Eur Radiol. 2025 Apr 19. doi: 10.1007/s00330-025-11588-z. Online ahead of print.

ABSTRACT

OBJECTIVES: To propose and evaluate a novel deep learning model for directly estimating pharmacokinetic (PK) parameter maps and uncertainty estimation from DCE-MRI.

METHODS: In this single-center study, patients with adult-type diffuse gliomas who underwent preoperative DCE-MRI from Apr 2010 to Feb 2020 were retrospectively enrolled. A spatiotemporal probabilistic model was used to create synthetic PK maps. Structural Similarity Index Measure (SSIM) to ground truth (GT) maps were calculated. Reliability was evaluated using the intraclass correlation coefficient (ICC) for synthetic and GT PK maps. For clinical validation, Area Under the Receiver Operating Characteristic Curve (AUROC) was obtained for predicting WHO low vs high grade and IDH-wildtype vs mutant.

RESULTS: 329 patients (mean age, 55 ± 15 years, 197 men) were eligible. Synthetic Ktrans, Vp, Ve maps showed high SSIM (0.961, 0.962, 0.890) compared to the GT maps. The ICC of PK maps was significantly higher in synthetic PK maps compared to the conventional approach: 1.00 vs 0.68 (p < 0.001) for Ktrans, 1.00 vs 0.59 (p < 0.001) for Vp, 1.00 vs 0.64 (p < 0.001) for Ve. PK values of enhancing tumor portion obtained from synthetic and GT maps were comparable in AUROC: (1) Ktrans, 0.857 vs 0.842 (p = 0.57); Vp, 0.864 vs 0.835 (p = 0.31); and Ve, 0.835 vs 0.830 (p = 0.88) for mutation prediction. (2) Ktrans, 0.934 vs 0.907 (p = 0.50); Vp, 0.927 vs 0.899 (p = 0.24); and Ve, 0.945 vs 0.910 (p = 0.24) for glioma grading.

CONCLUSION: Synthetic PK maps generated from DCE-MRI using a spatiotemporal probabilistic deep-learning model showed improved reliability without compromising diagnostic performance in glioma grading.

KEY POINTS: Question Can a deep learning model enhance the reliability of dynamic contrast-enhanced MRI (DCE-MRI) for more consistent and clinically acceptable glioma imaging? Findings A spatiotemporal deep learning model outperformed the Tofts model in Ktrans reliability and preserved diagnostic performance for IDH mutation and glioma grade, bypassing arterial input function estimation. Clinical relevance Enhancing DCE-MRI reliability with deep learning improves imaging consistency, supports molecular tumor characterization through reproducible pharmacokinetic maps, and enables personalized treatment planning, which might lead to better clinical outcomes for patients with diffuse gliomas.

PMID:40252095 | DOI:10.1007/s00330-025-11588-z

Br J Dermatol. 2025 Apr 19:ljaf149. doi: 10.1093/bjd/ljaf149. Online ahead of print.

ABSTRACT

BACKGROUND: Hair density traits, including follicular unit density (FUD) and hairs per follicular unit (HFU), are influenced by both environmental and genetic factors. Understanding these determinants can refine our knowledge of hair growth patterns and enable more targeted interventions. Large-scale research has historically been constrained by difficulties in precise phenotyping.

OBJECTIVES: We aimed to identify environmental and genetic factors associated with hair density in East Asian populations and to explore shared genetic influences among other hair traits and hair disorders.

METHODS: We performed quantitative assessments of FUD and HFU using trichoscopic images from 5,735 East Asian individuals. Accurate phenotyping was achieved through deep learning-based analyses with manual correction. We used multiple regression to evaluate demographic, lifestyle, and reproductive factors, and conducted genome-wide association studies (GWAS), meta-analysis, and a combined GWAS (C-GWAS) for hair density, curliness, eyebrow thickness, and beard thickness. Significant associations were compared with published results on male pattern baldness (MPB). Gene-finasteride usage interactions were evaluated via mixed linear models in longitudinal UK Biobank data.

