Front Med (Lausanne). 2025 Apr 16;12:1555907. doi: 10.3389/fmed.2025.1555907. eCollection 2025.

ABSTRACT

BACKGROUND: Recent advances in machine learning are transforming medical image analysis, particularly in cancer detection and classification. Techniques such as deep learning, especially convolutional neural networks (CNNs) and vision transformers (ViTs), are now enabling the precise analysis of complex histopathological images, automating detection, and enhancing classification accuracy across various cancer types. This study focuses on osteosarcoma (OS), the most common bone cancer in children and adolescents, which affects the long bones of the arms and legs. Early and accurate detection of OS is essential for improving patient outcomes and reducing mortality. However, the increasing prevalence of cancer and the demand for personalized treatments create challenges in achieving precise diagnoses and customized therapies.

METHODS: We propose a novel hybrid model that combines convolutional neural networks (CNN) and vision transformers (ViT) to improve diagnostic accuracy for OS using hematoxylin and eosin (H&E) stained histopathological images. The CNN model extracts local features, while the ViT captures global patterns from histopathological images. These features are combined and classified using a Multi-Layer Perceptron (MLP) into four categories: non-tumor (NT), non-viable tumor (NVT), viable tumor (VT), and non-viable ratio (NVR).

RESULTS: Using the Cancer Imaging Archive (TCIA) dataset, the model achieved an accuracy of 99.08%, precision of 99.10%, recall of 99.28%, and an F1-score of 99.23%. This is the first successful four-class classification using this dataset, setting a new benchmark in OS research and offering promising potential for future diagnostic advancements.

PMID:40313555 | PMC:PMC12045028 | DOI:10.3389/fmed.2025.1555907

Natl Sci Rev. 2025 Jan 28;12(5):nwaf030. doi: 10.1093/nsr/nwaf030. eCollection 2025 May.

ABSTRACT

Unraveling complex cell-type-composition and gene-expression patterns at the cellular spatial resolution is crucial for understanding intricate cell functions in the brain. In this study, we developed Deep Neural Network-based Spatial Cell Typing (DSCT)-an innovative framework for spatial cell typing within spatial transcriptomic data sets. This approach utilizes a synergistic integration of an enhanced gene-selection strategy and a lightweight deep neural network for data training, offering a more rapid and accurate solution for the analysis of spatial transcriptomic data. Based on comprehensive analysis, DSCT achieved exceptional accuracy in cell-type identification across various brain regions, species and spatial transcriptomic platforms. It also performed well in mapping finer cell types, thereby showcasing its versatility and adaptability across diverse data sets. Strikingly, DSCT exhibited high efficiency and remarkable processing speed, with fewer computational resource demands. As such, this novel approach opens new avenues for exploring the spatial organization of cell types and gene-expression patterns, advancing our understanding of biological functions and pathologies within the nervous system.

PMID:40313458 | PMC:PMC12045154 | DOI:10.1093/nsr/nwaf030

Front Neurorobot. 2025 Apr 17;19:1582995. doi: 10.3389/fnbot.2025.1582995. eCollection 2025.

ABSTRACT

INTRODUCTION: Understanding human actions in complex environments is crucial for advancing applications in areas such as surveillance, robotics, and autonomous systems. Identifying actions from UAV-recorded videos becomes more challenging as the task presents unique challenges, including motion blur, dynamic background, lighting variations, and varying viewpoints. The presented work develops a deep learning system that recognizes multi-person behaviors from data gathered by UAVs. The proposed system provides higher recognition accuracy while maintaining robustness along with dynamic environmental adaptability through the integration of different features and neural network models. The study supports the wider development of neural network systems utilized in complicated contexts while creating intelligent UAV applications utilizing neural networks.

METHOD: The proposed study uses deep learning and feature extraction approaches to create a novel method to recognize various actions in UAV-recorded video. The proposed model improves identification capacities and system robustness by addressing motion dynamic problems and intricate environmental constraints, encouraging advancements in UAV-based neural network systems.

RESULTS: We proposed a deep learning-based framework with feature extraction approaches that may effectively increase the accuracy and robustness of multi-person action recognition in the challenging scenarios. Compared to the existing approaches, our system achieved 91.50% on MOD20 dataset and 89.71% on Okutama-Action. These results do, in fact, show how useful neural network-based methods are for managing the limitations of UAV-based application.

