Curr Med Imaging. 2025 Apr 22. doi: 10.2174/0115734056370620250403030638. Online ahead of print.

ABSTRACT

INTRODUCTION: To examine the integration of artificial intelligence (AI) into Picture Archiving and Communication Systems (PACS) and assess its impact on medical imaging, diagnostic workflows, and patient outcomes. This review explores the technological evolution, key advancements, and challenges associated with AI-enhanced PACS in healthcare settings.

METHODS: A comprehensive literature search was conducted in PubMed, Scopus, and Web of Science databases, covering articles from January 2000 to October 2024. Search terms included "artificial intelligence," "machine learning," "deep learning," and "PACS," combined with keywords related to diagnostic accuracy and workflow optimization. Articles were selected based on predefined inclusion and exclusion criteria, focusing on peerreviewed studies that discussed AI applications in PACS, innovations in medical imaging, and workflow improvements. A total of 183 studies met the inclusion criteria, comprising original research, systematic reviews, and meta-analyses.

RESULTS: AI integration in PACS has significantly enhanced diagnostic accuracy, achieving improvements of up to 93.2% in some imaging modalities, such as early tumor detection and anomaly identification. Workflow efficiency has been transformed, with diagnostic times reduced by up to 90% for critical conditions like intracranial hemorrhages. Convolutional neural networks (CNNs) have demonstrated exceptional performance in image segmentation, achieving up to 94% accuracy, and in motion artifact correction, further enhancing diagnostic precision. Natural language processing (NLP) tools have expedited radiology workflows, reducing reporting times by 30-50% and improving consistency in report generation. Cloudbased solutions have also improved accessibility, enabling real-time collaboration and remote diagnostics. However, challenges in data privacy, regulatory compliance, and interoperability persist, emphasizing the need for standardized frameworks and robust security protocols. Conclusions The integration of AI into PACS represents a pivotal transformation in medical imaging, offering improved diagnostic workflows and potential for personalized patient care. Addressing existing challenges and enhancing interoperability will be essential for maximizing the benefits of AIpowered PACS in healthcare.

PMID:40265427 | DOI:10.2174/0115734056370620250403030638

J Appl Clin Med Phys. 2025 Apr 23:e70098. doi: 10.1002/acm2.70098. Online ahead of print.

ABSTRACT

Prostate Stereotactic Ablative Body Radiotherapy (SABR) is an ultra-hypofractionated treatment where small setup errors can lead to higher doses to organs at risk (OARs). Although bowel and bladder preparation protocols reduce inter-fraction variability, inconsistent patient adherence still results in OAR variability. At many centers without online adaptive machines, radiation therapists use decision trees (DTs) to visually assess patient setup, yet their application varies. To evaluate our center's DTs, we employed deep learning-generated cone-beam computed tomography (CBCT) contours to estimate daily doses to the rectum and bladder, comparing these with planned dose-volume metrics to guide future personalized DT development. Two hundred pretreatment CBCT scans from 40 prostate SABR patients (each receiving 40 Gy in five fractions) were auto-contoured retrospectively, and daily rectum and bladder doses were estimated by overlaying the planned dose on the CBCT using online rigid registration data. Dose-volume metrics were classified as "no", "minor", or "major" violations based on meeting preferred or mandatory goals. Twenty-seven percent of fractions exhibited at least one major bladder violation (with an additional 34% minor), while 14% of fractions had a major rectum violation (10% minor). Across treatments, five patients had recurring bladder V37 Gy major violations and two had rectum V36 Gy major violations. Bowel and bladder preparation significantly influenced OAR position and volume, leading to unmet mandatory goals. Our retrospective analysis underscores the significant impact of patient preparation on dosimetric outcomes. Our findings highlight that DTs based solely on visual assessment miss dose metric violations due to human error; only 23 of 59 under-filled bladder fractions were flagged. In addition to the insensitivity of visual assessments, variability in DT application further compromises patient setup evaluation. These analyses confirm that reliance on visual inspection alone can overlook deviations, emphasizing the need for automated tools to ensure adherence to dosimetric constraints in prostate SABR.

PMID:40265325 | DOI:10.1002/acm2.70098

Front Med (Lausanne). 2025 Apr 8;12:1503229. doi: 10.3389/fmed.2025.1503229. eCollection 2025.

