Book of Abstracts the MCSB 2026. The International Conference on Cybernetic Modeling of Biological Systems (MCSB 2026) continues a long-standing interdisciplinary tradition initiated in 1978, bringing together researchers from medicine, life sciences, engineering, computer science, and biomedical informatics. The 2026 edition, organized in Kraków by the Jagiellonian University Medical College and the AGH University of Kraków, focuses on the rapidly evolving landscape of digital medicine and cybernetic approaches to biological systems. The conference addresses contemporary challenges and opportunities related to artificial intelligence, biomedical signal processing, bioinformatics, digital twins, virtual and mixed reality, telemedicine, personalized medical technologies, and clinical decision support systems. Special emphasis is placed on the integration of large language models, mobile health solutions, and data-driven approaches into healthcare, education, and biomedical research. The scientific program also explores physiological modeling, systems biology, robotics, bioengineering, and computational methods supporting diagnosis and therapy. MCSB 2026 serves as a platform for interdisciplinary collaboration and knowledge exchange between scientists, clinicians, engineers, and young researchers. The conference highlights both theoretical foundations and practical applications of cybernetic modeling in medicine and healthcare systems. In addition to presenting innovative research results, the meeting promotes discussion on ethical, educational, and organizational aspects of implementing advanced digital technologies in clinical practice. The Book of Abstracts reflects the broad thematic scope and international character of the conference, presenting current trends and emerging directions in biomedical engineering, digital health, and computational medicine.
Objective: Hypoxia is a critical factor in tumour aggressiveness, metastasis, and treatment resistance. Despite its clinical significance, a non-invasive, high-precision method for assessing and mapping tissue oxygenation in vivo remains a major challenge in medicine.This study explores the potential of positronium imaging and quantum entanglement (QE) imaging as next-generation biomarkers for hypoxia. This manuscript introduces a novel method to assess tissue oxygen concentration via the QE of photons originating from positronium – a bound state of an electron and a positron – which is produced within the patient’s body during positron emission tomography (PET). We also investigate the possibility of assessing hypoxia by simultaneously detecting positronium lifetime and the positronium decay rate ratio. Methods: We introduce two distinct quantum sensing approaches. Method 1 utilises the correlation between oxygen concentration and ortho-positronium (o-Ps) decay rates, relying on the simultaneous measurement of the mean o-Ps lifetime (τoPs) and the 3γ-to-2γ annihilation rate ratio of o-Ps (RoPs-3γ/2γ). Method 2 introduces a novel hypothesis based on the degree of QE of annihilation photons. This method leverages recent discoveries indicating that the degree of QE is sensitive to the relative contribution of annihilation mechanisms (pick-off vs. conversion), which in turn depends on the oxygen concentration. We estimate the rate of conversion processes as a function of the partial pressure of oxygen (pO2) in water, isopropanol, cyclohexane, isooctane, and adipose tissue. We consider dependence of RoPs-3γ/2γ and τoPs, as well as the degree of QE, on oxygen pressure in the studied substances. Finally, we derive a formula for pO2 as a function of RoPs-3γ/2γ and τoPs and estimate the measurement accuracy required for these parameters – and for the degree of QE – to sense in-vivo oxygen pressure in the range between hypoxic and physoxic conditions. Results: Theoretical models and quantitative estimates for RoPs-3γ/2γ, τoPs and for the degree