Sci Rep. 2025 Mar 3;15(1):7441. doi: 10.1038/s41598-025-91725-2.

ABSTRACT

Recent advancements in deep learning have revolutionized digital dentistry, highlighting the importance of precise dental segmentation. This study leverages active learning with the three-dimensional (3D) nnU-net and multi-labels to improve segmentation accuracy of dental anatomies, including the maxillary sinuses, maxilla, mandible, and inferior alveolar nerves (IAN), which are important for implant planning, in 3D cone-beam computed tomography (CBCT) scans. Segmentation accuracy was compared using single-label, adjacent pair-label, and multi-label relevant anatomic structures with 60 CBCT scans from Kooalldam Dental Hospital and externally validated using data from Seoul National University Dental Hospital. The dataset was divided into three training stages for active learning. The evaluation metrics were assessed through the Dice similarity coefficient (DSC) and mean absolute difference. The overall internal test set DSCs from the multi-label, single-label, and pair-label models were 95%, 91% (paired t-test; p = 0.01), and 93% (p = 0.03), respectively. The DSC of the IAN in the internal and external datasets increased from 83% to 79%, 87% and 81%, to 90% and 86% for the single-label, pair-label, and multi-label models, respectively (all p = 0.01). Prediction accuracy improved over time, significantly reducing the manual correction time. Our active learning and multi-label strategies facilitated accurate automatic segmentation.

PMID:40033040 | DOI:10.1038/s41598-025-91725-2

Sci Rep. 2025 Mar 3;15(1):7500. doi: 10.1038/s41598-025-91483-1.

ABSTRACT

Color is an important index for human visual evaluation of landscape, and it is also a key factor affecting people's recognition and experience of heritage landscape. In this study, five important sites of the Yangzhou Grand Canal were selected for the color quantification analysis by using the Deep Learning(DL) scene parsing algorithm. The color characteristics of the winter and summer landscape of the five sites were evaluated as well as the Scenic Beauty Estimation (SBE) value. Furthermore, the correlation analysis between the color characteristics and the SBE value was established in order to study the relationship between color characteristics and the landscape beauty. The main results are as follows: ①.The dominant color of the five sites is blue and green, the building color is mainly orange and yellow in both winter and summer. The dominant plant color in five sites is green in summer, whereas in winter, changes to yellow(Site5:YZJGD) or cyan(Site1:DGGD, Site3:GZGD); ②.The overall color saturation is low in winter with the percentages of Very Low Saturation in almost each site(except site5:YZJGD)reach 80-98%. Summer has Medium Saturation colors, the percentage of Mid Saturation of sky in Site 2(GMS) in summer is 44.87%. ③. The landscapes have low brightness in winter and higher brightness in summer in all sites, sky is the only category whose High Brightness value exceeds 50% in both seasons.And in winter, landscapes are most prevalent in Low Brightness and Medium Brightness. In summer, the percentages of Medium Brightness and High Brightness increase.④.The color diversity of the sites in winter varies significantly, whereas the color diversity of the sites in summer varies slightly.The highest color diversity of plants is found in DGGD(Diversity > 1.5). ⑤.In winter, the highest SBE value is found in Site2:GMS(0.5956), and the lowest SBE value is found in Site5:YZJGD(- 0.8216),which is a large gap(1.4172).The highest average SBE value is in Site2:GMS(0.5062), followed by Site3:GZGD (0.2091), which both have average values greater than zero. ⑥.Correlation analysis revealed that there is no significant correlation between the saturation and SBE values(p > 0.05).However, the Pearson correlation coefficients which are - 0.625(winter) and 0.689(summer) indicate strong correlation.Meanwhile, there is no significant correlation between the color diversity and SBE values(p > 0.05). However, the Pearson correlation coefficients are 0.807(winter) and - 0.747(summer), indicating strong correlation.This study provides an in-depth examination of the Canal landscape color, it is hoped to promote the systematic and scientific study of landscape colors and provide a theoretical basis for the scientific design of heritage landscape color.

