Int J Comput Assist Radiol Surg. 2025 Aug 1. doi: 10.1007/s11548-025-03483-1. Online ahead of print.

ABSTRACT

PURPOSE: In order to automate the centerline extraction of the coronary tree, three challenges must be addressed: tracking branches automatically, passing through plaques successfully, and detecting endpoints accurately. This study aims to develop a method to solve the three challenges.

METHODS: We propose a branch-endpoint-aware coronary centerline extraction framework. The framework consists of a deep reinforcement learning-based tracker and a 3D dilated CNN-based detector. The tracker is designed to predict the actions of an agent with the objective of tracking the centerline. The detector identifies bifurcation points and endpoints, assisting the tracker in tracking branches and terminating the tracking process automatically. The detector can also estimate the radius values of the coronary artery.

RESULTS: The method achieves the state-of-the-art performance in both the centerline extraction and radius estimate. Furthermore, the method necessitates minimal user interaction to extract a coronary tree, a feature that surpasses other interactive methods.

CONCLUSION: The method can track branches automatically, pass through plaques successfully and detect endpoints accurately. Compared with other interactive methods that require multiple seeds, our method only needs one seed to extract the entire coronary tree.

PMID:40751109 | DOI:10.1007/s11548-025-03483-1

Sci Rep. 2025 Aug 1;15(1):28135. doi: 10.1038/s41598-025-12685-1.

ABSTRACT

In drug discovery, identifying molecules with desired pharmacological properties remains challenging, as conventional methods often rely on exhaustive trial-and-error and limited exploration of chemical space. Here, we present STELLA, a metaheuristics-based generative molecular design framework that combines an evolutionary algorithm for fragment-based chemical space exploration with a clustering-based conformational space annealing method for efficient multi-parameter optimization. Additionally, it leverages deep learning models for accurate prediction of pharmacological properties. Our case study, which focuses on docking score and quantitative estimate of drug-likeness as primary objectives, demonstrates that STELLA generates 217% more hit candidates with 161% more unique scaffolds and achieves more advanced Pareto fronts compared to REINVENT 4. In performance evaluations optimizing 16 properties simultaneously for MolFinder, REINVENT 4, and STELLA, STELLA consistently outperforms the control methods by achieving better average objective scores and exploring a broader region of chemical space. The results highlight STELLA's superior performance in both efficient exploration of chemical space and multi-parameter optimization, indicating that STELLA is a powerful tool for de novo molecular design.

PMID:40750989 | DOI:10.1038/s41598-025-12685-1

Sci Rep. 2025 Aug 1;15(1):28124. doi: 10.1038/s41598-025-12058-8.

ABSTRACT

Soil fauna play a critical role in maintaining ecosystem functions and assessing environmental health, making accurate and efficient detection essential. Therefore, this paper proposes an improved algorithm based on You Only Look Once (YOLO) v9, which enhances feature capture capability while reducing parameters by 33.6%. First, a dynamic local shuffle module (DLSConv) is proposed, which utilizes convolutions and adaptive shuffling, effectively enhancing information interaction and feature richness. Additionally, different efficient modules with multi-branch fusion structures, integrating DLSConv, are adopted for the Backbone and Neck to enhance feature extraction and fusion, while optimizing the feature maps fed into the detection head, thereby improving the network's ability to extract features and detect targets. Ablation experiments demonstrate that the model achieves a 2.3% improvement in F-score and 1.8% increase in mean average precision (mAP)@50. On the soil fauna dataset, it attains 94.3% in mAP@75, significantly outperforming the baseline in challenging scenarios. These results highlight the model's efficiency and reliability for soil fauna detection on resource-constrained devices. And this capability can significantly enhance ecological monitoring through scalable biodiversity assessment and empowers precision agriculture applications via actionable insights into soil health and faunal activity, underpinning sustainable land management practices.

PMID:40750964 | DOI:10.1038/s41598-025-12058-8

Sci Rep. 2025 Aug 1;15(1):28068. doi: 10.1038/s41598-025-10531-y.

