Eur Radiol. 2025 Apr 16. doi: 10.1007/s00330-025-11546-9. Online ahead of print.

ABSTRACT

OBJECTIVES: To evaluate the performance of high-resolution deep learning-based hip MR imaging (CSAI) compared to standard-resolution compressed sense (CS) sequences using hip arthroscopy as standard of reference.

METHODS: Thirty-two patients (mean age, 37.5 years (± 11.7), 24 men) with femoroacetabular impingement syndrome underwent 3-T MR imaging prior to hip arthroscopy. Coronal and sagittal intermediate-weighted TSE sequences with fat saturation were obtained using CS (0.6 × 0.8 mm) and high-resolution CSAI (0.3 × 0.4 mm), with 3 mm slice thickness and similar acquisition times (3:55-4:12 min). MR scans were independently assessed by three radiologists and a hip arthroscopy specialist for labral and cartilage abnormalities. Sensitivity, specificity, and accuracy were calculated using arthroscopy as reference standard. Statistical comparisons between CS and CSAI were performed using McNemar's test.

RESULTS: Labral abnormality detection showed excellent sensitivity for radiologists (CS and CSAI: 97-100%) and the surgeon (CS: 81%, CSAI: 90%, p = 0.08), with 100% specificity. Overall cartilage lesion sensitivity was significantly higher with CSAI versus CS (42% vs. 37%, p < 0.001). Highest sensitivity was observed in superolateral acetabular cartilage (CS: 81%, CSAI: 88%, p < 0.001), while highest specificity was found for the anteroinferior acetabular cartilage (CS and CSAI: 99%). Sensitivity was lowest for the assessment of the anteroinferior and posterior acetabular zones, and inferior and posterior femoral zones (CS and CSAI < 6%).

CONCLUSION: CS and CSAI MR imaging showed excellent diagnostic performance for labral abnormalities. Despite CSAI's improved cartilage lesion detection, overall diagnostic performance for cartilage assessment remained suboptimal.

KEY POINTS: Question Accurate preoperative detection of labral and cartilage lesions in femoroacetabular impingement remains challenging, with current MRI protocols showing variable diagnostic performance. Findings High-resolution deep learning-based and standard-resolution compressed sense MRI demonstrate comparable diagnostic performance, with high accuracy for labral defects but limited sensitivity for cartilage lesions. Clinical relevance Current MRI protocols, regardless of resolution optimization, show persistent limitations in cartilage evaluation, indicating the need for further technical advancement to improve diagnostic confidence in presurgical planning.

PMID:40240555 | DOI:10.1007/s00330-025-11546-9

Comput Biol Med. 2025 Apr 15;191:110124. doi: 10.1016/j.compbiomed.2025.110124. Online ahead of print.

ABSTRACT

The increasing digitalization of multi-modal data in medicine and novel artificial intelligence (AI) algorithms opens up a large number of opportunities for predictive models. In particular, deep learning models show great performance in the medical field. A major limitation of such powerful but complex models originates from their 'black-box' nature. Recently, a variety of explainable AI (XAI) methods have been introduced to address this lack of transparency and trust in medical AI. However, the majority of such methods have solely been evaluated on single data modalities. Meanwhile, with the increasing number of XAI methods, integrative XAI frameworks and benchmarks are essential to compare their performance on different tasks. For that reason, we developed BenchXAI, a novel XAI benchmarking package supporting comprehensive evaluation of fifteen XAI methods, investigating their robustness, suitability, and limitations in biomedical data. We employed BenchXAI to validate these methods in three common biomedical tasks, namely clinical data, medical image and signal data, and biomolecular data. Our newly designed sample-wise normalization approach for post-hoc XAI methods enables the statistical evaluation and visualization of performance and robustness. We found that the XAI methods Integrated Gradients, DeepLift, DeepLiftShap, and GradientShap performed well over all three tasks, while methods like Deconvolution, Guided Backpropagation, and LRP-α1-β0 struggled for some tasks. With acts such as the EU AI Act the application of XAI in the biomedical domain becomes more and more essential. Our evaluation study represents a first step towards verifying the suitability of different XAI methods for various medical domains.