ABSTRACT

Deoxyribonucleic acid (DNA) serves as fundamental genetic blueprint that governs development, functioning, growth, and reproduction of all living organisms. DNA can be altered through germline and somatic mutations. Germline mutations underlie hereditary conditions, while somatic mutations can be induced by various factors including environmental influences, chemicals, lifestyle choices, and errors in DNA replication and repair mechanisms which can lead to cancer. DNA sequence analysis plays a pivotal role in uncovering the intricate information embedded within an organism's genetic blueprint and understanding the factors that can modify it. This analysis helps in early detection of genetic diseases and the design of targeted therapies. Traditional wet-lab experimental DNA sequence analysis through traditional wet-lab experimental methods is costly, time-consuming, and prone to errors. To accelerate large-scale DNA sequence analysis, researchers are developing AI applications that complement wet-lab experimental methods. These AI approaches can help generate hypotheses, prioritize experiments, and interpret results by identifying patterns in large genomic datasets. Effective integration of AI methods with experimental validation requires scientists to understand both fields. Considering the need of a comprehensive literature that bridges the gap between both fields, contributions of this paper are manifold: It presents diverse range of DNA sequence analysis tasks and AI methodologies. It equips AI researchers with essential biological knowledge of 44 distinct DNA sequence analysis tasks and aligns these tasks with 3 distinct AI-paradigms, namely, classification, regression, and clustering. It streamlines the integration of AI into DNA sequence analysis tasks by consolidating information of 36 diverse biological databases that can be used to develop benchmark datasets for 44 different DNA sequence analysis tasks. To ensure performance comparisons between new and existing AI predictors, it provides insights into 140 benchmark datasets related to 44 distinct DNA sequence analysis tasks. It presents word embeddings and language models applications across 44 distinct DNA sequence analysis tasks. It streamlines the development of new predictors by providing a comprehensive survey of 39 word embeddings and 67 language models based predictive pipeline performance values as well as top performing traditional sequence encoding-based predictors and their performances across 44 DNA sequence analysis tasks.

PMID:40265190 | PMC:PMC12011883 | DOI:10.3389/fmed.2025.1503229

Front Med (Lausanne). 2025 Apr 8;12:1562818. doi: 10.3389/fmed.2025.1562818. eCollection 2025.

ABSTRACT

OBJECTIVE: This study conducts a bibliometric analysis of artificial intelligence (AI) applications in cervical cancer to provide a comprehensive overview of the research landscape and current advancements.

METHODS: Relevant publications on cervical cancer and AI were retrieved from the Web of Science Core Collection. Bibliometric analysis was performed using CiteSpace and VOSviewer to assess publication trends, authorship, country and institutional contributions, journal sources, and keyword co-occurrence patterns.

RESULTS: From 1996 to 2024, our analysis of 770 publications on cervical cancer and AI showed a surge in research, with 86% published in the last 5 years. China (315 pubs, 32%) and the US (155 pubs, 16%) were the top contributors. Key institutions were the Chinese Academy of Sciences, Southern Medical University, and Huazhong University of Science and Technology. Research hotspots included disease prediction, image analysis, and machine learning in cervical cancer. Schiffman led in publications (12) and citations (207). China had the highest citations (3,819). Top journals were "Diagnostics," "Scientific Reports," and "Frontiers in Oncology." Keywords like "machine learning" and "deep learning" indicated current research trends. This study maps the field's growth, highlighting key contributors and topics.

CONCLUSION: This bibliometric analysis provides valuable insights into research trends and hotspots, guiding future studies and fostering collaboration to enhance AI applications in cervical cancer.

PMID:40265176 | PMC:PMC12011737 | DOI:10.3389/fmed.2025.1562818

J Natl Cancer Cent. 2024 Dec 27;5(2):113-131. doi: 10.1016/j.jncc.2024.12.006. eCollection 2025 Apr.

