Smart Health (Amst). 2024 Sep;33:100498. doi: 10.1016/j.smhl.2024.100498. Epub 2024 Jun 5.

ABSTRACT

This work focuses on the development and manufacturing of a wireless, wearable, low-power fall detection sensor (FDS) designed to predict and detect falls in elderly at-risk individuals. Unintentional falls are a significant risk in this demographic, often resulting from diminished physical capabilities such as reduced hand grip strength and complications from conditions like arthritis, vertigo, and neuromuscular issues. To address this, we utilize advanced low-power field-programmable gate arrays (FPGAs) to implement a fixed-function neural network capable of categorizing activities of daily life (ADLs), including the detection of falls. This system employs a Convolutional Neural Network (CNN) model, trained and validated using the Caffe deep learning framework with data collected from human subjects experiments. This system integrates an ST Microelectronics LSM6DSOX inertial measurement unit (IMU) sensor, embedded with an ultra-low-power Lattice iCE40UP FPGA, which samples and stores joint acceleration and orientation rate. Additionally, we have acquired and published a dataset of 3D accelerometer and gyroscope measurements from predefined ADLs and falls, using volunteer human subjects. This innovative approach aims to enhance the safety and well-being of older adults by providing timely and accurate fall detection and prediction. In this paper, we present an innovative approach to utilizing a compact Convolutional Neural Network (CNN) core for accelerating convolutional operations on a machine learning model, suitable for deployment on an ultra-low power FPGA.

PMID:40777999 | PMC:PMC12327353 | DOI:10.1016/j.smhl.2024.100498

Digit Health. 2025 Aug 6;11:20552076251366855. doi: 10.1177/20552076251366855. eCollection 2025 Jan-Dec.

ABSTRACT

OBJECTIVES: Accurate segmentation of medical images is vital for effective disease diagnosis and treatment planning. This is especially important in resource-constrained environments. This study aimed to evaluate the performance of various U-Net-based deep learning architectures for chest X-ray (CXR) segmentation and identify the most effective model in terms of both accuracy and computational efficiency.

METHODS: We assessed the segmentation performance of eight U-Net variants: U-Net7, U-Net9, U-Net11, U-Net13, U-Net16, U-Net32, U-Net64, and U-Net128. The evaluation was conducted using a publicly available CXR dataset categorized into normal, COVID-19, and viral pneumonia classes. Each image was paired with a corresponding segmentation mask. Image preprocessing involved resizing, noise filtering, and normalization to standardize input quality. All models were trained under identical experimental conditions to ensure a fair comparison. Performance was evaluated using two key metrics: Intersection over Union (IoU) and Dice Coefficient (DC). Additionally, computational efficiency was measured by comparing the total number of trainable parameters and the training time for each model.

RESULTS: U-Net9 achieved the highest performance among all tested models. It recorded a DC of 0.98 and an IoU of 0.96, outperforming both shallower and deeper U-Net architectures. Models with increased depth or filter width, such as U-Net128, showed diminishing returns in accuracy. These models also incurred significantly higher computational costs. In contrast, U-Net16 and U-Net32 demonstrated reduced segmentation accuracy compared to U-Net9. Overall, U-Net9 provided the optimal balance between precision and computational efficiency for CXR segmentation tasks.

CONCLUSION: The U-Net9 architecture offers a superior solution for CXR image segmentation. It combines high segmentation accuracy with computational practicality, making it suitable for real-world applications. Its implementation can support radiologists by enabling faster and more reliable diagnoses. This can lead to improved clinical decision-making and reduced diagnostic delays. Future work will focus on integrating U-Net9 with multimodal imaging data, such as combining CXR with computerized tomography or MRI scans. Additionally, exploration of advanced architectures, including attention mechanisms and hybrid models, is planned to further enhance segmentation performance.