DISCUSSION: Results how that the proposed framework is indeed effective at multi-person action recognition under difficult UAV conditions.

PMID:40313416 | PMC:PMC12043872 | DOI:10.3389/fnbot.2025.1582995

Neural Regen Res. 2025 Apr 29. doi: 10.4103/NRR.NRR-D-24-00794. Online ahead of print.

ABSTRACT

Retinal ganglion cells, a crucial component of the central nervous system, are often affected by irreversible visual impairment due to various conditions, including trauma, tumors, ischemia, and glaucoma. Studies have shown that the optic nerve crush model and glaucoma model are commonly used to study retinal ganglion cell injury. While these models differ in their mechanisms, both ultimately result in retinal ganglion cell injury. With advancements in high-throughput technologies, techniques such as microarray analysis, RNA sequencing, and single-cell RNA sequencing have been widely applied to characterize the transcriptomic profiles of retinal ganglion cell injury, revealing underlying molecular mechanisms. This review focuses on optic nerve crush and glaucoma models, elucidating the mechanisms of optic nerve injury and neuron degeneration induced by glaucoma through single-cell transcriptomics, transcriptome analysis, and chip analysis. Research using the optic nerve crush model has shown that different retinal ganglion cell subtypes exhibit varying survival and regenerative capacities following injury. Single-cell RNA sequencing has identified multiple genes associated with retinal ganglion cell protection and regeneration, such as Gal, Ucn, and Anxa2. In glaucoma models, high-throughput sequencing has revealed transcriptomic changes in retinal ganglion cells under elevated intraocular pressure, identifying genes related to immune response, oxidative stress, and apoptosis. These genes are significantly upregulated early after optic nerve injury and may play key roles in neuroprotection and axon regeneration. Additionally, CRISPR-Cas9 screening and ATAC-seq analysis have identified key transcription factors that regulate retinal ganglion cell survival and axon regeneration, offering new potential targets for neurorepair strategies in glaucoma. In summary, single-cell transcriptomic technologies provide unprecedented insights into the molecular mechanisms underlying optic nerve injury, aiding in the identification of novel therapeutic targets. Future researchers should integrate advanced single-cell sequencing with multi-omics approaches to investigate cell-specific responses in retinal ganglion cell injury and regeneration. Furthermore, computational models and systems biology methods could help predict molecular pathways interactions, providing valuable guidance for clinical research on optic nerve regeneration and repair.

PMID:40313107 | DOI:10.4103/NRR.NRR-D-24-00794

Int J Med Inform. 2025 Sep;201:105942. doi: 10.1016/j.ijmedinf.2025.105942. Epub 2025 Apr 21.

ABSTRACT

OBJECTIVE: Dravet Syndrome (DS) is a developmental and epileptic encephalopathy that is characterized by severe, prolonged motor seizures and high resistance to multiple antiseizure medications (ASMs) with multiple comorbidities. Evaluating the efficacy of new drugs in DS preclinical models and mapping them to human phenotypes of DS through analysis of published literature is an important goal for improving outcomes in this rare pediatric epilepsy.

MATERIALS AND METHODS: Large language models (LLM) have demonstrated great promise in parsing published literature; however, the performance of LLMs falls short in medical applications. In this study, we investigate the effectiveness of domain ontology developed by human experts to optimize LLMs for medical text processing in a rare disease. Utilizing a benchmark dataset that describes the efficacy of 17 ASMs tested in preclinical models and DS patients, we define a new ontology-augmented phased in-context learning (PCL) approach to process 4935 full-text DS articles. We expand this analysis to 7 new drugs that demonstrate efficacy in reducing seizures to identify gaps in current knowledge for designing new experimental studies for drug discovery in DS.

RESULTS: Few-shot or in-context learning is a foundational capability of LLMs and the few-shot learning capability of the Gemini 1.0 Pro version LLM dramatically increases when we augment prompts with the DS epilepsy ontology. The DS epilepsy ontology is the largest epilepsy and seizure ontology in clinical use that was developed by DS basic scientists and clinical neurologists. The ontology-augmented PCL prompt achieves 100% accuracy in reproducing the benchmark drug efficacy dataset for 17 ASMs with only two examples for in-context learning.

