Mol Omics. 2025 Mar 17. doi: 10.1039/d4mo00241e. Online ahead of print.

ABSTRACT

Genome architecture in eukaryotes exhibits a high degree of complexity. Amidst the numerous intricacies, the existence of genes as non-continuous stretches composed of exons and introns has garnered significant attention and curiosity among researchers. Accurate identification of exon-intron (EI) boundaries is crucial to decipher the molecular biology governing gene expression and regulation. This includes understanding both normal and aberrant splicing, with aberrant splicing referring to the abnormal processing of pre-mRNA that leads to improper inclusion or exclusion of exons or introns. Such splicing events can result in dysfunctional or non-functional proteins, which are often associated with various diseases. The currently employed frameworks for genomic signals, which aim to identify exons and introns within a genomic segment, need to be revised primarily due to the lack of a robust consensus sequence and the limitations posed by the training on available experimental datasets. To tackle these challenges and capitalize on the understanding that DNA exhibits function-dependent local physicochemical variations, we present ChemEXIN, an innovative novel method for predicting EI boundaries. The method utilizes a deep-learning (DL) architecture alongside tri- and tetra-nucleotide-based structural and energy features. ChemEXIN outperforms existing methods with notable accuracy and precision. It achieves an accuracy of 92.5% for humans, 79.9% for mice, and 92.0% for worms, along with precision values of 92.0%, 79.6%, and 91.8% for the same organisms, respectively. These results represent a significant advancement in EI boundary annotations, with potential implications for understanding gene expression, regulation, and cellular functions.

PMID:40094442 | DOI:10.1039/d4mo00241e

Diagn Interv Radiol. 2025 Mar 17. doi: 10.4274/dir.2025.242999. Online ahead of print.

ABSTRACT

PURPOSE: Established methods for bone age assessment (BAA), such as the Greulich and Pyle atlas, suffer from variability due to population differences and observer discrepancies. Although automated BAA offers speed and consistency, limited research exists on its performance across different populations using deep learning. This study examines deep learning algorithms on the Turkish population to enhance bone age models by understanding demographic influences.

METHODS: We analyzed reports from Bağcılar Hospital's Health Information Management System between April 2012 and September 2023 using "bone age" as a keyword. Patient images were re-evaluated by an experienced radiologist and anonymized. A total of 2,730 hand radiographs from Bağcılar Hospital (Turkish population), 12,572 from the Radiological Society of North America (RSNA), and 6,185 from the Radiological Hand Pose Estimation (RHPE) public datasets were collected, along with corresponding bone ages and gender information. A random set of 546 radiographs (273 from Bağcılar, 273 from public datasets) was initially randomly split for an internal test set with bone age stratification; the remaining data were used for training and validation. BAAs were generated using a modified InceptionV3 model on 500 × 500-pixel images, selecting the model with the lowest mean absolute error (MAE) on the validation set.

RESULTS: Three models were trained and tested based on dataset origin: Bağcılar (Turkish), public (RSNA-RHPE), and a Combined model. Internal test set predictions of the Combined model estimated bone age within less than 6, 12, 18, and 24 months at rates of 44%, 73%, 87%, and 94%, respectively. The MAE was 9.2 months in the overall internal test set, 7 months on the public test set, and 11.5 months on the Bağcılar internal test data. The Bağcılar-only model had an MAE of 12.7 months on the Bağcılar internal test data. Despite less training data, there was no significant difference between the combined and Bağcılar models on the Bağcılar dataset (P > 0.05). The public model showed an MAE of 16.5 months on the Bağcılar dataset, significantly worse than the other models (P < 0.05).

CONCLUSION: We developed an automatic BAA model including the Turkish population, one of the few such studies using deep learning. Despite challenges from population differences and data heterogeneity, these models can be effectively used in various clinical settings. Model accuracy can improve over time with cumulative data, and publicly available datasets may further refine them. Our approach enables more accurate and efficient BAAs, supporting healthcare professionals where traditional methods are time-consuming and variable.

CLINICAL SIGNIFICANCE: The developed automated BAA model for the Turkish population offers a reliable and efficient alternative to traditional methods. By utilizing deep learning with diverse datasets from Bağcılar Hospital and publicly available sources, the model minimizes assessment time and reduces variability. This advancement enhances clinical decision-making, supports standardized BAA practices, and improves patient care in various healthcare settings.

PMID:40094318 | DOI:10.4274/dir.2025.242999

EClinicalMedicine. 2025 Feb 26;81:103128. doi: 10.1016/j.eclinm.2025.103128. eCollection 2025 Mar.

ABSTRACT

BACKGROUND: Misdiagnosis of hemorrhage secondary to cerebral venous sinus thrombosis (CVST-ICH) as arterial-origin spontaneous intracerebral hemorrhage (sICH) can lead to inappropriate treatment and the potential for severe adverse outcomes. The current practice for identifying CVST-ICH involves venography, which, despite being increasingly utilized in many centers, is not typically used as the initial imaging modality for ICH patients. The study aimed to develop an explainable deep learning model to quickly identify ICH caused by CVST based on non-contrast computed tomography (NCCT).

