Sci Rep. 2025 Apr 22;15(1):13914. doi: 10.1038/s41598-025-93975-6.

ABSTRACT

Accurate anatomical measurements of the eyelids are essential in periorbital plastic surgery for both disease treatment and procedural planning. Recent researches in eye diseases have adopted deep learning works to measure MRD. However, such works encounter challenges in practical implementation, and the model accuracy needs to be improved. Here, we have introduced a deep learning-based adaptive and automatic measurement (DeepAAM) by employing the U-Net architecture enhanced through attention mechanisms and multiple algorithms. DeepAAM enables adaptive image recognition and adjustment in practical application, and improves the measurement accuracy of Marginal Reflex Distance (MRD). Meanwhile, it for the first time measures the Margin Iris Intersectant Angle (MIA) as an innovative evaluation index. Besides, this fully automated method surpasses other models in terms of accuracy for iris and sclera segmentation. DeepAAM offers a novel, comprehensive, and objective approach to the quantification of ocular morphology.

PMID:40263452 | DOI:10.1038/s41598-025-93975-6

J Xray Sci Technol. 2025 Apr 22:8953996251325092. doi: 10.1177/08953996251325092. Online ahead of print.

ABSTRACT

BackgroundPathological images play a crucial role in the diagnosis of critically ill cancer patients. Since cancer patients often seek medical assistance when their condition is severe, doctors face the urgent challenge of completing accurate diagnoses and developing surgical plans within a limited timeframe. The complexity and diversity of pathological images require a significant investment of time from specialized physicians for processing and analysis, which can lead to missing the optimal treatment window.PurposeCurrent medical decision support systems are challenged by the high computational complexity of deep learning models, which demand extensive data training, making it difficult to meet the real-time needs of emergency diagnostics.MethodThis study addresses the issue of emergency diagnosis for malignant bone tumors such as osteosarcoma by proposing a Lightened Boundary-enhanced Digital Pathological Image Recognition Strategy (LB-DPRS). This strategy optimizes the self-attention mechanism of the Transformer model and innovatively implements a boundary segmentation enhancement strategy, thereby improving the recognition accuracy of tissue backgrounds and nuclear boundaries. Additionally, this research introduces row-column attention methods to sparsify the attention matrix, reducing the computational burden of the model and enhancing recognition speed. Furthermore, the proposed complementary attention mechanism further assists convolutional layers in fully extracting detailed features from pathological images.ResultsThe DSC value of LB-DPRS strategy reached 0.862, the IOU value reached 0.749, and the params was only 10.97 M.ConclusionExperimental results demonstrate that the LB-DPRS strategy significantly improves computational efficiency while maintaining prediction accuracy and enhancing model interpretability, providing powerful and efficient support for the emergency diagnosis of malignant bone tumors such as osteosarcoma.

PMID:40262109 | DOI:10.1177/08953996251325092

J Chem Inf Model. 2025 Apr 22. doi: 10.1021/acs.jcim.5c00296. Online ahead of print.

ABSTRACT

In chemical reaction network theory, ordinary differential equations are used to model the temporal change of chemical species concentration. As the functional form of these ordinary differential equation systems is derived from an empirical model of the reaction network, it may be incomplete. Our approach aims to elucidate these hidden insights in the reaction network by combining dynamic modeling with deep learning in the form of neural ordinary differential equations. Our contributions not only help to identify the shortcomings of existing empirical models but also assist the design of future reaction networks.

PMID:40262040 | DOI:10.1021/acs.jcim.5c00296

Arch Insect Biochem Physiol. 2025 Aug;118(4):e70060. doi: 10.1002/arch.70060.

