Ústav biomedicínského inženýrství
Browse
Recent Submissions
Now showing 1 - 5 of 163
- ItemEnhanced metabolomic predictions using concept drift analysis: identification and correction of confounding factors(Oxford Academic, 2025-04-04) Schwarzerová, Jana; Olešová, Dominika; Šabatová, Kateřina; Kvasnička, Aleš; Koštoval, Aleš; Friedecký, David; Sekora, Jiří; Dluhá, Jitka; Provazník, Valentýna; Popelinsky, Lubos; Weckwerth, WolframMotivation The increasing use of big data and optimized prediction methods in metabolomics requires techniques aligned with biological assumptions to improve early symptom diagnosis. One major challenge in predictive data analysis is handling confounding factors—variables influencing predictions but not directly included in the analysis. Results Detecting and correcting confounding factors enhances prediction accuracy, reducing false negatives that contribute to diagnostic errors. This study reviews concept drift detection methods in metabolomic predictions and selects the most appropriate ones. We introduce a new implementation of concept drift analysis in predictive classifiers using metabolomics data. Known confounding factors were confirmed, validating our approach and aligning it with conventional methods. Additionally, we identified potential confounding factors that may influence biomarker analysis, which could introduce bias and impact model performance. Availability and implementation Based on biological assumptions supported by detected concept drift, these confounding factors were incorporated into correction of prediction algorithms to enhance their accuracy. The proposed methodology has been implemented in Semi-Automated Pipeline using Concept Drift Analysis for improving Metabolomic Predictions (SAPCDAMP), an open-source workflow available at https://github.com/JanaSchwarzerova/SAPCDAMP.
- ItemNo hyponatremia despite continuous plasma sodium decline in female runners during a seven stage ultramarathon(Nature Portfolio, 2025-04-03) Chlíbková, Daniela; Filipenská, Marina; Knechtle, Beat; Rauter, Samo; Trnka, Martin; Weiss, Katja; Rosemann, ThomasThe role of sodium supplements and sex in the occurrence of exercise-associated hyponatremia (EAH) remains controversial. This study investigated hydration status in ultrarunners (19 males and 9 females) who completed seven marathons over seven consecutive days. Due to the limited number of female participants, no statistical comparison between sexes was performed. Plasma sodium concentration ([Na+]) and multiple hydration markers were assessed before, during, and after the race. Reported sodium supplement consumption showed no association with plasma [Na+]. An overall decline in plasma [Na+] was observed in females (regression slope = -1.278, p=0.02) across the event, whereas no significant change was detected in males (slope = -0.325, p=0.57). Additionally, no significant associations were found between plasma [Na+] and other monitored variables, including sodium supplement intake, pre-race hydration strategy, body mass, total body water, plasma osmolality, hematocrit, hemoglobin, urine specific gravity, urinary [Na+], thirst rating, or fluid intake reported pre-, during, and post-stage. No cases of symptomatic or asymptomatic hyponatremia were identified, suggesting that total fluid and sodium intake were adequate to maintain fluid-electrolyte balance and prevent EAH in both sexes.
- ItemAnalyzing the performance of biomedical time-series segmentation with electrophysiology data(NATURE PORTFOLIO, 2025-04-06) Ředina, Richard; Hejč, Jakub; Filipenská, Marina; Stárek, ZdeněkAccurate segmentation of biomedical time-series, such as intracardiac electrograms, is vital for understanding physiological states and supporting clinical interventions. Traditional rule-based and feature engineering approaches often struggle with complex clinical patterns and noise. Recent deep learning advancements offer solutions, showing various benefits and drawbacks in segmentation tasks. This study evaluates five segmentation algorithms, from traditional rule-based methods to advanced deep learning models, using a unique clinical dataset of intracardiac signals from 100 patients. We compared a rule-based method, a support vector machine (SVM), fully convolutional semantic neural network (UNet), region proposal network (Faster R-CNN), and recurrent neural network for electrocardiographic signals (DENS-ECG). Notably, Faster R-CNN has never been applied to 1D signals segmentation before. Each model underwent Bayesian optimization to minimize hyperparameter bias. Results indicated that deep learning models outperformed traditional methods, with UNet achieving the highest segmentation score of 88.9 % (root mean square errors for onset and offset of 8.43 ms and 7.49 ms), closely followed by DENS-ECG at 87.8 %. Faster R-CNN and SVM showed moderate performance, while the rule-based method had the lowest accuracy (77.7 %). UNet and DENS-ECG excelled in capturing detailed features and handling noise, highlighting their potential for clinical application. Despite greater computational demands, their superior performance and diagnostic potential support further exploration in biomedical time-series analysis.
- ItemSpeech production under stress for machine learning: multimodal dataset of 79 cases and 8 signals(Springer Nature, 2024-11-12) Pešán, Jan; Juřík, Vojtěch; Růžičková, Alexandra; Svoboda, Vojtěch; Janoušek, Oto; Němcová, Andrea; Bojanovská, Hana; Aldabaghová, Jasmína; Kyslík, Filip; Vodičková, Kateřina; Sodomová, Adéla; Bartys, Patrik; Chudý, Peter; Černocký, JanEarly identification of cognitive or physical overload is critical in fields where human decision making matters when preventing threats to safety and property. Pilots, drivers, surgeons, and operators of nuclear plants are among those affected by this challenge, as acute stress can impair their cognition. In this context, the significance of paralinguistic automatic speech processing increases for early stress detection. The intensity, intonation, and cadence of an utterance are examples of paralinguistic traits that determine the meaning of a sentence and are often lost in the verbatim transcript. To address this issue, tools are being developed to recognize paralinguistic traits effectively. However, a data bottleneck still exists in the training of paralinguistic speech traits, and the lack of high-quality reference data for the training of artificial systems persists. Regarding this, we present an original empirical dataset collected using the BESST experimental protocol for capturing speech signals under induced stress. With this data, our aim is to promote the development of pre-emptive intervention systems based on stress estimation from speech.
- ItemMethod Comparison for Bone Density in Multiple Myeloma Patients(Czech Society for Biomedical Engineering and Medical Informatics, 2024-09-30) Nohel, Michal; Mézl, Martin; Válek, Vlastimil; Dostál, Marek; Chmelík, JiříBone mineral density (BMD) is an important indicator of bone health, particularly in patients with conditions such as multiple myeloma. This study aims to compare three methodologies for quantifying BMD in vertebral regions affected by lytic lesions: two using data from conventional CT with different corrections for tissue composition, and one using data acquired on a dual-energy CT system. Method 1 is based on conventional CT with corrections using reference values for muscle and fat, Method 2 uses conventional CT with corrections based on the measured CT values of paraspinal muscle, and Method 3 is based on dual-energy CT. The Wilcoxon signed-rank test was used for statistical comparison, as the dataset did not follow a normal distribution. The results indicated significant differences between Methods 1 and 2 for BMD in regions of interest (ROIs) within lytic lesions, while no significant differences were found for other comparisons in this group. For vertebrae affected by multiple myeloma, significant differences were found between Methods 1 and 2, and Methods 2 and 3, but not between Methods 1 and 3. In healthy vertebrae, a significant difference was found only between Methods 2 and 3. When all ROIs were combined, significant differences were found between Methods 1 and 2, and Methods 2 and 3, with no difference between Methods 1 and 3. Future research will focus on objectively assessing the accuracy of these methods by comparing their results with a calibration phantom.