Biocybernetics and Biomedical Engineering最新文献

{"title":"Cytoplasm and nuclei as a basis for Bethesda cell cluster classification in cervical smears","authors":"Antonina Pater ,&nbsp;Lukasz Roszkowiak ,&nbsp;Krzysztof Siemion ,&nbsp;Jakub Zak ,&nbsp;Karol Deptuch ,&nbsp;Anna Korzynska","doi":"10.1016/j.bbe.2025.04.004","DOIUrl":"10.1016/j.bbe.2025.04.004","url":null,"abstract":"<div><div>Population screening in the form of cervical smears was introduced to reduce cervical cancer morbidity. However, the manual evaluation of cervical smears is a labour-intensive and meticulous task. This challenge has led to the development of various computer-aided cell identification methods aimed at improving the quality of evaluations and reducing the time required for slide analysis. These supportive tools for pathologists should adhere to the Bethesda classification system for cell types, facilitating integration with established clinical practices. Despite advances, the automatic classification of clustered cells in cervical smears remains a significant challenge for both automated and semiautomated methods that utilize classical image processing and deep learning techniques.</div><div>This study introduces a novel method for fragmenting clusters to improve the classification of clustered cells in digital images of Papanicolaou smears. The proposed method integrates explainable AI and marker-guided watershed segmentation ensuring both interpretability and versatility of an overall pipeline for diagnostician support in smear evaluation. Validation of the method was performed on a recently developed Papanicolaou cytology Bialystok dataset, as well as the publicly available CRIC dataset, demonstrating the method’s generalizability across different datasets.</div><div>The cell classification pipeline, enhanced by the proposed declustering method, improved the overall harmonic mean of recall and precision (F1 score) by 13.27 percentage points compared with the results obtained without this additional processing. The improvement in classifying the most critical cell type according to the Bethesda system (HSIL cell class) was even more significant, with an increase of 35.72 percentage points compared with classifying the entire cluster.</div></div>","PeriodicalId":55381,"journal":{"name":"Biocybernetics and Biomedical Engineering","volume":"45 3","pages":"Pages 414-425"},"PeriodicalIF":5.3,"publicationDate":"2025-05-28","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144166979","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Hemodynamic simulation for plantar arteries by anatomical structure based mathematical model","authors":"Lei Sun ,&nbsp;Xiang Geng ,&nbsp;Yichong Wang ,&nbsp;Daoyuan Wang ,&nbsp;Xin Ma ,&nbsp;Gangmin Ning","doi":"10.1016/j.bbe.2025.05.006","DOIUrl":"10.1016/j.bbe.2025.05.006","url":null,"abstract":"<div><div>Diabetic foot is a kind of common complications in diabetic patients, manifesting as ischemic lesions in plantar tissue. The plantar tissue ischemia can be estimated by the perfusion of plantar blood flow. Currently, it lacks effective methods to monitor and assess plantar blood flow perfusion in practice and detailed information is not available. This study intends to address this problem from the view of mathematical modelling. It aims to provide quantitative approach to investigate the risk of ischemia in diabetic foot by simulating the hemodynamic status along the plantar vascular network. A simplified topological structure map of the main plantar arteries was established considering the anatomical structure and physical data of each vessel. Aided by bond graph technology, a Poiseuille’s law based hemodynamic model of the plantar vessels was constructed. The model simulated the blood flow distribution in the feet of healthy individuals and diabetic patients. Compared with healthy individuals, diabetic patients show increased plantar blood pressure and decreased blood supply. The reduction in blood flow on the lateral side of the foot is greater than on the medial side in diabetic patients. The study marks the first application of vascular modeling in the plantar region, providing a quantitative method for investigating foot hemodynamics in diabetic patients and offering new potential for assessing the progression of diabetic foot and the risk of tissue ulceration.