RESULTS: Age, sex, body mass index, and menopausal status were significantly associated with both FUD and HFU. Three genetic loci, rs11940736 at 4q28.1 (near SPRY1), rs10908366 at 1p34.3 (near RSPO1), and rs3771033 at 2q23.3 (intron NRP2), showed significant associations with hair density, with functional annotations implicating these genes in hair follicle development. In particular, rs3771033 was a significant eQTL for NRP2, where the T allele correlated with lower NRP2 expression but higher FUD. We also observed substantial genetic overlap among hair density traits, curliness, eyebrow thickness, beard thickness, and MPB. In UK Biobank analyses, rs3771033 exhibited allele-specific treatment effects on finasteride response for MPB.

CONCLUSION: We identified three loci that shape hair density in East Asian populations. Our results clarify the genetic and environmental architecture underlying hair density traits and suggest that genotype-specific responses to finasteride may open new avenues for personalized management of hair disorders.

PMID:40251992 | DOI:10.1093/bjd/ljaf149

Brief Bioinform. 2025 Mar 4;26(2):bbaf151. doi: 10.1093/bib/bbaf151.

ABSTRACT

Accurate prediction of pathogenic variants in human disease-associated genes would have a profound effect on clinical decision-making; however, it remains a significant challenge due to the overwhelming number of these variants. We propose graph neural network for multimodal annotation-based pathogenicity prediction (GNN-MAP), a novel deep learning framework that effectively integrates multimodal annotations and similarity relationships among variants to predict the pathogenicity of multi-type variants. Trained on the ClinVar dataset, GNN-MAP exhibits superior predictive performance in internal validation and orthogonal test datasets, accurately predicting variant pathogenicity. Notably, GNN-MAP enables accurate prediction of the pathogenicity of rare variants and highly imbalanced datasets. Furthermore, it achieves high performance in the pathogenicity prediction of inherited retinal disease-specific variants, highlighting its effectiveness in disease-specific variant prediction. These findings suggest that the robust capability of GNN-MAP to predict pathogenicity across multiple variant types and datasets holds significant potential for applications in research and clinical settings.

PMID:40251830 | DOI:10.1093/bib/bbaf151

Brief Bioinform. 2025 Mar 4;26(2):bbaf179. doi: 10.1093/bib/bbaf179.

ABSTRACT

Non-coding RNAs (ncRNAs) play crucial roles in drug resistance and sensitivity, making them important biomarkers and therapeutic targets. However, predicting ncRNA-drug associations is challenging due to issues such as dataset imbalance and sparsity, limiting the identification of robust biomarkers. Existing models often fall short in capturing local and global sequence information, limiting the reliability of predictions. This study introduces DMGAT (diffusion map and heterogeneous graph attention network), a novel deep learning model designed to predict ncRNA-drug associations. DMGAT integrates diffusion maps for sequence embedding, graph convolutional networks for feature extraction, and GAT for heterogeneous information fusion. To address dataset imbalance, the model incorporates sensitivity associations and employs a random forest classifier to select reliable negative samples. DMGAT embeds ncRNA sequences and drug SMILES using the word2vec technique, capturing local and global sequence information. The model constructs a heterogeneous network by combining sequence similarity and Gaussian Interaction Profile kernel similarity, providing a comprehensive representation of ncRNA-drug interactions. Evaluated through five-fold cross-validation on a curated dataset from NoncoRNA and ncDR, DMGAT outperforms seven state-of-the-art methods, achieving the highest area under the receiver operating characteristic curve (0.8964), area under the precision-recall curve (0.8984), recall (0.9576), and F1-score (0.8285). The raw data are released to Zenodo with identifier 13929676. The source code of DMGAT is available at https://github.com/liutingyu0616/DMGAT/tree/main.

PMID:40251829 | DOI:10.1093/bib/bbaf179

Brief Funct Genomics. 2025 Jan 15;24:elaf003. doi: 10.1093/bfgp/elaf003.