CONCLUSION: The new ontology-augmented PCL approach significantly accelerates the few-shot learning capabilities of the Gemini LLM, thereby reducing the number of required examples and time needed to optimize LLMs for medical applications.

PMID:40311258 | DOI:10.1016/j.ijmedinf.2025.105942

BMC Geriatr. 2025 May 1;25(1):300. doi: 10.1186/s12877-025-05965-y.

ABSTRACT

BACKGROUND: Drug-drug interactions (DDIs) are significant causes of adverse drug reactions among patients with cancer. We aimed to identify the prevalence, severity, and predictors of potential DDIs among geriatric oncology patients.

METHODS: A cross-sectional, retrospective study was conducted at two tertiary medical centers. Geriatric patients (≥ 65 years) who were diagnosed with solid tumors and received outpatient prescriptions with a minimum of two drugs between January 2018 and December 2022 were included in the study. Patients' medications were screened for DDIs using Lexi-Interact. Univariate and multivariable logistic regression models were used to explore factors associated with DDIs.

RESULTS: The study included 247 geriatric patients with a mean age of 74.0 ± 7.3 years, and 48.6% of the patients were female. The most common type of cancer was gastrointestinal cancer (35.6%), followed by genitourinary cancer (20.6%), and 50.6% of the patients had metastasized tumors. Approximately one-half of the patients (49.0%) received anticancer therapy, and hormonal therapy (21.9%) or chemotherapy (16.6%) was the most common therapy. The mean number of medications used per patient was 6.9 ± 3.5. The majority of patients (79.4%) had at least one DDI, with a mean of 5.6 ± 5.3 DDIs per patient. Most of the interactions were classified as moderate (58.9%), and only 19.3% were classified as major. Multiple logistic regression revealed that females were more vulnerable to DDIs than their male counterparts were (adjusted odds ratio (AOR) = 37.4; 95% CI 4.13-338.3). The number of medications used was significantly associated with the risk of DDIs (AOR = 4.07; 95% CI 2.53-6.54). Compared with patients with gastrointestinal cancers, patients with breast or gynecologic cancers had lower odds of experiencing DDI (AOR = 0.02; 95% CI < 0.01-0.24 and AOR = 0.04; 95% CI < 0.01-0.29, respectively).

CONCLUSION: This study revealed a high prevalence of DDIs among geriatric oncology patients, with most interactions classified as moderate. Female patients and patients taking multiple medications had a greater risk of experiencing DDIs. Routine screening for potential DDIs is essential for this vulnerable population, and the factors identified in this study should be carefully considered.

PMID:40312689 | DOI:10.1186/s12877-025-05965-y

Trials. 2025 May 1;26(1):145. doi: 10.1186/s13063-024-08671-z.

ABSTRACT

BACKGROUND: We define clinical study reports (CSRs) as standardized full reports of the protocols, results, and other pertinent details of clinical studies that are typically submitted by pharmaceutical companies to regulatory authorities when they apply for marketing authorization.

METHODS: In this systematic review we searched various databases (Clarivate Web of Science, EMBASE and Ovid Medline, Google Scholar, and PubMed) for publications containing the term "clinical study report/s", without restrictions.

THEMATIC SYNTHESIS: In the first part of this review we discussed the history of CSRs, their contents and structure, definitions, and relevant terminology. In this second part we discuss the uses of CSRs, concentrating on the individual benefits and harms of pharmacological interventions, and thus the benefit to harm balance. We also discuss adherence to interventions, prepublication of protocols of clinical trials, and how CSRs are written, factors that can all affect estimation of the benefit-harm balance.

CONCLUSIONS: When clinical trial data from CSRs are compared with the data in published trial reports, the apparent benefits of pharmacological interventions are less impressive, and more information emerges about harms they can cause. Both of these effects change how the benefit-harm balance of a pharmacological intervention is estimated, generally making it less favourable than was otherwise thought. For more accurate assessment of the benefit-harm balance of an intervention, full, not abbreviated or synoptic, clinical study reports should continue to be made publicly available by regulatory authorities and manufacturers. Authorities that do not currently make them available should do so. CSRs should be introduced for assessment of surgical operations, therapeutic devices, and other non-pharmacological interventions in clinical trials.

