Eur J Dermatol. 2024 Dec 1;34(6):595-603. doi: 10.1684/ejd.2024.4804.

ABSTRACT

Advancements in machine learning (ML) are making artificial intelligence more feasible in dermatology, with promising results for diagnosing skin cancers, though few studies cover common or inflammatory dermatoses. To evaluate the diagnostic accuracy for common non-cancerous skin diseases and the clinical applicability of an ML model in practical telemedicine. A prospective, multi-centre, diagnostic accuracy study including patients with common dermatoses, between October 2022 and July 2023, was performed. The top three diagnoses (Top 1, Top 2 and Top 3) from the AI system, trained to recognize 25 common dermatoses based on skin lesion images and medical data, were compared to diagnoses by two dermatologists (gold standard) to calculate the AI model's diagnostic accuracy, sensitivity, and specificity. Two versions of the AI software were evaluated: version 1 (V1) and version 2 (V2) with and without medical supervision (MS), referring to the use of metadata to control diagnostic predictions. Seventy participants and 195 photographs were included. The sensitivity and specificity of the Top 3 algorithm were 88% and 90%, respectively, for V2, with a significant improvement compared with V1. For V1, diagnostic accuracy was 0.57 (0.46;0.69) for Top 1, 0.70 (0.59;0.81) for Top 2, and 0.81 (0.72;0.91) for Top 3. For V2, diagnostic accuracy was 0.69 (0.58;0.79) and 0.71 (0.61;0.82) without and with MS, respectively, for Top 1; 0.87 (0.79;0.95) for Top 2; and 0.90 (0.83;0.97) for Top 3. Our AI model appears to be a promising tool for triaging and diagnosing skin lesions, especially for non-specialist physicians.

PMID:39912464 | DOI:10.1684/ejd.2024.4804

ACS Appl Mater Interfaces. 2025 Feb 6. doi: 10.1021/acsami.4c21563. Online ahead of print.

ABSTRACT

The way people interact with machines through flexible acoustic sensors is revolutionizing the way we live. However, developing a human-machine interaction acoustic sensor that simultaneously offers low cost, high stability, high fidelity, and high sensitivity remains a significant challenge. In this study, a sensor based on a sound-driven triboelectric nanogenerator was proposed. A poly(vinylidene fluoride) (PVDF)/graphene oxide (GO) composite nanofiber film was obtained as the dielectric layer through electrospinning, and copper-nickel alloy conductive fabric was used as the electrode. An imitation embroidery shed structure was designed in the shape of a ring to secure the upper and lower electrodes and the dielectric layer as a whole. Due to the porosity of the electrode, the large contact area of the dielectric layer, and the high stability of the imitation embroidery shed structure, the sensor achieves a sensitivity of 4.76 V·Pa-1 and a frequency response range of 20-2000 Hz. A multilayer attention convolutional network (MLACN) was designed for speech recognition. The designed speech recognition system achieved a 99.5% accuracy rate in recognizing common word pronunciations. The integration of sound-driven triboelectric nanogenerator-based flexible acoustic sensors with deep learning techniques holds great promise in the field of human-machine interaction.

PMID:39912319 | DOI:10.1021/acsami.4c21563

Front Bioeng Biotechnol. 2025 Jan 22;13:1526478. doi: 10.3389/fbioe.2025.1526478. eCollection 2025.

ABSTRACT

BACKGROUND: The high prevalence of low back pain has led to an increasing demand for the analysis of lumbar magnetic resonance (MR) images. This study aimed to develop and evaluate a deep-learning-assisted automated system for diagnosing and grading lumbar intervertebral disc degeneration based on lumbar T2-weighted sagittal and axial MR images.

METHODS: This study included a total of 472 patients who underwent lumbar MR scans between January 2021 and November 2023, with 420 in the internal dataset and 52 in the external dataset. The MR images were evaluated and labeled by experts according to current guidelines, and the results were considered the ground truth. The annotations included the Pfirrmann grading of disc degeneration, disc herniation, and high-intensity zones (HIZ). The automated diagnostic model was based on the YOLOv5 network, modified by adding an attention module in the Cross Stage Partial part and a residual module in the Spatial Pyramid Pooling-Fast part. The model's diagnostic performance was evaluated by calculating the precision, recall, F1 score, and area under the receiver operating characteristic curve.