ABSTRACT

Super-enhancers (SEs) are typically located in the regulatory regions of genes, driving high-level gene expression. Identifying SEs is crucial for a deeper understanding of gene regulatory networks, disease mechanisms, and the development and physiological processes of organisms, thus exerting a profound impact on research and applications in the life sciences field. Traditional experimental methods for identifying SEs are costly and time-consuming. Existing methods for predicting SEs based solely on sequence data use deep learning for feature representation and have achieved good results. However, they overlook biological features related to physicochemical properties, leading to low interpretability. Additionally, the complex model structure often requires extensive labeled data for training, which limits their further application in biological data. In this paper, we integrate the strengths of different models and proposes an ensemble model based on an integration strategy to enhance the model's generalization ability. It designs a multi-angle feature representation method that combines local structure and global information to extract high-dimensional abstract relationships and key low-dimensional biological features from sequences. This enhances the effectiveness and interpretability of the model's input features, providing technical support for discovering cell-specific and species-specific patterns of SEs. We evaluated the performance on both mouse and human datasets using five metrics, including area under the receiver operating characteristic curve accuracy, and others. Compared to the latest models, EnsembleSE achieved an average improvement of 4.5% in F1 score and an average improvement of 8.05% in recall, demonstrating the robustness and adaptability of the model on a unified test set. Source codes are available at https://github.com/2103374200/EnsembleSE-main.

PMID:40251827 | DOI:10.1093/bfgp/elaf003

Plant Cell. 2025 Apr 19:koaf086. doi: 10.1093/plcell/koaf086. Online ahead of print.

ABSTRACT

Photosynthetic organisms coordinate their metabolism and growth with diurnal light, which can range in intensity from limiting to excessive. Little is known about how light intensity impacts the diurnal program in Chlamydomonas reinhardtii, nor how diurnal rhythms in gene expression and metabolism shape photoprotective responses at different times of day. To address these questions, we performed a systems analysis of synchronized Chlamydomonas populations acclimated to low, moderate, and high diurnal light. Transcriptomic and proteomic data revealed that the Chlamydomonas rhythmic gene expression program is resilient to limiting and excess light: genome-wide, waves of transcripts and proteins peak at the same times in populations acclimated to stressful light intensities as in populations acclimated to moderate light. Yet, diurnal photoacclimation gives rise to hundreds of gene expression changes, even at night. Time-course measurements of photosynthetic efficiency and pigments responsive to excess light showed that high-light-acclimated cells partially overcome photodamage in the latter half of the day prior to cell division. Although gene expression and photodamage are dynamic over the diurnal cycle, Chlamydomonas populations acclimated to low and high diurnal light maintain altered photosystem abundance, thylakoid architecture, and non-photochemical quenching capacity through the night phase. This suggests that cells remember or anticipate the light intensities that they have typically encountered during the day. The integrated data constitute an excellent resource for understanding photoacclimation in eukaryotes under environmentally relevant conditions.

PMID:40251989 | DOI:10.1093/plcell/koaf086

New Phytol. 2025 Apr 18. doi: 10.1111/nph.70154. Online ahead of print.

NO ABSTRACT

PMID:40251862 | DOI:10.1111/nph.70154

Sci Rep. 2025 Apr 18;15(1):13426. doi: 10.1038/s41598-025-87409-6.

ABSTRACT

Cervical cancer, a prevalent gynecological malignancy, poses significant threats to women's health. Despite advances in treatment modalities, radiotherapy remains a cornerstone in managing cervical cancer. However, radiotherapy-induced complications, such as radiation proctitis, present substantial diagnostic and prognostic challenges. Accurate diagnosis are crucial for optimizing treatment strategies and improving patient outcomes. Deep learning has shown remarkable success in medical image segmentation, aiding clinicians in assessing patient conditions. In the other hand, radiomics excels in extracting diagnostically valuable features from medical images but requires extensive manual annotation and often lacks generalizability. Therefore, combining the strengths of deep learning and radiomics is pivotal in addressing these challenges. In this study, we propose a novel paradigm that leverages deep learning models for initial segmentation, followed by detailed radiomics analysis. Specifically, we utilize the Transformer-based SAM-Med2D model to extract visual features from CT images of cervical cancer patients. We apply T-tests and Lasso regression to identify features most correlated with radiation proctitis and build predictive models using logistic regression, random forest, and naive Gaussian Bayesian algorithms. Experimental results demonstrate that our method effectively extracts CT imaging features and exhibits excellent performance in diagnosis radiation proctitis. This approach not only enhances predictive accuracy but also provides a valuable tool for personalizing treatment plans and improving patient outcomes in cervical cancer radiotherapy.