PMID:40312342 | DOI:10.1186/s13063-024-08671-z

Annu Rev Biomed Data Sci. 2025 May 1. doi: 10.1146/annurev-biodatasci-103123-094521. Online ahead of print.

ABSTRACT

Skin cancer is the most common cancer in the United States, with incidence rates continuing to rise both nationally and globally, posing significant health and economic burdens. These challenges are compounded by shortages in dermatological care and barriers to insurance access. To address these gaps, artificial intelligence (AI) and deep learning technologies offer promising solutions, enhancing skin cancer screening and diagnosis. AI has the potential to improve diagnostic accuracy and expand access to care, but significant challenges restrict its deployment. These challenges include clinical validation, algorithmic bias, regulatory oversight, and patient acceptance. Ethical concerns, such as disparities in access and fairness of AI algorithms, also require attention. In this review, we explore these limitations and outline future directions, including advancements in teledermatology and vision-language models (VLMs). Future research should focus on improving VLM reliability and interpretability and developing systems capable of integrating clinical context with dermatological images in a way that assists, rather than replaces, clinicians in making more accurate, timely diagnoses.

PMID:40312261 | DOI:10.1146/annurev-biodatasci-103123-094521

J Voice. 2025 Apr 30:S0892-1997(25)00158-4. doi: 10.1016/j.jvoice.2025.04.010. Online ahead of print.

ABSTRACT

Artificial Intelligence (AI) is transforming speech-language pathology (SLP) and dysphagia management, offering innovative solutions for assessment, diagnosis, and rehabilitation. This narrative review examines AI applications in these fields from 2014 to 2024, with particular focus on implementation challenges in Latin America. We analyze key AI technologies-including deep learning, machine learning algorithms, and natural language processing-that have demonstrated high accuracy in detecting voice disorders, analyzing swallowing function, and supporting personalized rehabilitation. The review identifies three primary domains of AI application: diagnostic tools with improved sensitivity for speech-language disorders, rehabilitation technologies that enable customized therapy, and telehealth platforms that expand access to specialized care in underserved regions. However, significant barriers persist, particularly in Latin America, where limited infrastructure, insufficient linguistic adaptation, and scarce regional datasets hamper widespread implementation. Our pilot study evaluating commercially available large language models for rehabilitation planning demonstrates their potential utility in generating structured therapy activities, especially in resource-constrained settings. While AI shows promise in enhancing clinical workflows and expanding service delivery, the evidence suggests that current applications remain predominantly focused on diagnosis rather than integrated rehabilitation. This review highlights the need for culturally and linguistically adapted AI models, expanded regional research collaborations, and regulatory frameworks that ensure ethical AI integration into SLP and dysphagia care, positioning these technologies as complementary tools that enhance rather than replace clinical expertise.

PMID:40312192 | DOI:10.1016/j.jvoice.2025.04.010

Lancet Digit Health. 2025 Apr 30:100868. doi: 10.1016/j.landig.2025.02.008. Online ahead of print.

ABSTRACT

BACKGROUND: Improving the accessibility of screening diabetic kidney disease (DKD) and differentiating isolated diabetic nephropathy from non-diabetic kidney disease (NDKD) are two major challenges in the field of diabetes care. We aimed to develop and validate an artificial intelligence (AI) deep learning system to detect DKD and isolated diabetic nephropathy from retinal fundus images.

METHODS: In this population-based study, we developed a retinal image-based AI-deep learning system, DeepDKD, pretrained using 734 084 retinal fundus images. First, for DKD detection, we used 486 312 retinal images from 121 578 participants in the Shanghai Integrated Diabetes Prevention and Care System for development and internal validation, and ten multi-ethnic datasets from China, Singapore, Malaysia, Australia, and the UK (65 406 participants) for external validation. Second, to differentiate isolated diabetic nephropathy from NDKD, we used 1068 retinal images from 267 participants for development and internal validation, and three multi-ethnic datasets from China, Malaysia, and the UK (244 participants) for external validation. Finally, we conducted two proof-of-concept studies: a prospective real-world study with 3 months' follow-up to evaluate the effectiveness of DeepDKD in screening DKD; and a longitudinal analysis of the effectiveness of DeepDKD in differentiating isolated diabetic nephropathy from NDKD on renal function changes with 4·6 years' follow-up.