</div></div>","PeriodicalId":55381,"journal":{"name":"Biocybernetics and Biomedical Engineering","volume":"45 3","pages":"Pages 390-398"},"PeriodicalIF":5.3,"publicationDate":"2025-05-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144131213","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Rhythmic physical exercise as a protocol to study synchronization phenomenon","authors":"Judyta Sobiech,&nbsp;Teodor Buchner","doi":"10.1016/j.bbe.2025.05.007","DOIUrl":"10.1016/j.bbe.2025.05.007","url":null,"abstract":"<div><div>The phenomenon of synchronization, which appears during the coupling of two or more oscillators, is often observed in nature and many fields of science. The case of several rhythms is particularly interesting, as the synchronizations may compete. In this work, we report the occurrence of the phase coupling between three biological rhythms: heart rate, respiration and locomotor rhythm. We analyzed its appearance during cycling on a stationary bike with stepwise increasing speed. Presence of synchronization was investigated using the synchrogram in order to detect phase-locking epochs.</div><div>Study group consisted of 21 young healthy volunteers (15 women and 6 men). Specific subgroups were also defined.</div><div>During exercise synchronization between all rhythms appeared in all subjects. Cardiolocomotor synchronization occurred significantly more often, which underlines the role of cardiac entrainment to phasic baroreflex input. In cardiolocomotor synchronization we observed a direct phase coupling (1:1) lasting at least 20 s. It occurred in 76 % of subjects, which manifests a dominant role of this type of synchronization. Additionally, we observed subjects, in whom epochs of simultaneous synchronization of all rhythms appeared.</div><div>Total duration of synchronization was significantly larger in men than in women and in less experienced cyclists. No significant difference in age or BMI was observed, however the group was relatively uniform. During the post-exercise epoch, cardiorespiratory synchronization appeared in nine subjects. Due to its frequent occurence, synchronization is a good candidate for an objective estimator of physical stress, applicable in diagnostic, therapeutical and lifestyle scenarios.</div></div>","PeriodicalId":55381,"journal":{"name":"Biocybernetics and Biomedical Engineering","volume":"45 3","pages":"Pages 380-389"},"PeriodicalIF":5.3,"publicationDate":"2025-05-21","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144107325","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Viscoelastic modeling of optic nerve head biomechanics: Effects of intraocular and cerebrospinal fluid pressure","authors":"Alireza Karimi ,&nbsp;Reza Razaghi ,&nbsp;Seyed Mohammadali Rahmati ,&nbsp;J. Crawford Downs","doi":"10.1016/j.bbe.2025.05.008","DOIUrl":"10.1016/j.bbe.2025.05.008","url":null,"abstract":"<div><div>Although the optic nerve head (ONH) has demonstrated a significant viscoelastic response to changes in intraocular pressure (IOP), existing computational models of ONH biomechanics have yet to fully account for these viscoelastic properties. In this study, we introduce and evaluate a mesh-free beam-in-solid coupling algorithm to model the complex viscoelastic behavior of anisotropic collagen fibers within a viscoelastic scleral matrix. We also incorporated viscoelastic formulations for the retina, lamina cribrosa, and optic nerve to provide a more comprehensive understanding of ONH tissue mechanics. We compared the biomechanics of the ONH resulting from hyperelastic and viscoelastic scleral material formulations using an eye-specific finite element model of the posterior human eye. This model integrates the detailed 3D microstructure of the load-bearing lamina cribrosa, including interspersed laminar neural tissues, as well as the heterogeneous, anisotropic behavior of the collagenous sclera and pia. The viscoelastic material properties were validated against published experimental tensile tests of human scleral and retinal tissue samples. Simulations of ONH biomechanical responses were conducted by applying changes in IOP and cerebrospinal fluid pressure (CSFP) typical of body position transitions, such as moving from sitting to supine, over a 250 ms period. In both simulations, the ONH tissues exhibited greater stresses and strains in the supine position compared to sitting, as anticipated. The laminar surface showed posterior deformation (+6 µm) during the transition from sitting to supine when using the hyperelastic material model, whereas it deformed anteriorly (−5.7 µm) with the viscoelastic model. Furthermore, the radial scleral canal expansion at the anterior laminar insertion was significantly smaller in the viscoelastic formulation (9 µm) compared to the hyperelastic formulation (19.8 µm). All results aligned with experimental observations. While the stresses, strains, and deformations remained within physiological ranges for both models, there were substantial differences between the two formulations, particularly in terms of deformation. Improving the accuracy of material formulations in ONH models is expected to enhance our understanding of ONH biomechanics. However, further experimental validation is needed to confirm these results and strengthen their applicability.