RESULTS: In the internal test set, the model achieved precisions of 0.78-0.91, 0.90-0.92, and 0.82 and recalls of 0.86-0.91, 0.90-0.93, and 0.81-0.88 for disc degeneration grading, disc herniation diagnosis, and HIZ detection, respectively. In the external test set, the precision values for disc degeneration grading, herniation diagnosis, and HIZ detection were 0.73-0.87, 0.86-0.92, and 0.74-0.84 and recalls were 0.79-0.87, 0.88-0.91, and 0.77-0.78, respectively.

CONCLUSION: The proposed model demonstrated a relatively high diagnostic and classification performance and exhibited considerable consistency with expert evaluation.

PMID:39912111 | PMC:PMC11794261 | DOI:10.3389/fbioe.2025.1526478

MethodsX. 2025 Jan 15;14:103172. doi: 10.1016/j.mex.2025.103172. eCollection 2025 Jun.

ABSTRACT

Plant Disease Detection in the early stage is paramount. Traditionally, it was done manually by the farmers, which is a laborious and time-intensive task. With the advent of AI, Machine Learning and Deep Learning methods are used to detect and categorize plant diseases. However, they rely on extensive datasets for accurate prediction, which is impracticable to acquire and annotate. Thus, Few Shot Learning is the state-of-the-art model in machine learning, which requires minimum examples to train the model for generalization. As humans need a few examples to recognize things, Few-shot Learning mimics the same human brain process. The proposed work uses a pre-trained convolution neural network, VGG16, as the backbone, fine-tuned on the corn disease dataset. An attention module is integrated with the backbone, and further, prototypical few-shot learning is used for corn disease prediction and classification with an accuracy of 98.25 %.•The proposed model intends to identify the diseases early, so the insights generated would be relevant for farmers, and probable losses can be reduced.•By applying Few-Shot Learning, the system avoids the significant challenges of requiring extensively annotated datasets, making it feasible for real-world agricultural applications.•Incorporating a fine-tuned VGG16 backbone along with an attention mechanism and prototypical Few-Shot Learning results in a robust and scalable solution with high accuracy for classifying corn diseases.

PMID:39911906 | PMC:PMC11795141 | DOI:10.1016/j.mex.2025.103172

R Soc Open Sci. 2025 Feb 5;12(2):241687. doi: 10.1098/rsos.241687. eCollection 2025 Feb.

ABSTRACT

The swift development of artificial intelligence (AI) technology has triggered substantial changes, particularly evident in the emergence of chat-based services driven by large language models. With the increasing number of users utilizing these services, understanding and analysing user satisfaction becomes crucial for service improvement. While previous studies have explored leveraging online reviews as indicators of user satisfaction, efficiently collecting and analysing extensive datasets remain a challenge. This research aims to address this challenge by proposing a framework to handle extensive review datasets from the Google Play Store, employing natural language processing with machine learning techniques for sentiment analysis. Specifically, the authors collect review data of chat-based AI applications and perform filtering through majority voting of multiple unsupervised sentiment analyses. This framework is a proposed methodology for eliminating inconsistencies between ratings and contents. Subsequently, the authors conduct supervised sentiment analysis using various machine learning and deep learning algorithms. The experimental results confirm the effectiveness of the proposed approach showing improvement in prediction accuracy with cost efficiency. In summary, the findings of this study enhance the predictive performance of user satisfaction for improving service quality in chat-based AI applications and provide valuable insights for the advancement of next-generation chat-based AI services.

PMID:39911884 | PMC:PMC11793979 | DOI:10.1098/rsos.241687

Comput Struct Biotechnol J. 2025 Jan 22;27:467-477. doi: 10.1016/j.csbj.2025.01.016. eCollection 2025.