PMID:40251184 | DOI:10.1038/s41598-025-87409-6

Sci Rep. 2025 Apr 18;15(1):13445. doi: 10.1038/s41598-025-98629-1.

ABSTRACT

Traditional drug design faces significant challenges due to inherent chemical and biological complexities, often resulting in high failure rates in clinical trials. Deep learning advancements, particularly generative models, offer potential solutions to these challenges. One promising algorithm is DrugGPT, a transformer-based model, that generates small molecules for input protein sequences. Although promising, it generates both chemically valid and invalid structures and does not incorporate the features of approved drugs, resulting in time-consuming and inefficient drug discovery. To address these issues, we introduce DrugGen, an enhanced model based on the DrugGPT structure. DrugGen is fine-tuned on approved drug-target interactions and optimized with proximal policy optimization. By giving reward feedback from protein-ligand binding affinity prediction using pre-trained transformers (PLAPT) and a customized invalid structure assessor, DrugGen significantly improves performance. Evaluation across multiple targets demonstrated that DrugGen achieves 100% valid structure generation compared to 95.5% with DrugGPT and produced molecules with higher predicted binding affinities (7.22 [6.30-8.07]) compared to DrugGPT (5.81 [4.97-6.63]) while maintaining diversity and novelty. Docking simulations further validate its ability to generate molecules targeting binding sites effectively. For example, in the case of fatty acid-binding protein 5 (FABP5), DrugGen generated molecules with superior docking scores (FABP5/11, -9.537 and FABP5/5, -8.399) compared to the reference molecule (Palmitic acid, -6.177). Beyond lead compound generation, DrugGen also shows potential for drug repositioning and creating novel pharmacophores for existing targets. By producing high-quality small molecules, DrugGen provides a high-performance medium for advancing pharmaceutical research and drug discovery.

PMID:40251288 | DOI:10.1038/s41598-025-98629-1

Int J Biol Macromol. 2025 Apr 16:143217. doi: 10.1016/j.ijbiomac.2025.143217. Online ahead of print.

ABSTRACT

Flaviviridae family comprises some of the most vulnerable viruses known for causing widespread outbreaks, high mortality rates, and severe long-term health complications in humans. Viruses like Dengue (DENV), Zika (ZIKV) and Hepatitis C (HCV) are endemic across the globe, especially in tropical and subtropical regions, infecting multiple tissues and leading to significant health crises. Investigating virus-host interactions across tissues can reveal tissue-specific drug targets and aid antiviral drug repurposing. In this study, we employed a multi-step computational approach to construct a comprehensive virus-human interactome by integrating virus-host protein-protein interactions (PPIs) with tissue-specific gene expression profiles to study key protein targets associated with Flaviviridae infections. Mapping drug-target predictions revealed druggable proteins - CCNA2 in peripheral blood mononuclear cells (PBMC) and EIF2S2, CDK7 and CARS in the liver, with Tamoxifen, Sirolimus, Entrectinib and L-cysteine as potential repurposable drugs, respectively. Further protein-ligand docking and molecular dynamics (MD) simulations were conducted to assess the stability of the complexes. These findings highlight common druggable human targets exploited by DENV, ZIKV and HCV, providing a foundation for broad-spectrum antiviral therapies. By focusing on shared host pathways and targets in viral replication, we propose promising drug candidates, supporting the development of unified therapeutic strategies against Flaviviridae viruses.