FINDINGS: For detecting DKD, DeepDKD achieved an area under the receiver operating characteristic curve (AUC) of 0·842 (95% CI 0·838-0·846) on the internal validation dataset and AUCs of 0·791-0·826 across external validation datasets. For differentiating isolated diabetic nephropathy from NDKD, DeepDKD achieved an AUC of 0·906 (0·825-0·966) on the internal validation dataset and AUCs of 0·733-0·844 across external validation datasets. In the prospective study, compared with the metadata model, DeepDKD could detect DKD with higher sensitivity (89·8% vs 66·3%, p<0·0001). In the longitudinal study, participants with isolated diabetic nephropathy and participants with NDKD identified by DeepDKD had a significant difference in renal function outcomes (proportion of estimated glomerular filtration rate decline: 27·45% vs 52·56%, p=0·0010).

INTERPRETATION: Among diverse multi-ethnic populations with diabetes, a retinal image-based AI-deep learning system showed its potential for detecting DKD and differentiating isolated diabetic nephropathy from NDKD in clinical practice.

FUNDING: National Key R & D Program of China, National Natural Science Foundation of China, Beijing Natural Science Foundation, Shanghai Municipal Key Clinical Specialty, Shanghai Research Centre for Endocrine and Metabolic Diseases, Innovative research team of high-level local universities in Shanghai, Noncommunicable Chronic Diseases-National Science and Technology Major Project, Clinical Special Program of Shanghai Municipal Health Commission, and the three-year action plan to strengthen the construction of public health system in Shanghai.

PMID:40312169 | DOI:10.1016/j.landig.2025.02.008

Gastroenterology. 2025 Apr 29:S0016-5085(25)00648-1. doi: 10.1053/j.gastro.2025.04.011. Online ahead of print.

ABSTRACT

This review explores the transformative potential of artificial intelligence in the diagnosis and management of biliopancreatic disorders. By leveraging cutting-edge techniques such as deep learning and convolutional neural networks, artificial intelligence has significantly advanced gastroenterology, particularly in endoscopic procedures such as colonoscopy, upper endoscopy, and capsule endoscopy. These applications enhance adenoma detection rates, and improve lesion characterization and diagnostic accuracy. Artificial intelligence's integration in cross-sectional imaging modalities, such as computed tomography and magnetic resonance imaging, has shown remarkable potential. Models have demonstrated high accuracy in identifying pancreatic ductal adenocarcinoma, pancreatic cystic lesions, and pancreatic neuroendocrine tumors, aiding in early diagnosis, resectability assessment, and personalized treatment planning. In advanced endoscopic procedures, like digital single-operator cholangioscopy and endoscopic ultrasound, artificial intelligence enhances anatomical recognition, improves lesion classification, with potential reduction in procedural variability, enabling more consistent diagnositc and therapeutic outcomes. Promising applications in biliopancreatic endoscopy include the detection of biliary stenosis, classification of dysplastic precursor lesions, and assessment of pancreatic abnormalities. This review aims to capture the current state of artificial intelligence application in biliopancreatic disorders, summarizing the results of early studies, and paving the path for future directions.

PMID:40311821 | DOI:10.1053/j.gastro.2025.04.011

Am J Pathol. 2025 Apr 29:S0002-9440(25)00143-9. doi: 10.1016/j.ajpath.2025.04.004. Online ahead of print.

ABSTRACT

Conventional ascitic fluid cytology for detecting ovarian cancer is limited by its low sensitivity. To address this issue, our multicenter study developed patch image (PIs)-based fully supervised convolutional neural network (CNN) models and clustering-constrained attention multiple-instance learning (CLAM) algorithms for detecting ovarian cancer using ascitic fluid cytology. We collected 356 benign and 147 cancer whole-slide images (WSIs), from which 14,699 benign and 8,025 cancer PIs were extracted. Additionally 131 WSIs (44 benign and 87 cancer) were used for external validation. Six CNN algorithms were developed for cancer detection using PIs. Subsequently, two CLAM algorithms, single branch (CLAM-SB) and multiple branch (CLAM-MB), were developed. ResNet50 demonstrated the best performance, achieving an accuracy of 0.973. The performance when interpreting internal WSIs was an area under the curve (AUC) of 0.982. CLAM-SB outperformed CLAM-MB with an AUC of 0.944 in internal WSIs. Notably, in the external test, CLAM-SB exhibited superior performance with an AUC of 0.866 compared to ResNet50's AUC of 0.804. Analysis of the heatmap revealed that cases frequently misinterpreted by AI were easily interpreted by humans, and vice versa. Because AI and humans were found to function complementarily, implementing computer-aided diagnosis is expected to significantly enhance diagnostic accuracy and reproducibility. Furthermore, the WSI-based learning in CLAM, eliminating the need for patch-by-patch annotation, offers an advantage over the CNN model.