of QE (CQE and RQE) as a function of pO2 are provided for water, organic solvents (isopropanol, cyclohexane, isooctane), and adipose tissue. Applying the formulas derived under the working hypothesis that in pick-off process the photons are not entangled, we estimated that for pO2 = 0, the degree of QE CQE is equal to 0.890 for adipose, 0.886 for isopropanol, 0.867 for water, 0.818 for cyclohexane, and 0.784 for isooctane. These results indicate that distinguishing between various tissue types requires a precision of σ(CQE) ≈ 0.01. We also estimated the values of τoPs, RoPs-3γ/2γ, and R3γ/2γ, as well as the changes in τoPs, RoPs-3γ/2γ and CQE between physoxic (pO2 ≈ 50 mmHg) and hypoxic (pO2 = 10 mmHg) conditions amount to ΔτoPs = 5 ps (water), 48 ps (adipose), 182 ps (isopropanol), 187 ps (cyclohexane), 444 ps (isooctane), ΔRoPs-3γ/2γ = 0.000045 (water), 0.0004 (adipose), 0.0015 (isopropanol), 0.0015 (cyclohexane), 0.0036 (isooctane), and ΔCQE = 0.0003 (water), 0.0019 (adipose), 0.0054 (isopropanol), 0.0106 (cyclohexane), 0.0243 (isooctane). To distinguish between physoxic and hypoxic conditions in vivo, the measurement precision σ(τoPs), σ(RoPs-3γ/2γ) and σ(CQE) must be several times higher than the predicted changes in these values. Conclusions: This work establishes a theoretical and methodological foundation for using the QE of photons and the properties of positronium as transformative diagnostic tools for the non-invasive assessment and mapping of tissue oxygenation in the human body. It demonstrates that QE can serve not only as a tool for improving PET image quality through noise reduction, but also as a completely new category of biomarker. Quantitative estimations of the effect of oxygen pressure on positronium parameters, as well as the event statistics required for hypoxia assessment, are feasible using the multi-photon total-body J-PET scanner based on plastic scintillators. Furthermore, such assessments will be possible with next-generation, high-sensitivity crystal-based PET scanners, provided they are upgraded to support multi-photon signal acquisition for τoPs and 3γ-to-2γ imaging, and double Compton scattering for imaging CQE – the degree of QE.
Objective: This study aims to demonstrate that cellular heterogeneity, typically associated with cancer cells, also occurs in normal cells. The study also investigates the impact of the concentration of fibronectin in a solution used to functionalise polyacrylamide (PA) substrates on cell behaviour. Methods: Mouse embryonic fibroblasts (MEF 3T3) were cultured on PA with elasticities of 20 and 40 kPa, functionalised with fibronectin in concentrations of 1, 5, 10, and 20 μg/ml. Two-hour time-lapse brightfield microscopy was employed to acquire images of migrating cells. The resulting sets of images were used to quantify the cell-spread area and migration velocity. Based on these parameters and observed migration strategies, the cells were classified into distinct subpopulations. Results: Four distinct cell subpopulations were identified: mesenchymal, amoeboid, slow amoeboid and polygonal/bigonal. Cells grown on softer substrates showed greater sensitivity to fibronectin concentration than those on stiffer substrates. The cell area was maximal and the velocity minimal for cells grown on PA substrates functionalised with 10-μg/ml fibronectin, making their parameters most similar to cells observed on glass. Cells underwent spontaneous transitions between subpopulations, although no direct transitions between mesenchymal and amoeboid or slow amoeboid phenotypes were observed. Conclusions: Our findings highlight that cellular heterogeneity is not exclusive to cancerous cells and can also be observed in normal fibroblasts. Cells also can transit between them spontaneously. Moreover, cell behaviour is significantly influenced by the concentration of fibronectin in the functionalising solution, with a greater impact on softer substrates.
