BioMedical Engineering OnLine最新文献

Background: Chagas disease, caused by Trypanosoma cruzi (T. cruzi), affects millions, mainly in Latin America, but is spreading globally due to migration and climate change. Early identification of infection is vital for preventing chronic complications, and analyzing multimodal cardiac function data may help detect T. cruzi infection early. This study presents a hybrid method based on late multimodal fusion for integrating machine learning (ML) and deep learning (DL) algorithms using cardiac biomarkers and echocardiography (ECHO) video to classify individuals with T. cruzi infection.

Methods: An experimental cohort of 96 ICR mice was utilized to study cardiac functionality in infected individuals. Ensemble feature selection (EFS) and weighted multiple kernel learning (MKL) methods were proposed to classify unimodal and multimodal cardiac biomarkers using an ML approach. In addition, two DL-based architectures were implemented for ECHO video classification. Finally, we integrated the ML and DL algorithms in a hybrid method based on late multimodal fusion.

Results: From 64 biomarkers, we identified 17 biomarkers as the most relevant using EFS. For ML, we trained algorithms with these selected biomarkers and obtained 73% accuracy (ACC), 84% area under the ROC curve (AUC), and an F1 score (F1) of 69% using unweighted MKL, and we noted that these results improved with weighted MKL, achieving ACC, AUC, and F1 of 80% on the test set. For the DL approach, we used ECHO for video classification, obtaining 65% ACC, 60% AUC, and F1 of 58%. Then, we integrated the ML and DL algorithms using the proposed hybrid method, which achieved 84% AUC, and 80% in ACC and F1.

Conclusions: We presented a hybrid method for fusion cardiac biomarkers and ECHO video using late multimodal fusion (ML + DL). This work has the potential to assist in the diagnosis and monitoring of T. cruzi infection by providing an automated tool capable of accurately identifying patients with CD.

Introduction: This study aimed to systematically assess the impact of compressive stress on the growth of osteoblasts and osteoclasts through a meta-analysis of existing literature. The focus was on understanding how compressive stress affects cell proliferation, differentiation, and overall bone metabolism.

Methods: A comprehensive Literature search was conducted in multiple databases, including PubMed, Web of Science, CNKI, and ScienceDirect, to identify studies published between January 2000 and April 2025. The selection criteria focused on experimental studies examining the effects of compressive stress on osteoblasts and osteoclasts. Data from 16 high-quality studies were extracted and analyzed using RevMan 5.4 and R 4.1.4, with subgroup analyses based on study type, stress type, and cell response.

Results: The meta-analysis found a significant positive effect of compressive stress on the growth of both osteoblasts and osteoclasts. In vitro studies demonstrated a stronger and more consistent effect compared to animal studies. Osteoblasts responded more significantly to compressive stress than osteoclasts. Different stress types, including compression stress and fluid shear stress, showed varying levels of impact, with compression stress having the most pronounced effects on cell growth.

Discussion: Compressive stress plays a critical role in promoting osteoblast and osteoclast growth, which has implications for bone health and metabolism. These findings highlight the potential of compressive stress as a therapeutic tool for bone-related conditions. Further research is needed to explore the molecular mechanisms and clinical applications of compressive stress in treating bone diseases.

Background and purpose: Quantitative assessment of left atrial volume (LAV) is an important factor in the study of the pathogenesis of atrial fibrillation. However, automated left atrial segmentation with quantitative assessment usually faces many challenges. The main objective of this study was to find the optimal left atrial segmentation model based on cardiac computed tomography angiography (CTA) and to perform quantitative LAV measurement.

Method: A multi-center left atrial study cohort containing 182 cardiac CTAs with atrial fibrillation was created, each case accompanied by expert image annotation by a cardiologist. Then, based on this left atrium dataset, five recent states-of-the-art (SOTA) models in the field of medical image segmentation were used to train and validate the left atrium segmentation model, including DAResUNet, nnFormer, xLSTM-UNet, UNETR, and VNet, respectively. Further, the optimal segmentation model was used to assess the consistency validation of the LAV.