METHODS: The study population included patients diagnosed with CVST-ICH and other spontaneous ICH from January 2016 to March 2023 at the Second Affiliated Hospital of Zhejiang University, Taizhou First People's Hospital, Taizhou Hospital, Quzhou Second People's Hospital, and Longyan First People's Hospital. A transfer learning-based 3D U-Net with segmentation and classification was proposed and developed only on admission plain CT. Model performance was assessed using the area under the curve (AUC), sensitivity, and specificity metrics. For further evaluation, the average diagnostic performance of nine doctors on plain CT was compared with model assistance. Interpretability methods, including Grad-CAM++, SHAP, IG, and occlusion, were employed to understand the model's attention.

FINDINGS: An internal dataset was constructed using propensity score matching based on age, initially including 102 CVST-ICH patients (median age: 44 [29, 61] years) and 683 sICH patients (median age: 65 [52, 73] years). After matching, 102 CVST-ICH patients and 306 sICH patients (median age: 50 [40, 62] years) were selected. An external dataset consisted of 38 CVST-ICH and 119 sICH patients from four other hospitals. Validation showed AUC 0·94, sensitivity 0·96, and specificity 0·8 for the internal testing subset; AUC 0·85, sensitivity 0·87, and specificity 0·82 for the external dataset, respectively. The discrimination performance of nine doctors interpreting CT images significantly improved with the assistance of the proposed model (accuracy 0·79 vs 0·71, sensitivity 0·88 vs 0·81, specificity 0·75 vs 0·68, p < 0·05). Interpretability methods highlighted the attention of model to the features of hemorrhage edge appearance.

INTERPRETATION: The present model demonstrated high-performing and robust results on discrimination between CVST-ICH and spontaneous ICH, and aided doctors' diagnosis in clinical practice as well. Prospective validation with larger-sample size is required.

FUNDING: The work was funded by the National Key R&D Program of China (2023YFE0118900), National Natural Science Foundation of China (No.81971155 and No.81471168), the Science and Technology Department of Zhejiang Province (LGJ22H180004), Medical and Health Science and Technology Project of Zhejiang Province (No.2022KY174), the 'Pioneer' R&D Program of Zhejiang (No. 2024C03006 and No. 2023C03026) and the MOE Frontier Science Center for Brain Science & Brain-Machine Integration, Zhejiang University.

PMID:40093990 | PMC:PMC11909457 | DOI:10.1016/j.eclinm.2025.103128

EClinicalMedicine. 2025 Feb 26;81:103125. doi: 10.1016/j.eclinm.2025.103125. eCollection 2025 Mar.

ABSTRACT

BACKGROUND: Accurate prediction of early recurrence is essential for disease management of patients with non-muscle-invasive bladder cancer (NMIBC). We aimed to develop and validate a deep learning-based early recurrence predictive model (ERPM) and a treatment response predictive model (TRPM) on whole slide images to assist clinical decision making.

METHODS: In this retrospective, multicentre study, we included consecutive patients with pathology-confirmed NMIBC who underwent transurethral resection of bladder tumour from five centres. Patients from one hospital (Sun Yat-sen Memorial Hospital of Sun Yat-sen University, Guangzhou, China) were assigned to training and internal validation cohorts, and patients from four other hospitals (the Third Affiliated Hospital of Sun Yat-sen University, and Zhujiang Hospital of Southern Medical University, Guangzhou, China; the Fifth Affiliated Hospital of Sun Yat-sen University, Zhuhai, China; Shenshan Medical Centre, Shanwei, China) were assigned to four independent external validation cohorts. Based on multi-instance and ensemble learning, the ERPM was developed to make predictions on haematoxylin and eosin (H&E) staining and immunohistochemistry staining slides. Sharing the same architecture of the ERPM, the TRPM was trained and evaluated by cross validation on patients who received Bacillus Calmette-Guérin (BCG). The performance of the ERPM was mainly evaluated and compared with the clinical model, H&E-based model, and integrated model through the area under the curve. Survival analysis was performed to assess the prognostic capability of the ERPM.

FINDINGS: Between January 1, 2017, and September 30, 2023, 4395 whole slide images of 1275 patients were included to train and validate the models. The ERPM was superior to the clinical and H&E-based model in predicting early recurrence in both internal validation cohort (area under the curve: 0.837 vs 0.645 vs 0.737) and external validation cohorts (area under the curve: 0.761-0.802 vs 0.626-0.682 vs 0.694-0.723) and was on par with the integrated model. It also stratified recurrence-free survival significantly (p < 0.0001) with a hazard ratio of 4.50 (95% CI 3.10-6.53). The TRPM performed well in predicting BCG-unresponsive NMIBC (accuracy 84.1%).

INTERPRETATION: The ERPM showed promising performance in predicting early recurrence and recurrence-free survival of patients with NMIBC after surgery and with further validation and in combination with TRPM could be used to guide the management of NMIBC.

FUNDING: National Natural Science Foundation of China, the Science and Technology Planning Project of Guangdong Province, the National Key Research and Development Programme of China, the Guangdong Provincial Clinical Research Centre for Urological Diseases, and the Science and Technology Projects in Guangzhou.