Introduction: This study aimed to develop a novel therapeutic strategy for treatment-resistant cancers based on nano-brachytherapy, using gold nanoparticles as carriers forthe Auger-electron-emitting radionuclides 197mHg and 197Hg. Methods: Five-nanometre PEGylated gold nanoparticles were functionalised with 197mHg/197Hg via surface amalgamation and evaluated in vitro in triple-negative breast cancer (MDA-MB-231) and glioblastoma (T98G) cell lines. Cellular internalisation and subcellular distribution were assessed by uptake studies and fractionation. Cytotoxicity was evaluated using MTS assays, while therapeutic efficacy in three-dimensional models was investigated in tumour spheroids. Flow cytometry was employed to analyse apoptosis and cell-cycle distribution. DNA double-strand breaks were quantified by gamma-H2AX phosphorylation.Ex vivo biodistribution studies compared intratumoural and intravenous administration routes. Results: The radio-conjugate underwent straightforward synthesis, exhibited very high radio-labelling efficiency at low mercury loading, and maintained excellent colloidal stability. Efficient cellularuptake and pronounced nuclearaccumulation were observed in both cell lines. In vitro studies demonstrated strong, time-and dose-dependent cytotoxicity predominantly mediated by apoptosis, with minimal necrosis, accompanied by significant induction of DNA double-strand breaks. In three-dimensional cultures, MDA-MB-231 spheroids underwent rapid, dose-dependent disintegration, whereas T98G spheroids displayed increased resistance. Ex vivo biodistribution revealed high tumour retention following intratumoural administration, minimal systemic exposure and predominantly renal clearance. Conclusions:197mHg/197Hg-functionalised gold nanoparticles represent a promising receptor-independent platform for localised nano-brachytherapy and warrant further preclinical evaluation in aggressive and therapy-resistant tumours.
Objective. In the fight against cancer, improving the detectors performance is crucial to enhance diagnostic accuracy and optimize therapeutic monitoring. Current clinical SPECT systems predominantly rely on NaI(Tl) crystals, which, despite their advantages, suffer from limited count-rate performance due to long scintillation decay times. Our goal is to address this limitation by developing an innovative class of plastic scintillators doped with high-Z elements. These materials combine the fast timing characteristics of organic scintillators with improved gamma-ray detection efficiency via enhanced photoelectric interaction probability. Methods. Our research has focused on synthesizing novel organic fluorophores to fabricate plastic scintillators doped with high-Z elements with concentrations up to 10%. Moreover, we explored different fabrication techniques, including thermal polymerisation, photoinitiated polymerisation and resin 3D-printing. Results. The resulting prototypes show promising characteristics in terms of optical transparency, dopant homogeneity, light yield and timing performance, reaching levels comparable with commercial standards. Conclusions. These novel scintillators will be at the core of a next-generation SPECT detector, in which the doped scintillators are polymerized directly into the holes of a 3D-printed tungsten collimator, with signal readout performed by tiled CMOS sensors to fully exploit the plastics’ timing properties, and FPGA modules for data pre-processing. Moreover, they will serve for the development of a compact portable dosimeter tailored for metastatic castration-resistant prostate cancer (mCRPC) patients undergoing Lu-177-PSMA-617 radio-metabolic therapy. This device is designed to retrieve the radiopharmaceutical washout curve without requiring multiple SPECT scans, in order to customize the radiopharmaceutical prescription by determining the patient-specific radiometabolic parameters.
Objective: Oxford Nanopore long-read sequencing enhanced the investigation of DNA and RNA modifications, provoking the rapid development of various computational methods for their detection. In this review, we aimed to present the selection of those tools as well as the existing benchmarking studies and application guidelines. Methods: We conducted a comprehensive literature review on the state-of-the-art tools for mapping of various modification types in DNA and RNA. Moreover, we referred to the benchmarking studies to sum up the existing recommendations for pipeline tailoring and discuss the current challenges. Results: Over the last few years, many modification detection algorithms have emerged, and this collection continues to grow. These methods can be categorized based on the features they rely on and the approach they use, including machine learning or deep learning, and statistical testing. For both DNA and RNA modification mapping, the choice of detection tool depends mainly on the type of modification of interest and the availability of reference samples. However, the unambiguous guidelines and gold standard protocol remain undefined. Conclusions: Research on the epigenome and epitranscriptome is likely to be permanently transformed by nanopore sequencing. In the near future, further development of methods for detecting modifications can be expected. There is also enormous potential for much-needed benchmarking studies, which are currently lagging behind tool updates. In addition, new biological datasets with established modifications are needed.