PMID:40239236 | DOI:10.1016/j.compbiomed.2025.110124

J Am Acad Orthop Surg Glob Res Rev. 2025 Apr 10;9(4). doi: 10.5435/JAAOSGlobal-D-24-00405. eCollection 2025 Apr 1.

ABSTRACT

BACKGROUND: A comprehensive review on the application of artificial intelligence (AI) within spine surgery as a specialty remains lacking.

METHODS: This scoping review was conducted upon PubMed and EMBASE databases according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines. Our analysis focused on publications from January 1, 2020, to March 31, 2024, with a specific focus on AI in the field of spine surgery. Review articles and articles predominantly concerning secondary validation of algorithms, medical physics, electronic devices, biomechanics, preclinical, and with a lack of clinical emphasis were excluded.

RESULTS: One hundred five studies were included after our inclusion/exclusion criteria were applied. Most studies (n = 100) were conducted through supervised learning upon prelabeled data sets. Overall, 38 studies used conventional machine learning methods upon predefined features, whereas 67 used deep learning methods, predominantly for medical image analyses. Only 25.7% of studies (27/105) collected data from more than 1,000 patients for model development and validation. Data originated from only a single center in 72 studies. The most common application was prognostication (38/105), followed by diagnosis (35/105), medical image processing (29/105), and surgical assistance (3/105).

CONCLUSION: The application of AI within the domain of spine surgery has significant potential to advance patient-specific diagnosis, management, and surgical execution.

PMID:40239218 | DOI:10.5435/JAAOSGlobal-D-24-00405

Invest Radiol. 2025 Apr 16. doi: 10.1097/RLI.0000000000001192. Online ahead of print.

ABSTRACT

The past decade has witnessed notable advancements in clinical neuroimaging facilitated by technological innovations and significant scientific discoveries. In conjunction with Investigative Radiology's 60th anniversary, this review examines key contributions from the past 10 years, emphasizing the journal's most accessed articles and their impact on clinical practice and research. Advances in imaging technologies, including photon-counting computed tomography, and innovations in low-field and high-field magnetic resonance imaging systems have expanded diagnostic capabilities. Progress in the development and translation of contrast media and rapid quantitative imaging techniques has further improved diagnostic accuracy. Additionally, the integration of advanced data analysis methods, particularly deep learning and medical informatics, has improved image interpretation and operational efficiency. Beyond technological developments, this review highlights basic neuroscience findings, such as the discovery and characterization of the glymphatic system. These insights have provided a deeper understanding of central nervous system physiology and pathology, bridging the gap between research and clinical applications. This review integrates these advancements to provide an overview of the progress and ongoing challenges in clinical neuroimaging, offering insights into its current state and potential future directions within the broader field of radiology.

PMID:40239043 | DOI:10.1097/RLI.0000000000001192

PLoS One. 2025 Apr 16;20(4):e0319329. doi: 10.1371/journal.pone.0319329. eCollection 2025.

ABSTRACT

Deep learning (DL) has become a powerful tool for the recognition and classification of biological sequences. However, conventional single-architecture models often struggle with suboptimal predictive performance and high computational costs. To address these challenges, we present EnsembleDL-Lipo, an innovative ensemble deep learning framework that combines Convolutional Neural Networks (CNNs) and Deep Neural Networks (DNNs) to enhance the identification of lipocalin sequences. Lipocalins are multifunctional extracellular proteins involved in various diseases and stress responses, and their low sequence similarity and occurrence in the 'twilight zone' of sequence alignment present significant hurdles for accurate classification. These challenges necessitate efficient computational methods to complement traditional, labor-intensive experimental approaches. EnsembleDL-Lipo overcomes these issues by leveraging a set of PSSM-based features to train a large ensemble of deep learning models. The framework integrates multiple feature representations derived from position-specific scoring matrices (PSSMs), optimizing classification performance across diverse sequence patterns. The model achieved superior results on the training dataset, with an accuracy (ACC) of 97.65%, recall of 97.10%, Matthews correlation coefficient (MCC) of 0.95, and area under the curve (AUC) of 0.99. Validation on an independent test set further confirmed the robustness of the model, yielding an ACC of 95.79%, recall of 90.48%, MCC of 0.92, and AUC of 0.97. These results demonstrate that EnsembleDL-Lipo is a highly effective and computationally efficient tool for lipocalin sequence identification, significantly outperforming existing methods and offering strong potential for applications in biomarker discovery.