PMID:40033036 | DOI:10.1038/s41598-025-91483-1

Sci Rep. 2025 Mar 3;15(1):7393. doi: 10.1038/s41598-025-92202-6.

ABSTRACT

This research explores the acquisition and analysis of vibroacoustic signals generated during tissue-tool interactions, using a conventional aspiration needle enhanced with a proximally mounted MEMS audio sensor, to extract temperature information. Minimally invasive temperature monitoring is critical in thermotherapy applications, but current methods often rely on additional sensors or simulations of typical tissue behavior. In this study, a commercially available needle was inserted into water-saturated foams with temperatures ranging from 25 to 55 °C, varied in 5° increments. Given that temperature affects the speed of sound, water's heat capacity, and the mechanical properties of most tissues, it was hypothesized that the vibroacoustic signals recorded during needle insertion would carry temperature-dependent information. The acquired signals were segmented, processed, and analyzed using signal processing techniques and a deep learning algorithm. Results demonstrated that the audio signals contained distinct temperature-dependent features, enabling temperature prediction with a root mean squared error of approximately 3 °C. We present these initial laboratory findings, highlighting significant potential for refinement. This novel approach could pave the way for a real-time, minimally invasive method for thermal monitoring in medical applications.

PMID:40032997 | DOI:10.1038/s41598-025-92202-6

BMJ Case Rep. 2025 Mar 3;18(3):e263966. doi: 10.1136/bcr-2024-263966.

ABSTRACT

Idiopathic pulmonary fibrosis (IPF) is a fibrosing pneumonia of unknown causation with a chronic, progressive course that may be modified by treatment with the antifibrotic agents, pirfenidone and nintedanib. Both drugs have been shown to slow disease progression, but, in rare cases, pirfenidone has been shown to stabilise and even improve lung function. We present a case of a patient whose lung function and pathognomonic features on CT imaging improved significantly on commencement of treatment with pirfenidone. Withholding pirfenidone was associated with a functional and morphological deterioration on imaging that subsequently reversed and stabilised following recommencement of this treatment. We discuss potential mechanisms that might explain this treatment response, compare our case to others described previously and the potential consequences that restricted prescribing within a specified range of vital capacity may have on the opportunity to influence the natural history of IPF early before irreversible fibrosis develops.

PMID:40032576 | DOI:10.1136/bcr-2024-263966

Nat Struct Mol Biol. 2025 Mar 3. doi: 10.1038/s41594-025-01482-z. Online ahead of print.

ABSTRACT

Spermatogenesis is a unidirectional differentiation process that generates haploid sperm, but how the gene expression program that directs this process is established is largely unknown. Here we determine the high-resolution three-dimensional (3D) chromatin architecture of mouse male germ cells during spermatogenesis and show that CTCF-mediated 3D chromatin dictates the gene expression program required for spermatogenesis. In undifferentiated spermatogonia, CTCF-mediated chromatin interactions between meiosis-specific super-enhancers (SEs) and their target genes precede activation of these SEs on autosomes. These meiotic SEs recruit the master transcription factor A-MYB (MYBL1) in meiotic spermatocytes, which strengthens their 3D contacts and instructs a burst of meiotic gene expression. We also find that at the mitosis-to-meiosis transition, the germline-specific Polycomb protein SCML2 facilitates the resolution of chromatin loops that are specific to mitotic spermatogonia. Moreover, SCML2 and A-MYB help shape the unique 3D chromatin organization of sex chromosomes during meiotic sex chromosome inactivation. We propose that CTCF-mediated 3D chromatin organization regulates epigenetic priming that directs unidirectional differentiation, thereby determining the cellular identity of the male germline.

PMID:40033153 | DOI:10.1038/s41594-025-01482-z

Nat Microbiol. 2025 Mar 3. doi: 10.1038/s41564-025-01938-4. Online ahead of print.