PMID:40250655 | DOI:10.1016/j.ijbiomac.2025.143217

Nat Commun. 2025 Apr 18;16(1):3355. doi: 10.1038/s41467-025-58464-4.

ABSTRACT

Therapeutic target identification is challenging in drug discovery, particularly for rare and orphan diseases. Here, we propose a disease signature, TRESOR, which characterizes the functional mechanisms of each disease through genome-wide association study (GWAS) and transcriptome-wide association study (TWAS) data, and develop machine learning methods for predicting inhibitory and activatory therapeutic targets for various diseases from target perturbation signatures (i.e., gene knockdown and overexpression). TRESOR enables highly accurate identification of target candidate proteins that counteract disease-specific transcriptome patterns, and the Bayesian optimization with omics-based disease similarities achieves the performance enhancement for diseases with few or no known targets. We make comprehensive predictions for 284 diseases with 4345 inhibitory target candidates and 151 diseases with 4040 activatory target candidates, and elaborate the promising targets using several independent cohorts. The methods are expected to be useful for understanding disease-disease relationships and identifying therapeutic targets for rare and orphan diseases.

PMID:40251160 | DOI:10.1038/s41467-025-58464-4

Urol Clin North Am. 2025 May;52(2):287-296. doi: 10.1016/j.ucl.2025.01.010. Epub 2025 Feb 28.

ABSTRACT

This article covers rare adrenal tumors including functional adenomas, myelolipomas, ganglioneuromas and neuroblastomas, and metastasis to the adrenal gland. It explores their clinical presentation and behavior, hormonal activity, imaging features, other diagnostic considerations, and approaches to management. The variety of rare tumors and their unique behaviors covered in this article underscores the need to maintain up-to-date knowledge and surgical skills, as well as the importance of a multidisciplinary approach to patient care.

PMID:40250895 | DOI:10.1016/j.ucl.2025.01.010

Thorax. 2025 Apr 18:thorax-2025-223225. doi: 10.1136/thorax-2025-223225. Online ahead of print.

NO ABSTRACT

PMID:40250987 | DOI:10.1136/thorax-2025-223225

Int J Radiat Oncol Biol Phys. 2025 Apr 16:S0360-3016(25)00372-4. doi: 10.1016/j.ijrobp.2025.04.008. Online ahead of print.

ABSTRACT

PURPOSE: Combining immune checkpoint inhibitors (ICIs) with radiation therapy (RT) has led to significant advancements in cancer treatment. However, evidence from clinical and experimental studies suggests that this combination may increase hematopoietic and lymphatic toxicity. This study aims to investigate the effects of the concurrent application of ICIs (anti-PD-1 and anti-CTLA-4) on radiation-induced hematopoietic and lymphatic injuries under standardized and controlled experimental conditions.

MATERIALS AND METHODS: We utilized various experimental models in C57BL/6 and BALB/c mice to evaluate the impact of ICIs combined with RT on the hematopoietic system. These models involved different RT doses, regimens, and target sites in both healthy and tumor-bearing mice.

RESULTS: Our findings showed that the concurrent use of ICIs did not meaningfully affect post-RT pancytopenia kinetics or the regeneration of specific blood cell lineages over time. Consistently, combining RT with ICIs did not significantly enhance DNA damage in immune cells within the bloodstream. This outcome was comparable across different RT doses, regimens, and target sites and was reproducible in both tumor-bearing and non-tumor-bearing mice. Additionally, there were no significant increases in late side effects, including reductions in bone marrow cell counts or megakaryocyte numbers, after combined radioimmunotherapy.

CONCLUSION: These findings suggest that combining ICIs with RT does not exacerbate hematological toxicity. This information is valuable for interpreting adverse events in clinical trials involving radioimmunotherapy and for predicting potential hematological side effects in cancer patients receiving these treatments.

PMID:40250771 | DOI:10.1016/j.ijrobp.2025.04.008

Cell. 2025 Apr 17;188(8):2043-2062. doi: 10.1016/j.cell.2025.03.011.