ABSTRACT

Goji berry is an important economic crop, yet pest infestations pose a significant threat to its yield and quality. Traditional pest identification mainly relies on manual inspection by experts with specialized knowledge, which is subjective, time-consuming, and labor-intensive. To address these issues, this experiment proposes an improved convolutional neural network (CNN) for accurate identification of 17 types of goji berry pests. Firstly, the original data set is augmented using a multi-graph-occlusion data augmentation method. Subsequently, the augmented data set is imported into the improved CNN for training. Based on the original ResNet18 model, a new CNN, named GojiNet, is constructed by embedding multi-attention fusion modules at appropriate locations. Experimental results demonstrate that GojiNet achieves an average recognition accuracy of 95.35%, representing a 2.60% improvement over the ResNet18 network. Notably, compared to the original network, the training time of this model increases only slightly, while its size is reduced, and the recognition accuracy is enhanced. The experiment verifies the performance of the GojiNet model through a series of evaluation indicators. This study confirms the tremendous potential and application prospects of deep learning in pest identification, providing a referential solution for intelligent and precise pest identification.

PMID:40262026 | DOI:10.1002/arch.70060

PLoS One. 2025 Apr 22;20(4):e0321849. doi: 10.1371/journal.pone.0321849. eCollection 2025.

ABSTRACT

Metallized Ceramic Ring is a novel electronic apparatus widely applied in communication, new energy, aerospace and other fields. Due to its complicated technique, there would be inevitably various defects on its surface; among which, the tiny pinhole defects with complex texture are the most difficult to detect, and there is no reliable method of automatic detection. This Paper proposes a method of detecting micro-pinhole defects on surface of metallized ceramic ring combining Improved Detection Transformer (DETR) Network with morphological operations, utilizing two modules, namely, deep learning-based and morphology-based pinhole defect detection to detect the pinholes, and finally combining the detection results of such two modules, so as to obtain a more accurate result. In order to improve the detection performance of DETR Network in aforesaid module of deep learning, EfficientNet-B2 is used to improve ResNet-50 of standard DETR network, the parameter-free attention mechanism (SimAM) 3-D weight attention mechanism is used to improve Sequeeze-and-Excitation (SE) attention mechanism in EfficientNet-B2 network, and linear combination loss function of Smooth L1 and Complete Intersection over Union (CIoU) is used to improve regressive loss function of training network. The experiment indicates that the recall and the precision of the proposed method are 83.5% and 86.0% respectively, much better than current mainstream methods of micro defect detection, meeting requirements of detection at industrial site.

PMID:40261923 | DOI:10.1371/journal.pone.0321849

PLoS One. 2025 Apr 22;20(4):e0321697. doi: 10.1371/journal.pone.0321697. eCollection 2025.

ABSTRACT

Autism Spectrum Disorder (ASD) is a neurodevelopmental disorder leading to an inability to socially communicate and in extreme cases individuals are completely dependent on caregivers. ASD detection at early ages is crucial as early detection can reduce the effect on social impairment. Deep learning models have shown capability to detect ASD earlier compared to traditional detection methods used by clinics and experts. Ensemble models, renowned for their ability to enhance predictive performance by combining multiple models, have emerged as a powerful tool in machine learning. This study harnesses the strength of ensemble learning to address the critical challenge of ASD diagnosis. This study proposed a deep ensemble model leveraging the strengths of VGG16 and Xception net trained on Facial Images for ASD detection overcoming limitations in existing datasets through extensive preprocessing. Proposed model preprocessed the training dataset of facial images by converting side posed images into frontal face images, using Histogram Equalization (HE) to enhance colors, data augmentation techniques application, and using the Hue Saturation Value (HSV) color model. By integrating the feature extraction strengths of VGG16 and Xception with fully connected layers, our model has achieved a notable 97% accuracy on the Kaggle ASD Face Image Dataset. This approach supports early detection of ASD and aligns with Sustainable Development Goal 3, which focuses on improving health and well-being.

PMID:40261913 | DOI:10.1371/journal.pone.0321697

IEEE J Biomed Health Inform. 2025 Apr 22;PP. doi: 10.1109/JBHI.2025.3563433. Online ahead of print.