ABSTRACT

Diet, microbiome, inflammation and host genetics have been linked to colorectal cancer development; however, it is not clear whether and how these factors interact to promote carcinogenesis. Here we used Il10-/- mice colonized with bacteria previously associated with colorectal cancer: enterotoxigenic Bacteroides fragilis, Helicobacter hepaticus or colibactin-producing (polyketide synthase-positive (pks+)) Escherichia coli and fed either a low-carbohydrate (LC) diet deficient in soluble fibre, a high-fat and high-sugar diet, or a normal chow diet. Colonic polyposis was increased in mice colonized with pks+ E. coli and fed the LC diet. Mechanistically, mucosal inflammation was increased in the LC-diet-fed mice, leading to diminished colonic PPAR-γ signalling and increased luminal nitrate levels. This promoted both pks+ E. coli growth and colibactin-induced DNA damage. PPAR-γ agonists or supplementation with dietary soluble fibre in the form of inulin reverted inflammatory and polyposis phenotypes. The pks+ E. coli also induced more polyps in mismatch-repair-deficient mice by inducing a senescence-associated secretory phenotype. Moreover, oncogenic effects were further potentiated by inflammatory triggers in the mismatch-repair-deficient model. These data reveal that diet and host genetics influence the oncogenic potential of a common bacterium.

PMID:40033140 | DOI:10.1038/s41564-025-01938-4

Commun Biol. 2025 Mar 3;8(1):355. doi: 10.1038/s42003-025-07768-8.

ABSTRACT

Plants cope with the environment by displaying large phenotypic variation. Two spectra of global plant form and function have been identified: a size spectrum from small to tall species with increasing stem tissue density, leaf size, and seed mass; a leaf economics spectrum reflecting slow to fast returns on investments in leaf nutrients and carbon. When species assemble to communities it is assumed that these spectra are filtered by the environment to produce community level functional composition. It is unknown what are the main drivers for community functional composition in a large area such as Amazonia. We use 13 functional traits, including wood density, seed mass, leaf characteristics, breeding system, nectar production, fruit type, and root characteristics of 812 tree genera (5211 species), and find that they describe two main axes found at the global scale. At community level, the first axis captures not only the 'fast-slow spectrum', but also most size-related traits. Climate and disturbance explain a minor part of this variance compared to soil fertility. Forests on poor soils differ largely in terms of trait values from those on rich soils. Trait composition and soil fertility exert a strong influence on forest functioning: biomass and relative biomass production.

PMID:40033015 | DOI:10.1038/s42003-025-07768-8

Nat Rev Rheumatol. 2025 Mar 3. doi: 10.1038/s41584-025-01231-y. Online ahead of print.

NO ABSTRACT

PMID:40032952 | DOI:10.1038/s41584-025-01231-y

Genome Res. 2025 Mar 3:gr.280021.124. doi: 10.1101/gr.280021.124. Online ahead of print.

ABSTRACT

SQANTI-reads leverages SQANTI3, a tool for the analysis of the quality of transcript models, to develop a read-level quality control framework for replicated long-read RNA-seq experiments. The number and distribution of reads, as well as the number and distribution of unique junction chains (transcript splicing patterns), in SQANTI3 structural categories are informative of raw data quality. Multisample visualizations of QC metrics are presented by experimental design factors to identify outliers. We introduce new metrics for 1) the identification of potentially under-annotated genes and putative novel transcripts and for 2) quantifying variation in junction donors and acceptors. We applied SQANTI-reads to two different datasets, a Drosophila developmental experiment and a multiplatform dataset from the LRGASP project and demonstrate that the tool effectively reveals the impact of read coverage on data quality, and readily identifies strong and weak splicing sites.

PMID:40032587 | DOI:10.1101/gr.280021.124

Genome Res. 2025 Mar 3:gr.279865.124. doi: 10.1101/gr.279865.124. Online ahead of print.