PMID:40093987 | PMC:PMC11909458 | DOI:10.1016/j.eclinm.2025.103125

Research (Wash D C). 2025 Mar 14;8:0603. doi: 10.34133/research.0603. eCollection 2025.

ABSTRACT

Takeover safety draws increasing attention in the intelligent transportation as the new energy vehicles with cutting-edge autopilot capabilities vigorously blossom on the road. Despite recent studies highlighting the importance of drivers' emotions in takeover safety, the lack of emotion-aware takeover datasets hinders further investigation, thereby constraining potential applications in this field. To this end, we introduce ViE-Take, the first Vision-driven (Vision is used since it constitutes the most cost-effective and user-friendly solution for commercial driver monitor systems) dataset for exploring the Emotional landscape in Takeovers of autonomous driving. ViE-Take enables a comprehensive exploration of the impact of emotions on drivers' takeover performance through 3 key attributes: multi-source emotion elicitation, multi-modal driver data collection, and multi-dimensional emotion annotations. To aid the use of ViE-Take, we provide 4 deep models (corresponding to 4 prevalent learning strategies) for predicting 3 different aspects of drivers' takeover performance (readiness, reaction time, and quality). These models offer benefits for various downstream tasks, such as driver emotion recognition and regulation for automobile manufacturers. Initial analysis and experiments conducted on ViE-Take indicate that (a) emotions have diverse impacts on takeover performance, some of which are counterintuitive; (b) highly expressive social media clips, despite their brevity, prove effective in eliciting emotions (a foundation for emotion regulation); and (c) predicting takeover performance solely through deep learning on vision data not only is feasible but also holds great potential.

PMID:40093973 | PMC:PMC11908832 | DOI:10.34133/research.0603

Theranostics. 2025 Feb 18;15(8):3223-3233. doi: 10.7150/thno.100786. eCollection 2025.

ABSTRACT

Retinal images provide a non-invasive and accessible means to directly visualize human blood vessels and nerve fibers. Growing studies have investigated the intricate microvascular and neural circuitry within the retina, its interactions with other systemic vascular and nervous systems, and the link between retinal biomarkers and various systemic diseases. Using the eye to study systemic health, based on these connections, has been given a term as oculomics. Advancements in artificial intelligence (AI) technologies, particularly deep learning, have further increased the potential impact of this study. Leveraging these technologies, retinal analysis has demonstrated potentials in detecting numerous diseases, including cardiovascular diseases, central nervous system diseases, chronic kidney diseases, metabolic diseases, endocrine disorders, and hepatobiliary diseases. AI-based retinal imaging, which incorporates established modalities such as digital color fundus photographs, optical coherence tomography (OCT) and OCT angiography, as well as emerging technologies like ultra-wide field imaging, shows great promises in predicting systemic diseases. This provides a valuable opportunity for systemic diseases screening, early detection, prediction, risk stratification, and personalized prognostication. As the AI and big data research field grows, with the mission of transforming healthcare, they also face numerous challenges and limitations both in data and technology. The application of natural language processing framework, large language model, and other generative AI techniques presents both opportunities and concerns that require careful consideration. In this review, we not only summarize key studies on AI-enhanced retinal imaging for predicting systemic diseases but also underscore the significance of these advancements in transforming healthcare. By highlighting the remarkable progress made thus far, we provide a comprehensive overview of state-of-the-art techniques and explore the opportunities and challenges in this rapidly evolving field. This review aims to serve as a valuable resource for researchers and clinicians, guiding future studies and fostering the integration of AI in clinical practice.

PMID:40093903 | PMC:PMC11905132 | DOI:10.7150/thno.100786

Eur J Radiol Open. 2025 Feb 24;14:100639. doi: 10.1016/j.ejro.2025.100639. eCollection 2025 Jun.

ABSTRACT

OBJECTIVES: This study aimed to develop a predictive model for lymph node metastasis (LNM) in papillary thyroid carcinoma (PTC) patients by deep learning radiomic (DLRad) and clinical features.

METHODS: This study included 271 thyroid lobes from 228 PTC patients who underwent preoperative neck non-contrast CT at Center 1 (May 2021-April 2024). LNM status was confirmed via postoperative pathology, with each thyroid lobe labeled accordingly. The cohort was divided into training (n = 189) and validation (n = 82) cohorts, with additional temporal (n = 59 lobes, Center 1, May-August 2024) and external (n = 66 lobes, Center 2) test cohorts. Thyroid lobes were manually segmented from the isthmus midline, ensuring interobserver consistency (ICC ≥ 0.8). Deep learning and radiomics features were selected using LASSO algorithms to compute DLRad scores. Logistic regression identified independent predictors, forming DLRad, clinical, and combined models. Model performance was evaluated using AUC, calibration, decision curves, and the DeLong test, compared against radiologists' assessments.

RESULTS: Independent predictors of LNM included age, gender, multiple nodules, tumor size group, and DLRad. The combined model demonstrated superior diagnostic performance with AUCs of 0.830 (training), 0.799 (validation), 0.819 (temporal test), and 0.756 (external test), outperforming the DLRad model (AUCs: 0.786, 0.730, 0.753, 0.642), clinical model (AUCs: 0.723, 0.745, 0.671, 0.660), and radiologist evaluations (AUCs: 0.529, 0.606, 0.620, 0.503). It also achieved the lowest Brier scores (0.167, 0.184, 0.175, 0.201) and the highest net benefit in decision-curve analysis at threshold probabilities > 20 %.