PMID:40238838 | DOI:10.1371/journal.pone.0319329

PLoS One. 2025 Apr 16;20(4):e0320578. doi: 10.1371/journal.pone.0320578. eCollection 2025.

ABSTRACT

Existing research on decision-making of autonomous vehicles (AVs) has mainly focused on normal road sections, with limited exploration of decision-making in complex traffic environments without lane markings. Taking toll plaza diverging area as an example, this study proposes a lateral motion strategy for AVs based on deep reinforcement learning (DRL) algorithms. First, a microscopic simulation platform is developed to simulate the realistic diverging trajectories of human-driven vehicles (HVs), providing AVs with a high-fidelity training environment. Next, a DRL-based self-efficient lateral motion strategy for AVs is proposed, with state and reward functions tailored to the environmental features of the diverging area. Simulation results indicate that the strategy can significantly reduce the diverging time of single vehicles. In addition, considering the long-term coexistence of AVs and HVs, the study further explores how the varying penetration of AVs with self-efficient strategy impacts traffic flow in the diverging area. Findings reveal that a moderate increase in AV penetration can improve overall traffic efficiency and safety. But an excessive penetration of AVs with self-efficient strategy leads to intense competition for limited road resources, further deteriorating operational conditions in the diverging area.

PMID:40238783 | DOI:10.1371/journal.pone.0320578

PLoS One. 2025 Apr 16;20(4):e0318599. doi: 10.1371/journal.pone.0318599. eCollection 2025.

ABSTRACT

BACKGROUND: Quantitative muscle water T2 (T2w) mapping is regarded as a biomarker for disease activity and response to treatment in neuromuscular diseases (NMD). However, the implementation in clinical settings is limited due to long scanning times and low resolution. Using artificial intelligence (AI) to accelerate MR image acquisition offers a possible solution. Combining compressed sensing and parallel imaging with AI-based reconstruction, known as CSAI (SmartSpeed, Philips Healthcare), allows for the generation of high-quality, weighted MR images in a shorter scan time. However, CSAI has not yet been investigated for quantitative MRI. Therefore, in the present work we assessed the performance of CSAI acceleration for T2w mapping compared to standard acceleration with SENSE.

METHODS: T2w mapping of the thigh muscles, based on T2-prepared 3D TSE with SPAIR fat suppression, was performed using standard SENSE (acceleration factor of 2; 04:35 min; SENSE) and CSAI (acceleration factor of 5; 01:57 min; CSAI 5x) in ten patients with facioscapulohumeral muscular dystrophy (FSHD). Subjects were scanned in two consecutive sessions (14 days in between). In each dataset, six regions of interest were placed in three thigh muscles bilaterally. SENSE and CSAI 5x acceleration were compared for i) image quality using apparent signal- and contrast-to-noise ratio (aSNR/aCNR), ii) diagnostic agreement of T2w values, and iii) intra- and inter-session reproducibility.

RESULTS: aSNR and aCNR of SENSE and CSAI 5x scans were not significantly different (p > 0.05). An excellent agreement of SENSE and CSAI 5x T2w values was shown (r = 0.99; ICC = 0.992). T2w mapping with both acceleration methods showed excellent, matching intra-method reproducibility.

CONCLUSION: AI-based acceleration of CS data allows for scan time reduction of more than 50% for T2w mapping in the thigh muscles of NMD patients without compromising quantitative validity.

PMID:40238781 | DOI:10.1371/journal.pone.0318599

PLoS One. 2025 Apr 16;20(4):e0319968. doi: 10.1371/journal.pone.0319968. eCollection 2025.