PMID:40777837 | PMC:PMC12329272 | DOI:10.1177/20552076251366855

Fundam Res. 2024 Mar 30;5(4):1725-1736. doi: 10.1016/j.fmre.2024.03.019. eCollection 2025 Jul.

ABSTRACT

Lensless imaging is an innovative and swiftly advancing technology at the intersection of optics, imaging technology, and computational science. It captures scene by directly recording the interference or diffraction patterns of light, subsequently utilizing algorithms to reconstruct the original image from these patterns. Lensless imaging transforms traditional imaging paradigms, offering newfound design flexibility and the capacity to seamlessly integrate within diverse imaging ecosystems. This paper aims to provide an overview of significant developments in optical modulation elements, image sensors, and reconstruction algorithms. The novel application scenarios that benefit from lensless computational imaging are presented. The opportunities and challenges associated with lensless camera are discussed for further improving its performance.

PMID:40777808 | PMC:PMC12327861 | DOI:10.1016/j.fmre.2024.03.019

Cureus. 2025 Jul 8;17(7):e87536. doi: 10.7759/cureus.87536. eCollection 2025 Jul.

ABSTRACT

We present the case of a 72-year-old man with idiopathic pulmonary fibrosis (IPF) who presented with stiffness and pelvic girdle pain and was initially treated for polymyalgia rheumatica (PMR). Despite transient improvement on steroids, persistent symptoms and atypical MRI findings led to muscle and renal biopsies, revealing necrotizing vasculitis and pauci-immune glomerulonephritis. Positive myeloperoxidase and perinuclear antineutrophil cytoplasmic antibody titers confirmed the diagnosis of antineutrophil cytoplasmic antibody-associated vasculitis (AAV). The patient's course was complicated by diffuse alveolar hemorrhage requiring intubation, which improved with pulse steroids and cyclophosphamide. The patient stabilized on methotrexate and methylprednisolone. This case highlights the complexity of diagnosing AAV, especially with overlapping PMR-like symptoms and underlying interstitial lung disease.

PMID:40777707 | PMC:PMC12331196 | DOI:10.7759/cureus.87536

bioRxiv [Preprint]. 2025 Jul 24:2025.07.23.666333. doi: 10.1101/2025.07.23.666333.

ABSTRACT

Idiopathic pulmonary fibrosis (IPF) is a progressive interstitial lung disease marked by aberrant fibroblast-to-myofibroblast differentiation, a process that requires metabolic reprogramming. We identify alanine as a critical metabolite that confers metabolic flexibility to support differentiation. TGF-β increases alanine by activating both its synthesis and import in normal and IPF lung fibroblasts. Alanine is synthesized primarily by GPT2, which is regulated by a glutamine-glutamate-α-ketoglutarate axis. Inhibiting GPT2 depletes alanine and suppresses TGF-β-induced expression of α-SMA and COL1A1, an effect reversed by alanine supplementation. We also identify SLC38A2 as a key transporter of both alanine and glutamine that is upregulated by TGF-β and alanine deprivation. Together, SLC38A2 and GPT2 activities converge to maintain intracellular alanine levels to support myofibroblast differentiation. Mechanistically, alanine deficiency suppresses glycolysis and depletes tricarboxylic acid cycle intermediates, while supplementation provides carbon and nitrogen for intracellular glutamate and proline biosynthesis, particularly in the absence of glutamine. Combined inhibition of GPT2 and SLC38A2 suppresses fibrogenic responses in fibroblasts and in human precision-cut lung slices, highlighting a potential therapeutic strategy for fibrotic lung disease.

PMID:40777505 | PMC:PMC12330655 | DOI:10.1101/2025.07.23.666333

bioRxiv [Preprint]. 2025 Jul 22:2025.07.22.665497. doi: 10.1101/2025.07.22.665497.