ABSTRACT

Long-read sequencing (LRS) technologies have revolutionized transcriptomic research by enabling the comprehensive sequencing of full-length transcripts. Using these technologies, researchers have reported tens of thousands of novel transcripts, even in well-annotated genomes, while developing new algorithms and experimental approaches to handle the noisy data. The LRGASP community effort benchmarked LRS methods in transcriptomics and validated many novel, lowly-expressed, sample-specific transcripts identified by long reads. These molecules represent deviations of the major transcriptional program, that were easily overlooked by short-read sequencing methods but are now captured by the full-length, single-molecule approach. This Perspective discusses the challenges and opportunities associated with LRS' capacity to unravel this fraction of the transcriptome, both in terms of transcriptome biology and genome annotation. For transcriptome biology, we need to develop novel experimental and computational methods to effectively differentiate technology errors from rare but real molecules. For genome annotation, we must agree on the strategy to capture molecular variability while still defining reference annotations that are useful for genome research.

PMID:40032585 | DOI:10.1101/gr.279865.124

Environ Pollut. 2025 Mar 1:125914. doi: 10.1016/j.envpol.2025.125914. Online ahead of print.

ABSTRACT

Microplastics are an increasingly prevalent form of pollution in coastal ecosystems. Current research focuses on understanding the impacts of such synthetic particles on the health and functioning of aquatic organisms. Recent studies have shown that invertebrates can accumulate microplastics in their tissue, impacting key functions such as growth, reproduction, feeding activity, and metabolism. Owing to their chemical composition, microplastics accumulating in the digestive tract of animals may alter the diversity and abundance of microbiota. Despite the important implications of such microbiota shifts on digestive ability and fitness, investigations on microplastics as causative agents are so far limited. In this study, we tested the effect of microfibers, on the digestive gland microbiota of the blue mussel Mytilus edulis after a 52-day exposure. Our findings show that exposure to microplastics can alter the composition of the digestive gland microbiota, with significant decreases in the classes of Actinobacteria, Bacteroidia, and significant increases for Alphaproteobacteria and Gammaproteobacteria. Furthermore, an increase in the number of genera containing potential pathogenic species for bivalves, such as Francisella and Vibrio, was detected. This suggests that accumulated microplastics pose a dual threat to filter-feeding organisms and the ecosystem services they provide. Further comparative studies are necessary to establish whether the microbiota shift is linked to the specific chemical composition of microplastics or whether there is an indirect link such as physiological stress resulting from ingestion.

PMID:40032227 | DOI:10.1016/j.envpol.2025.125914

Bone. 2025 Mar 1:117444. doi: 10.1016/j.bone.2025.117444. Online ahead of print.

ABSTRACT

Fracture healing is a complex biological process involving orchestrated interactions among cells, growth factors, and transcriptional pathways. Despite significant advancements in understanding bone repair, non-union and delayed healing remain prevalent, especially in patients with comorbidities such as aging, diabetes, or substance use. Murine models serve as a cornerstone in fracture healing research, offering genetic manipulability, cost-effectiveness, and biological relevance to humans. This scoping review consolidates findings from studies conducted between 2010 and 2024, focusing on mechanistic pathways derived from transcriptional analysis in secondary bone healing as identified through bulk RNA sequencing of murine models. Key mechanistic pathways were categorized and analyzed across the distinct phases of fracture healing-reactive, reparative, and remodeling-highlighting their unique roles in inflammation, ECM remodeling, cell proliferation, and tissue mineralization. The most recurrent mechanistic pathways included ECM-receptor interaction, focal adhesion, and signaling mechanisms such as MAPK and TGF-beta. Variability in methodologies and limited overlap among studies underscore the need for standardized protocols in RNA sequencing analysis. Additionally, comparisons across long bone fractures, hole defects, and craniofacial bone healing revealed shared molecular mechanisms alongside unique challenges, particularly in craniofacial models. This scoping review underscores the promise of integrating systems biology approaches with transcriptomic data to elucidate the intricate regulatory networks governing fracture repair. Addressing the identified gaps in early-phase healing and harmonizing research methodologies will advance therapeutic strategies to reduce non-union rates and improve clinical outcomes.