Results: DAResUNet achieved the best performance in DSC (0.924 ± 0.023) and JI (0.859 ± 0.065) among all models, while VNet is the best performer in HD (12.457 ± 6.831) and ASD (1.034 ± 0.178). The Bland-Altman plot demonstrated the extremely strong agreement (mean bias - 5.69 mL, 95% LoA - 19-7.6 mL) between the model's automatic prediction and manual measurements.

Conclusion: Deep learning models based on a study cohort of 182 CTA left atrial images were capable of achieving competitive results in left atrium segmentation. LAV assessment based on deep learning models may be useful for biomarkers of the onset of atrial fibrillation.

To reduce the burden of screening of colorectal polyps, *leave-in-situ* management of diminutive (≤ 5 mm) polyps is being considered. However, such an approach requires increased diagnostic efficacy (PIVI-1 criterion). The aim of this study was to use Optical Coherence Tomography (OCT) for the classification of pre-malignant polyps as benign (i.e., normal or hyperplastic) vs. those with a malignant potential (i.e., adenomas and sessile serrated adenomas) at an accuracy that would enable clinical screening of colorectal cancer. The OCT raw data, from volume imaging of resected polyps, were used to extract features that can serve as biomarkers of disease. In addition to intensity and textural measures, novel biomarkers, such as scatterer size, group velocity dispersion, and spectral information, were also estimated. Their statistical properties were combined to produce scores which, in turn, were used to classify the images or polyps. This approach yielded 79.6% accuracy (72.3% NPV) for the classification of individual images and 97.3% accuracy (95.5% NPV) when combining the feature values to classify whole polyp sections. The results of this study confirm the potential of OCT imaging of colorectal polyps as a viable adjunct to colonoscopy that could enable leave-in-situ management strategies.

Objectives: To construct a pathomics-based machine learning model to enhance the diagnostic efficacy of LungPro navigational bronchoscopy for peripheral pulmonary lesions and to optimize the management strategy for LungPro-diagnosed negative lesions.

Methods: Clinical data and hematoxylin and eosin (H&E)-stained whole slide images (WSIs) were collected from 144 consecutive patients undergoing LungPro virtual bronchoscopy at a single institution between January 2022 and December 2023. Patients were stratified into diagnosis-positive and diagnosis-negative cohorts based on histopathological or etiological confirmation. An artificial intelligence (AI) model was developed and validated using 94 diagnosis-positive cases. Logistic regression (LR) identified associations between clinical/imaging characteristics and malignant pulmonary lesion risk factors. We implemented a convolutional neural network (CNN) with weakly supervised learning to extract image-level features, followed by multiple instance learning (MIL) for patient-level feature aggregation. Multiple machine learning (ML) algorithms were applied to model the extracted features. A multimodal diagnostic framework integrating clinical, imaging, and pathomics data were subsequently developed and evaluated on 50 LungPro-negative patients to assess the framework's diagnostic performance and predictive validity.

Results: Univariable and multivariable logistic regression analyses identified that age, lesion boundary and mean computed tomography (CT) attenuation were independent risk factors for malignant peripheral pulmonary lesions (P < 0.05). A histopathological model using a MIL fusion strategy showed strong diagnostic performance for lung cancer, with area under the curve (AUC) values of 0.792 (95% CI 0.680-0.903) in the training cohort and 0.777 (95% CI 0.531-1.000) in the test cohort. Combining predictive clinical features with pathological characteristics enhanced diagnostic yield for peripheral pulmonary lesions to 0.848 (95% CI 0.6945-1.0000). In patients with initially negative LungPro biopsy results, the model identified 20 of 28 malignant lesions (sensitivity: 71.43%) and 15 of 22 benign lesions (specificity: 68.18%). Class activation mapping (CAM) validated the model by highlighting key malignant features, including conspicuous nucleoli and nuclear atypia.