</div></div>","PeriodicalId":55381,"journal":{"name":"Biocybernetics and Biomedical Engineering","volume":"45 3","pages":"Pages 357-379"},"PeriodicalIF":5.3,"publicationDate":"2025-05-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144090192","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Multi-scale CNN-Swin transformer network with boundary supervision for multiclass biomarker segmentation in retinal OCT images","authors":"Zhanpeng Fan ,&nbsp;Xiaoming Liu ,&nbsp;Ying Zhang ,&nbsp;Jia Zhang","doi":"10.1016/j.bbe.2025.05.002","DOIUrl":"10.1016/j.bbe.2025.05.002","url":null,"abstract":"<div><div>Retinal biomarker morphology is closely associated with a variety of chronic ophthalmic diseases, in which biomarker localization and segmentation in optical coherence tomography (OCT) play a key role in the diagnosis of retina-related diseases. Although great progress has been made in deep learning based OCT biomarker segmentation, several challenges still exist. Due to issues such as image noise or class imbalance, retinal biomarkers affect the model’s recognition of other biomarkers. Moreover, small biomarkers are prone to lose accuracy during downsampling. And most existing methods rely on convolutional neural networks, which make it challenging to obtain the global context due to locality of convolution. Benefiting from the Swin Transformer with powerful modeling capabilities, we propose MSCS-Net (Multi-scale CNN-Swin Network), a network for OCT biomarker segmentation, which effectively combines CNN and Swin Transformer and integrates them in parallel into a dual-encoder structure. Specifically, an edge detection path is added alongside to enhance the localization of biomarkers at the edges. For the Swin Transformer branch, considering the irregular distribution of most OCT biomarkers, a new windowing partition is performed in the Swin Transformer to capture the features more efficiently. Meanwhile, we design a Feature Dimensionality Reduction Module to extensively collect the information of small-scale biomarkers. To effectively integrate information from two scales, we design a Transformer Cross Fusion Module to finely fuse the global and local feature information from the two-branch encoders. We validate the proposed approach on local and public datasets, and the experimental results demonstrate the effectiveness of the proposed framework.</div></div>","PeriodicalId":55381,"journal":{"name":"Biocybernetics and Biomedical Engineering","volume":"45 3","pages":"Pages 340-356"},"PeriodicalIF":5.3,"publicationDate":"2025-05-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144090191","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Robust dual-hormone controller for full closed-loop glucose regulation in people with type 1 diabetes: An in silico study","authors":"Emilia Fushimi ,&nbsp;Fernando Daniel Bianchi ,&nbsp;Hernán De Battista ,&nbsp;Fabricio Garelli","doi":"10.1016/j.bbe.2025.05.004","DOIUrl":"10.1016/j.bbe.2025.05.004","url":null,"abstract":"<div><div>Current advanced methods for glucose control in people with type 1 diabetes (T1D), often referred to as artificial pancreas (AP) or automated insulin delivery (AID) systems, rely on the administration of a single hormone (insulin) to regulate blood glucose (BG). In general, these systems depend on patient-specific information usually obtained from the conventional insulin therapy to account for inter-patient variability. On the other hand, dual-hormone (DH) systems that use insulin and its counterregulatory hormone, glucagon, have the potential of further improving BG control. However, DH systems are still under development or in earlier testing stages. Since glucagon is not used in the traditional therapy for T1D, the sensitivity of each individual to this hormone is typically unknown. Here, a DH controller based on robust control is proposed. The controller in charge of glucagon dosing, based on H∞ optimal control, does not require any individualization, thus overcoming one of the challenges faced by DH approaches. The strategy is evaluated <em>in silico</em> and compared to previous works involving a personalized glucagon controller and its single-hormone counterpart. Results suggest that the robust control strategy allows satisfactory glucose outcomes without the need for individualization.</div></div>","PeriodicalId":55381,"journal":{"name":"Biocybernetics and Biomedical Engineering","volume":"45 3","pages":"Pages 329-339"},"PeriodicalIF":5.3,"publicationDate":"2025-05-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144072655","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A novel approach for remote ultrasonography of knee osteoarthritis","authors":"Angela Sorriento ,&nbsp;Lorena Guachi-Guachi ,&nbsp;Giorgia Marola ,&nbsp;Paolo Spinnato ,&nbsp;Fabien Rabusseau ,&nbsp;Paolo Cabras ,&nbsp;Erik Dumont ,&nbsp;Leonardo Ricotti ,&nbsp;Andrea Cafarelli","doi":"10.1016/j.bbe.2025.05.009","DOIUrl":"10.1016/j.bbe.2025.05.009","url":null,"abstract":"<div><div>In this study, we propose innovative technologies and a novel standardized procedure, enabling at-home, periodic and reliable ultrasound monitoring of knee osteoartrithis.