PMID:40311756 | DOI:10.1016/j.ajpath.2025.04.004

Physiol Behav. 2025 Apr 29:114934. doi: 10.1016/j.physbeh.2025.114934. Online ahead of print.

ABSTRACT

Social buffering is a phenomenon whereby the stress response of anyone exposed to a distressing stimulus is alleviated by the presence of conspecific(s). In this study, we aimed to determine whether brief buffering (only 3 min) with conspecific immediately after fear conditioning can produce social buffering in adolescent Sprague-Dawley rats (4-5 weeks, male and female) and whether close partner interaction can impact brief social buffering in adolescent female rats. The rats received an electric shock in the black room of shuttle box, followed by a 3 min buffering period. After two times of learning, the rats performed passive avoidance test individually, both immediately and 24 hr later. To reduce human bias and analyze variables not accessible to humans, data were analyzed using YOLOv8 and BoT-SORT, deep learning-based algorithm. As a result, Toy group, tested with an object resembling a rat, showed a significant increase in fear-related behavior for both sexes. Pair group, tested with a partner, showed a significant decrease in fear-related behavior in both sexes during the learning check, but only females maintained this decrease in the retention. In Pair female group, the longer the rat and its partner spent in the same room and the longer they stayed close, the higher the black room preference; this was a significant correlation. Therefore, we demonstrated that immediate brief social contact is sufficient to induce social buffering especially in female rats. In addition, social contact appears to be a key factor increasing the efficiency of social buffering.

PMID:40311725 | DOI:10.1016/j.physbeh.2025.114934

Prog Neuropsychopharmacol Biol Psychiatry. 2025 Apr 29:111386. doi: 10.1016/j.pnpbp.2025.111386. Online ahead of print.

ABSTRACT

The development of new drugs is a complex, expensive, and time-consuming process, often fraught with a high likelihood of failure. Amid these challenges, drug repurposing, which identifies new therapeutic applications for already existing medications, offers a more economical and time-saving approach, particularly in the challenging field of neurological and psychiatric disorders. This narrative review explores both preclinical and clinical studies to examine the potential of anesthetics such as ketamine, nitrous oxide, isoflurane, sevoflurane, propofol, dexmedetomidine, and sodium oxybate in treating central nervous system disorders. Various research highlights the potential of anesthetics to provide rapid antidepressant effects, enhance learning and memory, improve synaptic plasticity, and offer neuroprotective benefits, demonstrating promise for treating depression, post-traumatic stress disorder, cognitive decline, traumatic brain injury, and neurodegenerative disorders. Anesthetics appear to alleviate symptoms in neurological conditions, likely by modulating GABAergic and glutamatergic pathways. However, challenges such as dose-dependent neurotoxicity, variability in preclinical and clinical outcomes, as well as environmental concerns remain significant issues. Future research is essential to optimize dosing strategies, ensure long-term safety, and gain a deeper understanding of the precise mechanisms of action. The concept of anesthetics' repurposing presents a unique solution to tackle the challenges in neurological and psychiatric therapy by providing a platform for the development of new and improved therapies.

PMID:40311741 | DOI:10.1016/j.pnpbp.2025.111386

Cancer Radiother. 2025 Apr 30;29(2):104622. doi: 10.1016/j.canrad.2025.104622. Online ahead of print.

ABSTRACT

PURPOSE: Lung cancers associated with interstitial lung disease are challenging to diagnose and manage. We investigated the prevalence of interstitial lung disease among consecutively irradiated cancer patients, and the tolerance and prognosis of patients with or without interstitial lung disease after thoracic radiotherapy.