ABSTRACT

Honey bees play a crucial role in natural ecosystems, mainly through their pollination services. Within a hive, they exhibit intricate social behaviors and communicate among thousands of individuals. Accurate detection and segmentation of honey bees are crucial for automated behavior analysis, as they significantly enhance object tracking and behavior recognition by yielding high-quality results. This study is specifically centered on the detection and segmentation of individual bees, particularly Apis cerana, within a hive environment, employing the Mask R-CNN deep learning model. We used transfer learning weights from our previously trained Apis mellifera model and explored data preprocessing techniques, such as brightness and contrast enhancement, to enhance model performance. Our proposed approach offers an optimal solution with a minimal dataset size and computational time while maintaining high model performance. Mean average precision (mAP) served as the evaluation metric for both detection and segmentation tasks. Our solution for A. cerana segmentation achieves the highest performance with a mAP of 0.728. Moreover, the number of training and validation sets was reduced by 85% compared to our previous study on the A. mellifera segmentation model.

PMID:40238729 | DOI:10.1371/journal.pone.0319968

IEEE Trans Ultrason Ferroelectr Freq Control. 2025 Apr 16;PP. doi: 10.1109/TUFFC.2025.3561599. Online ahead of print.

ABSTRACT

Shear wave elastography (SWE) provides quantitative markers for tissue characterization by measuring shear wave speed (SWS), which reflects tissue stiffness. SWE uses an acoustic radiation force pulse sequence to generate shear waves that propagate laterally through tissue with transient displacements. These waves travel perpendicular to the applied force, and their displacements are tracked using high-frame-rate ultrasound. Estimating the SWS map involves two main steps: speckle tracking and SWS estimation. Speckle tracking calculates particle velocity by measuring RF/IQ data displacement between adjacent firings, while SWS estimation methods typically compare particle velocity profiles of samples that are laterally a few millimeters apart. Deep learning (DL) methods have gained attention for SWS estimation, often relying on supervised training using simulated data. However, these methods may struggle with real-world data, which can differ significantly from simulated training data, potentially leading to artifacts in the estimated SWS map. To address this challenge, we propose a physics-inspired learning approach that utilizes real data without known SWS values. Our method employs an adaptive unsupervised loss function, allowing the network to train with real noisy data to minimize the artifacts and improve the robustness. We validate our approach using experimental phantom data and in vivo liver data from two human subjects, demonstrating enhanced accuracy and reliability in SWS estimation compared to conventional and supervised methods. This hybrid approach leverages the strengths of both data-driven and physics-inspired learning, offering a promising solution for more accurate and robust SWS mapping in clinical applications.

PMID:40238602 | DOI:10.1109/TUFFC.2025.3561599

IEEE Trans Pattern Anal Mach Intell. 2025 Apr 16;PP. doi: 10.1109/TPAMI.2025.3562051. Online ahead of print.

ABSTRACT

Although supervised deep normal estimators have recently shown impressive results on synthetic benchmarks, their performance deteriorates significantly in real-world scenarios due to the domain gap between synthetic and real data. Building high-quality real training data to boost those supervised methods is not trivial because point-wise annotation of normals for varying-scale real-world 3D scenes is a tedious and expensive task. This paper introduces PointNorm-Net, the first self-supervised deep learning framework to tackle this challenge. The key novelty of PointNorm-Net is a three-stage multi-modal normal distribution estimation paradigm that can be integrated into either deep or traditional optimization-based normal estimation frameworks. Extensive experiments show that our method achieves superior generalization and outperforms state-of-the-art conventional and deep learning approaches across three real-world datasets that exhibit distinct characteristics compared to the synthetic training data.

PMID:40238601 | DOI:10.1109/TPAMI.2025.3562051

Foods. 2025 Apr 4;14(7):1269. doi: 10.3390/foods14071269.