Conclusions: The fusion diagnostic model that incorporates clinical and pathomic features markedly enhances the diagnostic accuracy of LungPro in this retrospective cohort. This model aids in the detection of subtle malignant characteristics, thereby offering evidence to support precise and targeted therapeutic interventions for lesions that LungPro classifies as negative in clinical settings.

Background: Dislocation of an endotracheal tube (ETT) during invasive ventilation can lead to serious events such as unilateral ventilation or unintentional extubation. The correct position of the endotracheal tube is determined visually. X-ray imaging or invasive procedures such as bronchoscopy are established for repeated position verification. However, these measures are time-consuming and only provide a limited number of snapshots. A new monitoring method can recognize dislocations of the ETT. The proposed system operates automatically without the need for continuous staff awareness or interaction.

Materials and methods: A ring-shaped permanent magnet is attached to the ETT. A small device is placed extracorporeally on the patient to detect the magnetic field. This device uses 64 magnetic sensors arranged as a sensor array in an 8x8 matrix. The sensor signals are digitally converted, enabling the position of the ETT with the attached magnet to be determined by software. Two processing methods (image similarity and localization) are tested for monitoring. The prototype system detects displacements with millimeter scale positioning deviations in our tests.

Results: Our system triggers an alarm upon detecting an impermissible dislocation, complete extubation, or unintended bronchial intubation. The proposed methods were validated on a sensor array prototype and assessed through a dedicated experimental setup. The results are promising and could lead to further development towards clinical usability.

Conclusion: Early warnings would be particularly advantageous, even for minor or beginning dislocations of the ETT. An automated continuous tube monitoring process could help reduce the workload of the staff and improve patient safety.

Background: Coronary artery calcification (CAC) represents a major cardiovascular risk in patients with end-stage renal disease (ESRD) undergoing hemodialysis. Given that radial artery pulse waveforms can reflect vascular status, this study aimed to evaluate their utility in the non-invasive assessment of CAC severity.

Methods: 58 patients with ESRD undergoing hemodialysis were enrolled. CAC severity was assessed using low-dose computed tomography (LDCT) and classified into four groups based on Agatston scores: no calcification (0), mild (1-100), moderate (101-400), and severe (> 400). Radial artery pulse waveforms were recorded before, hourly during, and after hemodialysis. Key features were extracted based on morphological differences among groups. Statistical inter-group comparisons and intra-group trend analyses were performed. A gradient boosting decision tree (GBDT) model was trained to classify CAC severity using waveform features.

Results: Clear morphological differences were observed among the four CAC groups. The non-calcified group showed a distinct main wave followed by identifiable tidal waves. With increasing CAC severity, the tidal waves became progressively attenuated and less distinguishable, resulting in a smoother overall waveform, particularly in the severe calcification group. Pulse waveform features exhibited significant variation across CAC groups and over the hemodialysis process, including parameters related to waveform morphology, descending limb, complexity and distribution, mean value, and full-process stereoscopic pulse wave features. The GBDT model demonstrated robust and consistent performance, with an average accuracy of 84.1% and a macro-AUC of 0.962 in fivefold cross-validation, and comparable results (83.9% accuracy, 0.961 macro-AUC) on the independent test set. Notably, the model performed particularly well in identifying Severe Calcification cases.

Conclusions: Radial artery pulse wave analysis may serve as a non-invasive adjunct for assessing CAC severity in patients with ESRD undergoing hemodialysis.

Background: Colorectal cancer (CRC) is a highly aggressive and extensive malignancy. Although long noncoding RNAs (lncRNAs) are often used as diagnostic biomarkers, their diagnostic effectiveness in CRC remains uncertain.

Methods: From January 1, 2015, to April 1, 2024, we conducted a comprehensive search of Embase, China National Knowledge Infrastructure (CNKI), Wanfang, PubMed, Cochrane Library, and Web of Science (WoS). The pooled sensitivity, specificity, positive likelihood ratio (PLR), negative likelihood ratio (NLR), diagnostic odds ratio (DOR), area under the receiver operating characteristic curve (AUC) and Fagan plot analysis were used to assess the overall test performance of lncRNAs. Moreover, we evaluated the publication bias using the Deeks' funnel plot asymmetry test.