ABSTRACT

Major progress has been made in elucidating the molecular, cellular, and supracellular mechanisms underlying aging. This has spurred the birth of geroscience, which aims to identify actionable hallmarks of aging. Aging can be viewed as a process that is promoted by overactivation of gerogenes, i.e., genes and molecular pathways that favor biological aging, and alternatively slowed down by gerosuppressors, much as cancers are caused by the activation of oncogenes and prevented by tumor suppressors. Such gerogenes and gerosuppressors are often associated with age-related diseases in human population studies but also offer targets for modeling age-related diseases in animal models and treating or preventing such diseases in humans. Gerogenes and gerosuppressors interact with environmental, behavioral, and psychological risk factors to determine the heterogeneous trajectory of biological aging and disease manifestation. New molecular profiling technologies enable the characterization of gerogenic and gerosuppressive pathways, which serve as biomarkers of aging, hence inaugurating the era of precision geromedicine. It is anticipated that, pending results from randomized clinical trials and regulatory approval, gerotherapeutics will be tailored to each person based on their genetic profile, high-dimensional omics-based biomarkers of aging, clinical and digital biomarkers of aging, psychosocial profile, and past or present exposures.

PMID:40250404 | DOI:10.1016/j.cell.2025.03.011

Radiat Oncol. 2025 Apr 18;20(1):59. doi: 10.1186/s13014-025-02642-7.

ABSTRACT

BACKGROUND: Delineating the internal gross tumor volume (IGTV) is crucial for the treatment of non-small cell lung cancer (NSCLC). Deep learning (DL) enables the automation of this process; however, current studies focus mainly on multiple phases of four-dimensional (4D) computed tomography (CT), which leads to indirect results. This study proposed a DL-based method for automatic IGTV delineation using maximum and average intensity projections (MIP and AIP, respectively) from 4D CT.

METHODS: We retrospectively enrolled 124 patients with NSCLC and divided them into training (70%, n = 87) and validation (30%, n = 37) cohorts. Four-dimensional CT images were acquired, and the corresponding MIP and AIP images were generated. The IGTVs were contoured on 4D CT and used as the ground truth (GT). The MIP or AIP images, along with the corresponding IGTVs (IGTVMIP-manu and IGTVAIP-manu, respectively), were fed into the DL models for training and validation. We assessed the performance of three segmentation models-U-net, attention U-net, and V-net-using the Dice similarity coefficient (DSC) and the 95th percentile of the Hausdorff distance (HD95) as the primary metrics.

RESULTS: The attention U-net model trained on AIP images presented a mean DSC of 0.871 ± 0.048 and mean HD95 of 2.958 ± 2.266 mm, whereas the model trained on MIP images achieved a mean DSC of 0.852 ± 0.053 and mean HD95 of 3.209 ± 2.136 mm. Among the models, attention U-net and U-net achieved similar results, considerably surpassing V-net.

CONCLUSIONS: DL models can automate IGTV delineation using MIP and AIP images, streamline contouring, and enhance the accuracy and consistency of lung cancer radiotherapy planning to improve patient outcomes.

PMID:40251610 | DOI:10.1186/s13014-025-02642-7

BMC Oral Health. 2025 Apr 18;25(1):592. doi: 10.1186/s12903-025-05989-1.

ABSTRACT

BACKGROUND: Artificial intelligence (AI) holds immense potential in revolutionizing restorative dentistry, offering transformative solutions for diagnostic, prognostic, and treatment planning tasks. Traditional restorative dentistry faces challenges such as clinical variability, resource limitations, and the need for data-driven diagnostic accuracy. AI's ability to address these issues by providing consistent, precise, and data-driven solutions is gaining significant attention. This comprehensive literature review explores AI applications in caries detection, endodontics, dental restorations, tooth surface loss, tooth shade determination, and regenerative dentistry. While this review focuses on restorative dentistry, AI's transformative impact extends to orthodontics, prosthodontics, implantology, and dental biomaterials, showcasing its versatility across various dental specialties. Emerging trends such as AI-powered robotic systems, virtual assistants, and multi-modal data integration are paving the way for groundbreaking innovations in restorative dentistry.