Objective. The goal of the work is to develop methods of calibrating the positron emission tomography system built from plastic scintillators, and to present results of the modular J-PET scanner calibration. Methods. Measurements with radionuclide 22Na and 44Sc (used as a point-like source and enclosed in a collimator) were performed using the modular J-PET scanner, and the data were analysed with a dedicated software framework. The detection modules were synchronised using signals from annihilation photons and prompt gamma. Results. The application of the time calibration methods yields a fully synchronised detector. Time-of-Flight resolution for modular J-PET is determined to be about 490 ps (FWHM). Conclusions. J-PET scanner built from plastic scintillators can be calibrated using β+γ emitters and taking advantage of the fact that the direction of propagation of annihilation and prompt photons are not correlated.
Objective: Positron Emission Tomography enables non-invasive imaging of metabolic processes. Standard PET reconstructs radiotracer distribution using two back-to-back photons from electron-positron annihilation. However, about 1% of annihilations results in creation of three photons. Three photons may be created in direct annihilation and in annihilation via formation of metastable ortho-positronium carrying additional information not used by conventional PET. The Jagiellonian-PET detector allows us to use this information by applying positronium imaging based on ortho-positronium lifetime or the 3γ/2γ decay ratio. Accurate tomographic images require attenuation correction. This study investigates photon absorption in phantom models to form a basis for future attenuation maps for three-gamma decays. Methods: Monte Carlo simulations in ROOT and GATE were performed for water sphere, cylinder, a simplified head model, and the mesh50_XCAT phantom. Both p-Ps and o-Ps decays were simulated as uniformly distributed sources. Absorption probabilities were calculated by checking whether any photon in a multiplet interacted within the phantom. Emission-point-specific absorption maps were generated for all models. Toy Monte Carlo and GATE simulations showed good overall agreement. Results: Photon triplets from o-Ps decays experienced higher absorption than photon pairs from p-Ps due to lower individual energies and higher attenuation. Absorption maps showed dependence of photon survival probability on the decay location. In the mesh50_XCAT phantom, 24.9% of p-Ps pairs and 10.3% of o-Ps triplets escaped without interaction. Conclusions: Gamma absorption depends strongly on positronium decay mode and location, with o-Ps events experiencing higher attenuation. The generated absorption maps provide the first step toward dedicated attenuation correction for three-gamma positronium imaging, enabling accurate reconstruction in novel 3γ/2γ positronium imaging technique. A Study presented indicates that the absorption of 3γ in the head is only about 2.5 times higher than for the 2γ, which is encouraging for further development of the 3γ/2γ rate ratio imaging.
MicroRNAs (miRNAs) are scalable biomarkers and therapeutic nodes. In acute ischemic stroke, in which reperfusion triggers rapid pathway shifts, we built a standardised enrichment workflow to recover robust, clinically relevant biology and to clarify tool performance, quantify the impact of target estimation, and provide practical, reproducible recommendations that non-bioinformatic end-users can adopt. We prioritized miR-19a-3p based on repeated signals in external datasets and links to endothelial/rt-PA biology. Predicted targets defined DIANA, miRDB, and intersection lists. KEGG and GO:BP enrichment was run in STRING, ShinyGO, and miRNET under standardised parameters; cross-tool correlation used Spearman rank, and GO:BP terms were collapsed into nine themes. Across KEGG, we observed shared patterns with tool-specific emphasis. ShinyGO and STRING were highly concordant (r≈0.85-0.95), consistently recovering signalling and adhesion/trafficking. miRNet often diverged for miRDB and the intersection (|r|≤0.06), preferentially amplifying cancer terms; concordance improved only on DIANA. Input choice shaped stability: miRDB and the intersection yielded lower FDRs and higher fold-enrichment, whereas DIANA showed flatter effects and greater dispersion. GO:BP mirrored this: ShinyGO/STRING aligned broadly, miRNET inflated nucleic-acid metabolism terms, yet “negative regulation of translation” was a cross-tool consensus peak. Thus, shared core biology is robust, while tool/input biases drive differences. Our recommendations are: favour the intersection to stabilize ranks/FDR; if a single predictor is required, prefer miRDB; treat DIANA-only findings as exploratory and confirm on miRDB or the intersection; if constrained to DIANA, rely on ShinyGO/STRING; and for any inter-tool comparison, standardise the background and FE/FDR, reporting both alongside the tool used.