ABSTRACT

Time-lapse imaging and deep-learning algorithms are promising tools to assess the most viable embryos and improve embryo selection in IVF laboratories. Here, we developed and validated a deep learning model based on self-supervised contrastive learning. The model was developed with a new approach based on matched KID (Known Implantation Data) embryos derived from the same cohort of a stimulation cycle, both judged to be of good quality according to classical morphological criteria and morphokinetics, transferred fresh or frozen, but with a different implantation fate (clinical pregnancy vs. failure of implantation). We used self-supervised contrastive learning to train convolutional neural networks to ensure an unbiased and comprehensive learning of the morphokinetics features of the embryos, followed by a Siamese neural network fine-tuning and an XGBoost final prediction model to prevent overfitting. 1580 embryo videos of 460 patients were included between January 2020 and February 2023. With the knowledge of the implantation outcome of a previous transfer of an embryo derived from the same stimulation cycle, this model could predict the pregnancy outcome of the subsequent transfer with an AUC of 0.57. Without any knowledge of transfer history, the model achieved a satisfactory performance in predicting implantation (AUC = 0.64). This model could be considered as an adjunct tool for biologists to better select embryos and reduce the number of useless transfers per patient, when a cohort with several embryos classified as good quality by classical criteria is obtained.

PMID:40750959 | DOI:10.1038/s41598-025-10531-y

Insights Imaging. 2025 Aug 1;16(1):165. doi: 10.1186/s13244-025-02045-y.

ABSTRACT

OBJECTIVES: Subvariants of testicular germ cell tumor (TGCT) significantly affect therapeutic strategies and patient prognosis. However, preoperatively distinguishing seminoma (SE) from non-seminoma (n-SE) remains a challenge. This study aimed to evaluate the performance of a deep learning-based super-resolution (SR) US radiomics model for SE/n-SE differentiation.

MATERIALS AND METHODS: This international multicenter retrospective study recruited patients with confirmed TGCT between 2015 and 2023. A pre-trained SR reconstruction algorithm was applied to enhance native resolution (NR) images. NR and SR radiomics models were constructed, and the superior model was then integrated with clinical features to construct clinical-radiomics models. Diagnostic performance was evaluated by ROC analysis (AUC) and compared with radiologists' assessments using the DeLong test.

RESULTS: A total of 486 male patients were enrolled for training (n = 338), domestic (n = 92), and international (n = 59) validation sets. The SR radiomics model achieved AUCs of 0.90, 0.82, and 0.91, respectively, in the training, domestic, and international validation sets, significantly surpassing the NR model (p < 0.001, p = 0.031, and p = 0.001, respectively). The clinical-radiomics model exhibited a significantly higher across both domestic and international validation sets compared to the SR radiomics model alone (0.95 vs 0.82, p = 0.004; 0.97 vs 0.91, p = 0.031). Moreover, the clinical-radiomics model surpassed the performance of experienced radiologists in both domestic (AUC, 0.95 vs 0.85, p = 0.012) and international (AUC, 0.97 vs 0.77, p < 0.001) validation cohorts.

CONCLUSIONS: The SR-based clinical-radiomics model can effectively differentiate between SE and n-SE.

CRITICAL RELEVANCE STATEMENT: This international multicenter study demonstrated that a radiomics model of deep learning-based SR reconstructed US images enabled effective differentiation between SE and n-SE.

KEY POINTS: Clinical parameters and radiologists' assessments exhibit limited diagnostic accuracy for SE/n-SE differentiation in TGCT. Based on scrotal US images of TGCT, the SR radiomics models performed better than the NR radiomics models. The SR-based clinical-radiomics model outperforms both the radiomics model and radiologists' assessment, enabling accurate, non-invasive preoperative differentiation between SE and n-SE.

PMID:40750949 | DOI:10.1186/s13244-025-02045-y

Commun Eng. 2025 Aug 1;4(1):140. doi: 10.1038/s44172-025-00477-4.