CONCLUSIONS: The combined model integrating DLRad and clinical features exhibits good performance in predicting LNM in PTC patients.

PMID:40093877 | PMC:PMC11908562 | DOI:10.1016/j.ejro.2025.100639

J Immunother Cancer. 2025 Mar 15;13(3):e011149. doi: 10.1136/jitc-2024-011149.

ABSTRACT

BACKGROUND: Accurate prediction of pathologic complete response (pCR) following neoadjuvant immunotherapy combined with chemotherapy (nICT) is crucial for tailoring patient care in esophageal squamous cell carcinoma (ESCC). This study aimed to develop and validate a deep learning model using a novel voxel-level radiomics approach to predict pCR based on preoperative CT images.

METHODS: In this multicenter, retrospective study, 741 patients with ESCC who underwent nICT followed by radical esophagectomy were enrolled from three institutions. Patients from one center were divided into a training set (469 patients) and an internal validation set (118 patients) while the data from the other two centers was used as external validation sets (120 and 34 patients, respectively). The deep learning model, Vision-Mamba, integrated voxel-level radiomics feature maps and CT images for pCR prediction. Additionally, other commonly used deep learning models, including 3D-ResNet and Vision Transformer, as well as traditional radiomics methods, were developed for comparison. Model performance was evaluated using accuracy, area under the curve (AUC), sensitivity, specificity, and prognostic stratification capabilities. The SHapley Additive exPlanations analysis was employed to interpret the model's predictions.

RESULTS: The Vision-Mamba model demonstrated robust predictive performance in the training set (accuracy: 0.89, AUC: 0.91, sensitivity: 0.82, specificity: 0.92) and validation sets (accuracy: 0.83-0.91, AUC: 0.83-0.92, sensitivity: 0.73-0.94, specificity: 0.84-1.0). The model outperformed other deep learning models and traditional radiomics methods. The model's ability to stratify patients into high and low-risk groups was validated, showing superior prognostic stratification compared with traditional methods. SHAP provided quantitative and visual model interpretation.

CONCLUSIONS: We present a voxel-level radiomics-based deep learning model to predict pCR to neoadjuvant immunotherapy combined with chemotherapy based on pretreatment diagnostic CT images with high accuracy and robustness. This model could provide a promising tool for individualized management of patients with ESCC.

PMID:40090670 | DOI:10.1136/jitc-2024-011149

Environ Pollut. 2025 Mar 14:125993. doi: 10.1016/j.envpol.2025.125993. Online ahead of print.

ABSTRACT

Airborne microplastics (AMPs) are prevalent in both indoor and outdoor environments, posing potential health risks to humans. Automating the process of spotting them in micrographs can significantly enhance research and monitoring. Although deep learning has shown substantial promise in microplastic analysis, existing studies have primarily focused on high-resolution images of samples collected from marine and freshwater environments. In contrast, this work introduces a novel approach by employing enhanced U-Net models (Attention U-Net and Dynamic RU-NEXT) along with the Mask Region Convolutional Neural Network (Mask R-CNN) to identify and classify AMPs in lower-resolution micrographs (256 × 256 pixels) obtained from outdoor environments. A key innovation involves integrating classification directly within the U-Net-based segmentation frameworks, thereby streamlining the workflow and improving computational efficiency which is an advancement over previous work where segmentation and classification were performed separately. The enhanced U-Net models attained average classification F1-scores exceeding 85% and segmentation scores above 77%. Additionally, the Mask R-CNN model achieved an average bounding box precision of 73.32% on the test set, a classification F1-score of 84.29%, and a mask precision of 71.31%, demonstrating robust performance. The proposed method provides a faster and more accurate means of identifying AMPs compared to thresholding techniques. It also functions effectively as a pre-screening tool, substantially reducing the number of particles requiring labour-intensive chemical analysis. By integrating advanced deep learning strategies into AMPs research, this study paves the way for more efficient monitoring and characterisation of microplastics.

PMID:40090454 | DOI:10.1016/j.envpol.2025.125993

J Dent. 2025 Mar 14:105689. doi: 10.1016/j.jdent.2025.105689. Online ahead of print.

ABSTRACT

OBJECTIVES: To develop and validate an explainable Artificial Intelligence (AI) model for classifying and quantifying upper airway obstruction related to adenoid hypertrophy using three-dimensional (3D) shape analysis of cone-beam computed tomography (CBCT) scans.

METHODS: 400 CBCT scans of patients aged 5-18 years were analyzed. Nasopharyngeal airway obstruction (NAO) ratio was calculated to label scans into four grades of obstruction severity, used as the ground truth. Upper airway surface meshes were used to train a deep learning model combining multiview and point-cloud approaches for 3D shape analysis and obstruction severity classification and quantification. Surface Gradient-weighted Class Activation Mapping (SurfGradCAM) generated explainability heatmaps. Performance was evaluated using area under the curve (AUC), precision, recall, F1-score, mean absolute error, root mean squared error, and correlation coefficients.

