamically captures hierarchical and multi-scale dependencies in MRI data, enabling more accurate feature representation and contextual analysis. Additionally, the Spatio-Temporal Transformer Attention Mechanism (STTAM) effectively models both spatial MRI patterns and temporal evolution by incorporating cross-frame dependencies, enhancing the model's sensitivity to subtle disease progression. By analyzing multi-modal MRI sequences, ACGDM dynamically adjusts its focus across spatial and temporal dimensions, enabling precise identification of salient features. Simulations are conducted using Python and standard libraries to evaluate the model on the BRATS 2018, 2019, 2020 datasets and the Br235H dataset, encompassing diverse MRI scans with expert annotations. Extensive experimentation demonstrates 99.8% accuracy in detecting various tumor types, showcasing its potential to revolutionize diagnostic practices and improve patient outcomes.

PMID:40471519 | DOI:10.1007/s12539-025-00718-2

Med Biol Eng Comput. 2025 Jun 5. doi: 10.1007/s11517-025-03386-y. Online ahead of print.

ABSTRACT

The integration of brain-computer interface (BCI) and virtual reality (VR) systems offers transformative potential for spatial cognition training and assessment. By leveraging artificial intelligence (AI) to analyze electroencephalogram (EEG) data, brain activity patterns during spatial tasks can be decoded with high precision. In this context, a hybrid neural network named MSFHNet is proposed, optimized for extracting spatiotemporal features from spatial cognitive EEG signals. The model employs a hierarchical architecture where its temporal module uses multi-scale dilated convolutions to capture dynamic EEG variations, while its spatial module integrates channel-spatial attention mechanisms to model inter-channel dependencies and spatial distributions. Cross-stacked modules further refine discriminative features through deep-level fusion. Evaluations demonstrate the superiority of MSFHNet in the beta2 frequency band, achieving 98.58% classification accuracy and outperforming existing models. This innovation enhances EEG signal representation, advancing AI-powered BCI-VR systems for robust spatial cognitive training.

PMID:40471491 | DOI:10.1007/s11517-025-03386-y

Clin Rheumatol. 2025 Jun 5. doi: 10.1007/s10067-025-07518-5. Online ahead of print.

ABSTRACT

OBJECTIVES: This study aimed to develop a deep learning-based model to predict the risk of high-risk extra-glandular organ involvement (HR-OI) in patients with Sjogren's syndrome (SS) using whole-slide images (WSI) from labial gland biopsies.

METHODS: We collected WSI data from 221 SS patients. Pre-trained models, including ResNet50, InceptionV3, and EfficientNet-B5, were employed to extract image features. A classification model was constructed using multi-instance learning and ensemble learning techniques.

RESULTS: The ensemble model achieved high area under the receiver operating characteristic (ROC) curve values on both internal and external validation sets, indicating strong predictive performance. Moreover, the model was able to identify key pathological features associated with the risk of HR-OI.

CONCLUSIONS: This study demonstrates that a deep learning-based model can effectively predict the risk of HR-OI in SS patients, providing a novel basis for clinical decision-making. Key Points 1. What is already known on this topic? • Sjogren's syndrome (SS) is a chronic autoimmune disease affecting the salivary and lacrimal glands. • Accurate prediction of high-risk extra-glandular organ involvement (HR-OI) is crucial for timely intervention and improved patient outcomes in SS. • Traditional methods for HR-OI prediction rely on clinical data and lack objectivity. 2. What this study adds? • This study proposes a novel deep learning-based model using whole-slide images (WSI) from labial gland biopsies for predicting HR-OI in SS patients. • Our model utilizes pre-trained convolutional neural networks (CNNs) and a Vision Transformer (ViT) module to extract informative features from WSI data. • The ensemble model achieves high accuracy in predicting HR-OI, outperforming traditional methods. • The model can identify key pathological features in WSI data associated with HR-OI risk. 3. How this study might affect research, practice or policy? • This study provides a novel and objective approach for predicting HR-OI in SS patients, potentially leading to improved clinical decision-making and personalized treatment strategies. • Our findings encourage further investigation into the role of deep learning and WSI analysis in SS diagnosis and risk stratification. • The development of a non-invasive and objective diagnostic tool based on WSI analysis could benefit clinical practice and inform policy decisions regarding patient care for SS.The development of a non-invasive and objective diagnostic tool based on WSI analysis could benefit clinical practice and inform policy decisions regarding patient care for SS.

PMID:40471393 | DOI:10.1007/s10067-025-07518-5

Eur Spine J. 2025 Jun 5. doi: 10.1007/s00586-025-08946-2. Online ahead of print.

ABSTRACT

OBJECTIVES: To develop a deep learning model to automatically measure the curvature-related sagittal parameters on cervical spinal X-ray images.