ABSTRACT

In addition to its flavor and nutritional value, the origin of kelp has become a crucial factor influencing consumer choices. Nevertheless, research on kelp's origin traceability by volatile organic compound (VOC) analysis is lacking, and the application of deep learning in this field remains scarce due to its black-box nature. To address this gap, we attempted to identify the origin of kelp by analyzing its VOCs in conjunction with explainable deep learning. In this work, we identified 115 distinct VOCs in kelp samples using gas chromatography coupled with ion mobility spectroscopy (GC-IMS), of which 68 categories were discernible. Consequently, we developed a comprehensible one-dimensional convolutional neural network (1D-CNN) model that incorporated 107 VOCs exhibiting significant regional disparities (p < 0.05). The model successfully discerns the origin of kelp, achieving perfect metrics across accuracy (100%), precision (100%), recall (100%), F1 score (100%), and AUC (1.0). SHapley Additive exPlanations (SHAP) analysis highlighted the impact of features such as 1-Octen-3-ol-M, (+)-limonene, allyl sulfide-D, 1-hydroxy-2-propanone-D, and (E)-2-hexen-1-al-M on the model output. This research provides deeper insights into how critical product features correlate with specific geographic information, which in turn boosts consumer trust and promotes practical utilization in actual settings.

PMID:40238501 | DOI:10.3390/foods14071269

Dentomaxillofac Radiol. 2025 Apr 16:twaf031. doi: 10.1093/dmfr/twaf031. Online ahead of print.

ABSTRACT

OBJECTIVE: This study explores the application of artificial intelligence (AI), specifically deep learning, in the detection and classification of mandibular fractures using CT scans.

MATERIALS AND METHODS: Data from 459 patients were retrospectively obtained from West China Hospital of Stomatology, Sichuan University, spanning from 2020 to 2023. The CT scans were divided into training, testing, and independent validation sets. This research focuses on training and validating a deep learning model using the nnU-Net segmentation framework for pixel-level accuracy in identifying fracture locations. Additionally, a 3D-ResNet with pre-trained weights was employed to classify fractures into three types based on severity. Performance metrics included sensitivity, precision, specificity, and area under the receiver operating characteristic curve (AUC).

RESULTS: The study achieved high diagnostic accuracy in mandibule fracture detection, with sensitivity>0.93, precision>0.79, and specificity>0.80. For mandibular fracture classification, accuracies were all above 0.718, with a mean AUC of 0.86.

CONCLUSION: Detection and classification of mandibular fractures in CT images can be significantly enhanced using the nnU-Net segmentation framework, aiding in clinical diagnosis.

PMID:40238181 | DOI:10.1093/dmfr/twaf031

JAMA Cardiol. 2025 Apr 16. doi: 10.1001/jamacardio.2025.0498. Online ahead of print.

ABSTRACT

IMPORTANCE: Accurate assessment of tricuspid regurgitation (TR) is necessary for identification and risk stratification.

OBJECTIVE: To design a deep learning computer vision workflow for identifying color Doppler echocardiogram videos and characterizing TR severity.

DESIGN, SETTING, AND PARTICIPANTS: An automated deep learning workflow was developed using 47 312 studies (2 079 898 videos) from Cedars-Sinai Medical Center (CSMC) between 2011 and 2021. Data analysis was performed in 2024. The pipeline was tested on a temporally distinct test set of 2462 studies (108 138 videos) obtained in 2022 at CSMC and a geographically distinct cohort of 5549 studies (278 377 videos) from Stanford Healthcare (SHC). Training and validation cohorts contained data from 31 708 patients at CSMC receiving care between 2011 and 2021. Patients were chosen for parity across TR severity classes, with no exclusion criteria based on other clinical or demographic characteristics. The 2022 CSMC test cohort and SHC test cohorts contained studies from 2170 patients and 5014 patients, respectively.

EXPOSURE: Deep learning computer vision model.

MAIN OUTCOMES AND MEASURES: The main outcomes were area under the receiver operating characteristic curve (AUC), sensitivity, and specificity in identifying apical 4-chamber (A4C) videos with color Doppler across the tricuspid valve and AUC in identifying studies with moderate to severe or severe TR.