PMID:40032014 | DOI:10.1016/j.bone.2025.117444

Sci Rep. 2025 Mar 3;15(1):7411. doi: 10.1038/s41598-025-91753-y.

ABSTRACT

Individuals with latent tuberculosis infection (LTBI) are at risk of progressing to active tuberculosis (TB), which remains a significant cause of death globally. Although various antiTB medications-rifampin and isoniazid-exist for treating for both LTBI and active TB, pharmacovigilance studies evaluating their adverse effects are especially scare for LTBI. Given the continued status of South Korea as having the highest TB incidence among Organization for Economic Cooperation and Development countries, this study examines drug-related adverse events (AEs) and identifies novel signals associated with rifampin or isoniazid in TB prevention and treatment in South Korea using the national AE reporting system. Analyzing data from the Korea Institute of Drug Safety and Risk Management-Korea Adverse Event Reporting System Database (KIDS-KAERS DB, 2301A0006) between 2017 and 2021, we observed that rifampin was frequently listed as a suspected drug in AE reports. Serious adverse events (SAEs), including life-threatening events and hospitalizations, were observed in LTBI as well as active TB cases when rifampin was the suspected drug. Novel signals, including QT prolongation and acne, were also identified, underscoring the importance of AE monitoring in LTBI or active TB treatment.

PMID:40032948 | DOI:10.1038/s41598-025-91753-y

BMJ Open Qual. 2025 Mar 3;14(1):e003022. doi: 10.1136/bmjoq-2024-003022.

ABSTRACT

BACKGROUND: Access to reliable data about patient's medications before surgery represents a challenge for reducing the risk of postoperative adverse events (AE) potentially related to preoperative treatment.

OBJECTIVE: To evaluate the impact on AE of a nationwide ambulatory electronic pharmaceutical record (EPR) used by a pharmacist for best possible medication history (BPMH), associated with the preoperative evaluation.

METHODS: This quasi-experimental comparative interventional study included 750 adult patients with an available EPR, admitted to the preoperative clinic for elective orthopaedic surgery, between April 2014 and April 2017. Data analysis was completed in September 2022. In the intervention group, a pharmacist performed the BPMH using the EPR, before the patient's medical evaluation. In the control group, there was conventional preoperative evaluation. The primary outcome was the number of patients with at least one AE collected by using the trigger tool method, within 30 days after surgery. Secondary outcomes were the number of medications reported in the medical record and the number of patients with at least one documented adverse drug event (ADE) by an independent committee within 30 days after surgery.

RESULTS: Of 1924 patients admitted to the preoperative clinic, 750 patients who had a record (39%) were included (153 (41%) men; median age 61 (49-71 and 50-70) years in both groups), 375 in each group. There was a 29% reduction in the proportion of patients with at least one AE in the intervention group (110/374 patients (29%) with 165 AE vs 156/372 patients (42%) with 233 AE) (OR 0.58 (0.43-0.78), p<0.01). There were significantly more drugs reported on the medical record in the intervention group (3 (1-5) vs 2 (1-4), p<0.01). There was no significant difference between the two groups in the number of patients with ADE (71/374 patients (19%) with 96 ADE vs 80/372 patients (22%) with 108 ADE, p=0.44).

CONCLUSIONS AND RELEVANCE: A BPMH performed by a pharmacist using a nationwide EPR at the time of preoperative evaluation contributed to reducing AE, potentially preventing harm to patients.

TRIAL REGISTRATION NUMBER: NCT02071472.