ABSTRACT

AlphaFold 2 (AF2) has revolutionised protein structure prediction but, like any new tool, its performance on specific classes of targets, especially those potentially under-represented in its training data, merits attention. Prompted by a highly confident prediction for a biologically meaningless, randomly permuted repeat sequence, we assessed AF2 performance on sequences composed of perfect repeats of random sequences of different lengths. AF2 frequently folds such sequences into β-solenoids which, while ascribed high confidence, contain unusual and implausible features such as internally stacked and uncompensated charged residues. A number of sequences confidently predicted as β-solenoids are predicted by other advanced methods as intrinsically disordered. The instability of some predictions is demonstrated by molecular dynamics. Importantly, other deep learning-based structure prediction tools predict different structures or β-solenoids with much lower confidence suggesting that AF2 alone has an unreasonable tendency to predict confident but unrealistic β-solenoids for perfect repeat sequences. The potential implications for structure prediction of natural (near-)perfect sequence repeat proteins are also explored.

PMID:39911842 | PMC:PMC11795689 | DOI:10.1016/j.csbj.2025.01.016

Front Microbiol. 2025 Jan 22;15:1516667. doi: 10.3389/fmicb.2024.1516667. eCollection 2024.

ABSTRACT

Microbiome research, the study of microbial communities in diverse environments, has seen significant advances due to the integration of deep learning (DL) methods. These computational techniques have become essential for addressing the inherent complexity and high-dimensionality of microbiome data, which consist of different types of omics datasets. Deep learning algorithms have shown remarkable capabilities in pattern recognition, feature extraction, and predictive modeling, enabling researchers to uncover hidden relationships within microbial ecosystems. By automating the detection of functional genes, microbial interactions, and host-microbiome dynamics, DL methods offer unprecedented precision in understanding microbiome composition and its impact on health, disease, and the environment. However, despite their potential, deep learning approaches face significant challenges in microbiome research. Additionally, the biological variability in microbiome datasets requires tailored approaches to ensure robust and generalizable outcomes. As microbiome research continues to generate vast and complex datasets, addressing these challenges will be crucial for advancing microbiological insights and translating them into practical applications with DL. This review provides an overview of different deep learning models in microbiome research, discussing their strengths, practical uses, and implications for future studies. We examine how these models are being applied to solve key problems and highlight potential pathways to overcome current limitations, emphasizing the transformative impact DL could have on the field moving forward.

PMID:39911715 | PMC:PMC11794229 | DOI:10.3389/fmicb.2024.1516667

Front Neurosci. 2025 Jan 22;19:1522227. doi: 10.3389/fnins.2025.1522227. eCollection 2025.

ABSTRACT

PURPOSE: To test the feasibility and consistency of a deep-learning (DL) accelerated QSM method for deep brain nuclei evaluation.

METHODS: Participants were scanned with both parallel imaging (PI)-QSM and DL-QSM methods. The PI- and DL-QSM scans had identical imaging parameters other than acceleration factors (AF). The DL-QSM employed Poisson disk style under-sampling scheme and a previously developed cascaded CNN based reconstruction model, with acquisition time of 4:35, 3:15, and 2:11 for AF of 3, 4, and 5, respectively. For PI-QSM acquisition, the AF was 2 and the acquisition time was 6:46. The overall image similarity was assessed between PI- and DL-QSM images using the structural similarity index (SSIM) and peak signal-to-noise ratio (PSNR). QSM values from 7 deep brain nuclei were extracted and agreements between images with different Afs were assessed. Finally, the correlations between age and QSM values in the selected deep brain nuclei were evaluated.

RESULTS: 59 participants were recruited. Compared to PI-QSM images, the mean SSIM of DL images were 0.87, 0.86, and 0.85 for AF of 3, 4, and 5. The mean PSNR were 44.56, 44.53, and 44.23. Susceptibility values from DL-QSM were highly consistent with routine PI-QSM images, with differences of less than 5% at the group level. Furthermore, the associations between age and QSM values could be consistently revealed.

CONCLUSION: DL-QSM could be used for measuring susceptibility values of deep brain nucleus. An AF up to 5 did not significantly impact the correlation between age and susceptibility in deep brain nuclei.

PMID:39911700 | PMC:PMC11794186 | DOI:10.3389/fnins.2025.1522227

Front Plant Sci. 2025 Jan 22;15:1518294. doi: 10.3389/fpls.2024.1518294. eCollection 2024.

ABSTRACT

INTRODUCTION: Due to the limited computing power and fast flight speed of the picking of unmanned aerial vehicles (UAVs), it is important to design a quick and accurate detecting algorithm to obtain the fruit position.