ABSTRACT

BACKGROUND: Idiopathic pulmonary fibrosis is a fatal lung disease of progressive lung parenchymal scarring caused by the aberrant response of an alveolar epithelium repeatedly exposed to injury. Understanding epithelial dysfunction has been hampered by the lack of physiological alveolar type 2 (AT2) cell models and defined disease triggers. Monogenic forms of familial pulmonary fibrosis (FPF) caused by toxic gain-of-function variants provide an opportunity to investigate early pathogenic events. One such variant, surfactant protein C (SFTPC)-I73T, abnormally localises within AT2 cells and causes their dysfunction.

METHODS: We used base editing of fetal lung-derived AT2 (fdAT2) organoids to create a heterozygous disease model of endogenous SFTPC-I73T expression. We also created an inducible overexpression system to interrogate temporal changes associated with SFTPC-I73T expression. We cultured fdAT2 both in 3D culture and at air-liquid interface to understand the importance of polarity cues and air exposure on disease phenotypes.

RESULTS: In our heterozygous endogenous expression system, we found that fdAT2 expressing SFTPC-I73T grew without a lumen and were unable to correctly polarise. SFTPC-I73T accumulated with time and caused gross enlargement of early endosomes, preventing correct apico-basal trafficking of multiple endosomally trafficked cargoes including polarity markers and cell adhesion proteins. This phenotype was exacerbated by air exposure and led to loss of epithelial monolayer integrity and abnormal wound healing after injury.

CONCLUSION: Using endogenous gene editing for the first time in differentiated alveolar organoids, we have demonstrated that the pathogenic effects of SFTPC-I73T are mediated through endosomal dysfunction and abnormal epithelial organisation. This has important implications for AT2 function in vivo .

PMID:40777458 | PMC:PMC12330739 | DOI:10.1101/2025.07.22.665497

bioRxiv [Preprint]. 2025 Jul 26:2025.07.24.665373. doi: 10.1101/2025.07.24.665373.

ABSTRACT

BACKGROUND: Idiopathic pulmonary fibrosis (IPF) is a fatal lung disease for which novel therapeutic approaches are desperately needed. Inhibitor of DNA binding (ID) proteins are regulated by Transforming Growth Factor-β. However, the regulation and the effects of ID proteins in IPF remain poorly understood. We aimed to assess the expression of ID proteins in IPF and determine the effects of ID proteins on human lung fibroblasts (HLF) in vitro and pulmonary fibrosis in vivo .

METHODS: The expression of ID proteins in lungs and lung fibroblasts from mice and human patients with pulmonary fibrosis was evaluated. The effects of ID1/ID3 inhibition and overexpression on HLF were assessed. Genetic and pharmacological approaches were used in vivo to determine the role of ID1/ID3 in pulmonary fibrosis.

RESULTS: ID1/ID3 levels were elevated in HLFs isolated from pulmonary fibrosis-diseased patients and mice. ID1/ID3 knockdown decreased IPF-diseased HLF proliferation and differentiation into myofibroblasts. Bleomycin-exposed ID1/ID3 KO mice displayed improved lung function and presented with decreased lung fibrosis when compared to WT mice. A pharmacological inhibitor of ID1/ID3 decreased IPF-diseased HLF proliferation and differentiation in vitro and attenuated pulmonary fibrosis in vivo . A lung specific inhibition of ID1/ID3, using adeno-associated viruses expressing short hairpins targeting ID1 and ID3, reversed pulmonary fibrosis in mice. Mechanistically, ID1/ID3 inhibition decreased fibroblast proliferation through cell cycle genes and inhibited fibroblast differentiation through the MEK/ERK pathway.

CONCLUSIONS: Our data indicate that a simultaneous inhibition of ID1 and ID3 attenuates pulmonary fibrosis. ID1/ID3 inhibition holds potential as a novel therapeutic treatment for IPF.

PMID:40777426 | PMC:PMC12330687 | DOI:10.1101/2025.07.24.665373

Data Brief. 2025 Jul 10;61:111871. doi: 10.1016/j.dib.2025.111871. eCollection 2025 Aug.