ABSTRACT

Upper gastrointestinal cancers, mainly comprising esophageal and gastric cancers, are among the most prevalent cancers worldwide. There are many new cases of upper gastrointestinal cancers annually, and the survival rate tends to be low. Therefore, timely screening, precise diagnosis, appropriate treatment strategies, and effective prognosis are crucial for patients with upper gastrointestinal cancers. In recent years, an increasing number of studies suggest that artificial intelligence (AI) technology can effectively address clinical tasks related to upper gastrointestinal cancers. These studies mainly focus on four aspects: screening, diagnosis, treatment, and prognosis. In this review, we focus on the application of AI technology in clinical tasks related to upper gastrointestinal cancers. Firstly, the basic application pipelines of radiomics and deep learning in medical image analysis were introduced. Furthermore, we separately reviewed the application of AI technology in the aforementioned aspects for both esophageal and gastric cancers. Finally, the current limitations and challenges faced in the field of upper gastrointestinal cancers were summarized, and explorations were conducted on the selection of AI algorithms in various scenarios, the popularization of early screening, the clinical applications of AI, and large multimodal models.

PMID:40265096 | PMC:PMC12010392 | DOI:10.1016/j.jncc.2024.12.006

Front Oncol. 2025 Apr 8;15:1567577. doi: 10.3389/fonc.2025.1567577. eCollection 2025.

ABSTRACT

The segmentation and classification of breast ultrasound (BUS) images are crucial for the early diagnosis of breast cancer and remain a key focus in BUS image processing. Numerous machine learning and deep learning algorithms have shown their effectiveness in the segmentation and diagnosis of BUS images. In this work, we propose a multi-task learning network with an object contextual attention module (MTL-OCA) for the segmentation and classification of BUS images. The proposed method utilizes the object contextual attention module to capture pixel-region relationships, enhancing the quality of segmentation masks. For classification, the model leverages high-level features extracted from unenhanced segmentation masks to improve accuracy. Cross-validation on a public BUS dataset demonstrates that MTL-OCA outperforms several current state-of-the-art methods, achieving superior results in both classification and segmentation tasks.

PMID:40265029 | PMC:PMC12011763 | DOI:10.3389/fonc.2025.1567577

Front Oncol. 2025 Apr 8;15:1549803. doi: 10.3389/fonc.2025.1549803. eCollection 2025.

ABSTRACT

BACKGROUND: Uterine broad ligament fibroids present unique surgical challenges due to their proximity to vital pelvic structures. This study aimed to evaluate artificial intelligence (AI)-guided MRI instance segmentation for optimizing laparoscopic myomectomy outcomes.

METHODS: In this trial, 120 patients with MRI-confirmed broad ligament fibroids were allocated to either AI-assisted group (n=60) or conventional MRI group (n=60). A deep learning model was developed to segment fibroids, uterine walls, and uterine cavity from preoperative MRI.

RESULT: Compared to conventional MRI guidance, AI assistance significantly reduced operative time (118 [112.25-125.00] vs. 140 [115.75-160.75] minutes; p<0.001). The AI group also demonstrated lower intraoperative blood loss (50 [50-100] vs. 85 [50-100] ml; p=0.01) and faster postoperative recovery (first flatus within 24 hours: (15[25.00%] vs. 29[48.33%], p=0.01).

CONCLUSION: This multidisciplinary AI system enhances surgical precision through millimeter-level anatomical delineation, demonstrating transformative potential for complex gynecologic oncology procedures. Clinical adoption of this approach could reduce intraoperative blood loss and iatrogenic complications, thereby promoting postoperative recovery.

PMID:40265020 | PMC:PMC12011577 | DOI:10.3389/fonc.2025.1549803

Front Oncol. 2025 Apr 8;15:1538854. doi: 10.3389/fonc.2025.1538854. eCollection 2025.

ABSTRACT

PURPOSE: This study aimed to establish and evaluate a model utilizing bi-parametric ultrasound-based deep learning radiomics (DLR) in conjunction with clinical factors to anticipate clinically significant prostate cancer (csPCa).

METHODS: We retrospectively analyzed 232 participants from our institution who underwent both B-mode ultrasound and shear wave elastography (SWE) prior to prostate biopsy between June 2022 and December 2023. A random allocation placed the participants into training and test cohorts with a 7:3 distribution. We developed a nomogram that integrates DLR with clinical factors within the training cohort, which was subsequently validated using the test cohort. The diagnostic performance and clinical applicability were evaluated with receiver operating characteristic (ROC) curve analysis and decision curve analysis.