METHODS: This paper proposes a lightweight deep learning algorithm, named YOLOv8s-Longan, to improve the detection accuracy and reduce the number of model parameters for fruitpicking UAVs. To make the network lightweight and improve its generalization performance, the Average and Max pooling attention (AMA) attention module is designed and integrated into the DenseAMA and C2f-Faster-AMA modules on the proposed backbone network. To improve the detection accuracy, a crossstage local network structure VOVGSCSPC module is designed, which can help the model better understand the information of the image through multiscale feature fusion and improve the perception and expression ability of the model. Meanwhile, the novel Inner-SIoU loss function is proposed as the loss function of the target bounding box.

RESULTS AND DISCUSSION: The experimental results show that the proposed algorithm has good detection ability for densely distributed and mutually occluded longan string fruit under complex backgrounds with a mAP@0.5 of 84.3%. Compared with other YOLOv8 models, the improved model of mAP@0.5 improves by 3.9% and reduces the number of parameters by 20.3%. It satisfies the high accuracy and fast detection requirements for fruit detection in fruit-picking UAV scenarios.

PMID:39911656 | PMC:PMC11794187 | DOI:10.3389/fpls.2024.1518294

Front Plant Sci. 2025 Jan 22;15:1501612. doi: 10.3389/fpls.2024.1501612. eCollection 2024.

ABSTRACT

The leaf area index (LAI) is a critical parameter for characterizing plant foliage abundance, canopy structure changes, and vegetation productivity in ecosystems. Traditional phenological measurements are often destructive, time-consuming, and labor-intensive. This paper proposes a high-throughput 3D point cloud data processing pipeline to segment field soybean plants and estimate their LAI. The 3D point cloud data is obtained from a UAV equipped with a LiDAR camera. First, The PointNet++ model was applied to simplify the segmentation process by isolating field soybean plants from their surroundings and eliminating environmental complexities. Subsequently, individual segmentation was achieved using the Watershed approach and k-means clustering algorithms, segmenting the field soybeans into individual plants. Finally, the LAI of soybean plant was estimated using a machine learning method and validated against measured values. The PointNet++ model improved segmentation accuracy by 6.73%, and the watershed algorithm achieved F1 scores of 0.89-0.90, outperforming k-means in complex adhesion cases. For LAI estimation, the SVM model showed the highest accuracy (R² = 0.79, RMSE = 0.47), with RF and XGBoost also performing well (R² > 0.69, RMSE< 0.65). This indicates that the individual segmentation algorithm, Watershed-based approach combined with PointNet++, can serve as a crucial foundation for extracting high-throughput plant phenotypic data. The experimental results confirm that the proposed method can rapidly calculate the morphological parameters of each soybean plant, making it suitable for high-throughput soybean phenotyping.

PMID:39911650 | PMC:PMC11794303 | DOI:10.3389/fpls.2024.1501612

Front Psychiatry. 2025 Jan 22;15:1512818. doi: 10.3389/fpsyt.2024.1512818. eCollection 2024.

ABSTRACT

Autism Spectrum Disorder (ASD) affects millions of individuals worldwide, presenting challenges in social communication, repetitive behaviors, and sensory processing. Despite its prevalence, diagnosis can be lengthy, and access to appropriate treatment varies greatly. This project utilizes the power of Artificial Intelligence (AI), particularly Machine Learning (ML) and Deep Learning (DL), to improve Autism Spectrum Disorder diagnosis and treatment. A central data hub, the Master Data Plan (MDP), will aggregate and analyze information from diverse sources, feeding AI algorithms that can identify risk factors for ASD, personalize treatment plans based on individual needs, and even predict potential relapses. Furthermore, the project incorporates a patient-facing chatbot to provide information and support. By integrating patient data, empowering individuals with ASD, and supporting healthcare professionals, this platform aims to transform care accessibility, personalize treatment approaches, and optimize the entire care journey. Rigorous data governance measures will ensure ethical and secure data management. This project will improve access to care, personalize treatments for better outcomes, shorten wait times, boost patient involvement, and raise ASD awareness, leading to better resource allocation. This project marks a transformative shift toward data-driven, patient-centred ASD care in Italy. This platform enhances treatment outcomes for individuals with ASD and provides a scalable model for integrating AI into mental health, establishing a new benchmark for personalized patient care. Through AI integration and collaborative efforts, it aims to redefine mental healthcare standards, enhancing the well-being for individuals with ASD.