PMID:40032596 | DOI:10.1136/bmjoq-2024-003022

Drug Ther Bull. 2025 Mar 3;63(3):34. doi: 10.1136/dtb.2025.000007.

NO ABSTRACT

PMID:40032363 | DOI:10.1136/dtb.2025.000007

IEEE J Biomed Health Inform. 2024 Nov 13;PP. doi: 10.1109/JBHI.2024.3497591. Online ahead of print.

ABSTRACT

Prediction of drug-target interactions (DTIs) is one of the crucial steps for drug repositioning. Identifying DTIs through bio-experimental manners is always expensive and time-consuming. Recently, deep learning-based approaches have shown promising advancements in DTI prediction, but they face two notable challenges: (i) how to explicitly capture local interactions between drug-target pairs and learn their higher-order substructure embeddings; (ii) How to filter out redundant information to obtain effective embeddings for drugs and targets. Results: In this study, we propose a novel approach, termed DSANIB, to infer potential interactions between drugs and targets. DSANIB comprises two primary components: (1) DSAN component: The Inter-view Attention Network Module explicitly learns the local interactions between drugs and targets, while the Intra-view Attention Network Module aggregates information from local interaction features to obtain their higher-order substructure embeddings. (2) Information Bottleneck (IB) component: DSANIB adopts the IB strategy, which could retain relevant information while minimizing the redundant features to obtain their discriminative representations. Extensive experimental results demonstrate that DSANIB outperforms other SOTA prediction models. In addition, visualization of drug and target embeddings learned through DSANIB could provide interpretable insights for the prediction results. Availability: The source code has been made publicly available on GitHub https://github.com/Zzz-Soar/DSANIB.

PMID:40030194 | DOI:10.1109/JBHI.2024.3497591

Proc Natl Acad Sci U S A. 2025 Mar 11;122(10):e2419992122. doi: 10.1073/pnas.2419992122. Epub 2025 Mar 3.

ABSTRACT

Variants with large effect contribute to congenital heart disease (CHD). To date, recessive genotypes (RGs) have commonly been implicated through anecdotal ascertainment of consanguineous families and candidate gene-based analysis; the recessive contribution to the broad range of CHD phenotypes has been limited. We analyzed whole exome sequences of 5,424 CHD probands. Rare damaging RGs were estimated to contribute to at least 2.2% of CHD, with greater enrichment among laterality phenotypes (5.4%) versus other subsets (1.4%). Among 108 curated human recessive CHD genes, there were 66 RGs, with 54 in 11 genes with >1 RG, 12 genes with 1 RG, and 85 genes with zero. RGs were more prevalent among offspring of consanguineous union (4.7%, 32/675) than among nonconsanguineous probands (0.7%, 34/4749). Founder variants in GDF1 and PLD1 accounted for 74% of the contribution of RGs among 410 Ashkenazi Jewish probands. We identified genome-wide significant enrichment of RGs in C1orf127, encoding a likely secreted protein expressed in embryonic mouse notochord and associated with laterality defects. Single-cell transcriptomes from gastrulation-stage mouse embryos revealed enrichment of RGs in genes highly expressed in the cardiomyocyte lineage, including contractility-related genes MYH6, UNC45B, MYO18B, and MYBPC3 in probands with left-sided CHD, consistent with abnormal contractile function contributing to these malformations. Genes with significant RG burden account for 1.3% of probands, more than half the inferred total. These results reveal the recessive contribution to CHD, and indicate that many genes remain to be discovered, with each likely accounting for a very small fraction of the total.

PMID:40030011 | DOI:10.1073/pnas.2419992122

IEEE Trans Med Imaging. 2025 Jan 14;PP. doi: 10.1109/TMI.2025.3529621. Online ahead of print.