RESULTS: In our study, the bi-parametric ultrasound-based DLR model demonstrated an area under the curve (AUC) of 0.80 (95%CI: 0.70-0.91) in the test set, surpassing the performance of both the radiomics and deep learning models individually. By integrating clinical factors, a composite model, presented as the nomogram, was developed and exhibited superior diagnostic performance, achieving an AUC of 0.87 (95%CI: 0.77-0.95) in the test set. The performance exceeded that of the DLR (P = 0.049) and the clinical model (AUC = 0.79, 95%CI: 0.69-0.86, P = 0.041). Furthermore, the decision curve analysis indicated that the composite model provided a greater net benefit across a various high-risk threshold than the DLR or the clinical model alone.

CONCLUSION: To our knowledge, this is the first proposal of a nomogram integrating ultrasound-based DLR with clinical indicators for predicting csPCa. This nomogram can improve the accuracy of csPCa prediction and may help physicians make more confident decisions regarding interventions, particularly in settings where MRI is unavailable.

PMID:40265019 | PMC:PMC12011619 | DOI:10.3389/fonc.2025.1538854

Patterns (N Y). 2025 Mar 3;6(4):101186. doi: 10.1016/j.patter.2025.101186. eCollection 2025 Apr 11.

ABSTRACT

The ever-increasing stream of big data offers potential for deep decarbonization in the transportation sector but also presents challenges in extracting interpretable insights due to its complexity and volume. This overview discusses the application of transportation big data to help understand carbon dioxide emissions and introduces how artificial intelligence models, including machine learning (ML) and deep learning (DL), are used to assimilate and understand these data. We suggest using ML to interpret low-dimensional data and DL to enhance the predictability of data with spatial connections across multiple timescales. Overcoming challenges related to algorithms, data, and computation requires interdisciplinary collaboration on both technology and data.

PMID:40264962 | PMC:PMC12010448 | DOI:10.1016/j.patter.2025.101186

Patterns (N Y). 2025 Apr 11;6(4):101228. doi: 10.1016/j.patter.2025.101228. eCollection 2025 Apr 11.

ABSTRACT

Perturbation analysis in single-cell RNA sequencing (scRNA-seq) data is challenging due to the complexity of cellular responses. To address this, Xu and Fleming et al. developed CellCap, a generative deep-learning model that decodes the perturbation effect on a particular cell state. CellCap extracts interpretable latent representations of perturbation response modules, identifying key cellular pathways activated under various conditions. This allows for a deeper understanding of cell-state-specific responses to genetic, chemical, or biological perturbations.

PMID:40264958 | PMC:PMC12010446 | DOI:10.1016/j.patter.2025.101228

Front Mol Biosci. 2025 Apr 8;12:1542267. doi: 10.3389/fmolb.2025.1542267. eCollection 2025.

ABSTRACT

Intrinsically Disordered Proteins (IDPs) challenge traditional structure-function paradigms by existing as dynamic ensembles rather than stable tertiary structures. Capturing these ensembles is critical to understanding their biological roles, yet Molecular Dynamics (MD) simulations, though accurate and widely used, are computationally expensive and struggle to sample rare, transient states. Artificial intelligence (AI) offers a transformative alternative, with deep learning (DL) enabling efficient and scalable conformational sampling. They leverage large-scale datasets to learn complex, non-linear, sequence-to-structure relationships, allowing for the modeling of conformational ensembles in IDPs without the constraints of traditional physics-based approaches. Such DL approaches have been shown to outperform MD in generating diverse ensembles with comparable accuracy. Most models rely primarily on simulated data for training and experimental data serves a critical role in validation, aligning the generated conformational ensembles with observable physical and biochemical properties. However, challenges remain, including dependence on data quality, limited interpretability, and scalability for larger proteins. Hybrid approaches combining AI and MD can bridge the gaps by integrating statistical learning with thermodynamic feasibility. Future directions include incorporating physics-based constraints and learning experimental observables into DL frameworks to refine predictions and enhance applicability. AI-driven methods hold significant promise in IDP research, offering novel insights into protein dynamics and therapeutic targeting while overcoming the limitations of traditional MD simulations.

PMID:40264953 | PMC:PMC12011600 | DOI:10.3389/fmolb.2025.1542267

Lin Chuang Er Bi Yan Hou Tou Jing Wai Ke Za Zhi. 2025 May;39(5):496-500. doi: 10.13201/j.issn.2096-7993.2025.05.020.