</div><div>Anatomical and ultrasound data were collected on a population of 27 volunteers with different anthropometric features, leading to the development of two technological solutions: i) three probe positioning systems (wearable brace, adhesive film, temporary tattoo), provided with predefined openings to ensure accurate and repeatable placement of an ultrasound probe; ii) an image selection algorithm for automatic identification of the correct ultrasound images acquired by non-expert users.</div><div>The effectiveness and usability of these solutions were tested on 20 volunteers and 20 caregivers, with no prior experience in ultrasound imaging. In the first step, an expert radiologist acquired reference ultrasound images. In the second step, all volunteers exploited the positioning systems to place the probe and acquire informative images (automatically extracted by the algorithm from a video).</div><div>The average usability percentage score was highest for the wearable brace (98.7%), followed by the tattoo (94.8%), and film (89.6%). Overall, the three positioning systems proved to be effective in guiding ultrasound acquisitions for inexperienced subjects. The highest average rate of informative images successffully acquired among all volunteers was achieved by using the brace (70%), while the highest average score among the caregivers was obtained using the tattoo (76%).</div><div>These findings highlight the possibility to perform reliable remote ultrasound acquisitions for non-expert users and opens the way to novel tele-ultrasound procedures, enabling a precise, asynchronous, and repeatable monitoring of knee osteoarthritis.</div></div>","PeriodicalId":55381,"journal":{"name":"Biocybernetics and Biomedical Engineering","volume":"45 3","pages":"Pages 319-328"},"PeriodicalIF":5.3,"publicationDate":"2025-05-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"144070282","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Assessment of the intermuscular coherence for the early detection of diabetic peripheral neuropathy: a cross-sectional study","authors":"I. Junquera-Godoy ,&nbsp;J.L. Martinez-De-Juan ,&nbsp;G. González Lorente ,&nbsp;J.M. Carot-Sierra ,&nbsp;J. Gomis-Tena ,&nbsp;J. Saiz ,&nbsp;G.C. Mas Penalva ,&nbsp;E. Soler Climent ,&nbsp;G. Prats-Boluda","doi":"10.1016/j.bbe.2025.05.003","DOIUrl":"10.1016/j.bbe.2025.05.003","url":null,"abstract":"<div><div>Diabetes mellitus (DM) is a global epidemic marked by chronic hyperglycaemia due to insulin insufficiency or resistance. Diabetic peripheral neuropathy (DPN) is a common complication of diabetes, affecting up to 50 % of patients. It typically presents as a distal, symmetric, length-dependent neuropathy. Early detection of DPN is crucial to mitigate its impact on quality of life and healthcare costs. While current diagnostic methods like nerve conduction studies have limitations, surface electromyography (sEMG) shows promise for its non-invasive, real-time neuromuscular assessments. This study explores the potential of sEMG, particularly through intermuscular coherence, to serve as a sensitive biomarker for early DPN detection. Two coherence parameters, partial coherence (PC) and partial directed coherence (PDC), based on the Multivariate Autoregressive (MVAR) models have been analysed in three population groups: 33 healthy volunteers (CT), 10 diabetic patients with a low risk of DPN (LW), and 17 moderate/high-risk patients (MH). In early-stage DPN, synergistic muscle pairs showed increased PC and PDC values compared to controls, likely due to neuronal hyperexcitability and compensatory mechanisms within the nervous system. In contrast, advanced DPN stages exhibited reduced coherence, reflecting nerve fiber loss and central nervous system (CNS) impairment, possibly exacerbated by structural CNS changes like spinal cord atrophy, affecting neural plasticity and adaptation. Our study found that the dorsiflexor muscle pair, especially in the bandwidth 10–50 Hz of the PDC parameter, effectively discriminated between DPN stages, distinguishing all three groups with statistical significance. This suggests its potential as an early detection biomarker for DPN and for monitoring disease progression.