RESULTS: The explainable AI model demonstrated strong performance in both classification and quantification tasks. The AUC values for the classification task ranged from 0.77 to 0.94, with the highest values of 0.88 and 0.94 for Grades 3 and 4, respectively, indicating excellent discriminative ability for identifying more severe cases of obstruction. The SurfGradCAM-generated heatmaps consistently highlighted the most relevant regions of the upper airway influencing the AI's decision-making process. In the quantification task, the regression model successfully predicted the NAO ratio, with a strong correlation coefficient of 0.854 (p<0.001) and R2= 0.728, explaining a substantial proportion of the variance in NAO ratios.

CONCLUSIONS: The proposed explainable AI model, using 3D shape analysis, demonstrated strong performance in classifying and quantifying adenoid hypertrophy-related upper airway obstruction in CBCT scans.

CLINICAL SIGNIFICANCE: This AI model provides clinicians with a reliable, automated tool for standardized adenoid hypertrophy assessment. The model's explainable nature enhances clinical confidence and patient communication, potentially improving diagnostic workflow and treatment planning.

PMID:40090403 | DOI:10.1016/j.jdent.2025.105689

Can J Ophthalmol. 2025 Mar 13:S0008-4182(25)00070-5. doi: 10.1016/j.jcjo.2025.02.013. Online ahead of print.

ABSTRACT

PURPOSE: To evaluate the association between localized features of diabetic macular edema (DME) and point-wise retinal sensitivity (RS) assessed with microperimetry (MP) using deep learning (DL)-based automated quantification on optical coherence tomography (OCT) scans.

DESIGN: Cross-sectional study.

PARTICIPANTS: Twenty eyes of 20 subjects with clinically significant DME were included in this study.

METHODS: Patients with DME visible on OCT scans (Spectralis HRA+OCT) completed 2 MP examinations using a custom 45 stimuli grid on MAIA (CenterVue). MP stimuli were coregistered with the corresponding OCT location using image registration algorithms. DL-based algorithms were used to quantify intraretinal fluid (IRF) and ellipsoid zone (EZ) thickness. Hard exudates (HEs) were quantified semiautomatically. Multivariable mixed-effect models were calculated to investigate the association between DME-specific OCT features and point-wise RS. As EZ thickness values below HEs were excluded, the models included either EZ thickness or HEs.

RESULTS: A total of 1800 MP stimuli from 20 eyes of 20 patients were analyzed. Stimuli with IRF (n = 568) showed significantly decreased RS compared to areas without (estimate [95% CI]: -1.11 dB [-1.69, -0.52]; p = 0.0002). IRF volume was significantly negatively (-0.45 dB/nL [-0.71; -0.18]; p = 0.001) and EZ thickness positively (0.14 dB/µm [0.1; 0.19]; p < 0.0001) associated with localized point-wise RS. In the multivariable mixed model, including HE volume instead of EZ thickness, a negative impact on RS was observed (-0.43/0.1 nL [-0.81; -0.05]; p = 0.027).

CONCLUSIONS: DME-specific features, as analyzed on OCT, have a significant impact on point-wise RS. IRF and HE volume showed a negative and EZ thickness, a positive association with localized RS.

PMID:40090368 | DOI:10.1016/j.jcjo.2025.02.013

Comput Biol Med. 2025 Mar 15;189:109957. doi: 10.1016/j.compbiomed.2025.109957. Online ahead of print.

ABSTRACT

The availability of large-scale electrocardiogram (ECG) databases and advancements in machine learning have facilitated the development of automated diagnostic systems for cardiac arrhythmias. Deep learning models, despite their potential for high accuracy, have had limited clinical adoption due to their inherent lack of interpretability. This study bridges this gap by proposing a feature-based approach that maintains comparable performance to deep learning while providing enhanced interpretability for clinical utility. The method extracts a total of 654 individual features, classified into 60 feature types from each ECG. The features use mathematical techniques such as the Fourier transform, wavelet transform, and cross-correlation for rigorous evaluation of ECG characteristics. The eXtreme Gradient Boosting framework was employed to classify each ECG into one of nine diagnostic classes. Shapley Additive Explanations (SHAP) value analysis was used to downselect the feature set to the minimal set without incurring a performance reduction (159 features). Overall, the proposed method demonstrated performance comparable to state-of-the-art deep learning classifiers, achieving a weighted F1 score of 81% during cross-validation and 68% on the external test dataset, while offering greater ease of implementation and adaptability to diverse clinical applications. Notably, the proposed method demonstrated superior accuracy in identifying atrial fibrillation and block-related abnormalities, achieving overall F1 scores of 89% and 87% during cross-validation and 79% and 75% on the external test dataset, respectively. SHAP value analysis of the testing results revealed the top-performing features for each diagnostic class aligned with standard clinical diagnostic processes, highlighting the clinical interpretability of our approach.

PMID:40090185 | DOI:10.1016/j.compbiomed.2025.109957

Med Phys. 2025 Mar 16. doi: 10.1002/mp.17759. Online ahead of print.