PMID:39911557 | PMC:PMC11795314 | DOI:10.3389/fpsyt.2024.1512818

J Cheminform. 2025 Feb 5;17(1):17. doi: 10.1186/s13321-025-00959-9.

ABSTRACT

Recently, advancements in cheminformatics such as representation learning for chemical structures, deep learning (DL) for property prediction, data-driven discovery, and optimization of chemical data handling, have led to increased demands for handling chemical simplified molecular input line entry system (SMILES) data, particularly in text analysis tasks. These advancements have driven the need to optimize components like positional encoding and positional embeddings (PEs) in transformer model to better capture the sequential and contextual information embedded in molecular representations. SMILES data represent complex relationships among atoms or elements, rendering them critical for various learning tasks within the field of cheminformatics. This study addresses the critical challenge of encoding complex relationships among atoms in SMILES strings to explore various PEs within the transformer-based framework to increase the accuracy and generalization of molecular property predictions. The success of transformer-based models, such as the bidirectional encoder representations from transformer (BERT) models, in natural language processing tasks has sparked growing interest from the domain of cheminformatics. However, the performance of these models during pretraining and fine-tuning is significantly influenced by positional information such as PEs, which help in understanding the intricate relationships within sequences. Integrating position information within transformer architectures has emerged as a promising approach. This encoding mechanism provides essential supervision for modeling dependencies among elements situated at different positions within a given sequence. In this study, we first conduct pretraining experiments using various PEs to explore diverse methodologies for incorporating positional information into the BERT model for chemical text analysis using SMILES strings. Next, for each PE, we fine-tune the best-performing BERT (masked language modeling) model on downstream tasks for molecular-property prediction. Here, we use two molecular representations, SMILES and DeepSMILES, to comprehensively assess the potential and limitations of the PEs in zero-shot learning analysis, demonstrating the model's proficiency in predicting properties of unseen molecular representations in the context of newly proposed and existing datasets.Scientific contributionThis study explores the unexplored potential of PEs using BERT model for molecular property prediction. The study involved pretraining and fine-tuning the BERT model on various datasets related to COVID-19, bioassay data, and other molecular and biological properties using SMILES and DeepSMILES representations. The study details the pretraining architecture, fine-tuning datasets, and the performance of the BERT model with different PEs. It also explores zero-shot learning analysis and the model's performance on various classification and regression tasks. In this study, newly proposed datasets from different domains were introduced during fine-tuning in addition to the existing and commonly used datasets. The study highlights the robustness of the BERT model in predicting chemical properties and its potential applications in cheminformatics and bioinformatics.

PMID:39910649 | DOI:10.1186/s13321-025-00959-9

Plant Methods. 2025 Feb 5;21(1):12. doi: 10.1186/s13007-024-01321-0.

ABSTRACT

Genomic Selection (GS) predicts traits using genome-wide markers, speeding up genetic progress and enhancing breeding efficiency. Recent emphasis has been placed on deep learning models to enhance prediction accuracy. However, current deep learning models focus on learning specific phenotypes for the given task, overlooking the inter-correlations among different phenotypes. In response, we introduce MtCro, a multi-task learning approach that simultaneously captures diverse plant phenotypes within a shared parameter space. Extensive experiments reveal that MtCro outperforms mainstream models, including DNNGP and SoyDNGP, with performance gains of 1-9% on the Wheat2000 dataset, 1-8% on Wheat599, and 1-3% on Maize8652. Furthermore, comparative analysis shows a consistent 2-3% improvement in multi-phenotype predictions, emphasizing the impact of inter-phenotype correlations on accuracy. By leveraging multi-task learning, MtCro efficiently captures diverse plant phenotypes, enhancing both model training efficiency and prediction accuracy, ultimately accelerating the progress of plant genetic breeding. Our code is available on https://github.com/chaodian12/mtcro .

PMID:39910577 | DOI:10.1186/s13007-024-01321-0

BMC Oral Health. 2025 Feb 5;25(1):187. doi: 10.1186/s12903-025-05482-9.