</div></div>","PeriodicalId":55381,"journal":{"name":"Biocybernetics and Biomedical Engineering","volume":"45 3","pages":"Pages 305-318"},"PeriodicalIF":5.3,"publicationDate":"2025-05-09","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143922904","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"Classifying developmental delays with EEG: A comparative study of machine learning and deep learning approaches","authors":"Muhammad Usman ,&nbsp;Wen-Yi Lin ,&nbsp;Yi-Yin Lin ,&nbsp;Sheng-Ta Hsieh ,&nbsp;Yao-Tien Chen ,&nbsp;Yu-Chun Lo ,&nbsp;Chun-Ling Lin","doi":"10.1016/j.bbe.2025.04.001","DOIUrl":"10.1016/j.bbe.2025.04.001","url":null,"abstract":"<div><div>Early detection of developmental delays is crucial for improving children’s cognitive, social, and emotional outcomes through timely interventions. This study explores the potential of machine learning (ML) and deep learning (DL) to classify Electroencephalography (EEG) data from an oddball task, distinguishing between children with and without developmental delays. Participants underwent language assessments and EEG recordings, with subsequent analysis using Event-Related Potentials (ERPs), Event-Related Spectral Perturbations (ERSPs), and functional connectivity to characterize group differences. Three methodologies were employed in this research to classify EEG data. Firstly, statistical features are extracted from the EEG data and various ML algorithms are applied for classification, with feature selection techniques utilized to identify the most relevant features and enhance classification accuracy. Secondly, brain dynamics is utilized to incorporate ERP, ERSP, and functional connectivity measures as features for developmental delay detection. Similar to the first approach, feature selection techniques are again employed to enhance classification accuracy. Lastly, DL approaches are explored by implementing multiple convolutional neural networks (CNNs), including a 2D CNN (EEGNet), various hybrid CNN architectures integrating LSTM, GRU, and attention mechanisms, and a novel 1D CNN with a standardized convolutional layer (SCL) for improved stability and training performance. The effectiveness of each approach in accurately classifying EEG data for developmental delay detection is independently analyzed. The results demonstrate that the proposed 1D convolutional neural network outperforms both EEGNet and the employed ML classifiers. This model achieves an impressive accuracy of 96.4% and an F1 score of 96.6%, underscoring its potential as a valuable tool for early and accurate developmental delay detection using EEG data.</div></div>","PeriodicalId":55381,"journal":{"name":"Biocybernetics and Biomedical Engineering","volume":"45 2","pages":"Pages 229-246"},"PeriodicalIF":5.3,"publicationDate":"2025-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143799347","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

{"title":"A sepsis diagnosis method based on Chain-of-Thought reasoning using Large Language Models","authors":"Weimin Zhang ,&nbsp;Mengfei Wu ,&nbsp;Luyao Zhou ,&nbsp;Min Shao ,&nbsp;Cui Wang ,&nbsp;Yu Wang","doi":"10.1016/j.bbe.2025.04.002","DOIUrl":"10.1016/j.bbe.2025.04.002","url":null,"abstract":"<div><div>Sepsis is a severe infectious disease with high incidence and mortality rates globally. Early diagnosis of sepsis is crucial for improving patient outcomes. Previous diagnostic methods heavily relied on subjective clinical experience, while the machine learning-based methods can only learn knowledge from a specific dataset. Recently, the rapid development of Large Language Models (LLMs) has significantly enhanced various downstream dialogue tasks by leveraging prior semantic knowledge. Therefore, it is of great interest to explore the potential of LLMs in sepsis diagnosis. This study proposed an early sepsis diagnosis method based on the Chain of Thought (CoT) reasoning using LLMs. First, the clinical data of a patients were transformed into a textual representation to form the prompt. Subsequently, a CoT was created to simulate the reasoning process of human medical experts and utilized the prior semantic knowledge in LLMs to achieve sepsis diagnosis. The proposed method was validated using real clinical data, demonstrating high classification performance with an accuracy of 0.87, recall of 0.98, and F1 score of 0.88. These metrics showed an improvement in F1 score by 7 to 8 percentage points compared to commonly used machine learning classifiers. The experimental results indicated that the proposed method can enhance the performance of early sepsis diagnosis, and the introduction of CoT enhanced the interpretability of diagnostic results, contributing to the application of LLMs in clinical diagnosis.</div></div>","PeriodicalId":55381,"journal":{"name":"Biocybernetics and Biomedical Engineering","volume":"45 2","pages":"Pages 269-277"},"PeriodicalIF":5.3,"publicationDate":"2025-04-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143848345","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}