ABSTRACT

Polygonum minus Huds., commonly referred to as kesum, is a traditional medicinal plant in Malaysia known for its unique fragrance attributed to its volatile compounds. The essential oil extracted from P. minus displayed various benefits, comprised of antioxidant, antimicrobial, and anticancer properties. Therefore, to identify volatile metabolites associated with such biological activities, we performed untargeted metabolite analysis on the essential oil extracted from P. minus leaf tissue. After the leaf samples were collected from the INBIOSIS experimental plot, a six-hour hydro-distillation procedure was performed to extract the essential oil, which was then subjected to GC-MS analysis. Thus, this data could provide a reference and benchmark for researchers conducting analyses on volatile compounds in P. minus.

PMID:40777556 | PMC:PMC12329244 | DOI:10.1016/j.dib.2025.111871

Front Immunol. 2025 Jul 24;16:1597862. doi: 10.3389/fimmu.2025.1597862. eCollection 2025.

ABSTRACT

INTRODUCTION: Mathematical models are powerful tools that can be used to advance our understanding of complex diseases. Autoimmune disorders such as systemic lupus erythematosus (SLE) are highly heterogeneous and require high-resolution mechanistic approaches. In this work, we present ONIDsc, a single-cell regulatory network inference model designed to elucidate immune-related disease mechanisms in SLE.

METHODS: ONIDsc enhances SINGE's Generalized Lasso Granger (GLG) causality model used in Single-cell Inference of Networks using Granger ensembles (SINGE) by finding the optimal lambda penalty with cyclical coordinate descent. We benchmarked ONIDsc against existing models and found it consistently outperforms SINGE and other methods when gold standards are generated from chromatin immunoprecipitation sequencing (ChIP-seq) and ChIP-chip experiments. We then applied ONIDsc to three large-scale datasets, one from control patients and the two from SLE patients, to reconstruct networks common to different immune cell types.

RESULTS: ONIDsc identified four gene transcripts: matrix remodelling-associated protein 8 (MXRA8), nicotinamide adenine dinucleotide kinase (NADK), RNA Polymerase III Subunit GL (POLR3GL) and Ultrabithorax Domain Protein 11 (UBXN11) in CD4+ T-lymphocytes, CD8+ Regulatory T-Lymphocytes, CD8+ T-lymphocytes 1 and Low Density Granulocytes that were present in SLE patients but absent in controls.

DISCUSSION: These genes were significantly related to nicotinate metabolism, ribonucleic acid (RNA) transcription, protein phosphorylation and the Rho family GTPase (RND) 1-3 signaling pathways, previously associated with immune regulation. Our results highlight ONIDsc's potential as a powerful tool for dissecting physiological and pathological processes in immune cells using high-dimensional single-cell data.

PMID:40777011 | PMC:PMC12328306 | DOI:10.3389/fimmu.2025.1597862

Chem Biodivers. 2025 Aug 8:e01727. doi: 10.1002/cbdv.202501727. Online ahead of print.

ABSTRACT

Pulmonary tuberculosis (TB), caused by the Mycobacterium tuberculosis (MTB) bacterium, remains a significant health problem worldwide, intensified by the emergence of multidrug-resistant (MDR) and highly resistant (XDR) strains. The current treatment protocols, related side effects, and the increasing incidence of drug resistance limit the efficacy of conventional therapeutic strategies. Traditional medicinal constituents rich in diverse phytoconstituents offer multi-target action with reduced toxicity, minimal risk of resistance, and immunomodulatory properties. Network pharmacology (NP), an integrated approach merging systems biology and computational modeling, facilitates understanding complex interactions among phytochemicals, molecular targets, and signaling pathways. Integrating modern pharmacology principles with traditional wisdom, NP provides a logical framework for developing new plant-based anti-TB agents and advancing adjunctive therapies. Combining protein-protein interaction networks, pathway enrichment analyses, multi-combinational data, and molecular docking studies offers insights into how phytoconstituents affect the immune response, block efflux pumps, and reduce resistance. This review provides a detailed analysis of NP-based methods for the identification of active compounds (e.g., alkaloids, flavonoids, terpenoids, polyphenols) and their related molecular targets involved in the pathogenesis of TB, including tumor necrosis factor-alpha (TNF-α), Toll-like receptors (TLR), nucleotide-binding oligomerization domain (NOD)-like receptor, and Janus kinase/signal transducer and activator of transcription (JAK-STAT) pathway.