ABSTRACT

BACKGROUND: Magnetic resonance-guided adaptive radiation therapy (MRgART) systems combine Magnetic resonance imaging (MRI) technology with linear accelerators (LINAC) to enhance the precision and efficacy of cancer treatment. These systems enable real-time adjustments of treatment plans based on the latest patient anatomy, creating an urgent need for accurate and rapid dose calculation algorithms. Traditional CT-based dose calculations and ray-tracing (RT) processes are time-consuming and may not be feasible for the online adaptive workflow required in MRgART. Recent advancements in deep learning (DL) offer promising solutions to overcome these limitations.

PURPOSE: This study aims to develop a DL-based dose calculation engine for MRgART that relies solely on MR images. This approach addresses the critical need for accurate and rapid dose calculations in the MRgART workflow without relying on CT images or time-consuming RT processes.

METHODS: We used a deep residual network inspired by U-Net to establish a direct connection between distance-corrected conical (DCC) fluence maps and dose distributions in the image domain. The study utilized data from 30 prostate cancer patients treated with fixed-beam Intensity-Modulated Radiation Therapy (IMRT) on an MR-guided LINAC system. We trained, validated, and tested the model using a total of 120 online treatment plans, which encompassed 1080 individual beams. We extensively evaluated the network's performance by comparing its dose calculation accuracy against Monte Carlo (MC)-based methods using metrics such as mean absolute error (MAE) of pixel-wise dose differences, 3D gamma analysis, dose-volume histograms (DVHs), dosimetric indices, and isodose line similarity.

RESULTS: The proposed DL model demonstrated high accuracy in dose calculations. The median MAE of pixel-wise dose differences was 1.2% for the whole body, 1.9% for targets, and 1.1% for organs at risk (OARs). The median 3D gamma passing rates for the 3%/3 mm criterion were 94.8% for the whole body, 95.7% for targets, and 98.7% for OARs. Additionally, the Dice similarity coefficient (DSC) of isodose lines between the DL-based and MC-based dose calculations averaged 0.94 ± 0.01. There were no big differences between the DL-based and MC-based calculations in the DVH curves and clinical dosimetric indices. This proved that the two methods were clinically equivalent.

CONCLUSION: This study presents a novel MR-only dose calculation engine that eliminates the need for CT images and complex RT processes. By leveraging DL, the proposed method significantly enhances the efficiency and accuracy of the MRgART workflow, particularly for prostate cancer treatment. This approach holds potential for broader applications across different cancer types and MR-linac systems, paving the way for more streamlined and precise radiation therapy planning.

PMID:40089982 | DOI:10.1002/mp.17759

Med Phys. 2025 Mar 16. doi: 10.1002/mp.17747. Online ahead of print.

ABSTRACT

BACKGROUND: Quantitative susceptibility mapping (QSM) is a post-processing magnetic resonance imaging (MRI) technique that extracts the distribution of tissue susceptibilities and holds significant promise in the study of neurological diseases. However, the ill-conditioned nature of dipole inversion often results in noise and artifacts during QSM reconstruction from the tissue field. Deep learning methods have shown great potential in addressing these issues; however, most existing approaches rely on basic U-net structures, leading to limited performances and reconstruction artifacts sometimes.

PURPOSE: This study aims to develop a novel deep learning-based method, IR2QSM, for improving QSM reconstruction accuracy while mitigating noise and artifacts by leveraging a unique network architecture that enhances latent feature utilization.

METHODS: IR2QSM, an advanced U-net architecture featuring four iterations of reverse concatenations and middle recurrent modules, was proposed to optimize feature fusion and improve QSM accuracy, and comparative experiments based on both simulated and in vivo datasets were carried out to compare IR2QSM with two traditional iterative methods (iLSQR, MEDI) and four recently proposed deep learning methods (U-net, xQSM, LPCNN, and MoDL-QSM).

RESULTS: In this work, IR2QSM outperformed all other methods in reducing artifacts and noise in QSM images. It achieved on average the lowest XSIM (84.81%) in simulations, showing improvements of 12.80%, 12.68%, 18.66%, 10.49%, 25.57%, and 19.78% over iLSQR, MEDI, U-net, xQSM, LPCNN, and MoDL-QSM, respectively, and yielded results with the least artifacts on the in vivo data and present the most visually appealing results. In the meantime, it successfully alleviated the over-smoothing and susceptibility underestimation in LPCNN results.

CONCLUSION: Overall, the proposed IR2QSM showed superior QSM results compared to iterative and deep learning-based methods, offering a more accurate QSM solution for clinical applications.

PMID:40089979 | DOI:10.1002/mp.17747

Hepatol Int. 2025 Mar 16. doi: 10.1007/s12072-025-10793-8. Online ahead of print.

ABSTRACT

PURPOSE: Existing prognostic staging systems depend on expensive manual extraction by pathologists, potentially overlooking latent patterns critical for prognosis, or use black-box deep learning models, limiting clinical acceptance. This study introduces a novel deep learning-assisted paradigm that complements existing approaches by generating interpretable, multi-view risk scores to stratify prognostic risk in hepatocellular carcinoma (HCC) patients.