MATERIAL AND METHODS: This bicentric study was designed as a case-control study of patients with interstitial lung disease prior to radiotherapy (cases) and controls without interstitial lung disease. Patients were irradiated with curative intent for localized, locally advanced or oligometastatic non-small cell lung cancer. Consecutive lung cancer patients undergoing curative radiotherapy between January 2018 and December 2020 had centralized review of their baseline and 6-month CT scans by a multidisciplinary board. Functional evaluation, radiological scores, clinical toxicities, best objective response, progression-free survival and overall survival were assessed.

RESULTS: Twelve cases were detected out of 166 patients (7.2 %), including six diagnosed a posteriori by central review (50 %). Initial patient, tumour and lung cancer treatment characteristics were similar between cases and controls except for performance status (P=0.004), Kazerooni scores of fibrosis and ground glass patterns (P<0.001). Cases and controls underwent three-dimensional radiotherapy in 0 and 37 (24.2 %), intensity-modulated radiotherapy in eight (66.7 %) and 60 (39.2 %), stereotactic body radiotherapy in four (33.3 %) and 56 (36.6 %), respectively (P=0.079). Grade≥2 pneumonitis occurred in 41.7 % of cases versus 11 %, of controls (P=0.01). Hospitalization rates were 16 % in cases versus 2 % in controls and one case died of lung toxicity. Best objective response was worse for cases (P=0.046). Median progression-free survival was 9.35 months for cases and 18.56 months for controls. Median overall survival was 17 months for cases and not reached for controls (P=0.002). Sex, tumour stage, histology, and interstitial pulmonary fibrosis were prognostic factors for overall survival on univariate analysis.

CONCLUSION: Interstitial lung disease was present in 7 % of the patients with lung cancer. Patients with interstitial lung disease had higher risks of toxicity events and poorer prognosis, suggesting the lungs should be assessed carefully and that specific management strategies are warranted.

PMID:40311519 | DOI:10.1016/j.canrad.2025.104622

J Cyst Fibros. 2025 May 1:S1569-1993(25)01462-6. doi: 10.1016/j.jcf.2025.04.007. Online ahead of print.

ABSTRACT

BACKGROUND: Inhaled antibiotics are frequently used as chronic Pseudomonas aeruginosa (Pa) suppressive therapy among people with cystic fibrosis (PwCF). However, their use might increase the risk of developing treatment-emergent respiratory organisms. This study aimed to describe the proportion of PwCF utilizing inhaled antibiotics, determine factors associated with inhaled antibiotic prescription, and determine if chronic inhaled antibiotic use is associated with an increased risk of Aspergillus fumigatus, Stenotrophomonas maltophilia, or Achromobacter spp.

METHODS: This retrospective cohort study utilized Canadian CF Registry data. Pa status (chronic, intermittent, and negative) was defined per calendar year. The risk of developing A. fumigatus, S. maltophilia, or Achromobacter spp was compared between PwCF prescribed versus not prescribed inhaled antibiotics, adjusting for confounding by indication using inverse probability of treatment weighting.

RESULTS: This analysis included data from 2800 PwCF. >75 % of PwCF with chronic Pa were prescribed inhaled antibiotics, while up to 13 % of PwCF negative for Pa received inhaled antibiotics during the study period. There was an increased risk of developing A. fumigatus among PwCF with intermittent Pa (HR 1.43, 95 %CI; 1.08-1.88; p = 0.01) and who were Pa negative (HR 2.44, 95 %CI; 1.65-3.61; p < 0.001), but not for PwCF with chronic Pa (HR 1.36, 95 %CI; 0.94-1.95; p = 0.10). No association was seen between inhaled antibiotics and developing either S. maltophilia or Achromobacter spp.

CONCLUSIONS: Inhaled antibiotic use among Canadian PwCF was associated with an increased risk of A. fumigatus acquisition but not S. maltophilia or Achromobacter spp. Prospective studies are needed to better define this association.

PMID:40312233 | DOI:10.1016/j.jcf.2025.04.007

BMC Med Imaging. 2025 May 1;25(1):142. doi: 10.1186/s12880-025-01682-5.