ABSTRACT

Adaptive modeling is imperative for analyzing nonlinear systems deployed in natural dynamic environments. It facilitates filtering, prediction, and automatic control of the target object in real time to respond to unpredictable and non-repetitive sudden physical impairment caused by ambient impacts, such as corrosion, thermal drift, interference, etc. Existing nonlinear modeling approaches, however, are too complex for online training or fall short in rapid model recalibration under such conditions. To address this challenge, here we present a strategy that applies a regulator to the Koopman operator, enabling real-time model adaptation for nonlinear systems. In our approach, the regulator is directly implemented in nonlinear state-space without disrupting the pre-trained black-box predictor. The proposed technique demonstrates efficacy in capturing a broad spectrum of nonlinear dynamics and exhibits rapid adaptability to system changes without requiring offline retraining. Furthermore, its lightweight implementation and high-speed performance make it well-suited for embedded systems and applications demanding fast model recalibration and robustness.

PMID:40750921 | DOI:10.1038/s44172-025-00477-4

Med Image Anal. 2025 Jul 26;105:103732. doi: 10.1016/j.media.2025.103732. Online ahead of print.

ABSTRACT

Fast radial-MRI approaches based on compressed sensing (CS) and deep learning (DL) often use non-uniform fast Fourier transform (NUFFT) as the forward imaging operator, which might introduce interpolation errors and reduce image quality. Using the polar Fourier transform (PFT), we developed fully polar CS and DL algorithms for fast 2D cardiac radial-MRI. Our methods directly reconstruct images in polar spatial space from polar k-space data, eliminating frequency interpolation and ensuring an easy-to-compute data consistency term for the DL framework via the variable splitting (VS) scheme. Furthermore, PFT reconstruction produces initial images with fewer artifacts in a reduced field of view, making it a better starting point for CS and DL algorithms, especially for dynamic imaging, where information from a small region of interest is critical, as opposed to NUFFT, which often results in global streaking artifacts. In the cardiac region, PFT-based CS technique outperformed NUFFT-based CS at acceleration rates of 5x (mean SSIM: 0.8831 vs. 0.8526), 10x (0.8195 vs. 0.7981), and 15x (0.7720 vs. 0.7503). Our PFT(VS)-DL technique outperformed the NUFFT(GD)-based DL method, which used unrolled gradient descent with the NUFFT as the forward imaging operator, with mean SSIM scores of 0.8914 versus 0.8617 at 10x and 0.8470 versus 0.8301 at 15x. Radiological assessments revealed that PFT(VS)-based DL scored 2.9±0.30 and 2.73±0.45 at 5x and 10x, whereas NUFFT(GD)-based DL scored 2.7±0.47 and 2.40±0.50, respectively. Our methods suggest a promising alternative to NUFFT-based fast radial-MRI for dynamic imaging, prioritizing reconstruction quality in a small region of interest over whole image quality.

PMID:40749276 | DOI:10.1016/j.media.2025.103732

Med Image Anal. 2025 Jul 29;105:103744. doi: 10.1016/j.media.2025.103744. Online ahead of print.

ABSTRACT

Deep learning-driven 3D medical image segmentation generally necessitates dense voxel-wise annotations, which are expensive and labor-intensive to acquire. Cross-annotation, which labels only a few orthogonal slices per scan, has recently emerged as a cost-effective alternative that better preserves the shape and precise boundaries of the 3D object than traditional weak labeling methods such as bounding boxes and scribbles. However, learning from such sparse labels, referred to as barely-supervised learning (BSL), remains challenging due to less fine-grained object perception, less compact class features and inferior generalizability. To tackle these challenges and foster collaboration between model training and human expertise, we propose a Multi-Faceted ConSistency learning (MF-ConS) framework with a Diversity and Uncertainty Sampling-based Active Learning (DUS-AL) strategy, specifically designed for the active BSL scenario. This framework combines a cross-annotation BSL strategy, where only three orthogonal slices are labeled per scan, with an AL paradigm guided by DUS to direct human-in-the-loop annotation toward the most informative volumes under a fixed budget. Built upon a teacher-student architecture, MF-ConS integrates three complementary consistency regularization modules: (i) neighbor-informed object prediction consistency for advancing fine-grained object perception by encouraging the student model to infer complete segmentation from masked inputs; (ii) prototype-driven consistency, which enhances intra-class compactness and discriminativeness by aligning latent feature and decision spaces using fused prototypes; and (iii) stability constraint that promotes model robustness against input perturbations. Extensive experiments on three benchmark datasets demonstrate that MF-ConS (DUS-AL) consistently outperforms state-of-the-art methods under extremely limited annotation.