PMID:40776773 | DOI:10.1002/cbdv.202501727

Mol Nutr Food Res. 2025 Aug 8:e70195. doi: 10.1002/mnfr.70195. Online ahead of print.

ABSTRACT

Optimizing plant-based protein intake, such as pea protein hydrolysates (PPHs), may aid in obesity management. This study investigated whether PPHs with varying bitterness and degrees of hydrolysis (DH) differently affect satiety in healthy male participants. In a short-term randomized control trial, 19 moderately overweight men (BMI 25-30 kg/m2) consumed boluses of 75 g glucose plus 15 g PPH (control without PPH; PPH1: less bitter, DH = 35%; PPH2: more bitter, DH = 23%). Upon PPH administration, energy intake from an ad libitum breakfast was reduced by -126 ± 329 kcal (p < 0.05) in the PPH2 group compared to the control. PPH1 decreased plasma ghrelin and DPP-4 levels (AUC: -9.4 ± 19.6 and -12.5 ± 24.7, p < 0.05). Gastric emptying was delayed by a mean of 65% (p < 0.0001) after PPH2 consumption, assessed via 13C-Na-acetate breath test. Bitterness and DH of PPH influence satiety signals differently. PPH1 (less bitter, higher DH) reduces DPP-4 and ghrelin levels, promoting satiety. PPH2 (more bitter, lower DH) delays gastric emptying, enhancing satiation. These findings highlight the potential of PPHs as functional ingredients in weight management strategies.

PMID:40776629 | DOI:10.1002/mnfr.70195

Stud Health Technol Inform. 2025 Aug 7;329:32-36. doi: 10.3233/SHTI250796.

ABSTRACT

Despite the existence of ICD-O for tumor classification, the broader ICD-10 system is often used in practice. While OncoTree is significant in research and molecular tumor boards, it provides a more detailed classification based on molecular and histological characteristics, crucial for clinical trial enrollment and data comparison. Therefore, a mapping between ICD-10 and OncoTree was developed. The mapping uses SNOMED CT as an intermediary step because both ICD-10 and OncoTree are structured differently. During the mapping process, some challenges arose, such as differences in the structure of the coding systems and inaccurate mappings. Despite this, the approach achieved an accuracy rate of 86.18%, which is considered satisfactory. Future efforts will focus on refining the mapping process to enhance its integration into production systems.

PMID:40775814 | DOI:10.3233/SHTI250796

Front Public Health. 2025 Jul 24;13:1630412. doi: 10.3389/fpubh.2025.1630412. eCollection 2025.

ABSTRACT

BACKGROUND: Osteoporosis is a prevalent condition globally, often linked to a significant risk of fractures. Drug-induced osteoporosis (DIOP) is an increasingly recognized adverse effect of various medications, but the sex-specific risks and time-to-onset patterns remain inadequately understood. Addressing these gaps in knowledge is critical to improving patient safety and pharmacovigilance.

OBJECTIVE: This study aimed to explore sex-related differences in DIOP, identify high-risk medications, and assess the onset patterns of osteoporosis-related adverse events by analyzing data from the FDA Adverse Event Reporting System (FAERS) and validating the findings using the Canada Vigilance Adverse Reaction Online Database (Canada Vigilance ADR).