Introduction: Extracellular vesicles (EVs) are membrane-bound structures that playa crucial role in intercellular communication and molecular transport. Due to theirbiological functions, EVs hold great potential as a novel diagnostic tool or as the targeteddrug carriers in anticancer therapies. Elevated glucose concentrations affect cellularmetabolism, thereby modifying the composition of the EVs and their lipid bilayer. This, inturn, can influence EVs' properties and morphology, which is important in their function. Hypothesis: Hyperglycemic conditions, by altering the composition and metabolism ofEVs, influence their size distribution and physicochemical properties. Objective: Visualisation and size distribution comparison of EVs isolated fromnormoglycemic (NG) and hyperglycemic (HG) conditions. Methods: Two complementary methods based on different physical principles were appliedand compared in terms of their applicability: nanoparticle tracking analysis (NTA) and cryo--transmission electron microscopy (Cryo-TEM). In this study, 1.1B4 cells were cultured innormo- (NG, 5 mM glucose) and hyperglycemic (HG, 25 mM glucose) conditions. Results: The mean hydrodynamic EV size under hyperglycemic conditions (236 28 nm)was slightly higher than the control (209 43 nm; Paired Sample Wilcoxon Signed Test,p = 0.05). Physical EV sizes (Cryo-TEM), as well as mode hydrodynamic sizes (NTA) wereconsistent and not affected by glucose concentration. In contrast to NTA, Cryo-TEM isa more precise technique for real-size distribution analysis, whereas NTA provides a rapidand cost-effective assessment of hydrodynamic radii. The theoretical level of detection ofNTA for EVs was close to what was observed in results. Conclusions: Hyperglycemic conditions slightly influenced only the mean hydrodynamic radius of EVs, indicating effects on EV corona size and the biogenesis of large EVs (lEVs).
This study proposes a methodology for the automated localisation of lower limb bones on T1 weighted MRI scans, employing a deep learning (DL) approach. The primary objective is to facilitate precise identification of skeletal structures, thereby supporting radiomics based diagnostics of neuromuscular disorders. The developed framework is not confined to the recognition of lower limb bones. A dataset of 1,243 MRI scans was used, with a subset of 29 manually labelled bone segmentations of six key lower limb bone classes. Axial slices were divided into training (2,283), validation (300), and hold-out (378) sets. A two part segmentation pipeline was developed using a combination of U-Net and ResNet architectures, with a custom cost function to handle variable label presence across slices. A novel method for obtaining precise bone-related slice location on the MRI volume was developed. Segmentation quality for Tibia and Femur was high, achieving 86.04% and 86.97% Dice score on the hold-out subset. The ResNet classifier correctly identified the defined regions on the volume, achieving AUCs over 97% on the hold-out subset for most leg fragments except for the knee label. This work introduces a new method for anatomical spatial localisation estimation in MRI scans. Unlike previous studies, which could only recognise body parts, the proposed method also estimates the precise bone-related slice location within the scanned volume, providing an added layer of anatomical context. The developed pipeline can be retrained for other body fragments. This solution exhibits a strong potential for use in clinical workflows, especially for studies involving musculoskeletal diseases.
Objective: Glycans – structurally diverse carbohydrates that decorate proteins and lipids – are fundamental regulators of biological processes. They influence protein folding and stability, receptor signalling, immune modulation, cellular migration and tissue homeostasis. Despite being the most abundant class of biomolecules in living organisms, their biomedical relevance was long neglected. One major challenge lies in their structural complexity: glycan profiles integrate both genetic and environmental inputs, making them highly dynamic indicators of physiological and pathological states. Consequently, aberrant glycosylation is increasingly recognised as a hallmark of disease. Methods: This article presents a subjective review of recent literature on the significance of glycosylation in biomedicine. Results and conclusions: Characteristic glycomic alterations are observed in cancer, autoimmune disorders, inflammation, congenital disorders of glycosylation (CDGs) and neurodegenerative diseases, positioning glycans as promising biomarkers and prognostic tools. Concurrently, analytical advances, including high-resolution mass spectrometry (MS), glycan arrays and mass spectrometry imaging (MSI), are enabling clinical translation. These technologies are transforming glycan analysis from a descriptive discipline into a driver of biomarker discovery, patient stratification and therapeutic innovation. This overview highlights the biomedical potential of glycans, illustrated through selected examples of analytical approaches and emerging diagnostic and therapeutic strategies.