ABSTRACT

BACKGROUND: The adrenal glands are small retroperitoneal organs, few reference standards exist for adrenal CT measurements in clinical practice. This study aims to develop a deep learning (DL) model for automated adrenal gland segmentation on non-contrast CT images, and to conduct a preliminary large-scale study on age-related volume changes in normal adrenal glands using the model output values.

METHODS: The model was trained and evaluated on a development dataset of annotated non-contrast CT scans of bilateral adrenal glands, utilizing nnU-Net for segmentation task. The ground truth was manually established by two experienced radiologists, and the model performance was assessed using the Dice similarity coefficient (DSC). Additionally, five radiologists provided annotations on a subset of 20 randomly selected cases to measure inter-observer variability. Following validation, the model was applied to a large-scale normal adrenal glands dataset to segment adrenal glands.

RESULTS: The DL model development dataset contained 1301 CT examinations. In the test set, the median DSC scores for the segmentation model of left and right adrenal glands were 0.899 and 0.904 respectively, and in the independent test set were 0.900 and 0.896. Inter-observer DSC for radiologist manual segmentation did not differ from automated machine segmentation (P = 0.541). The large-scale normal adrenal glands dataset contained 2000 CT examinations, the graph shows that adrenal gland volume increases first and then decreases with age.

CONCLUSION: The developed DL model demonstrates accurate adrenal gland segmentation, and enables a comprehensive study of age-related adrenal gland volume variations.

PMID:40312690 | DOI:10.1186/s12880-025-01682-5

BMC Med Imaging. 2025 May 1;25(1):144. doi: 10.1186/s12880-025-01663-8.

ABSTRACT

BACKGROUND: Magnetic resonance imaging (MRI) is an essential tool for medical diagnosis. However, artifacts may degrade images obtained through MRI, especially owing to patient movement. Existing methods that mitigate the artifact problem are subject to limitations including extended scan times. Deep learning architectures, such as U-Net, may be able to address these limitations. Optimizing deep learning networks with batch normalization (BN) and dropout layers enhances their convergence and accuracy. However, the influence of this strategy on U-Net has not been explored for artifact removal.

METHODS: This study developed a U-Net-based regression network for the removal of motion artifacts and investigated the impact of combining BN and dropout layers as a strategy for this purpose. A Transformer-based network from a previous study was also adopted for comparison. In total, 1200 images (with and without motion artifacts) were used to train and test three variations of U-Net.

RESULTS: The evaluation results demonstrated a significant improvement in network accuracy when BN and dropout layers were implemented. The peak signal-to-noise ratio of the reconstructed images was approximately doubled and the structural similarity index was improved by approximately 10% compared with those of the artifact images.

CONCLUSIONS: Although this study was limited to phantom images, the same strategy may be applied to more complex tasks, such as those directed at improving the quality of MR and CT images. We conclude that the accuracy of motion artifact removal can be improved by integrating BN and dropout layers into a U-Net-based network, with due consideration of the correct location and dropout rate.

PMID:40312665 | DOI:10.1186/s12880-025-01663-8

Sci Rep. 2025 May 1;15(1):15295. doi: 10.1038/s41598-025-98033-9.

ABSTRACT

Effective teacher performance evaluation is important for enhancing the quality of educational systems. This study presents a novel approach that integrates deep learning and metaheuristics to assess the pedagogical quality of English as a foreign language (EFL) instruction in a classroom setting. A comprehensive index framework is developed, comprising five primary dimensions: instructional design, instructional materials, teaching methods and approaches, teaching effectiveness, and classroom management. Each dimension is further divided into secondary indicators that capture specific aspects of teaching quality, including pronunciation, content coverage, lesson objectives, and student engagement. The proposed approach uses a convolutional neural network (CNN) architecture optimized by a modified virus colony search (VCS) algorithm to analyze audio and video recordings of classroom interactions. The results demonstrate that the VCS/CNN algorithm can accurately evaluate EFL instruction based on multiple criteria and indicators, outperforming existing methods in terms of accuracy, robustness, flexibility, and efficiency. This study contributes to the development of a reliable and efficient teacher evaluation framework that can provide timely feedback, identify teacher strengths and weaknesses, and inform areas for professional development. The proposed approach has the potential to improve the quality of EFL instruction and administration by enhancing teacher performance and student learning outcomes.

PMID:40312557 | DOI:10.1038/s41598-025-98033-9