PMID:40749274 | DOI:10.1016/j.media.2025.103744

Eur J Med Chem. 2025 Jul 25;298:118002. doi: 10.1016/j.ejmech.2025.118002. Online ahead of print.

ABSTRACT

The orthopoxvirus genus, particularly the monkeypox virus (MPXV), continues to pose a significant global public health threat. Therefore, the development of novel anti-orthopoxvirus agents remains an urgent priority. Machine learning has proven to be an effective approach for identifying potential drug candidates. In this study, we implemented a dual-view deep learning model that combines BERT and a graph neural network to analyze molecular sequences and structural graphs. The model was trained following a pre-training-then-fine-tuning paradigm and was subsequently applied to identify new molecules with potential anti-orthopoxvirus activity. Notably, a cinnamoyl anthranilic acid derivative (compound 6) was successfully predicted and demonstrated potent anti-orthopoxvirus effects both in vitro and in vivo. Furthermore, integrin subunit beta 3 (ITGB3) has been validated as one of the direct target protein of 6. In conclusion, we established a robust dual-view deep learning model for the discovery of novel anti-orthopoxvirus agents, and compound 6 is a promising candidate for orthopoxvirus treatment via ITGB3 targeting.

PMID:40749255 | DOI:10.1016/j.ejmech.2025.118002

PLoS One. 2025 Aug 1;20(8):e0327743. doi: 10.1371/journal.pone.0327743. eCollection 2025.

ABSTRACT

INTRODUCTION: Retinal diseases, a significant global health concern, often lead to severe vision impairment and blindness, resulting in substantial functional and social limitations. This study explored a novel goal of developing and comparing the performance of multiple state-of-the-art convolutional neural network (CNN) models for the automated detection and classification of retinal diseases using optical coherence tomography (OCT) images.

METHOD: The study utilized several models, including DenseNet121, ResNet50, Inception V3, MobileNet, and OCT images obtained from the WATBORG Eye Clinic, to detect and classify multiple retinal diseases such as glaucoma, macular edema, posterior vitreous detachment (PVD), and normal eye cases. The preprocessing techniques employed included data augmentation, resizing, and one-hot encoding. We also used the Gaussian Process-based Bayesian Optimization (GPBBO) approach to fine-tune the hyperparameters. Model performance was evaluated using the F1-Score, precision, recall, and area under the curve (AUC).

RESULT: All the CNN models evaluated in this study demonstrated a strong capability to detect and classify various retinal diseases with high accuracy. MobileNet achieved the highest accuracy at 96% and AUC of 0.975, closely followed by DenseNet121, which had 95% accuracy and an AUC of 0.963. Inception V3 and ResNet50, while not as high in accuracy, showed potential in specific contexts, with 83% and 79% accuracy, respectively.

CONCLUSION: These results underscore the potential of advanced CNN models for diagnosing retinal diseases. With the exception of ResNet50, the other CNN models displayed accuracy levels that are comparable to other state-of-the-art deep learning models. Notably, MobileNet and DenseNet121 showed considerable promise for use in clinical settings, enabling healthcare practitioners to make rapid and accurate diagnoses of retinal diseases. Future research should focus on expanding datasets, integrating multi-modal data, exploring hybrid models, and validating these models in clinical environments to further enhance their performance and real-world applicability.

PMID:40748964 | DOI:10.1371/journal.pone.0327743

ACS Synth Biol. 2025 Aug 1. doi: 10.1021/acssynbio.5c00250. Online ahead of print.