ABSTRACT

Synergistic drug combinations have shown promising results in treating cancer cell lines by enhancing therapeutic efficacy and minimizing adverse reactions. The effects of a drug vary across cell lines, and cell lines respond differently to various drugs during treatment. Recently, many AI-based techniques have been developed for predicting synergistic drug combinations. However, existing computational models have not addressed this phenomenon, neglecting the refinement of features for the same drug and cell line in different scenarios. In this work, we propose a feature refinement deep learning framework, termed FRSynergy, to identify synergistic drug combinations. It can guide the refinement of drug and cell line features in different scenarios by capturing relationships among diverse drug-drug-cell line triplet features and learning feature contextual information. The heterogeneous graph attention network is employed to acquire topological information-based original features for drugs and cell lines from sampled sub-graphs. Then, the feature refinement network is designed by combining attention mechanism and context information, which can learn context-aware feature representations for each drug and cell line feature in diverse drug-drug-cell line triplet contexts. Extensive experiments affirm the strong performance of FRSynergy in predicting synergistic drug combinations and, more importantly, demonstrate the effectiveness of feature refinement network in synergistic drug combination prediction.

PMID:40261768 | DOI:10.1109/JBHI.2025.3563433

IEEE Trans Ultrason Ferroelectr Freq Control. 2025 Apr 22;PP. doi: 10.1109/TUFFC.2025.3562313. Online ahead of print.

ABSTRACT

Surgical tool tip localization and tracking are essential components of surgical and interventional procedures. The cross sections of tool tips can be considered as acoustic point sources to achieve these tasks with deep learning applied to photoacoustic channel data. However, source localization was previously limited to the lateral and axial dimensions of an ultrasound transducer. In this paper, we developed a novel deep learning-based three-dimensional (3D) photoacoustic point source localization system using an object detection-based approach extended from our previous work. In addition, we derived theoretical relationships among point source locations, sound speeds, and waveform shapes in raw photoacoustic channel data frames. We then used this theory to develop a novel deep learning instance segmentation-based 3D point source localization system. When tested with 4,000 simulated, 993 phantom, and 1,983 ex vivo channel data frames, the two systems achieved F1 scores as high as 99.82%, 93.05%, and 98.20%, respectively, and Euclidean localization errors (mean ± one standard deviation) as low as 1.46±1.11 mm, 1.58±1.30 mm, and 1.55±0.86 mm, respectively. In addition, the instance segmentation-based system simultaneously estimated sound speeds with absolute errors (mean ± one standard deviation) of 19.22±26.26 m/s in simulated data and standard deviations ranging 14.6-32.3 m/s in experimental data. These results demonstrate the potential of the proposed photoacoustic imaging-based methods to localize and track tool tips in three dimensions during surgical and interventional procedures.

PMID:40261767 | DOI:10.1109/TUFFC.2025.3562313

Database (Oxford). 2025 Apr 22;2025:baaf031. doi: 10.1093/database/baaf031.

ABSTRACT

As the importance of precision agriculture grows, scalable and efficient methods for real-time data collection and analysis have become essential. In this study, we developed a system to collect real-time crop images, focusing on physiological disorders in tomatoes. This system systematically collects crop images and related data, with the potential to evolve into a valuable tool for researchers and agricultural practitioners. A total of 58 479 images were produced under stress conditions, including bacterial wilt (BW), Tomato Yellow Leaf Curl Virus (TYLCV), Tomato Spotted Wilt Virus (TSWV), drought, and salinity, across seven tomato varieties. The images include front views at 0 degrees, 120 degrees, 240 degrees, and top views and petiole images. Of these, 43 894 images were suitable for labeling. Based on this, 24 000 images were used for AI model training, and 13 037 images for model testing. By training a deep learning model, we achieved a mean Average Precision (mAP) of 0.46 and a recall rate of 0.60. Additionally, we discussed data augmentation and hyperparameter tuning strategies to improve AI model performance and explored the potential for generalizing the system across various agricultural environments. The database constructed in this study will serve as a crucial resource for the future development of agricultural AI. Database URL: https://crops.phyzen.com/.