ABSTRACT

BACKGROUND: Artificial intelligence (AI) methods, including machine learning and deep learning, are increasingly applied in orthodontics for tasks like assessing skeletal maturity. Accurate timing of treatment is crucial, but traditional methods such as cervical vertebral maturation (CVM) staging have limitations due to observer variability and complexity. AI has the potential to automate CVM assessment, enhancing reliability and user-friendliness. This systematic review and meta-analysis aimed to evaluate the overall performance of artificial intelligence (AI) models in assessing cervical vertebrae maturation (CVM) in radiographs, when compared to clinicians.

METHODS: Electronic databases of Medline (via PubMed), Google Scholar, Scopus, Embase, IEEE ArXiv and MedRxiv were searched for publications after 2010, without any limitation on language. In the present review, we included studies that reported AI models' performance on CVM assessment. Quality assessment was done using Quality assessment and diagnostic accuracy Tool-2 (QUADAS-2). Quantitative analysis was conducted using hierarchical logistic regression for meta-analysis on diagnostic accuracy. Subgroup analysis was conducted on different AI subsets (Deep learning, and Machine learning).

RESULTS: A total of 1606 studies were screened of which 25 studies were included. The performance of the models was acceptable. However, it varied based on the methods employed. Eight studies had a low risk of bias in all domains. Twelve studies were included in the meta-analysis and their pooled values for sensitivity, specificity, positive and negative likelihood ratios, and diagnostic odds ratio (DOR) were calculated for each cervical stage (CS). The most accurate CVM evaluation was observed for CS1, boasting a sensitivity of 0.87, a specificity of 0.97, and a DOR of 213. Conversely, CS3 exhibited the lowest performance with a sensitivity of 0.64, and a specificity of 0.96, yet maintaining a DOR of 32.

CONCLUSION: AI has demonstrated encouraging outcomes in CVM assessment, achieving notable accuracy.

PMID:39910512 | DOI:10.1186/s12903-025-05482-9

BMC Med Imaging. 2025 Feb 5;25(1):40. doi: 10.1186/s12880-025-01579-3.

ABSTRACT

BACKGROUND: Long-term severe cholangitis can lead to dense adhesions and increased fragility of the bile duct, complicating surgical procedures and elevating operative risk in children with pancreaticobiliary maljunction (PBM). Consequently, preoperative diagnosis of moderate-to-severe chronic cholangitis is essential for guiding treatment strategies and surgical planning. This study aimed to develop and validate a deep learning radiomics nomogram (DLRN) based on contrast-enhanced CT images and clinical characteristics to preoperatively identify moderate-to-severe chronic cholangitis in children with PBM.

METHODS: A total of 323 pediatric patients with PBM who underwent surgery were retrospectively enrolled from three centers, and divided into a training cohort (n = 153), an internal validation cohort (IVC, n = 67), and two external test cohorts (ETC1, n = 58; ETC2, n = 45). Chronic cholangitis severity was determined by postoperative pathology. Handcrafted radiomics features and deep learning (DL) radiomics features, extracted using transfer learning with the ResNet50 architecture, were obtained from portal venous-phase CT images. Multivariable logistic regression was used to establish the DLRN, integrating significant clinical factors with handcrafted and DL radiomics signatures. The diagnostic performances were evaluated in terms of discrimination, calibration, and clinical usefulness.

RESULTS: Biliary stones and peribiliary fluid collection were selected as important clinical factors. 5 handcrafted and 5 DL features were retained to build the two radiomics signatures, respectively. The integrated DLRN achieved satisfactory performance, achieving area under the curve (AUC) values of 0.913 (95% CI, 0.834-0.993), 0.916 (95% CI, 0.845-0.987), and 0.895 (95% CI, 0.801-0.989) in the IVC, and two ETCs, respectively. In comparison, the clinical model, handcrafted signature, and DL signature had AUC ranges of 0.654-0.705, 0.823-0.857, and 0.840-0.872 across the same cohorts. The DLRN outperformed single-modality clinical, handcrafted radiomics, and DL radiomics models, with all integrated discrimination improvement values > 0 and P < 0.05. The Hosmer-Lemeshow test and calibration curves showed good consistency of the DLRN (P > 0.05), and the decision curve analysis and clinical impact curve further confirmed its clinical utility.