Objective: The integration of omics technologies has opened new opportunities in toxicological research. This article aims to explore how toxico-proteomics and toxico-metabolomics contribute to the understanding of xenobiotic mechanisms, biomarker discovery, and modern risk assessment frameworks. Methods: Relevant literature was analysed to highlight recent advances in proteomics and metabolomics applied to toxicology. Particular attention was given to mass spectrometry-based approaches, spatial omics, in silico modelling, and combined omics strategies. Case examples from drug- and environment-related toxicology were used to illustrate practical applications. Results: High-resolution mass-spectrometry-based proteomics enables the sensitive detection of changes in protein levels, post-translational modifications, and proteinprotein interactions. Toxico-proteomic studies have clarified mechanisms of cardio-, hepato-, and atd-neurotoxic effects. Metabolomics supports the profiling of low molecular weight compounds and early responses to toxicants. Toxico-metabolomic analyses identified changes related to energy metabolism and amino acid metabolism. In vitro models and zebrafish embryos provided organ-specific insights. Integrating omics data has led to the identification of candidate biomarkers of exposure and toxic effects. Conclusions: Toxico-proteomics and toxico-metabolomics represent powerful tools for toxicology. Their application enhances the sensitivity of toxicity detection, reduces reliance on animal models, and supports the development of predictive strategies. As analytical platforms and computational tools continue to evolve, these disciplines are expected to play an increasingly central role in environmental and biomedical toxicology, with implications for diagnostics, therapeutics, and regulatory demands.
Objective: The use of neural networks for disease classification based on medical imaging is susceptible to variations in results caused by even a single-pixel change, a phenomenon known as a one-pixel attack, which should be examined qualitatively and quantitatively. Methods: For an extended dataset of brain MRI images representing four diagnoses, the networks VGG-16, ResNet-50, DenseNet-121, MobileNetV2, EfficientNet-B0, NASNetMobile, and ViT Base were implemented. Each model was trained three times on 96 × 96 inputs, with the best-performing trial selected for adversarial testing (Phase 1). The three most robust models from Phase 1 (VGG-16, MobileNetV2, EfficientNet-B0) were then retrained on 224 × 224 inputs to assess the effect of higher resolution on susceptibility (Phase 2). The susceptibility of a diagnosis change to a single bright pixel alteration in the input image was assessed, and an average number of vulnerable pixels (ANVP) per image was carried out. Results: At 96 × 96 resolution, the least vulnerable model was MobileNetV2 (ANVP: 20.45, susceptibility: 0.22%). This was followed by ViT Base (22.20, 0.24%), EfficientNet-B0 (38.55, 0.42%), DenseNet-121 (43.52, 0.47%), ResNet-50 (69.11, 0.75%), and VGG-16 (78.66, 0.85%). The most vulnerable was NASNetMobile (119.52, 1.30%). At 224 × 224 resolution, robustness further improved for EfficientNet-B0 (37.53, 0.07%) and MobileNetV2 (49.51, 0.10%), while VGG-16 remained less stable (99.44, 0.20%). Conclusions: Implementing disease classification based on medical imaging using neural networks may pose a potential risk of misinterpretation due to changes in data irrelevant to the study, which are clearly noticeable to a human.