ABSTRACT

The 5'UTR sequence and N-terminal coding sequence (NCS) have been used to regulate gene expression in Corynebacterium glutamicum (C. glutamicum) microbial cell factories. However, there is currently insufficient research on the relationship between these expression element sequences and the protein expression rate in C. glutamicum. This study established a pattern between 5'UTR and NCS feature sequences and protein expression and validated their effects on protein expression. First, a 5'UTR library and a NCS library containing base N were constructed separately, and a continuous regulatory range across 5 orders of magnitude for the enhanced green fluorescent protein (eGFP) expression was achieved in both libraries by fluorescence activated cell sorting (FACS) and high-throughput sequencing. Next, the relationship between sequence information and protein expression was established based on the 5'UTR sequence and NCS sequence characteristics analysis in terms of CG content, minimum free energy (MFE), tRNA adaptability index, and deep learning. Moreover, four 5'UTR characteristic sequences and four NCS characteristic sequences were finally screened, which showed strong compatibility with different exogenous proteins. Furthermore, dynamic adjustment of eGFP fluorescence intensity from 45% to 511% was achieved through 16 different combinations of the screened four 5'UTR and four NCS sequences, confirming the synergistic effect of these two components. At the same time, these combinations also have a wide range of dynamic regulation of protein expression levels of other recombinant proteins such as mCherry and heavy chain antibody. This study provided a potential tool for finely regulating gene expression or protein production in C. glutamicum.

PMID:40748894 | DOI:10.1021/acssynbio.5c00250

PLoS One. 2025 Aug 1;20(8):e0328013. doi: 10.1371/journal.pone.0328013. eCollection 2025.

ABSTRACT

Financial data prediction and risk assessment represent a complex multi-task problem that requires effective handling of time-series data and multi-dimensional features. Traditional models struggle to simultaneously capture temporal dependencies, global information, and intricate nonlinear relationships, resulting in limited prediction accuracy. To address this challenge, we propose LTR-Net, a multi-module deep learning model that combines LSTM, Transformer, and ResNet. LTR-Net effectively processes the multi-dimensional features and dynamic changes in financial data by incorporating a temporal dependency modeling module, a global information capture module, and a deep feature extraction module. Experimental results demonstrate that LTR-Net significantly outperforms existing mainstream models, including LSTM, GRU, Transformer, and DeepAR, across multiple financial datasets. On the Kaggle Financial Distress Prediction Dataset and the Yahoo Finance Stock Market Data, LTR-Net exhibits higher accuracy, stability, and robustness across various metrics such as MSE, RMSE, MAE, and AUC. Ablation experiments further validate the indispensability of each module within LTR-Net, confirming the pivotal roles of the LSTM, Transformer, and ResNet modules in financial data analysis. LTR-Net not only enhances the accuracy of financial data prediction but also exhibits strong generalization capabilities, making it adaptable to data analysis and risk assessment tasks in other domains.

PMID:40748872 | DOI:10.1371/journal.pone.0328013

IEEE Trans Neural Netw Learn Syst. 2025 Aug 1;PP. doi: 10.1109/TNNLS.2025.3592788. Online ahead of print.

ABSTRACT

Predicting remaining useful life (RUL) plays a crucial role in the prognostics and health management of industrial systems that involve a variety of interrelated sensors. Given a constant stream of time-series sensory data from such systems, deep learning (DL) models have risen to prominence at identifying complex, nonlinear temporal dependencies in these data. In addition to the temporal dependencies of individual sensors, spatial dependencies emerge as important correlations among these sensors, which can be naturally modeled by a temporal graph that describes time-varying spatial relationships. However, the majority of existing studies have relied on capturing discrete snapshots of this temporal graph, a coarse-grained approach that leads to a loss of temporal information. Moreover, given the variety of heterogeneous sensors, it becomes vital that such inherent heterogeneity is leveraged for RUL prediction in temporal sensor graphs. To capture the nuances of the temporal and spatial relationships and heterogeneous characteristics in an interconnected graph of sensors, we introduce a novel model named temporal and heterogeneous graph neural networks (THGNNs). Specifically, THGNN aggregates historical data from neighboring nodes to accurately capture the temporal dynamics and spatial correlations within the stream of sensor data in a fine-grained manner. Moreover, the model leverages feature-wise linear modulation (FiLM) to address the diversity of sensor types, significantly improving the model's capacity to learn the heterogeneity in the data sources. Finally, we have validated the effectiveness of our approach through comprehensive experiments. Our empirical findings demonstrate significant advancements on the N-CMAPSS dataset, achieving improvements of up to 19.2% and 31.6% in terms of two different evaluation metrics over state-of-the-art methods.

PMID:40748812 | DOI:10.1109/TNNLS.2025.3592788

IEEE J Biomed Health Inform. 2025 Aug 1;PP. doi: 10.1109/JBHI.2025.3594014. Online ahead of print.