METHODS: This retrospective study collected a total of 700 lateral cervical spine X-ray images from three hospitals, consisting of 500 training sets, 100 internal test sets, and 100 external test sets. 6 measured parameters and 34 landmarks were measured and labeled by two doctors and averaged as the gold standard. A Convolutional neural network (CNN) model was built by training on 500 images and testing on 200 images. Statistical analysis is used to evaluate labeling differences and model performance.

RESULTS: The percentages of the difference in distance between landmarks within 4 mm were 96.90% (Dr. A vs. Dr. B), 98.47% (Dr. A vs. model), and 97.31% (Dr. B vs. model); within 3 mm were 94.88% (Dr. A vs. Dr. B), 96.43% (Dr. A vs. model), and 94.16% (Dr. B vs. model). The mean difference of the algorithmic model in labeling landmarks was 1.17 ± 1.14 mm. The mean absolute error (MAE) of the algorithmic model for the Borden method, Cervical curvature index (CCI), Vertebral centroid measurement cervical lordosis (CCL), C0-C7 Cobb, C1-C7 Cobb, C2-C7 Cobb in the test sets are 1.67 mm, 2.01%, 3.22°, 2.37°, 2.49°, 2.81°, respectively; symmetric mean absolute percentage error (SMAPE) was 20.06%, 21.68%, 20.02%, 6.68%, 5.28%, 20.46%, respectively. Also, the algorithmic model of the six cervical sagittal parameters is in good agreement with the gold standard (intraclass correlation efficiency was 0.983; p < 0.001).

CONCLUSION: Our deep learning algorithmic model had high accuracy in recognizing the landmarks of the cervical spine and automatically measuring cervical spine-related parameters, which can help radiologists improve their diagnostic efficiency.

PMID:40471336 | DOI:10.1007/s00586-025-08946-2

Eur J Nucl Med Mol Imaging. 2025 Jun 5. doi: 10.1007/s00259-025-07370-4. Online ahead of print.

ABSTRACT

PURPOSE: To investigate the effects of the deep learning model RaDynPET on fourfold reduced-count whole-body PET examinations.

METHODS: A total of 120 patients (84 internal cohorts and 36 external cohorts) undergoing 18F-FDG PET/CT examinations were enrolled. PET images were reconstructed using OSEM algorithm with 120-s (G120) and 30-s (G30) list-mode data. RaDynPET was developed to generate enhanced images (R30) from G30. Two experienced nuclear medicine physicians independently evaluated subjective image quality using a 5-point Likert scale. Standardized uptake values (SUV), standard deviations, liver signal-to-noise ratio (SNR), lesion tumor-to-background ratio (TBR), and contrast-to-noise ratio (CNR) were compared. Subgroup analyses evaluated performance across demographics, and lesion detectability were evaluated using external datasets. RaDynPET was also compared to other deep learning methods.

RESULTS: In internal cohorts, R30 demonstrated significantly higher image quality scores than G30 and G120. R30 showed excellent agreement with G120 for liver and lesion SUV values and surpassed G120 in liver SNR and CNR. Liver SNR and CNR of R30 were comparable to G120 in thin group, and the CNR of R30 was comparable to G120 in young age group. In external cohorts, R30 maintained strong SUV agreement with G120, with lesion-level sensitivity and specificity of 95.45% and 98.41%, respectively. There was no statistical difference in lesion detection between R30 and G120. RaDynPET achieved the highest PSNR and SSIM among deep learning methods.

CONCLUSION: The RaDynPET model effectively restored high image quality while maintaining SUV agreement for 18F-FDG PET scans acquired in 25% of the standard acquisition time.

PMID:40471320 | DOI:10.1007/s00259-025-07370-4

Angew Chem Int Ed Engl. 2025 Jun 5:e202507102. doi: 10.1002/anie.202507102. Online ahead of print.

ABSTRACT

Organic cocrystals, particularly the evolution from binary to higher-order structures, have garnered considerable attention due to their tunable intermolecular interactions and unique material properties. Binary cocrystals, formed through π-π stacking, charge transfer, and hydrogen/halogen bonding, allow for precise control over molecular packing and enhanced optoelectronic properties. In contrast, higher-order cocrystals, incorporating three or more components, enable greater complexity and functional diversity. Strategies such as homologation via isostructural substitution, hierarchical intermolecular interactions and Long-range Synthon Aufbau Modules facilitate the synthesis of these advanced materials. The shift toward higher-order cocrystals paves the way for novel applications in fields such as deep learning for cocrystal prediction, drug design, organic solar cells, and NIR-II photothermal conversion. However, challenges related to molecular screening, ratio optimization, scalable synthesis, and long-term stability remain critical hurdles for the broader implementation of these materials in practical applications.