Objective: To quantitatively assess how different preparation methods alter the nanoscale free-volume architecture of human plasma blood clots using positron annihilation lifetime spectroscopy (PALS), and to establish the relevance of PALS for fibrin network characterisation in biomedical research. Materials: Human plasma blood clots prepared in four distinct physical states: fresh, glutaraldehyde-fixed, desiccator-dried and critical-point-dried (CPD). Methods: High-resolution fast-coincidence PALS measurements were performed at 37 degrees C. Ortho-positronium (o-Ps) lifetimes and intensities were extracted using PALS Avalanche and independently verified by fitting in the Origin software package. Reported values represent mean +/- SD from n = 3 independent plasma clots prepared and measured under identical conditions. Comparative analysis across preparation states was conducted to assess preparation-dependent variations in o-Ps parameters. Results: Distinct o-Ps lifetime (tau 3) and intensity (I3) values were observed across clot states: fresh (1.95 +/- 0.01 ns; 11 +/- 0.07%), fixed (2.11 +/- 0.02 ns; 16 +/- 0.16%), desiccator-dried (2.15 +/- 0.04 ns; 6 +/- 0.11%), and CPD (1.95 +/- 0.01 ns; 9 +/- 0.07%). These differences indicate that preparation protocols substantially affect positronium (Ps) lifetime characteristics and formation probability within plasma clots. Conclusions: Sample preparation has a pronounced effect on PALS-derived o-Ps parameters in plasma blood clots, underscoring the sensitivity of PALS to preparation-induced nanoscale changes. To the best of our knowledge, this study provides the first systematic PALS-based comparison of plasma clots across multiple physical states, highlighting the importance of protocol selection when applying positronium-based techniques to biological materials.
Objective: The classification of high-frequency oscillations (HFOs) in epileptic brain signals remains a challenging task. As promising biomarkers, HFOs can assist in identifying epileptogenic zones in drug-resistant epilepsy, thereby supporting presurgical decision-making and seizure monitoring. This study aims to develop a robust framework for deep feature extraction and classification frameworks to improve automated HFO detection from intracranial EEG (iEEG) data. The framework was evaluated using iEEG recordings from 20 patients, with 28 intervals per patient, each segmented using a 300 ms time window, resulting in a total of 900 trials. Methods: A deep-feature extraction system based on the GoogLeNet architecture was implemented to obtain discriminative representations of iEEG segments. The extracted deep features were classified using four well-known machine-learning algorithms: support vector machine (SVM), multilayer perceptron (MLP), Gaussian na & iuml;ve Bayes (GNB) and random forest (RF). A suite of comparative experiments was conducted to evaluate the performance of the classification system and its robustness. Results: The proposed deep-feature extraction model, combined with the MLP classifier, achieved the best performance, with an overall accuracy of 95.18%, outperforming the other classifiers tested. Results confirm that deep features, extracted via GoogLeNet, significantly improve the precision of HFO classification compared to conventional feature-based methods. Conclusions: The proposed hybrid approach offers a reliable and efficient method for classifying HFOs in iEEG data. Its excellent performance demonstrates the potential for integration into clinical workflows to assist neurologists in identifying epileptogenic regions and monitoring seizures. This study contributes to advancing automated epilepsy diagnostics through the use of deep-learning-based feature extraction.
Objective: We present simulation results of a limited-angle time-of-flight positron emission tomography (TOF-PET) system designed for intraoperative surgical applications. The purpose of this study is to show the effects of various detector parameters on the resolution of reconstructed images. Methods: All simulated system configurations and parameters were performed in the GATE Monte Carlo package. In our simulation setup, detector modules typically used in whole-body PET (WB-PET) imaging were arranged in two parallel planes: one positioned above the patient's body and the other beneath the patient/surgery bed. We simulated a phantom consisting of a cold sphere with a diameter of 15 mm and four hot sphere regions with sphere diameters of 8 mm, 6 mm, 4 mm and 2 mm. Several parameters were simulated, including depth-of-interaction (DOI), full width at half maximum (FWHM) of the coincidence time resolution (CTR), crystal thickness, pixel size, uptake ratio and background water depth. To assess image resolution we employed simple back projection (SBP) reconstruction due to its fast speed compared to list-mode maximum likelihood expectation maximisation (MLEM). We evaluated the quality of the reconstructed images using contrast-to-noise ratio (CNR), the contrast recovery coefficient (CRC) and signal-to-noise ratio (SNR) metrics. The data acquisition length was set to 1 minute. Results and conclusions: Different parameters were simulated, and impact on the reconstructed image resolution is evaluated. Using detectors with 100-ps CTR it is possible to detect 4-mm spheres with a 12-cm thickness of warm background, even with a low uptake ratio (5:1). Additionally, in the same thickness, 2-mm spheres with a larger uptake ratio (10:1) can be resolved with a 50-ps CTR. Our results indicate that with one-minute acquisition and high timing resolution, use of intraoperative imaging with high resolution is achievable.