ABSTRACT

Parkinson's Disease (PD) treatment is challenging due to symptom heterogeneity and the lack of a definitive cure. Lifelong medication requires personalized treatment plans developed by physicians, but such approaches are constrained by high costs and limited physician capacity. Although deep learning (DL) methods have been explored, they lack interpretability and are restricted to numerical data inputs. In this study, we propose a novel framework that leverages large language models (LLMs) to design personalized PD treatment strategies, integrating both patient information in natural language form and external textual knowledge sources (e.g., medical guidelines). To enhance effectiveness, we use Monte Carlo Tree Search (MCTS) to refine strategies and establish a robust medication recommendation dataset. To enhance reliability and interpretability, we incorporate Retrieval-Augmented Generation (RAG) and Chain-of-Thought (CoT) reasoning within the LLM system, ensuring that each proposed strategy is accompanied by step-by-step explanations and references to similar historical cases. Experimental evaluations using the Parkinson's Progression Marking Initiative (PPMI) dataset show that our method surpasses physician-prescribed treatments, achieving an average reduction of over 1.4 points in the revised unified Parkinson's disease rating scale part III (MDS-UPDRS-III) scores. Our method also outperforms the RL-method by 1.01 points on average. Furthermore, over 43% of patients achieve more than 2 point-reduction of MDS-UPDRS-III scores. A detailed case study highlights the flexibility of LLMs in dynamically adjusting medication plans for patients at different disease stages, highlighting its potential to advance personalized PD management in real-world settings.

PMID:40748804 | DOI:10.1109/JBHI.2025.3594014

IEEE J Biomed Health Inform. 2025 Aug 1;PP. doi: 10.1109/JBHI.2025.3594210. Online ahead of print.

ABSTRACT

Accurate prediction of protein-ligand affinity (PLA) is critical for drug discovery. Recent deep learning approaches have adopted data-driven models for PLA prediction by learning intrinsic patterns from one-dimensional (1D) sequential or two-dimensional (2D) graph representations of proteins and ligands. However, these low-dimensional methods overlook the three-dimensional (3D) geometric features, which are hypothesized to be critical in binding interaction. To address the above problem, we present a Geometric deep learning approach with Hybrid message passing strategies-HybridGeo, for protein-ligand affinity prediction. We adopt dual-view graph learning to model the intra- and inter-molecular atomic interactions and propose to aggregate the spatial information with hybrid strategies. In addition, to fully model the inter-residue dependency upon message aggregation, we adopt a geometric graph transformer on the residue-scale graph of protein pockets. Extensive experiments on the PDBbind dataset show that HybridGeo achieves state-of-the-art performance with a Root Mean Square Error (RMSE) of 1.172. HybridGeo also achieves the best among all baseline models on three external test sets, showcasing good generalizability and robustness. Through systematic ablation experiments, we validated the effectiveness of the proposed modules, and further demonstrated the superior performance of HybridGeo in predicting the binding affinity of macrocyclic compound complexes through case studies. Visualization analysis further indicates the biological interpretability of the model predictions. Our code is publicly available at https://github.com/anxiangbiye1231/HybridGeo.

PMID:40748800 | DOI:10.1109/JBHI.2025.3594210

IEEE J Biomed Health Inform. 2025 Aug 1;PP. doi: 10.1109/JBHI.2025.3593825. Online ahead of print.

ABSTRACT

Deep learning-based molecular generation models have shown great potential in efficiently exploring vast chemical spaces by generating potential drug candidates with desired properties. However, these models often produce chemically invalid molecules, which limits the usable scope of the learned chemical space and poses significant challenges for practical applications. To address this issue, we propose ChemFixer, a framework designed to correct invalid molecules into valid ones. Chem- Fixer is built on a transformer architecture, pre-trained using masking techniques, and fine-tuned on a large-scale dataset of valid/invalid molecular pairs that we constructed. Through comprehensive evaluations across diverse generative models, ChemFixer improved molecular validity while effectively preserving the chemical and biological distributional properties of the original outputs. This indicates that ChemFixer can recover molecules that could not be previously generated, thereby expanding the diversity of potential drug candidates. Furthermore, ChemFixer was effectively applied to a drug-target interaction (DTI) prediction task using limited data, improving the validity of generated ligands and discovering promising ligand-protein pairs. These results suggest that ChemFixer is not only effective in data-limited scenarios, but also extensible to a wide range of downstream tasks. Taken together, ChemFixer shows promise as a practical tool for various stages of deep learning-based drug discovery, enhancing molecular validity and expanding accessible chemical space.