ABSTRACT

Objective:Ménière's disease（MD） is a common disorder of the inner ear. The fluctuating clinical symptoms and the absence of gold standards for diagnosis have posed serious problems for clinical diagnosis and treatment over the years. With the development of science and technology, artificial intelligence （AI） has been widely used in the field of medicine, and the potential of AI application to MD is demonstrated. The purpose of this review is to outline the use of AI in MD. Initially, specific instances where AI aids in differentiating MD from other causes of vertigo are presented. Furthermore, the role of AI in the evaluation of Endolymphatic Hydrops （EH）, particularly through imaging and biochemical assays, is highlighted due to its correlation with MD. Additionally, the effectiveness of AI in managing MD patients and forecasting disease progression is examined. In conclusion, the prevalent challenges hindering the clinical integration of AI in MD treatment are discussed, alongside potential strategies to surmount these barriers.

PMID:40263665 | DOI:10.13201/j.issn.2096-7993.2025.05.020

Sci Rep. 2025 Apr 22;15(1):13964. doi: 10.1038/s41598-025-99131-4.

ABSTRACT

Finite element analysis (FEA), a biomechanical simulation technique capable of providing direct mechanical visualization for CT-based digital models, has been extensively applied to fossil image datasets to address key evolutionary questions in paleontology. However, the rock matrix filling intertrabecular space of fossils often causes severe deviations in FEA results. Segmentation strategies such as thresholding and manual labeling have been employed to mitigate these disturbances. However, the efficiency of manual segmentation and the accuracy of thresholding remain questionable. In this study, we applied FEA combined with deep learning-based segregation on a femoral specimen of Jeholosaurus (a small bipedal dinosaur). This novel methodology efficiently generates the FE model with stress distribution that closely reflects the trabecular architecture in fossils of extinct taxa, reflecting a more natural state of biomechanical performance with high biological reality. Our approach provides a practical strategy for studying the biomechanics, functional morphology, and taxonomy of extinct species.

PMID:40263619 | DOI:10.1038/s41598-025-99131-4

Sci Rep. 2025 Apr 22;15(1):13942. doi: 10.1038/s41598-025-98195-6.

ABSTRACT

The fault of one of the key systems in artificial satellites, the Control Moment Gyroscope (CMG), can lead to significant economic losses and irreparable consequences. Therefore, it is crucial to diagnose its faults promptly. Traditional fault diagnosis methods, however, face challenges such as local feature traps and difficulty in feature extraction when dealing with CMG vibration signals, making it hard to meet the requirements for accuracy and robustness. Hence, it is essential to design a high-accuracy model to assess the health status of CMG on time. To address these issues, a fault diagnosis method that combines the Joint Attention Mechanism (JAM) with one-dimensional dilated convolutional networks and residual connections is proposed. The method efficiently learns feature information through the JAM, effectively addressing the time-varying characteristics of vibration signals and focusing more on fault-related features. The influence of rotational speed on the model is overcome to some extent through JAM. The three rotational speeds are mixed as datasets, and the model achieves high accuracy. The proposed method significantly enhances the accuracy and robustness of CMG fault diagnosis. Experimental results on a self-collected dataset demonstrate that the proposed method achieves excellent accuracy (98.14%) and robustness in CMG fault diagnosis.

PMID:40263562 | DOI:10.1038/s41598-025-98195-6

Sci Rep. 2025 Apr 22;15(1):13904. doi: 10.1038/s41598-025-98015-x.