METHODS: Methodologically, a systematic approach was employed, focusing on English-language studies published between 2020-2025(January), resulting in 63 peer-reviewed publications for analysis. Studies in caries detection, pedodontics, dental restorations, endodontics, tooth surface loss, and tooth shade determination highlighted AI trends and advancements. Inclusion criteria focused on AI applications in restorative dentistry, and publication timeframe. PRISMA guidelines were followed to ensure transparency in study selection, emphasizing on accuracy metrics and clinical relevance. The study selection process was carefully documented, and a flowchart of the stages, including identification, screening, eligibility, and inclusion, is shown in Fig. 1 to provide further clarity and reproducibility in the selection process.

RESULTS: The review identified significant advancements in AI-driven solutions across multiple domains of restorative dentistry. Notable studies demonstrated AI's ability to achieve high diagnostic accuracy, such as up to 95% accuracy in caries detection, and its capacity to improve treatment planning efficiency, thus reducing patient chair time. Predictive analytics for personalized treatments was another area where AI has shown substantial promise.

CONCLUSION: The review discussed trends, challenges, and future research directions in AI-driven dentistry, highlighting the transformative potential of AI in optimizing dental care. Key challenges include data privacy concerns, algorithmic bias, interpretability of AI decision-making processes, and the need for standardized AI training programs in dental education. Further research should focus on integrating AI with emerging technologies like 3D printing for personalized restorations, and developing AI training programs for dental professionals.

CLINICAL SIGNIFICANCE: The integration of AI into restorative dentistry offers precision-driven solutions for improved patient outcomes. By enabling faster diagnostics, personalized treatment approaches, and preventive care strategies, AI can significantly enhance patient-centered care and clinical efficiency. This review contributes to advancing the understanding and implementation of AI in dental practice by synthesizing key findings, identifying trends, and addressing challenges.

PMID:40251567 | DOI:10.1186/s12903-025-05989-1

BMC Med Imaging. 2025 Apr 18;25(1):128. doi: 10.1186/s12880-025-01667-4.

ABSTRACT

BACKGROUND: Medical imaging has been essential and has provided clinicians with useful information about the human body to diagnose various health issues. Early diagnosis of diseases based on medical imaging can mitigate the risk of severe consequences and enhance long-term health outcomes. Nevertheless, the task of diagnosing diseases based on medical imaging can be challenging due to the exclusive ability of clinicians to interpret the outcomes of medical imaging, which is time-consuming and susceptible to human fallibility. The ensemble model has the potential to enhance the accuracy of diagnoses of diseases based on medical imaging by analyzing vast volumes of data and identifying trends that may not be immediately apparent to doctors. However, it takes a lot of memory and processing resources to train and maintain several ensemble models. These challenges highlight the necessity of effective and scalable ensemble models that can manage the intricacies of medical imaging assignments.

METHODS: This study employed an SLR technique to explore the latest advancements and approaches. By conducting a thorough and systematic search of Scopus and Web of Science databases in accordance with the principles outlined in the PRISMA, employing keywords namely ensemble model and medical imaging.

RESULTS: This study included a total of 75 papers that were published between 2019 and 2024. The categorization, methodologies, and use of medical imaging were key factors examined in the analysis of the 30 cited papers included in this study, with a focus on diagnosing diseases.

CONCLUSIONS: Researchers have observed the emergence of an ensemble model for disease diagnosis using medical imaging since it has demonstrated improved accuracy and may guide future studies by highlighting the limitations of the ensemble model.

PMID:40251529 | DOI:10.1186/s12880-025-01667-4

BMC Bioinformatics. 2025 Apr 18;26(1):108. doi: 10.1186/s12859-025-06116-1.