PMID:40261733 | DOI:10.1093/database/baaf031

IET Syst Biol. 2025 Apr 22:e12107. doi: 10.1049/syb2.12107. Online ahead of print.

ABSTRACT

Single-cell sequencing (scRNA-seq) allows researchers to study cellular heterogeneity in individual cells. In single-cell transcriptomics analysis, identifying the cell type of individual cells is a key task. At present, single-cell datasets often face the challenges of high dimensionality, large number of samples, high sparsity and sample imbalance. The traditional methods of cell type recognition have been challenged. The authors propose a deep residual generation model based on semi-supervised learning (scRSSL) to address these challenges. ScRSSL creatively introduces residual networks into semi-supervised generative models. The authors take advantage of its semi-supervised learning to solve the problem of sample imbalance. During the training of the model, the authors use a residual neural network to accomplish the inference of cell types so that local features of single-cell data can be extracted. Because of the semi-supervised learning approach, it can automatically and accurately predict individual cell types in datasets, even with only a small number of cell labels. Experimentally, the authors' method has proven to have better performance compared to other methods.

PMID:40261690 | DOI:10.1049/syb2.12107

Radiol Phys Technol. 2025 Apr 22. doi: 10.1007/s12194-025-00905-2. Online ahead of print.

ABSTRACT

This study aimed to develop and evaluate a deep-learning model for attenuation and scatter correction in whole-body 64Cu-based PET imaging. A swinUNETR model was implemented using the MONAI framework. Whole-body PET-nonAC and PET-CTAC image pairs were used for training, where PET-nonAC served as the input and PET-CTAC as the output. Due to the limited number of Cu-based PET/CT images, a model pre-trained on 51 Ga-PSMA PET images was fine-tuned on 15 Cu-based PET images via transfer learning. The model was trained without freezing layers, adapting learned features to the Cu-based dataset. For testing, six additional Cu-based PET images were used, representing 1-h, 12-h, and 48-h time points, with two images per group. The model performed best at the 12-h time point, with an MSE of 0.002 ± 0.0004 SUV2, PSNR of 43.14 ± 0.08 dB, and SSIM of 0.981 ± 0.002. At 48 h, accuracy slightly decreased (MSE = 0.036 ± 0.034 SUV2), but image quality remained high (PSNR = 44.49 ± 1.09 dB, SSIM = 0.981 ± 0.006). At 1 h, the model also showed strong results (MSE = 0.024 ± 0.002 SUV2, PSNR = 45.89 ± 5.23 dB, SSIM = 0.984 ± 0.005), demonstrating consistency across time points. Despite the limited size of the training dataset, the use of fine-tuning from a previously pre-trained model yielded acceptable performance. The results demonstrate that the proposed deep learning model can effectively generate PET-DLAC images that closely resemble PET-CTAC images, with only minor errors.

PMID:40261572 | DOI:10.1007/s12194-025-00905-2

Macromol Biosci. 2025 Apr 22:e2500007. doi: 10.1002/mabi.202500007. Online ahead of print.

ABSTRACT

Computer-aided protein design (CAPD) has become a transformative field, harnessing advances in computational power and deep learning to deepen the understanding of protein structure, function, and design. This review provides a comprehensive overview of CAPD techniques, with a focus on their application to protein-based therapeutics such as monoclonal antibodies, protein drugs, antigens, and protein polymers. This review starts with key CAPD methods, particularly those integrating deep learning-based predictions and generative models. These approaches have significantly enhanced protein drug properties, including binding affinity, specificity, and the reduction of immunogenicity. This review also covers CAPD's role in optimizing vaccine antigen design, improving protein stability, and customizing protein polymers for drug delivery applications. Despite considerable progress, CAPD faces challenges such as model overfitting, limited data for rare protein families, and the need for efficient experimental validation. Nevertheless, ongoing advancements in computational methods, coupled with interdisciplinary collaborations, are poised to overcome these obstacles, advancing protein engineering and therapeutic development. In conclusion, this review highlights the future potential of CAPD to transform drug development, personalized medicine, and biotechnology.