This editorial article outlines the origins, development and scientific mission of Bio-Algorithms and Med-Systems on the occasion of its 20th anniversary. It reconstructs the historical context of early 21st-century Poland, when interdisciplinary collaboration between medicine, computer science and engineering was still uncommon and often met with scepticism. The text describes the pioneering role of the Jagiellonian University Medical College and the AGH University of Science and Technology in promoting biomedical informatics, cybernetics and biomedical engineering - fields that would later become essential to modern healthcare. It also recounts the establishment of the journal in 2005 as a response to the lack of publication venues for interdisciplinary work combining bio-phenomena, technical sciences and medical applications. The article presents the journal's contribution in shaping the newly emerging discipline of biomedical engineering in Poland, its early publishing philosophy, and its evolution through various editorial and publishing stages. Finally, the authors reflect on the journal's legacy, emphasising the importance of interdisciplinary cooperation, technological innovation and ethical frameworks as prerequisites for scientific progress.
Objective: The structure of proteins has encoded the record of biological activity in the form of a specific polarity-hydrophobicity relationship system. Structural changes related to the function change this record according to the process stages in which the given protein is used. Methods: The fuzzy oil drop model (FOD-M) was used to describe such phenomena in the aspect of the polarity/hydrophobicity relationship. Results: Application of this model allows quantitative assessment of the status reflecting the specific nature of individual proteins. Examples of this application are discussed in this publication The local exposure of hydrophobicity is a specific record of the possible hydrophobic interaction with another protein, providing an agent stabilising such a system. Conclusions: The local hydrophobicity deficit means the presence of a cavity, ready to interact with a substrate in the case of an enzyme. Other cavity types may be adapted to interact with a ligand comprising a permanent ingredient of a complex, often guaranteeing biological activity. In the case of membrane-anchored proteins, exposure of hydrophobic residues is a typical example, rendering the proteins stable in the membrane environment.
Objective: The aim of this study is to develop and evaluate a preprocessing pipeline aimed at reducing technical variability in histomic features extracted from whole-slide images (WSIs) of endometrial cancer tissue using foundation models, thus improving the reliability and generalisability of downstream computational pathology analyses. Methods: Haematoxylin and eosin (H&E) stained images from three datasets (TCGA UCEC, CPTAC UCEC and Cracow UCEC) were preprocessed using a pipeline that included filtration of artifacts and Vahadane-stain normalisation. Four histopathological foundation models (ResNet18 Histo, CTransPath, UNI, H-optimus-0) were used as feature extractors. Batch effects were evaluated before and after preprocessing using UMAP visualisation plots and LISI and ARI metrics. Results: For each foundation model tested a strong batch effect was observed. The proposed method allowed for the reduction of technical batch effect arising from differences between datasets (e.g., for ResNet18 Histo, ARI decreased from 0.7780 to 0.1428, and LISI increased from 0.0043 to 0.1393), but it did not reduce variability related to the tissue-embedding medium for most models. Additionally, biological diversity associated with tumour grade decreased slightly (e.g., for UNI, ARI decreased from 0.0391 to 0.0384, and LISI increased from 0.5658 to 0.6485). Conclusions: The introduced preprocessing approach effectively mitigates technical variability in histomic features extracted by foundation models, improving the data preparation for model building.