ABSTRACT

Global food security depends on tomato growing, but several fungal, bacterial, and viral illnesses seriously reduce productivity and quality, therefore causing major financial losses. Reducing these impacts depends on early, exact diagnosis of diseases. This work provides a deep learning-based ensemble model for tomato leaf disease classification combining MobileNetV2 and ResNet50. To improve feature extraction, the models were tweaked by changing their output layers with GlobalAverage Pooling2D, Batch Normalization, Dropout, and Dense layers. To take use of their complimentary qualities, the feature maps from both models were combined. This study uses a publicly available dataset from Kaggle for tomato leaf disease classification. Training on a dataset of 11,000 annotated pictures spanning 10 disease categories, including bacterial spot, early blight, late blight, leaf mold, septoria leaf spot, spider mites, target spot, yellow leaf curl virus, mosaic virus, and healthy leaves. Data preprocessing included image resizing and splitting, along with an 80-10-10 split, allocating 80% for training, 10% for testing, and 10% for validation to ensure a balanced evaluation. The proposed model with a 99.91% test accuracy, the suggested model was quite remarkable. Furthermore, guaranteeing strong classification performance across all disease categories, the model showed great precision (99.92%), recall (99.90%), and an F1-score of 99.91%. With few misclassifications, the confusion matrix verified almost flawless classification even further. These findings show how well deep learning can automate tomato disease diagnosis, therefore providing a scalable and quite accurate solution for smart agriculture. By means of early intervention and precision agriculture techniques, the suggested strategy has the potential to improve crop health monitoring, reduce economic losses, and encourage sustainable farming practices.

PMID:40263518 | DOI:10.1038/s41598-025-98015-x

Sci Rep. 2025 Apr 22;15(1):13830. doi: 10.1038/s41598-025-98571-2.

ABSTRACT

Driven by the rapid development of the Internet of Things (IoT), deploying deep learning models on resource-constrained hardware has become an increasingly critical challenge, which has propelled the emergence of TinyML as a viable solution. This study aims to deploy lightweight deep learning models for human activity recognition (HAR) using TinyML on edge devices. We designed and evaluated three models: a 2D Convolutional Neural Network (2D CNN), a 1D Convolutional Neural Network (1D CNN), and a DeepConv LSTM. Among these, the DeepConv LSTM outperformed existing lightweight models by effectively capturing both spatial and temporal features, achieving an accuracy of 98.24% and an F1 score of 98.23%. After performing full integer quantization on the best model, its size was reduced from 513.23 KB to 136.51 KB and was successfully deployed on the Arduino Nano 33 BLE Sense Rev2 using the Edge Impulse platform. The device's memory usage was 29.1 KB, flash usage was 189.6 KB, and the model's average inference time was 21 milliseconds, requiring approximately 0.01395 GOP, with a computational performance of around 0.664 GOPS. Even after quantization, the model maintained an accuracy of 97% and an F1 score of 97%, ensuring efficient utilization of computational resources. This deployment highlights the potential of TinyML in achieving low-latency and efficient HAR systems, making it suitable for real-time human activity recognition applications.

PMID:40263516 | DOI:10.1038/s41598-025-98571-2

Sci Rep. 2025 Apr 22;15(1):14000. doi: 10.1038/s41598-025-98269-5.

ABSTRACT

Habesha kemis, an Ethiopian attire traditionally donned by women belonging to the Habesha community, has undergone variations of designs over time. Initially, it comprised a lengthy dress with a fitted bodice and sleeves extending to the ankles. In the Amhara region, various ethnic groups such as Gojjam, Gondar, Shewa, Agew, and Wollo uphold their distinct cultural customs. While these Habesha garments may appear similar outwardly, their embroidered motifs exhibit unique patterns, shapes, and hues, symbolizing the rich cultural legacy of Gojjam, Gondar, Shewa, Agew, and Wollo. The study aimed to identify the most appropriate model for recognizing and classifying the quality of Habesha kemis embroidery design. Digital image processing methods and CNN models incorporating VGG16, VGG19, and ResNet50v2 classifiers were used. Following the gathering of datasets, image preprocessing and segmentation were employed to enhance the model's performance. In segmentation, we used canny edge detection, local binary pattern, and dilation with contour detection for segmenting and automatically cropping each habesha kemis. After applying the segmentation process, the individual habesha kemis and foreign matters are placed in a folder based on their corresponding categories. This resulted in 320 images before augmenting for each class amount representative. The performance of VGG16, VGG19, and ResNet50v2 for Agew, Gojjam, Gonder, Shewa, and Wollo was evaluated. This process resulted in an image size of 224 × 224 in the CNN model with a VGG16 architecture and a SoftMax classifier of course we try also 64 × 64 and 128 × 128. Augmentation techniques were applied to increase the dataset size from 1600 to 3,270. Finally, the model was evaluated and achieved an accuracy of 95.72% in test data and 99.62% in training data compared to the VGG19 and ResNet50v2 models.