METHODS: 510 HCC patients were enrolled in an internal dataset (SYSUCC) as training and validation cohorts to develop the Hybrid Deep Score (HDS). The Attention Activator (ATAT) was designed to heuristically identify tissues with high prognostic risk, and a multi-view risk-scoring system based on ATAT established HDS from microscopic to macroscopic levels. HDS was also validated on an external testing cohort (TCGA-LIHC) with 341 HCC patients. We assessed prognostic significance using Cox regression and the concordance index (c-index).

RESULTS: The ATAT first heuristically identified regions where necrosis, lymphocytes, and tumor tissues converge, particularly focusing on their junctions in high-risk patients. From this, this study developed three independent risk factors: microscopic morphological, co-localization, and deep global indicators, which were concatenated and then input into a neural network to generate the final HDS for each patient. The HDS demonstrated competitive results with hazard ratios (HR) (HR 3.24, 95% confidence interval (CI) 1.91-5.43 in SYSUCC; HR 2.34, 95% CI 1.58-3.47 in TCGA-LIHC) and c-index values (0.751 in SYSUCC; 0.729 in TCGA-LIHC) for Disease-Free Survival (DFS). Furthermore, integrating HDS into existing clinical staging systems allows for more refined stratification, which enables the identification of potential high-risk patients within low-risk groups.

CONCLUSION: This novel paradigm, from identifying high-risk tissues to constructing prognostic risk scores, offers fresh insights into HCC research. Additionally, the integration of HDS complements the existing clinical staging system by facilitating more detailed stratification in DFS and Overall Survival (OS).

PMID:40089963 | DOI:10.1007/s12072-025-10793-8

Ann Nucl Med. 2025 Mar 16. doi: 10.1007/s12149-025-02038-3. Online ahead of print.

ABSTRACT

OBJECTIVE: To provide fully automatic scanner-independent 5-level categorization of the [123I]FP-CIT uptake in striatal subregions in dopamine transporter SPECT.

METHODS: A total of 3500 [123I]FP-CIT SPECT scans from two in house (n = 1740, n = 640) and two external (n = 645, n = 475) datasets were used for this study. A convolutional neural network (CNN) was trained for the categorization of the [123I]FP-CIT uptake in unilateral caudate and putamen in both hemispheres according to 5 levels: normal, borderline, moderate reduction, strong reduction, almost missing. Reference standard labels for the network training were created automatically by fitting a Gaussian mixture model to histograms of the specific [123I]FP-CIT binding ratio, separately for caudate and putamen and separately for each dataset. The CNN was trained on a mixed-scanner subsample (n = 1957) and tested on one independent identically distributed (IID, n = 1068) and one out-of-distribution (OOD, n = 475) test dataset.

RESULTS: The accuracy of the CNN for the 5-level prediction of the [123I]FP-CIT uptake in caudate/putamen was 80.1/78.0% in the IID test dataset and 78.1/76.5% in the OOD test dataset. All 4 regional 5-level predictions were correct in 54.3/52.6% of the cases in the IID/OOD test dataset. A global binary score automatically derived from the regional 5-scores achieved 97.4/96.2% accuracy for automatic classification of the scans as normal or reduced relative to visual expert read as reference standard.

CONCLUSIONS: Automatic scanner-independent 5-level categorization of the [123I]FP-CIT uptake in striatal subregions by a CNN model is feasible with clinically useful accuracy.

PMID:40089953 | DOI:10.1007/s12149-025-02038-3

Adv Sci (Weinh). 2025 Mar 16:e2500859. doi: 10.1002/advs.202500859. Online ahead of print.

ABSTRACT

Evolutionary constraints significantly limit the diversity of naturally occurring enzymes, thereby reducing the sequence repertoire available for enzyme discovery and engineering. Recent breakthroughs in protein structure prediction and de novo design, powered by artificial intelligence, now enable to create enzymes with desired functions without solely relying on traditional genome mining. Here, a computational strategy is demonstrated for creating new-to-nature polyethylene terephthalate hydrolases (PET hydrolases) by leveraging the known catalytic mechanisms and implementing multiple deep learning algorithms and molecular computations. This strategy includes the extraction of functional motifs from a template enzyme (here leaf-branch compost cutinase, LCC, is used), regeneration of new protein sequences, computational screening, experimental validation, and sequence refinement. PET hydrolytic activity is successfully replicated with designer enzymes that are at least 30% shorter in sequence length than LCC. Among them, RsPETase1 stands out due to its robust expressibility. It exhibits comparable catalytic efficiency (kcat/Km) to LCC and considerable thermostability with a melting temperature of 56 °C, despite sharing only 34% sequence similarity with LCC. This work suggests that enzyme diversity can be expanded by recapitulating functional motifs with computationally built protein scaffolds, thus generating opportunities to acquire highly active and robust enzymes that do not exist in nature.

PMID:40089854 | DOI:10.1002/advs.202500859

Syst Rev. 2025 Mar 15;14(1):62. doi: 10.1186/s13643-025-02779-2.

ABSTRACT

BACKGROUND: Artificial intelligence (AI) has shown immense potential in the field of medicine, but its actual effectiveness and safety still need to be validated through clinical trials. Currently, the research themes, methodologies, and development trends of AI-related clinical trials remain unclear, and further exploration of these studies will be crucial for uncovering AI's practical application potential and promoting its broader adoption in clinical settings.