CONCLUSIONS: The integrated DLRN can be a useful and non-invasive tool for preoperatively identifying moderate-to-severe chronic cholangitis in children with PBM, potentially enhancing clinical decision-making and personalized management strategies.

PMID:39910477 | DOI:10.1186/s12880-025-01579-3

J Cheminform. 2025 Feb 5;17(1):18. doi: 10.1186/s13321-025-00952-2.

ABSTRACT

Accurate prediction of drug-target interactions is critical for advancing drug discovery. By reducing time and cost, machine learning and deep learning can accelerate this laborious discovery process. In a novel approach, BarlowDTI, we utilise the powerful Barlow Twins architecture for feature-extraction while considering the structure of the target protein. Our method achieves state-of-the-art predictive performance against multiple established benchmarks using only one-dimensional input. The use of our hybrid approach of deep learning and gradient boosting machine as the underlying predictor ensures fast and efficient predictions without the need for substantial computational resources. We also propose the use of an influence method to investigate how the model reaches its decision based on individual training samples. By comparing co-crystal structures, we find that BarlowDTI effectively exploits catalytically active and stabilising residues, highlighting the model's ability to generalise from one-dimensional input data. In addition, we further benchmark new baselines against existing methods. Together, these innovations improve the efficiency and effectiveness of drug-target interactions predictions, providing robust tools for accelerating drug development and deepening the understanding of molecular interactions. Therefore, we provide an easy-to-use web interface that can be freely accessed at https://www.bio.nat.tum.de/oc2/barlowdti . SCIENTIFIC CONTRIBUTION: Our computationally efficient and effective hybrid approach, combining the deep learning model Barlow Twins and gradient boosting machines, outperforms state-of-the-art methods across multiple splits and benchmarks using only one-dimensional input. Furthermore, we advance the field by proposing an influence method that elucidates model decision-making, thereby providing deeper insights into molecular interactions and improving the interpretability of drug-target interactions predictions.

PMID:39910404 | DOI:10.1186/s13321-025-00952-2

Sci Rep. 2025 Feb 5;15(1):4415. doi: 10.1038/s41598-025-88582-4.

ABSTRACT

Cavitation is a technical challenge for high-speed underwater vehicles, such as nuclear submarines and underwater robots, et al. The cavitation phenomena of hydrofoils are typically studied through water tunnel experiments or numerical simulations, which yield extensive cavitation images. To conveniently extract cavitation features from the massive images, a feature extraction method for hydrofoil cavitation was proposed in this work based on deep learning image semantic segmentation techniques. This method is employed to investigate the mechanism of the transition process from sheet cavitation to cloud cavitation on hydrofoils. The accuracy and generalization ability of the proposed method have been validated. The results indicate that, in addition to accurately obtaining the cavitation length automatically, the method can also derive more sensitive indicators such as area and position changes of the cavitation regions. This heightened sensitivity is invaluable for precisely pinpointing the transition from sheet-like cavitation to cloud cavitation, thereby aiding in a more effective analysis of the development mechanism of attached cavitation. In summary, our proposed method not only streamlines the extraction of cavitation features from massive images but also enhances the understanding of development mechanisms of attached cavitation by providing additional data and more sensitive indicators for analysis.

PMID:39910333 | DOI:10.1038/s41598-025-88582-4

Sci Rep. 2025 Feb 5;15(1):4416. doi: 10.1038/s41598-025-88982-6.