PMID:40263488 | DOI:10.1038/s41598-025-98269-5

Sci Rep. 2025 Apr 22;15(1):13854. doi: 10.1038/s41598-025-98281-9.

ABSTRACT

With the rapid advancement of medical informatics, the accumulation of electronic medical records and clinical diagnostic data provides unprecedented opportunities for intelligent medical text classification. However, challenges such as class imbalance, semantic heterogeneity, and data sparsity limit the effectiveness of traditional classification models. In this study, we propose an enhanced medical text classification framework by integrating a self-attentive adversarial augmentation network (SAAN) for data augmentation and a disease-aware multi-task BERT (DMT-BERT) strategy. The proposed SAAN incorporates adversarial self-attention, improving the generation of high-quality minority class samples while mitigating noise. Furthermore, DMT-BERT simultaneously learns medical text representations and disease co-occurrence relationships, enhancing feature extraction from rare symptoms. Extensive experiments on the private clinical datasets and the public CCKS 2017 dataset demonstrate that our approach significantly outperforms baseline models, achieving the highest F1-score and ROC-AUC values. The proposed innovations address key limitations in medical text classification and contribute to the development of robust clinical decision-support systems.

PMID:40263477 | DOI:10.1038/s41598-025-98281-9

Sci Rep. 2025 Apr 22;15(1):13825. doi: 10.1038/s41598-025-96918-3.

ABSTRACT

Leukocytes or white blood cells plays an important role in protecting the body from various contagious diseases and infectious agents. Different conventional leukocyte analysis approaches often face several problems like inaccuracies, demanding the need for sophisticated approaches to improve diagnostic precision. Therefore, a holistic structure namely a novel Attention-based Dual Channel U-shaped Network (ADCU-Net) utilizing three datasets is introduced in this paper for effective leukocyte classification. The image quality is boosted in the preprocessing phase through noise reduction, contrast enhancement, and background removal, significantly improving clarity. Then, the Dung Beetle Optimization (DBO) algorithm enhanced with Levy flight optimization is implemented for effective image segmentation processes. A dung beetle with a levy flight strategy assists in streamlined exploration of the search space thereby the detection and delineation of specific regions within images are improved, which results in higher boundary detection accuracy. The evaluation of major quantitative measures such as standard deviation, mean and entropy is comprised in the feature extraction process which offers crucial insights into the structural characteristics of leukocytes. Finally, a novel ADCU-Net model is utilized for the effective classification process. This ADCU-Net model is particularly selected to effectively capture various features and preserve spatial data, achieving98.4% accuracy. Overall, this paper highlights the performance of integrated sophisticated deep-learning structures for accurate leukocyte classification and segmentation, enabling the path for improved diagnostic tools in clinical settings.

PMID:40263470 | DOI:10.1038/s41598-025-96918-3

Sci Data. 2025 Apr 22;12(1):678. doi: 10.1038/s41597-025-05021-9.

ABSTRACT

Accurate mapping of vegetation canopy height and biomass distribution is essential for effective forest monitoring, climate change mitigation, and sustainable forestry. Here we present high-resolution remote sensing-based canopy height (10 m resolution) and above ground biomass (AGB, 50 m resolution) maps for the forests of the Iberian Peninsula from 2017 to 2021, using a deep learning framework that integrates Sentinel-1, Sentinel-2, and LiDAR data. Two UNET models were developed: one trained on Airborne Laser Scanning (ALS) data (MAE: 1.22 m), while another using Global Ecosystem Dynamics Investigation (GEDI) footprints (MAE: 3.24 m). External validation with 6,308 Spanish National Forest Inventory (NFI) plots (2017-2019) confirmed canopy height reliability, showing MAEs of 2-3 m in tree-covered areas. AGB estimates were obtained through Random Forest models that linked UNET derived height predictions to NFI AGB data, achieves an MAE of ~29 Mg/ha. The creation of high-resolution maps of canopy height and biomass across various forest landscapes in the Iberian Peninsula provides a valuable new tool for environmental researchers, policy makers, and forest management professionals, offering detailed insights that can inform conservation strategies, carbon sequestration efforts, and sustainable forest management practices.

PMID:40263468 | DOI:10.1038/s41597-025-05021-9