PMID:40260555 | DOI:10.1002/mabi.202500007

IEEE Access. 2025;13:57780-57797. doi: 10.1109/access.2025.3556713. Epub 2025 Apr 1.

ABSTRACT

Glioma is the most common malignant tumor of the central nervous system, and diffuse Glioma is classified as grades II-IV by world health organization (WHO). In the the cancer genome atlas (TCGA) Glioma dataset, grade II and III Gliomas are classified as low-grade glioma (LGG), and grade IV Gliomas as glioblastoma multiforme (GBM). In clinical practice, the survival and treatment process with Glioma patients depends on properly diagnosing the subtype. With this background, there has been much research on Glioma over the years. Among these researches, the origin and evolution of whole slide images (WSIs) have led to many attempts to support diagnosis by image analysis. On the other hand, due to the disease complexities of Glioma patients, multimodal analysis using various types of data rather than a single data set has been attracting attention. In our proposed method, multiple deep learning models are used to extract features from histopathology images, and the features of the obtained images are concatenated with those of the clinical data in a fusion approach. Then, we perform patch-level classification by machine learning (ML) using the concatenated features. Based on the performances of the deep learning models, we ensemble feature sets from top three models and perform further classifications. In the experiments with our proposed ensemble fusion AI (EFAI) approach using WSIs and clinical data of diffuse Glioma patients on TCGA dataset, the classification accuracy of the proposed multimodal ensemble fusion method is 0.936 with an area under the curve (AUC) value of 0.967 when tested on a balanced dataset of 240 GBM, 240 LGG patients. On an imbalanced dataset of 141 GBM, 242 LGG patients the proposed method obtained the accuracy of 0.936 and AUC of 0.967. Our proposed ensemble fusion approach significantly outperforms the classification using only histopathology images alone with deep learning models. Therefore, our approach can be used to support the diagnosis of Glioma patients and can lead to better diagnosis.

PMID:40260100 | PMC:PMC12011355 | DOI:10.1109/access.2025.3556713

Front Cell Neurosci. 2025 Apr 7;19:1523149. doi: 10.3389/fncel.2025.1523149. eCollection 2025.

ABSTRACT

More than 5% of the global population suffers from disabling hearing loss, primarily sensorineural hearing loss (SNHL). SNHL is often caused by factors such as vascular disorders, viral infections, ototoxic drugs, systemic inflammation, age-related labyrinthine membrane degeneration, and noise-induced hearing loss (NIHL). NIHL, in particular, leads to changes in blood-labyrinth-barrier (BLB) physiology, increased permeability, and various health issues, including cardiovascular disease, hypertension, diabetes, neurological disorders, and adverse reproductive outcomes. Recent advances in neuromodulation and vector-based approaches offer hope for overcoming biological barriers such as the BLB in the development of innovative treatments. Computational methods, including molecular docking, molecular dynamics simulations, QSAR/QSPR analysis with machine/deep learning algorithms, and network pharmacology, hold potential for identifying drug candidates and optimizing their interactions with BLB transporters, such as the glutamate transporter. This paper provides an overview of NIHL, focusing on its pathophysiology; its impact on membrane transporters, ion channels, and BLB structures; and associated symptoms, comorbidities, and emerging therapeutic approaches. Recent advancements in neuromodulation and vector-based strategies show great promise in overcoming biological barriers such as BLB, facilitating the development of innovative treatment options. The primary aim of this review is to examine NIHL in detail and explore its underlying mechanisms, physiological effects, and cutting-edge therapeutic strategies for its effective management and prevention.