METHODS: We analyzed adverse event reports from the FAERS database covering the period from Q1 2004 to Q4 2024. Drugs were standardized using the RxNorm drug terminology system, and adverse events were matched to MedDRA 27.1. Disproportionality analysis was conducted using Reporting Odds Ratio (ROR), Multi-item Gamma Poisson Shrinker (MGPS), and Bayesian Confidence Propagation Neural Network (BCPNN) methods. To validate our findings, we performed external validation using the Canada Vigilance ADR database. Stratified analyses by sex were performed to assess differences in drug-osteoporosis associations.

RESULTS: A total of 236,928 osteoporosis-related reports were identified, with 64.6% of the reports coming from females. We identified 68 drugs associated with DIOP, including 15 male-specific and 26 female-specific potential risk drugs. Notable drugs such as tenofovir disoproxil and esomeprazole were linked to both sexes. Drugs like upadacitinib exhibited early-onset failure patterns, while others like tenofovir demonstrated cumulative risk patterns over prolonged use. External validation with the Canada Vigilance ADR confirmed 32 drugs with potential osteoporosis risks.

CONCLUSIONS: This study highlights important sex-specific differences in the risk of drug-induced osteoporosis and underscores the need for targeted pharmacovigilance strategies. The findings contribute to a more personalized approach to drug safety, promoting more informed decision-making regarding medication use in osteoporosis-prone populations.

PMID:40777658 | PMC:PMC12328446 | DOI:10.3389/fpubh.2025.1630412

Stud Health Technol Inform. 2025 Aug 7;329:1922-1923. doi: 10.3233/SHTI251280.

ABSTRACT

This study identifies key risk factors for immune-related adverse events (irAEs) in older cancer adults receiving immune checkpoint inhibitors (ICIs). Using the Flatiron Health database, the XGBoost model and SHAP analysis identified BMI, SES, ECOG score, sex, comorbidities, and pain score as significant predictors.

PMID:40776297 | DOI:10.3233/SHTI251280

Stud Health Technol Inform. 2025 Aug 7;329:1140-1144. doi: 10.3233/SHTI251017.

ABSTRACT

In this study, we combined automatically generated labels from large language models (LLMs) with a small number of manual annotations to classify adverse event-related treatment discontinuations in Japanese EHRs. By fine-tuning JMedRoBERTa and T5 on 6,156 LLM-labeled records and 200 manually labeled samples and then evaluating on a 100-record test set, T5 achieved a precision of 0.83, albeit with a recall of only 0.25. We noted that when training solely on the 200 human-labeled samples (that contained significantly few positive cases), the model failed to detect any adverse events, making a reliable measurement of precision or recall infeasible (that is, N/A). This underscores the potential of large-scale LLM-driven labeling as well as the need to improve recall and label quality in practical clinical scenarios.

PMID:40776035 | DOI:10.3233/SHTI251017

Stud Health Technol Inform. 2025 Aug 7;329:784-788. doi: 10.3233/SHTI250947.

ABSTRACT

Extracting nuanced adverse drug reactions (ADRs) from patient self-reported messages using is pivotal but challenging, particularly given HIPAA constraints. We investigate locally deployable small LLMs-Mistral-7B, Llama-3-8B, and Gemma-7B-for ADR extraction using the PsyTAR dataset of self-reported messages. We implement in-context learning, demonstration selection, and fine-tuning with QLoRA. Results show Mistral-7B excels in few-shot settings, and Fine-tuning with F1 = 86%. These approaches safeguard data privacy and offer resource-efficient solutions for healthcare organizations. Our pipeline enables real-time ADR monitoring, helping clinicians address concerns more swiftly and enhance patient outcomes. Findings underscore that smaller LLMs can be effectively used under strict data privacy constraints, allowing healthcare teams to quickly identify and address patient-reported ADRs. Ultimately, these accessible solutions bolster patient safety.

PMID:40775965 | DOI:10.3233/SHTI250947

Stud Health Technol Inform. 2025 Aug 7;329:703-707. doi: 10.3233/SHTI250931.