ABSTRACT

This study aims to refine a radiomics-based diagnostic approach for detecting neonatal respiratory distress syndrome (NRDS) and examines the influence of rib suppression on the diagnostic precision of radiomics models using neonatal chest X-ray (CXR) images. A total of 138 CXR images were collected in this study. The data was partitioned into training and validation subsets based on chronological order. We applied rib suppression to the CXR images and extracted and analyzed radiomic features from lung regions both before and after rib suppression. This approach was designed to identify NRDS, develop radiomics models, and assess the impact of rib suppression on model performance. To establish these radiomics models, six machine learning models were utilized in the study. The performance was evaluated using the area under the receiver operating characteristic curve (AUC). On the validation set, the models demonstrated significant improvements after rib suppression. Specifically, the Gradient Boosting Machine (GBM) achieved an AUC of 0.781 post-suppression compared to 0.556 pre-suppression. Notably, Linear Discriminant Analysis (LDA) and Logistic Regression (LR) performed particularly well when combining features from both scenarios, achieving AUCs of 0.762 and 0.756. The results indicate the feasibility of developing radiomics models for diagnosing NRDS and highlight the enhancement in model performance due to rib suppression. This study provides a promising new method for the imaging diagnosis and prognosis evaluation of neonatal respiratory distress syndrome, showcasing the potential of radiomics in pediatric imaging.

PMID:39910276 | DOI:10.1038/s41598-025-88982-6

Sci Rep. 2025 Feb 5;15(1):4337. doi: 10.1038/s41598-025-88450-1.

ABSTRACT

Disabled persons demanding healthcare is a developing global occurrence. The support in longer-term care includes nursing, intricate medical, recovery, and social help services. The price is large, but advanced technologies can aid in decreasing expenditure by certifying effective health services and enhancing the superiority of life. The transformative latent of the Internet of Things (IoT) prolongs the existence of nearly one billion persons worldwide with disabilities. By incorporating smart devices and technologies, the IoT provides advanced solutions to tackle numerous tasks challenged by individuals with disabilities and promote equality. Human activity detection methods are the technical area which studies the classification of actions or movements an individual achieves over the recognition of signals directed by smartphones or wearable sensors or over images or video frames. They are efficient in certifying functions of detection of actions, observing crucial functions, and tracking. Conventional machine learning and deep learning approaches effectively detect human activity. This study develops and designs a metaheuristic optimization-driven ensemble model for smart monitoring of indoor activities for disabled persons (MOEM-SMIADP) model. The proposed MOEM-SMIADP model concentrates on detecting and classifying indoor activities using IoT applications for physically challenged people. First, data preprocessing is performed using min-max normalization to convert input data into useful format. Furthermore, the marine predator algorithm is employed in feature selection. For the detection of indoor activities, the proposed MOEM-SMIADP model utilizes an ensemble of three classifiers, namely the graph convolutional network model, long short-term memory sequence-to-sequence (LSTM-seq2seq) method, and convolutional autoencoder. Eventually, the hyperparameter tuning is accomplished by an improved coati optimization algorithm to enhance the classification outcomes of ensemble models. A wide range of experiments was accompanied to endorse the performance of the MOEM-SMIADP technique. The performance validation of the MOEM-SMIADP technique portrayed a superior accracy value of 99.07% over existing methods.

PMID:39910242 | DOI:10.1038/s41598-025-88450-1

Sci Rep. 2025 Feb 5;15(1):4406. doi: 10.1038/s41598-025-88907-3.

ABSTRACT

Employing two standard mammography views is crucial for radiologists, providing comprehensive insights for reliable clinical evaluations. This study introduces paired mammogram view based-network(PMVnet), a novel algorithm designed to enhance breast lesion detection by integrating relational information from paired whole mammograms, addressing the limitations of current methods. Utilizing 1,636 private mammograms, PMVnet combines cosine similarity and the squeeze-and-excitation method within a U-shaped architecture to leverage correlated information. Performance comparisons with single view-based models with VGGnet16, Resnet50, and EfficientnetB5 as encoders revealed PMVnet's superior capability. Using VGGnet16, PMVnet achieved a Dice Similarity Coefficient (DSC) of 0.709 in segmentation and a recall of 0.950 at 0.156 false positives per image (FPPI) in detection tasks, outperforming the single-view model, which had a DSC of 0.579 and a recall of 0.813 at 0.188 FPPI. These findings demonstrate PMVnet's effectiveness in reducing false positives and avoiding missed true positives, suggesting its potential as a practical tool in computer-aided diagnosis systems. PMVnet can significantly enhance breast lesion detection, aiding radiologists in making more precise evaluations and improving patient outcomes. Future applications of PMVnet may offer substantial benefits in clinical settings, improving patient care through enhanced diagnostic accuracy.

PMID:39910228 | DOI:10.1038/s41598-025-88907-3