RESULTS: In the CSMC test dataset, the view classifier demonstrated an AUC of 1.000 (95% CI, 0.999-1.000) and identified at least 1 A4C video with color Doppler across the tricuspid valve in 2410 of 2462 studies with a sensitivity of 0.975 (95% CI, 0.968-0.982) and a specificity of 1.000 (95% CI, 1.000-1.000). In the CSMC test cohort, moderate or severe TR was detected with an AUC of 0.928 (95% CI, 0.913-0.943), and severe TR was detected with an AUC of 0.956 (95% CI, 0.940-0.969). In the SHC cohort, the view classifier correctly identified at least 1 TR color Doppler video in 5268 of the 5549 studies, resulting in an AUC of 0.999 (95% CI, 0.998-0.999), a sensitivity of 0.949 (95% CI, 0.944-0.955), and a specificity of 0.999 (95% CI, 0.999-0.999). The artificial intelligence model detected moderate or severe TR with an AUC of 0.951 (95% CI, 0.938-0.962) and severe TR with an AUC of 0.980 (95% CI, 0.966-0.988).

CONCLUSIONS AND RELEVANCE: In this study, an automated pipeline was developed to identify clinically significant TR with excellent performance. With open-source code and weights, this project can serve as the foundation for future prospective evaluation of artificial intelligence-assisted workflows in echocardiography.

PMID:40238103 | DOI:10.1001/jamacardio.2025.0498

EJNMMI Phys. 2025 Apr 16;12(1):38. doi: 10.1186/s40658-025-00745-4.

ABSTRACT

BACKGROUND: A reduction of dose and/or acquisition duration of PET examinations is desirable in terms of radiation protection, patient comfort and throughput, but leads to decreased image quality due to poorer image statistics. Recently, different deep-learning based methods have been proposed to improve image quality of low-count PET images. For example, one such approach allows the generation of AI-enhanced PET images (AI-PET) based on ultra-low count PET/CT scans. The performance of this algorithm has so far only been clinically evaluated on patient data featuring limited scan statistics and unknown actual activity concentration. Therefore, this study investigates the performance of this deep-learning algorithm using PET measurements of a phantom resembling different lesion sizes and count statistics (from ultra-low to high) to understand the capabilities and limitations of AI-based post processing for improved image quality in ultra-low count PET imaging.

METHODS: A previously trained pix2pixHD Generative Adversarial Network was evaluated. To this end, a NEMA PET body phantom filled with two sphere-to-background activity concentration ratios (4:1 and 10:1) and two attenuation scenarios to investigate the effects of obese patients was scanned in list mode. Images were reconstructed with 13 different acquisition durations ranging from 5 s up to 900 s. Image noise, recovery coefficients, SUV-differences, image quality measurement metrics such as the Structural Similarity Index Metric, and the contrast-to-noise-ratio were assessed. In addition, the benefits of the deep-learning network over Gaussian smoothing were investigated.

RESULTS: The presented AI-algorithm is very well suitable for denoising ultra-low count PET images and for restoring structural information, but increases image noise in ultra-high count PET scans. The generated AI-PET scans strongly underestimate SUV especially in small lesions with a diameter ≤ 17 mm, while quantitative measures of large lesions ≥ 37 mm in diameter were accurately recovered. In ultra-low count or low contrast images, the AI algorithm might not be able to recognize small lesions ≤ 13 mm in diameter. In comparison to standardized image post-processing using a Gaussian filter, the deep-learning network is better suited to improve image quality, but at the same time degrades SUV accuracy to a greater extent than post-filtering and quantitative SUV accuracy varies for different lesion sizes.

CONCLUSIONS: Phantom-based validation of AI-based algorithms allows for a detailed assessment of the performance, limitations, and generalizability of deep-learning based algorithms for PET image enhancement. Here it was confirmed that the AI-based approach performs very well in denoising ultra-low count PET images and outperforms traditional Gaussian post-filtering. However, there are strong limitations in terms of quantitative accuracy and detectability of small lesions.

PMID:40237913 | DOI:10.1186/s40658-025-00745-4

Discov Oncol. 2025 Apr 16;16(1):537. doi: 10.1007/s12672-025-02329-1.