ABSTRACT

This study assessed the effectiveness of natural language processing (NLP) in detecting adverse events (AEs) from anticancer agents by analyzing data from over 39,000 cancer patients. A specialized machine learning model identified known AEs from anticancer agents like capecitabine, oxaliplatin, and anthracyclines, revealing a significantly higher incidence in the treatment groups compared to non-users. While the NLP approach effectively detected most symptomatic AEs requiring manual review, it struggled with rarely documented conditions and commonly used clinical terms. Overall, the method shows promise for automated AE detection in medical records, particularly for symptoms without laboratory markers or diagnosis codes.

PMID:40775949 | DOI:10.3233/SHTI250931

Stud Health Technol Inform. 2025 Aug 7;329:460-464. doi: 10.3233/SHTI250882.

ABSTRACT

Understanding cancer patients' experiences with medication safety events at home is crucial due to their impact on health outcomes. However, patients often face barriers to sharing this information with their clinicians. This study identified key factors influencing cancer patients' self-reporting behaviors through a mixed-methods approach, analyzing survey and interview data from 41 patients diagnosed with breast, lung, prostate, or colorectal cancer. Logistic regression analysis identified predictors such as beliefs about medication and perceived system usability, achieving an AUC of 0.85. Thematic analysis indicated that patients were more likely to report safety events when medications disrupted their daily routines, with perceived severity as the key trigger. System usability and relationships with clinicians also affected reporting behaviors. Our findings highlight the need for user-friendly reporting systems and supportive communication to improve patient engagement and medication safety, offering valuable insights for designing patient-centered reporting systems to facilitate patient-contributed data.

PMID:40775900 | DOI:10.3233/SHTI250882

Stud Health Technol Inform. 2025 Aug 7;329:450-454. doi: 10.3233/SHTI250880.

ABSTRACT

Adverse drug reactions are a significant concern in medical research and patient safety, and social media offers a valuable resource for broad and continuous monitoring of these reactions. However, extracting and analyzing relevant information about adverse drug reactions from social media poses several challenges for language models. Some challenges are linked to the characteristics of social media writing style, such as the use of hashtags, idiomatic expressions, and personal opinions. In this study, we identify these characteristics and investigate how language models perform in their presence in several languages. Our findings reveal that while current models can effectively classify some aspects of ADR-related content, significant challenges remain in accurately processing these social media-specific features. Further research is needed to enhance model performance and improve the reliability of ADR detection from social media data.

PMID:40775898 | DOI:10.3233/SHTI250880

Stud Health Technol Inform. 2025 Aug 7;329:352-356. doi: 10.3233/SHTI250860.

ABSTRACT

This study explores the potential of nursing records in improving pharmaceutical safety by analyzing adverse effects (AEs) using ChatGPT-4 for data categorization. A total of 927 anonymized records from a Japanese community hospital were analyzed through a three-stage protocol: identifying AEs using ChatGPT-4, constructing a co-occurrence network to analyze relationships among key terms, and identifying central patterns using network metrics. Results revealed a high proportion of entries classified as "Unknown" for both AE status (674 cases) and patient conditions (865 cases), highlighting documentation gaps. The most frequently documented medicines were Red Blood Cells (123 instances) and Fresh Frozen Plasma (87 instances), while Thermal Therapy (30 instances) and CT Scans (23 instances) were the most used medical devices. However, AE documentation was limited, with only 6 cases for medicines and 2 for devices. The co-occurrence network emphasized the importance of informed consent (IC) in communicating treatment risks, particularly in chemotherapy and antibiotic use. Despite challenges, this study underscores the critical role of nursing records in AE management and calls for standardized documentation and integration. Future efforts could explore combining large language models with human-in-the-loop feedback to further improve AE extraction accuracy from incomplete nursing notes. into clinical workflows to enhance patient safety.

PMID:40775878 | DOI:10.3233/SHTI250860