PMID:40260077 | PMC:PMC12009814 | DOI:10.3389/fncel.2025.1523149

J Biomed Phys Eng. 2025 Apr 1;15(2):137-158. doi: 10.31661/jbpe.v0i0.2408-1811. eCollection 2025 Apr.

ABSTRACT

BACKGROUND: Diabetic retinopathy (DR), a diabetes complication, causes blindness by damaging retinal blood vessels. While deep learning has advanced DR diagnosis, many models face issues like inconsistent performance, limited datasets, and poor interpretability, reducing their clinical utility.

OBJECTIVE: This research aimed to develop and evaluate a deep learning structure combining Convolutional Neural Networks (CNNs) and transformer architecture to improve the accuracy, reliability, and generalizability of DR detection and severity classification.

MATERIAL AND METHODS: This computational experimental study leverages CNNs to extract local features and transformers to capture long-range dependencies in retinal images. The model classifies five types of retinal images and assesses four levels of DR severity. The training was conducted on the augmented APTOS 2019 dataset, addressing class imbalance through data augmentation techniques. Performance metrics, including accuracy, Area Under the Curve (AUC), specificity, and sensitivity, were used for metric evaluation. The model's robustness was further validated using the IDRiD dataset under diverse scenarios.

RESULTS: The model achieved a high accuracy of 94.28% on the APTOS 2019 dataset, demonstrating strong performance in both image classification and severity assessment. Validation on the IDRiD dataset confirmed its generalizability, achieving a consistent accuracy of 95.23%. These results indicate the model's effectiveness in accurately diagnosing and assessing DR severity across varied datasets.

CONCLUSION: The proposed Artificial intelligence (AI)-powered diagnostic tool improves diabetic patient care by enabling early DR detection, preventing progression and reducing vision loss. The proposed AI-powered diagnostic tool offers high performance, reliability, and generalizability, providing significant value for clinical DR management.

PMID:40259941 | PMC:PMC12009469 | DOI:10.31661/jbpe.v0i0.2408-1811

Curr Med Imaging. 2025;21:e15734056332443. doi: 10.2174/0115734056332443241129113146.

ABSTRACT

BACKGROUND: The skin, being the largest organ in the human body, plays a vital protective role. Skin lesions are changes in the appearance of the skin, such as bumps, sores, lumps, patches, and discoloration. If not identified and treated promptly, skin lesion diseases would become a serious and worrisome problem for society due to their detrimental effects. However, visually inspecting skin lesions during medical examinations can be challenging due to their similarities.

OBJECTIVE: The proposed research aimed at leveraging technological advancements, particularly deep learning methods, to analyze dermoscopic images of skin lesions and make accurate predictions, thereby aiding in diagnosis.

METHODS: The proposed study utilized four pre-trained CNN architectures, RegNetX, EfficientNetB3, VGG19, and ResNet-152, for the multi-class classification of seven types of skin diseases based on dermoscopic images. The significant finding of this study was the integration of attention mechanisms, specifically channel-wise and spatial attention, into these CNN variants. These mechanisms allowed the models to focus on the most relevant regions of the dermoscopic images, enhancing feature extraction and improving classification accuracy. Hyperparameters of the models were optimized using Bayesian optimization, a probabilistic model-based technique that efficiently uses the hyperparameter space to find the optimal configuration for the developed models.

RESULTS: The performance of these models was evaluated, and it was found that RegNetX outperformed the other models with an accuracy of 98.61%. RegNetX showed robust performance when integrated with both channel-wise and spatial attention mechanisms, indicating its effectiveness in focusing on significant features within the dermoscopic images.

CONCLUSION: The results demonstrated the effectiveness of attention-aware deep learning models in accurately classifying various skin diseases from dermoscopic images. By integrating attention mechanisms, these models can focus on the most relevant features within the images, thereby improving their classification accuracy. The results confirmed that RegNetX, integrated with optimized attention mechanisms, can provide robust, accurate diagnoses, which is critical for early detection and treatment of skin diseases.