ABSTRACT

BACKGROUND: Breast cancer (BC) remains a leading cause of cancer-related mortality among women globally, with increasing incidence rates posing significant public health challenges. Recent advancements in artificial intelligence (AI) have revolutionized medical imaging, particularly in enhancing diagnostic accuracy and prognostic capabilities for BC. While multimodal imaging combined with AI has shown remarkable potential, a comprehensive analysis is needed to synthesize current research and identify emerging trends and hotspots in AI-assisted multimodal imaging for BC.

METHODS: This study analyzed literature on AI-assisted multimodal imaging in BC from January 2010 to November 2024 in Web of Science Core Collection (WoSCC). Bibliometric and visualization tools, including VOSviewer, CiteSpace, and the Bibliometrix R package, were employed to assess countries, institutions, authors, journals, and keywords.

RESULTS: A total of 80 publications were included, revealing a steady increase in annual publications and citations, with a notable surge post-2021. China led in productivity and citations, while Germany exhibited the highest citation average. The United States demonstrated the strongest international collaboration. The most productive institution and author are Radboud University Nijmegen and Xi, Xiaoming. Publications were predominantly published in Computerized Medical Imaging and Graphics, with Qian, XJ's 2021 study on BC risk prediction under deep learning frameworks being the most influential. Keyword analysis highlighted themes such as "breast cancer", "classification", and "deep learning".

CONCLUSIONS: AI-assisted multimodal imaging has significantly advanced BC diagnosis and management, with promising future developments. This study offers researchers a comprehensive overview of current frameworks and emerging research directions. Future efforts are expected to focus on improving diagnostic precision and refining therapeutic strategies through optimized imaging techniques and AI algorithms, emphasizing international collaboration to drive innovation and clinical translation.

PMID:40237900 | DOI:10.1007/s12672-025-02329-1

Radiol Artif Intell. 2025 Apr 16:e240097. doi: 10.1148/ryai.240097. Online ahead of print.

ABSTRACT

"Just Accepted" papers have undergone full peer review and have been accepted for publication in Radiology: Artificial Intelligence. This article will undergo copyediting, layout, and proof review before it is published in its final version. Please note that during production of the final copyedited article, errors may be discovered which could affect the content. Purpose To present and evaluate Dafne (deep anatomic federated network), a freely available decentralized, collaborative deep learning system for the semantic segmentation of radiologic images through federated incremental learning. Materials and Methods Dafne is free software with a client-server architecture. The client side is an advanced user interface that applies the deep learning models stored on the server to the user's data and allows the user to check and refine the prediction. Incremental learning is then performed at the client's side and sent back to the server, where it is integrated into the root model. Dafne was evaluated locally, by assessing the performance gain across model generations on 38 MRI datasets of the lower legs, and through the analysis of real-world usage statistics (n = 639 use-cases). Results Dafne demonstrated a statistically improvement in the accuracy of semantic segmentation over time (average increase of the Dice Similarity Coefficient by 0.007 points/generation on the local validation set, P < .001). Qualitatively, the models showed enhanced performance on various radiologic image types, including those not present in the initial training sets, indicating good model generalizability. Conclusion Dafne showed improvement in segmentation quality over time, demonstrating potential for learning and generalization. ©RSNA, 2025.

PMID:40237599 | DOI:10.1148/ryai.240097

Radiol Artif Intell. 2025 May;7(3):e250125. doi: 10.1148/ryai.250125.

NO ABSTRACT

PMID:40237597 | DOI:10.1148/ryai.250125

J Agric Food Chem. 2025 Apr 16. doi: 10.1021/acs.jafc.4c11492. Online ahead of print.