Results: Twenty-eight publications were identified and incorporated into this meta-analysis. The aggregated diagnostic data were as follows: The pooled sensitivity was 0.79 (95% CI, 0.75-0.83). The pooled specificity was 0.81 (95% CI, 0.78-0.84). The PLR was 3.68 (95% CI, 3.18-4.26). The NLR was 0.28 (95% CI, 0.24-0.33). The DOR was 15.01 (95% CI, 11.85-19.00). The AUC was 0.87 (95% CI, 0.84-0.90). Deeks' funnel plot asymmetry test indicated no significant evidence of publication bias (p > 0.05). The Fagan plot analysis showed that the post-test probability was 81% for positive results and 20% for negative results. Univariate meta-regression identified multiple sources of heterogeneity in the data, including year, sample size and specimen.

Conclusion: In summary, our findings demonstrate that lncRNAs have a promising diagnostic accuracy for CRC, underscoring their potential as effective non-invasive biomarkers.

Microdamage often occurs in biological hard tissues which mainly include bone tissue and tooth hard tissue, and it primarily comprises diffuse damage and microcracks. The unique microscopic structures of biological hard tissues directly influence the initiation and progression of microcracks. Mechanical forces, loading methods, macroscopic tissue characteristics, aging-related changes, diseases, and medication factors contribute to the complexity of analysis in studying microdamage of biological hard tissues. A large number of literatures have verified the detection and research methods of microcracks. The mechanisms underlying the absorption and repair of biological hard tissues caused by microdamage are still not completely clear. This article reviews the occurrence and development of various types of microdamage in biological hard tissues from microscopic to macroscopic scales, summarizes research approaches of microdamage, elucidates the mechanisms involved in absorption and repair of microdamage, analyzes existing gaps and controversies in current research findings, and proposes potential directions for future research. The study on microdamage of biological hard tissues is crucial for developing biomimetic materials. Such studies facilitate the prediction, control, prevention, and even restoration of microdamage in these materials.

Background: Spontaneous isolated superior mesenteric artery dissection (SISMAD) is a rare but potentially lethal vascular emergency with unclear pathogenesis. While hemodynamic forces are implicated in its development, current understanding remains limited by the lack of patient-specific data. This study aimed to characterize the detailed hemodynamic environment in SISMAD using patient-specific computational fluid dynamics modeling.

Results: Analysis of a three-dimensional model reconstructed from computed tomography angiography of a Yun Type I SISMAD revealed complex flow patterns with marked hemodynamic differences between the true lumen (TL) and false lumen (FL). The TL exhibited high-velocity flow concentrated near the entry tear and significantly elevated wall shear stress (WSS) and time-averaged wall shear stress (TAWSS) along the intimal flap. In contrast, the FL demonstrated markedly lower velocities, regions of flow stasis, and low WSS. A substantial pressure gradient existed across the intimal flap, with higher pressure in the TL compared to the FL. The FL also showed significantly higher oscillatory shear index (OSI) values, often exceeding 0.4 with a peak of 0.45. These findings provide quantitative confirmation of the theorized hemodynamic forces contributing to dissection progression and potential thrombosis formation, particularly the pro-thrombotic environment within the FL.

Conclusions: Patient-specific computational modeling reveals a complex and heterogeneous hemodynamic environment within the dissected superior mesenteric artery. The high-velocity flow, elevated WSS, and TAWSS in the TL may contribute to flap instability and inflammation, while the low-flow, stagnant conditions, low WSS, and high OSI in the FL likely promote thrombogenesis. This patient-specific approach provides valuable mechanistic insights into SISMAD pathophysiology and demonstrates potential for personalized risk assessment and data-driven treatment planning in this rare but serious vascular condition.