PMID:40259872 | DOI:10.2174/0115734056332443241129113146

J Magn Reson Imaging. 2025 Apr 22. doi: 10.1002/jmri.29798. Online ahead of print.

ABSTRACT

BACKGROUND: MRI plays a critical role in prostate cancer (PCa) detection and management. Bi-parametric MRI (bpMRI) offers a faster, contrast-free alternative to multi-parametric MRI (mpMRI). Routine use of mpMRI for all patients may not be necessary, and a tailored imaging approach (bpMRI or mpMRI) based on individual risk might optimize resource utilization.

PURPOSE: To develop and evaluate a deep learning (DL) model for classifying clinically significant PCa (csPCa) using bpMRI and to assess its potential for optimizing MRI protocol selection by recommending the additional sequences of mpMRI only when beneficial.

STUDY TYPE: Retrospective and prospective.

POPULATION: The DL model was trained and validated on 26,129 prostate MRI studies. A retrospective cohort of 151 patients (mean age 65 ± 8) with ground-truth verification from biopsy, prostatectomy, or long-term follow-up, alongside a prospective cohort of 142 treatment-naïve patients (mean age 65 ± 9) undergoing bpMRI, was evaluated.

FIELD STRENGTH/SEQUENCE: 3 T, Turbo-spin echo T2-weighted imaging (T2WI) and single shot EPI diffusion-weighted imaging (DWI).

ASSESSMENT: The DL model, based on a 3D ResNet-50 architecture, classified csPCa using PI-RADS ≥ 3 and Gleason ≥ 7 as outcome measures. The model was evaluated on a prospective cohort labeled by consensus of three radiologists and a retrospective cohort with ground truth verification based on biopsy or long-term follow-up. Real-time inference was tested on an automated MRI workflow, providing classification results directly at the scanner.

STATISTICAL TESTS: AUROC with 95% confidence intervals (CI) was used to evaluate model performance.

RESULTS: In the prospective cohort, the model achieved an AUC of 0.83 (95% CI: 0.77-0.89) for PI-RADS ≥ 3 classification, with 93% sensitivity and 54% specificity. In the retrospective cohort, the model achieved an AUC of 0.86 (95% CI: 0.80-0.91) for Gleason ≥ 7 classification, with 93% sensitivity and 62% specificity. Real-time implementation demonstrated a processing latency of 14-16 s for protocol recommendations.

DATA CONCLUSION: The proposed DL model identifies csPCa using bpMRI and integrates it into clinical workflows.

EVIDENCE LEVEL: 1.

TECHNICAL EFFICACY: Stage 2.

PMID:40259798 | DOI:10.1002/jmri.29798

Brain Behav. 2025 Apr;15(4):e70480. doi: 10.1002/brb3.70480.

ABSTRACT

BACKGROUND: The treatment surgery for hemorrhagic stroke typically involves a puncture drainage procedure to remove the hematoma. However, the puncture targets for puncture and the puncture trajectory significantly influence the therapeutic outcome. This study proposes a computational framework integrating artificial intelligence (AI)-driven segmentation, principal component analysis (PCA), and empirical optimization to automate puncture path generation.

METHODS: A software platform named Puncture Trajectory ToolKits (PTK) was developed using C++/Python with ITK/VTK libraries. Key innovations include hybrid segmentation that combines ResNet-50 deep learning and adaptive thresholding for robust hematoma detection. PCA-based longest axis extraction was enhanced by Laplacian mesh smoothing. Skull quadrant theory and safety corridor modeling were used to avoid critical structures. Five complex clinical cases were used to validate the framework's performance.