PMID:40748798 | DOI:10.1109/JBHI.2025.3593825

Oral Radiol. 2025 Aug 1. doi: 10.1007/s11282-025-00848-9. Online ahead of print.

ABSTRACT

OBJECTIVE: The purpose of this study is the development of a deep learning model based on nnU-Net v2 for the automated segmentation of sphenoid sinus and middle skull base anatomic structures in cone-beam computed tomography (CBCT) volumes, followed by an evaluation of the model's performance.

MATERIAL AND METHODS: In this retrospective study, the sphenoid sinus and surrounding anatomical structures in 99 CBCT scans were annotated using web-based labeling software. Model training was conducted using the nnU-Net v2 deep learning model with a learning rate of 0.01 for 1000 epochs. The performance of the model in automatically segmenting these anatomical structures in CBCT scans was evaluated using a series of metrics, including accuracy, precision, recall, dice coefficient (DC), 95% Hausdorff distance (95% HD), intersection on union (IoU), and AUC.

RESULTS: The developed deep learning model demonstrated a high level of success in segmenting sphenoid sinus, foramen rotundum, and Vidian canal. Upon evaluation of the DC values, it was observed that the model demonstrated the highest degree of ability to segment the sphenoid sinus, with a DC value of 0.96.

CONCLUSION: The nnU-Net v2-based deep learning model achieved high segmentation performance for the sphenoid sinus, foramen rotundum, and Vidian canal within the middle skull base, with the highest DC observed for the sphenoid sinus (DC: 0.96). However, the model demonstrated limited performance in segmenting other foramina of the middle skull base, indicating the need for further optimization for these structures.

PMID:40748555 | DOI:10.1007/s11282-025-00848-9

Acta Biotheor. 2025 Aug 1;73(3):11. doi: 10.1007/s10441-025-09500-9.

ABSTRACT

While RNA folding prediction remains challenging, even with machine and deep learning methods, it can also be approached from a topological mathematics perspective. The purpose of the present paper is to elucidate this problem for students and researchers in both the mathematical physics and biology fields, fostering interest in developing novel theoretical and applied solutions that could propel RNA research forward. With this intention, the mathematical method, based on matrix field theory, to compute the topological classification of RNA structures is reviewed. Similarly, McGenus, a computational software that exploits matrix field theory for topological and folding predictions, is examined. To further illustrate the outcomes of this mathematical approach, two types of analyses are performed: the prediction results from McGenus are compared with topological information extracted from experimentally-determined RNA structures, and the topology of RNA structures is investigated for biological significance, both in evolutionary and functional terms. Lastly, we advocate for more research efforts to be conducted at the intersection between physics, mathematics and biology, with a particular focus on the potential contributions that topology can make to the study of RNA folding and structure.

PMID:40748481 | DOI:10.1007/s10441-025-09500-9

Abdom Radiol (NY). 2025 Aug 1. doi: 10.1007/s00261-025-05100-w. Online ahead of print.

ABSTRACT

PURPOSE: This study explored the effects of using Intelligent Quick Magnetic Resonance (IQMR) image post-processing on image quality in Field of View Optimized and Constrained Single-Shot Diffusion-Weighted Imaging (FOCUS-DWI) sequences for prostate cancer detection, and assessed its efficacy in distinguishing malignant from benign lesions.