OBJECTIVE: To analyze the current status, hotspots, and trends of published clinical research on AI applications.

METHODS: Publications related to AI clinical applications were retrieved from the Web of Science database. Relevant data were extracted using VOSviewer 1.6.17 to generate visual cooperation network maps for countries, organizations, authors, and keywords. Burst citation detection for keywords and citations was performed using CiteSpace 5.8.R3 to identify sudden surges in citation frequency within a short period, and the theme evolution was analyzed using SciMAT to track the development and trends of research topics over time.

RESULTS: A total of 22,583 articles were obtained from the Web of Science database. Seven-hundred and thirty-five AI clinical application research were published by 1764 institutions from 53 countries. The majority of publications were contributed by the United States, China, and the UK. Active collaborations were noted among leading authors, particularly those from developed countries. The publications mainly focused on evaluating the application value of AI technology in the fields of disease diagnosis and classification, disease risk prediction and management, assisted surgery, and rehabilitation. Deep learning and chatbot technologies were identified as emerging research hotspots in recent studies on AI applications.

CONCLUSIONS: A total of 735 articles on AI in clinical research were analyzed, with publication volume and citation counts steadily increasing each year. Institutions and researchers from the United States contributed the most to the research output in this field. Key areas of focus included AI applications in surgery, rehabilitation, disease diagnosis, risk prediction, and health management, with emerging trends in deep learning and chatbots. This study also provides detailed and intuitive information about important articles, journals, core authors, institutions, and topics in the field through visualization maps, which will help researchers quickly understand the current status, hotspots, and trends of artificial intelligence clinical application research. Future clinical trials of artificial intelligence should strengthen scientific design, ethical compliance, and interdisciplinary and international cooperation and pay more attention to its practical clinical value and reliable application in diverse scenarios.

PMID:40089747 | DOI:10.1186/s13643-025-02779-2

Sci Rep. 2025 Mar 15;15(1):9015. doi: 10.1038/s41598-025-93011-7.

ABSTRACT

Physicians evaluate a patient's respiratory health during a physical examination by visual assessment of the work of breathing (WoB) to determine respiratory stability, and by detecting abnormal lung sounds via lung auscultation using a stethoscope to identify common pathological lung diseases, such as chronic obstructive pulmonary disease (COPD) and pneumonia. Since these assessment methods are subjective, a low-profile device used for an accurate and quantitative monitoring approach could provide valuable preemptive insights into respiratory health, proving to be clinically beneficial. To achieve this goal, we have developed a miniature patch consisting of a sensitive wideband multi-axis seismometer that can be placed on the anatomical areas of a patient's lungs to enable an effective quantification of a patient's WoB and lung sounds. When used on a patch, the seismometer captures chest wall vibrations due to respiratory muscle effort, known as high-frequency mechanomyogram (MMG), during tidal breathing as well as seismic pulmonary-induced vibrations (PIVs) during deep breathing due to normal and/or adventitious lung sounds like crackles, while simultaneously recording respiration rate and phase. A system comprised of multiple patches was evaluated on 124 patients in the hospital setting and shown to accurately assess and quantify a patent's physical signs of WoB by measuring the average respiratory effort extracted from high-frequency MMG signals, demonstrating statistical significance of this method in comparison to clinical bedside observation of WoB and respiration rate. A data fusion deep learning model was developed which combined the inputs of PIVs lung sounds and the corresponding respiration phase to detect crackle, wheeze and normal breath sound features. The model exhibited high accuracy, sensitivity, specificity, precision and F1 score of 93%, 93%, 97%, 93% and 93% respectively, with area under the curve (AUC) of precision recall (PR) of 0.97 on the test set. Additionally, the PIVs with corresponding respiration phase captured from each auscultation point generated an acoustic map of the patient's lung, which correlated with traditional lung radiographic findings.

PMID:40089574 | DOI:10.1038/s41598-025-93011-7

Sci Rep. 2025 Mar 15;15(1):9005. doi: 10.1038/s41598-025-88687-w.

ABSTRACT

Social media has become a powerful tool for public discourse, shaping opinions and the emotional landscape of communities. The extensive use of social media has led to a massive influx of online content. This content includes instances where negativity is amplified through hateful speech but also a significant number of posts that provide support and encouragement, commonly known as hope speech. In recent years, researchers have focused on the automatic detection of hope speech in languages such as Russian, English, Hindi, Spanish, and Bengali. However, to the best of our knowledge, detection of hope speech in Urdu and English, particularly using translation-based techniques, remains unexplored. To contribute to this area we have created a multilingual dataset in English and Urdu and applied a translation-based approach to handle multilingual challenges and utilized several state-of-the-art machine learning, deep learning, and transfer learning based methods to benchmark our dataset. Our observations indicate that a rigorous process for annotator selection, along with detailed annotation guidelines, significantly improved the quality of the dataset. Through extensive experimentation, our proposed methodology, based on the Bert transformer model, achieved benchmark performance, surpassing traditional machine learning models with accuracies of 87% for English and 79% for Urdu. These results show improvements of 8.75% in English and 1.87% in Urdu over baseline models (SVM 80% English and 78% in Urdu).

PMID:40089522 | DOI:10.1038/s41598-025-88687-w