RESULTS: The framework demonstrated high accuracy in puncture trajectory planning, with the optimized L2 path achieving a mean surgeon satisfaction score of 4.4/5 (Likert scale) compared to manual methods. The average angle difference between automatically generated and manually designed paths was 16.36°. These results highlight PTK's potential to enhance the efficiency and safety of robotic-assisted neurosurgery.

CONCLUSION: PTK establishes a systematic pipeline for trajectory planning assistance, demonstrating technical superiority over conventional methods. The high acceptance rate among surgeons and improved planning efficiency underscore its clinical applicability. Future integration with robotic systems and validation through clinical trials are warranted.

PMID:40259699 | DOI:10.1002/brb3.70480

NMR Biomed. 2025 Jun;38(6):e70045. doi: 10.1002/nbm.70045.

ABSTRACT

Magnetic resonance fingerprinting (MRF), as an emerging versatile and noninvasive imaging technique, provides simultaneous quantification of multiple quantitative MRI parameters, which have been used to detect changes in cartilage composition and structure in osteoarthritis. Deep learning (DL)-based methods for quantification mapping in MRF overcome the memory constraints and offer faster processing compared to the conventional dictionary matching (DM) method. However, limited attention has been given to the fine-tuning of neural networks (NNs) in DL and fair comparison with DM. In this study, we investigate the impact of training parameter choices on NN performance and compare the fine-tuned NN with DM for multiparametric mapping in MRF. Our approach includes optimizing NN hyperparameters, analyzing the singular value decomposition (SVD) components of MRF data, and optimization of the DM method. We conducted experiments on synthetic data, the NIST/ISMRM MRI system phantom with ground truth, and in vivo knee data from 14 healthy volunteers. The results demonstrate the critical importance of selecting appropriate training parameters, as these significantly affect NN performance. The findings also show that NNs improve the accuracy and robustness of T1, T2, and T1ρ mappings compared to DM in synthetic datasets. For in vivo knee data, the NN achieved comparable results for T1, with slightly lower T2 and slightly higher T1ρ measurements compared to DM. In conclusion, the fine-tuned NN can be used to increase accuracy and robustness for multiparametric quantitative mapping from MRF of the knee joint.

PMID:40259681 | DOI:10.1002/nbm.70045

J Med Syst. 2025 Apr 22;49(1):51. doi: 10.1007/s10916-025-02177-0.

ABSTRACT

Arrhythmias are common and can affect individuals with or without structural heart disease. Deep learning models (DLMs) have shown the ability to recognize arrhythmias using 12-lead electrocardiograms (ECGs). However, the limited types of arrhythmias and dataset robustness have hindered widespread adoption. This study aimed to develop a DLM capable of detecting various arrhythmias across diverse datasets. This algorithm development study utilized 22,130 ECGs, divided into development, tuning, validation, and competition sets. External validation was conducted on three open datasets (CODE-test, PTB-XL, CPSC2018) comprising 32,495 ECGs. The study also assessed the long-term risks of new-onset atrial fibrillation (AF), heart failure (HF), and mortality in individuals with false-positive AF detection by the DLM. In the validation set, the DLM achieved area under the receiver operating characteristic curve above 0.97 and sensitivity/specificity exceeding 90% across most arrhythmia classes. It demonstrated cardiologist-level performance, ranking first in balanced accuracy in a human-machine competition. External validation confirmed comparable performance. Individuals with false-positive AF detection had a significantly higher risk of new-onset AF (hazard ration [HR]: 1.69, 95% confidence interval [CI]: 1.11-2.59), HF (HR: 1.73, 95% CI: 1.20-2.51), and mortality (HR: 1.40, 95% CI: 1.02-1.92) compared to true-negative individuals after adjusting for age and sex. We developed an accurate DLM capable of detecting 23 cardiac arrhythmias across multiple datasets. This DLM serves as a valuable screening tool to aid physicians in identifying high-risk patients, with potential implications for early intervention and risk stratification.

PMID:40259136 | DOI:10.1007/s10916-025-02177-0