ABSTRACT

BACKGROUND: A gene regulatory network (GRN) is a graph-level representation that describes the regulatory relationships between transcription factors and target genes in cells. The reconstruction of GRNs can help investigate cellular dynamics, drug design, and metabolic systems, and the rapid development of single-cell RNA sequencing (scRNA-seq) technology provides important opportunities while posing significant challenges for reconstructing GRNs. A number of methods for inferring GRNs have been proposed in recent years based on traditional machine learning and deep learning algorithms. However, inferring the GRN from scRNA-seq data remains challenging owing to cellular heterogeneity, measurement noise, and data dropout.

RESULTS: In this study, we propose a deep learning model called graph representational learning GRN (GRLGRN) to infer the latent regulatory dependencies between genes based on a prior GRN and data on the profiles of single-cell gene expressions. GRLGRN uses a graph transformer network to extract implicit links from the prior GRN, and encodes the features of genes by using both an adjacency matrix of implicit links and a matrix of the profile of gene expression. Moreover, it uses attention mechanisms to improve feature extraction, and feeds the refined gene embeddings into an output module to infer gene regulatory relationships. To evaluate the performance of GRLGRN, we compared it with prevalent models and performed ablation experiments on seven cell-line datasets with three ground-truth networks. The results showed that GRLGRN achieved the best predictions in AUROC and AUPRC on 78.6% and 80.9% of the datasets, and achieved an average improvement of 7.3% in AUROC and 30.7% in AUPRC. The interpretation discussion and the network visualization were conducted.

CONCLUSIONS: The experimental results and case studies illustrate the considerable performance of GRLGRN in predicting gene interactions and provide interpretability for the prediction tasks, such as identifying hub genes in the network and uncovering implicit links.

PMID:40251476 | DOI:10.1186/s12859-025-06116-1

Eur Radiol. 2025 Apr 19. doi: 10.1007/s00330-025-11596-z. Online ahead of print.

ABSTRACT

OBJECTIVES: Deep learning reconstruction (DLR) has shown promising image denoising ability, but its radiation dose reduction potential remains unknown. The objective of this study was to investigate the diagnostic performance of DLR compared to iterative reconstruction (IR) in the detection of liver lesions at standard-dose and reduced-dose CT.

MATERIALS AND METHODS: Participants with known liver metastases from gastrointestinal and pancreatic adenocarcinoma were prospectively included from routine follow-up (October 2020 to March 2022). Participants received standard-dose CT and two additional reduced-dose scans during the same contrast administration, each reconstructed with IR and high-strength DLR. Two radiologists evaluated images for the presence of liver lesions, and a third established a reference standard. Diagnostic performance was compared using McNemar's test and mixed effects logistic regression.

RESULTS: Forty-four participants (mean age 66 years ± 11 [standard deviation], 28 men) were evaluated with 348 included liver lesions ≤ 20 mm (297 metastases, 51 benign; mean size 9.1 ± 4.3 mm). Mean volume CT dose index was 14.2, 7.8 mGy, and 5.1 mGy. Between algorithms, no significant difference in lesion detection was observed within dose levels. Detection of 233 lesions ≤ 10 mm was deteriorated with lower dose levels despite DLR denoising, with 185 detected at standard-dose IR (79.4%; 95% CI: 73.5-84.3) vs 128 at medium-dose DLR (54.9%; 95% CI: 48.3-61.4; p < 0.001) and 105 at low-dose DLR (45.1%; 95% CI: 38.6-51.7; p < 0.001).

CONCLUSION: Diagnostic performance for liver lesion detection was comparable between algorithms. When the detection of smaller lesions is important, DLR did not facilitate substantial dose reduction.

KEY POINTS: Question Methods to reduce CT radiation dose are desirable in clinical practice, and DLR has shown promising image denoising capabilities. Findings Liver lesion detection was comparable for DLR and IR across dose levels, but detection of smaller lesions deteriorated with lower dose levels. Clinical relevance Although potent in image noise reduction, the diagnostic performance of DLR is comparable to IR at standard-dose and reduced-dose CT. Care must be taken in pursuit of dose reduction when the detection and characterization of smaller liver lesions are of clinical importance.

PMID:40251443 | DOI:10.1007/s00330-025-11596-z