ABSTRACT

Near-infrared spectral tomography (NIRST) is a non-invasive imaging technique that provides functional information about biological tissues. Due to diffuse light propagation in tissue and limited boundary measurements, NIRST image reconstruction presents an ill-posed and ill-conditioned computational problem that is difficult to solve. To address this challenge, we developed a reconstruction algorithm (Model-CNN) that integrates a diffusion equation model with a convolutional neural network (CNN). The CNN learns a regularization prior to restrict solutions to the space of desirable chromophore concentration images. Efficacy of Model-CNN was evaluated by training on numerical simulation data, and then applying the network to physical phantom and clinical patient NIRST data. Results demonstrated the superiority of Model-CNN over the conventional Tikhonov regularization approach and a deep learning algorithm (FC-CNN) in terms of absolute bias error (ABE) and peak signal-to-noise ratio (PSNR). Specifically, in comparison to Tikhonov regularization, Model-CNN reduced average ABE by 55% for total hemoglobin (HbT) and 70% water (H2O) concentration, while improved PSNR by an average of 5.3 dB both for HbT and H2O images. Meanwhile, image processing time was reduced by 82%, relative to the Tikhonov regularization. As compared to FC-CNN, the Model-CNN achieved a 91% reduction in ABE for HbT and 75% for H2O images, with increases in PSNR by 7.3 dB and 4.7 dB, respectively. Notably, this Model-CNN approach was not trained on patient data; but instead, was trained on simulated phantom data with simpler geometrical shapes and optical source-detector configurations; yet, achieved superior image recovery when faced with real-world data.

PMID:40031020 | DOI:10.1109/TMI.2025.3529621

IEEE Trans Image Process. 2025 Jan 15;PP. doi: 10.1109/TIP.2025.3527886. Online ahead of print.

ABSTRACT

Denoising is one of the fundamental steps of the processing pipeline that converts data captured by a camera sensor into a display-ready image or video. It is generally performed early in the pipeline, usually before demosaicking, although studies swapping their order or even conducting them jointly have been proposed. With the advent of deep learning, the quality of denoising algorithms has steadily increased. Even so, modern neural networks still have a hard time adapting to new noise levels and scenes, which is indispensable for real-world applications. With those in mind, we propose a self-similarity-based denoising scheme that weights both a pre- and a post-demosaicking denoiser for Bayer-patterned CFA video data. We show that a balance between the two leads to better image quality, and we empirically find that higher noise levels benefit from a higher influence pre-demosaicking. We also integrate temporal trajectory prefiltering steps before each denoiser, which further improve texture reconstruction. The proposed method only requires an estimation of the noise model at the sensor, accurately adapts to any noise level, and is competitive with the state of the art, making it suitable for real-world videography.

PMID:40031011 | DOI:10.1109/TIP.2025.3527886

IEEE Trans Pattern Anal Mach Intell. 2025 Jan 13;PP. doi: 10.1109/TPAMI.2025.3528449. Online ahead of print.

ABSTRACT

Geometric deep learning (GDL) models have demonstrated a great potential for the analysis of non-Euclidian data. They are developed to incorporate the geometric and topological information of non-Euclidian data into the end-to-end deep learning architectures. Motivated by the recent success of discrete Ricci curvature in graph neural network (GNNs), we propose TorGNN, an analytic Torsion enhanced Graph Neural Network model. The essential idea is to characterize graph local structures with an analytic torsion based weight formula. Mathematically, analytic torsion is a topological invariant that can distinguish spaces which are homotopy equivalent but not homeomorphic. In our TorGNN, for each edge, a corresponding local simplicial complex is identified, then the analytic torsion (for this local simplicial complex) is calculated, and further used as a weight (for this edge) in message-passing process. Our TorGNN model is validated on link prediction tasks from sixteen different types of networks and node classification tasks from four types of networks. It has been found that our TorGNN can achieve superior performance on both tasks, and outperform various state-of-the-art models. This demonstrates that analytic torsion is a highly efficient topological invariant in the characterization of graph structures and can significantly boost the performance of GNNs.

PMID:40030998 | DOI:10.1109/TPAMI.2025.3528449