ABSTRACT

Quality inspection of fruits and vegetables linked to food safety monitoring and quality control. Traditional chemical analysis and physical measurement techniques are reliable, they are also time-consuming, costly, and susceptible to environmental and sample changes. Hyperspectral imaging technology combined with deep learning methods can effectively overcome these problems. Compared with human evaluation, automated inspection improves inspection efficiency, reduces subjective error, and promotes the intelligent and precise fruit and vegetable quality inspection. This paper reviews reports on the application of hyperspectral imaging technology combined to deep learning methods in various aspects of fruits and vegetables quality assessment. In addition, the latest applications of these technologies in the fields of fruit and vegetable safety, internal quality, and external quality inspection are reviewed, and the challenges and future development directions of hyperspectral imaging technology combined with deep learning in this field are prospected. Hyperspectral imaging combined with deep learning has shown significant advantages in fruit and vegetable quality inspection, especially in improving inspection accuracy and efficiency. Future research should focus on reducing costs, optimizing equipment, personalizing feature extraction, and model generalizability. In addition, the development of lightweight models and the balance of accuracy, the enhancement of the database and the importance of quantitative research should also be brought to attention. These efforts will promote the wide application of hyperspectral imaging technology in fruit and vegetable inspection, improve its practicability in the actual production environment, and bring important progress for food safety and quality management.

PMID:40237548 | DOI:10.1021/acs.jafc.4c11492

mSystems. 2025 Apr 16:e0145524. doi: 10.1128/msystems.01455-24. Online ahead of print.

ABSTRACT

Deep learning is revolutionizing biomedical research by facilitating the integration of multi-omics data sets while bridging classical bioinformatics with existing knowledge. Building on this powerful potential, Zhang et al. proposed a semi-supervised learning framework called Autoencoder-Based Subtypes Detector for Cancer (ASD-cancer) to improve the multi-omics data analysis (H. Zhang, X. Xiong, M. Cheng, et al., 2024, mSystems 9:e01395-24, https://doi.org/10.1128/msystems.01395-24). By utilizing autoencoders pre-trained on The Cancer Genome Atlas data, the ASD-cancer framework outperforms the baseline model. This approach also makes the framework scalable, enabling it to process new data sets through transfer learning without retraining. This commentary explores the methodological innovations and scalability of ASD-cancer while suggesting future directions, such as the incorporation of additional data layers and the development of adaptive AI models through continuous learning. Notably, integrating large language models into ASD-cancer could enhance its interpretability, providing more profound insights into oncological research and increasing its influence in cancer subtyping and further analysis.

PMID:40237527 | DOI:10.1128/msystems.01455-24

Clin Nucl Med. 2025 Apr 16. doi: 10.1097/RLU.0000000000005898. Online ahead of print.

ABSTRACT

PURPOSE: Convolutional neural networks (CNNs) have been studied for detecting bone metastases on bone scans; however, the application of ConvNeXt and transformer models has not yet been explored. This study aims to evaluate the performance of various deep learning models, including the ConvNeXt and transformer models, in diagnosing metastatic lesions from bone scans.

MATERIALS AND METHODS: We retrospectively analyzed bone scans from patients with cancer obtained at 2 institutions: the training and validation sets (n=4626) were from Hospital 1 and the test set (n=1428) was from Hospital 2. The deep learning models evaluated included ResNet18, the Data-Efficient Image Transformer (DeiT), the Vision Transformer (ViT Large 16), the Swin Transformer (Swin Base), and ConvNeXt Large. Gradient-weighted class activation mapping (Grad-CAM) was used for visualization.

RESULTS: Both the validation set and the test set demonstrated that the ConvNeXt large model (0.969 and 0.885, respectively) exhibited the best performance, followed by the Swin Base model (0.965 and 0.840, respectively), both of which significantly outperformed ResNet (0.892 and 0.725, respectively). Subgroup analyses revealed that all the models demonstrated greater diagnostic accuracy for patients with polymetastasis compared with those with oligometastasis. Grad-CAM visualization revealed that the ConvNeXt Large model focused more on identifying local lesions, whereas the Swin Base model focused on global areas such as the axial skeleton and pelvis.

CONCLUSIONS: Compared with traditional CNN and transformer models, the ConvNeXt model demonstrated superior diagnostic performance in detecting bone metastases from bone scans, especially in cases of polymetastasis, suggesting its potential in medical image analysis.

PMID:40237349 | DOI:10.1097/RLU.0000000000005898