METHODS: The clinical data and MRI images from 62 patients with prostate masses (31 benign and 31 malignant) were retrospectively analyzed. Axial T2-weighted imaging with fat saturation (T2WI-FS) and FOCUS-DWI sequences were acquired, and the FOCUS-DWI images were processed using the IQMR post-processing system to generate IQMR-FOCUS-DWI images. Two independent radiologists undertook subjective scoring, grading using the Prostate Imaging Reporting and Data System (PI-RADS), diagnosis of benign and malignant lesions, and diagnostic confidence scoring for images from the FOCUS-DWI and IQMR-FOCUS-DWI sequences. Additionally, quantitative analyses, specifically, the peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM), were conducted using T2WI-FS as the reference standard. The apparent diffusion coefficients (ADCs) of malignant and benign lesions were compared between the two imaging sequences. Spearman correlation coefficients were calculated to evaluate the associations between diagnostic confidence scores and diagnostic accuracy rates of the two sequence groups, as well as between the ADC values of malignant lesions and Gleason grading in the two sequence groups. Receiver operating characteristic (ROC) curves were utilized to assess the efficacy of ADC in distinguishing lesions.

RESULTS: The qualitative analysis revealed that IQMR-FOCUS-DWI images showed significantly better noise suppression, reduced geometric distortion, and enhanced overall quality relative to the FOCUS-DWI images (P < 0.001). There was no significant difference in the PI-RADS scores between IQMR-FOCUS-DWI and FOCUS-DWI images (P = 0.0875), while the diagnostic confidence scores of IQMR-FOCUS-DWI sequences were markedly higher than those of FOCUS-DWI sequences (P = 0.0002). The diagnostic results of the FOCUS-DWI sequences for benign and malignant prostate lesions were consistent with those of the pathological results (P < 0.05), as were those of the IQMR-FOCUS-DWI sequences (P < 0.05). The quantitative analysis indicated that the PSNR, SSIM, and ADC values were markedly greater in IQMR-FOCUS-DWI images relative to FOCUS-DWI images (P < 0.01). In both imaging sequences, benign lesions exhibited ADC values markedly greater than those of malignant lesions (P < 0.001). The diagnostic confidence scores of both groups of sequences were significantly positively correlated with the diagnostic accuracy rate. In malignant lesions, the ADC values of the FOCUS-DWI sequences showed moderate negative correlations with the Gleason grading, while the ADC values of the IQMR-FOCUS-DWI sequences were strongly negatively associated with the Gleason grading. ROC curves indicated the superior diagnostic performance of IQMR-FOCUS-DWI (AUC = 0.941) compared to FOCUS-DWI (AUC = 0.832) for differentiating prostate lesions (P = 0.0487).

CONCLUSION: IQMR-FOCUS-DWI significantly enhances image quality and improves diagnostic accuracy for benign and malignant prostate lesions compared to conventional FOCUS-DWI.

PMID:40748461 | DOI:10.1007/s00261-025-05100-w

Brief Bioinform. 2025 Jul 2;26(4):bbaf372. doi: 10.1093/bib/bbaf372.

ABSTRACT

Systematically investigating interactions among molecules of the same type across different contexts is crucial for unraveling disease mechanisms and developing potential therapeutic strategies. The "A-A-B" triplet paradigm provides a principled approach to model such context-specific interactions, and leveraging third-order tensor to capture such type ternary relationships is an efficient strategy. However, effectively modeling both multilinear and nonlinear characteristics to accurately identify such triplets using tensor-based methods remains a challenge. In this paper, we propose a novel Convolutional Neural Tensor Completion (ConvNTC) framework that collaboratively learns the multilinear and nonlinear representations to model triplet-based network interactions. ConvNTC consists of a multilinear module and a nonlinear module. The former is a tensor decomposition approach that integrates multiple constraints to learn the tensor factor embeddings. The latter contains three components: an embedding generator to produce position-specific index embeddings for each tensor entry in addition to the factor embeddings, a convolutional encoder to perform nonlinear feature mapping while preserving the tensor's rank-one property, and a Kolmogorov-Arnold Network (KAN) based predictor to effectively capture high-dimensional relationships aligned with the intrinsic structure of real-world data. We evaluate ConvNTC on two types triplet datasets of the "A-A-B" type: miRNA-miRNA-disease and drug-drug-cell. Comprehensive experiments against 11 state-of-the-art methods demonstrate the superiority of ConvNTC in terms of triplet prediction. ConvNTC reveals promising prognostic values of the miRNA-miRNA interactions on breast cancer and detects synergistic drug combinations in cancer cell lines.

PMID:40748325 | DOI:10.1093/bib/bbaf372