PMID:40471124 | DOI:10.1002/anie.202507102

Philos Trans A Math Phys Eng Sci. 2025 Jun 5;383(2298):20240236. doi: 10.1098/rsta.2024.0236. Epub 2025 Jun 5.

ABSTRACT

Solving computer vision problems through machine learning, one often encounters lack of sufficient training data. To mitigate this, we propose the use of ensembles of weak learners based on spectral total-variation (STV) features (Gilboa G. 2014 A total variation spectral framework for scale and texture analysis. SIAM J. Imaging Sci. 7, 1937-1961. (doi:10.1137/130930704)). The features are related to nonlinear eigenfunctions of the total-variation subgradient and can characterize well textures at various scales. It was shown (Burger M, Gilboa G, Moeller M, Eckardt L, Cremers D. 2016 Spectral decompositions using one-homogeneous functionals. SIAM J. Imaging Sci. 9, 1374-1408. (doi:10.1137/15m1054687)) that, in the one-dimensional case, orthogonal features are generated, whereas in two dimensions the features are empirically lowly correlated. Ensemble learning theory advocates the use of lowly correlated weak learners. We thus propose here to design ensembles using learners based on STV features. To show the effectiveness of this paradigm, we examine a hard real-world medical imaging problem: the predictive value of computed tomography (CT) data for high uptake in positron emission tomography (PET) for patients suspected of skeletal metastases. The database consists of 457 scans with 1524 unique pairs of registered CT and PET slices. Our approach is compared with deep-learning methods and to radiomics features, showing STV learners perform best (AUC=[Formula: see text]), compared with neural nets (AUC=[Formula: see text]) and radiomics (AUC=[Formula: see text]). We observe that fine STV scales in CT images are especially indicative of the presence of high uptake in PET.This article is part of the theme issue 'Partial differential equations in data science'.

PMID:40471027 | DOI:10.1098/rsta.2024.0236

Rev Sci Instrum. 2025 Jun 1;96(6):065202. doi: 10.1063/5.0256098.

ABSTRACT

The wide range of optical planes in underwater laser imaging results in the presence of a large number of noisy light bars in the background region. Since the shape and intensity of these noisy light bars are very similar to the target information, it is difficult to detect and eliminate them accurately. In this paper, a deep learning algorithm named YOLOv8-FWR is proposed, which can effectively improve the efficiency and quality of underwater laser imaging by combining with laser scanning imaging equipment. First, we introduce a novel pooling module called Focal_SPPF to mitigate the impact of background noise. Second, we propose a weighted feature Concat module to enhance the detection of small target light bars located at the object's edges. Finally, to enhance the model's adaptability for underwater deployment, we optimized the C2f module through structural reparameterization techniques. This approach effectively reduced the model's parameter count while enhancing its accuracy. We constructed a dataset containing a large amount of background noise by simulating the process of underwater laser scanning imaging and evaluated the effectiveness of the augmented model through ablation and comparison experiments. The experimental results indicate that our model outperforms the YOLOv8n by obtaining an 8.6% improvement on mAP50-95 and reducing the parameter count by 37%. A favorable balance between detection accuracy and number of parameters is achieved. Meanwhile, experiments on VOC2012 and the Underwater Detection Dataset (UDD) verify its good generalizability. Finally, we built a rotating line laser scanning imaging system and validated its effectiveness through underwater laser scanning experiments.

PMID:40471019 | DOI:10.1063/5.0256098

Adv Sci (Weinh). 2025 Jun 5:e17408. doi: 10.1002/advs.202417408. Online ahead of print.

ABSTRACT

This study presents BrainFusion, a unified software framework designed to improve reproducibility and support translational applications in multimodal brain-computer interface (BCI) and brain-body interaction research. While ​electroencephalography (EEG)​​-based BCIs have advanced considerably, integrating multimodal physiological signals remains hindered by analytical complexity, limited standardization, and challenges in real-world deployment. BrainFusion addresses these gaps through standardized data structures, automated preprocessing pipelines, cross-modal feature engineering, and integrated machine learning modules. Its application generator further enables streamlined deployment of workflows as standalone executables. Demonstrated in two case studies, BrainFusion achieves 95.5% accuracy in within-subject EEG-functional near-infrared spectroscopy (fNIRS)​​ motor imagery classification using ensemble modeling and 80.2% accuracy in EEG-electrocardiography (ECG)​​ sleep staging using deep learning, with the latter successfully deployed as an executable tool. Supporting EEG, fNIRS, electromyography (EMG)​, and ECG, BrainFusion provides a low-code, visually guided environment, facilitating accessibility and bridging the gap between multimodal research and application in real world.

PMID:40470749 | DOI:10.1002/advs.202417408