Selected scientific and technical publications
AI-supported Technologies
peer-reviewed
Authors: Alessandro Molani, Janos Palhalmi, Anna Mező, Francesca Pennati & Andrea Aliverti
Journal: Nature – Scientific Reports, 2026
DOI: https://doi.org/10.1038/s41598-026-61176-4
Abstract
The integration of artificial intelligence (AI) with digital holographic microscopy (DHM) is transforming optical methods for particle detection and classification, particularly in biosecurity and biosafety. However, existing AI-DHM methods are typically developed or fine-tuned on hardware-specific experimental datasets, limiting their generalizability across different configurations and applications. A standardized and hardware-agnostic benchmarking framework for systematically evaluating AI performance in DHM is currently lacking.
This study introduces a simulation and evaluation framework designed to generate reproducible, parameter-controlled synthetic datasets of raw holographic images, configurable to replicate specific optical setups while remaining independent of any particular hardware implementation. The framework provides a flexible environment for pre-training and testing open-source and proprietary machine learning (ML) and deep learning (DL) models, enabling transfer learning strategies and guiding the design and optimization of optical setups and computational pipelines. Here, the framework is demonstrated for the recognition of micrometric and submicrometric particles relevant to biosecurity scenarios, where particle size and concentration are critical operational parameters that directly influence sampling, filtration strategies and downstream AI analysis. By systematically varying these parameters alongside optical configurations, the framework is used to investigate their individual and combined effects on detection and classification accuracy.
Results demonstrate the robustness of DL models, even under challenging conditions with small particles and high concentrations, while ML approaches are more sensitive to fringe overlap. Overall, this work provides a reproducible and extensible environment for the development of AI-driven microscopy systems, offering quantitative insights for optimizing experimental design, algorithm development and sample preparations in both laboratory and field applications, prior to the deployment on specific optical setups.
Download full paper:
Biosensor Validation
Publications related to optical biosignal measurement, PPG raw signal quality, and sensor-performance comparison.
peer-reviewed
Authors: Dr. Janos Palhalmi (PhD) and Jan-Hein Broeders
Journal: Computing in Cardiology 2019, Vol 46
DOI: 10.22489/CinC.2019.165
Abstract
Biometric metrology is becoming increasingly important as the wearable application specific biosensors are capable of generating accurate raw signals representing certain vital states. A comparative statistical approach has been worked out to answer questions arising from a biosensor testing measurement technology perspective. In this specific work two high quality optical solutions (Analog Devices’ ADPD107 and ADPD188GGZ) were compared to analyse the deep data level similarities or differences between the photo-plethysmography (PPG) raw biomedical signals generated by the two individual systems.
Two minutes long parallel recordings were carried out with the evaluation boards of both optical systems (EVAL-HCRWATCH and EVAL-ADPD188GGZ) on 11 healthy human subjects. Recordings were repeated on both wrists to avoid side specific influence. Both systems have been controlled by Analog Devices’ user-interface called “Application WaveTool”. For the test, configuration settings were optimized to achieve the highest signal quality and lowest power consumption (5.1 mW at 100 Hz sampling frequency). The physiologically and biometrically irrelevant frequency bands (<0.25 Hz,>40 Hz) were filtered before the data analysis.
A wavelet coherence-based method was worked out to compare the relevant frequency bands and two different correlation-based methods were developed to compare the wave-to-wave stability and similarity of the two compared signals. Magnitude squared coherence values were above 0.9 in all the explored frequency bands (0.25-40 Hz). Correlation coefficients were never under 0.9 while p values were always under 0.001 in case of the PPG wave-to-wave comparisons. Based on the results we can conclude, that the quality of the PPG signal recorded by the new ADPD188GGZ integrated optical module, reaches the same performance level as the ADPD107 discrete module, with key benefits such as small PCB area, ease of use (especially for companies with limited Optical expertise) and shorter Time to Market.
Download full paper:
technical article
Authors: Dr. Janos Palhalmi (PhD) and Jan-Hein Broeders
Journal: Analog Devices Instruments, Technical Articles
Abstract
Photoplethysmography (PPG) is a common technology for measuring oxygen saturation (SPO2) levels in blood. Light is sent by a light emitter into the body, and the amount of reflective or unabsorbed light is measured with a photoreceiver. Depending on the ratio between two wavelengths, the amount of oxygenated hemoglobin can be measured. Comparable technologies are also used to measure heart rate with an optical technology or heart rate variability.
All these systems require one or more photoemitters, which need to be controlled, and a photodetector to measure the amount of photocurrent as a measure for the received light. This receive signal finally needs to be amplified, conditioned, and digitized. Such an optical system might sound straightforward; however, with a missing dose of optical knowledge, it is very easy to retrieve an optical signal, which doesn’t have anything to do with the signals the user is looking for.
To help companies achieve their optical objectives, a new, fully integrated optical module has been introduced. It has been tested and compared to a well-proven discrete optical system with outstanding results. You will read more about the results and methodologies behind this exercise.
Download full paper:
Biosignal Metrology and Digital Biomarker Methodology
Publications related to physiological measurements, uncertainty, biometric features, and biomarker-oriented computational analysis.
peer-reviewed
Authors: Dr. Janos Palhalmi (PhD)
Journal: Computing in Cardiology 2020, Vol 47
DOI: 10.22489/CinC.2020.359
Abstract
The most common hemodynamic measurement technologies are the classical auscultatoric sphygmomanometric method and numerous solutions for automated sphygmomanometry. The classical method is the reference for testing the automated solutions according to the ISO81060-2:2018 guideline.
Both of the above mentioned blood pressure estimations are indirect measurement methods according to the approach of metrology.
According to the ISO81060-2:2018 guideline, two indirect measurement and estimation methods are compared with each other without the theoretical and technical consideration of the possible effect of the measurement uncertainty on the results.
Within this simulation study several components of the measurement uncertainty were implemented to test the reliability and confidence limit of the ISO guideline recommended process flow, experimental design and statistical analysis.
Download full paper:
peer-reviewed
Authors: Dr. Janos Palhalmi (PhD)
Journal: Honvédorvos 2019, (71)/3-4. (Journal of Military Medicine)
DOI: 10.29068/HO.2019.3-4.42-51
Abstract
In the field of blood pressure management, the estimated ratio of false positive and false negative diagnoses caused by inappropriate measurement technology is remarkable. Due to the above reason, several attempts have recently been made to harmonize the related ISO/IEC and AAMI protocols, in order to support the development of more accurate evaluation and classification methods.
The fundamental standards of the classical auscultatoric sphygmomanometric and oscillometric blood pressure measurement methods have been in use for decades regarding both the clinical routine applications and the validation standards.
While these classical methods can detect prominent pathological changes under clinical circumstances, their accuracy is questionable under continuous performance monitoring applications.
The author proposes a new approach based on the classical measurement technologies but applying newly developed bioinformatics algorithms. This approach provides the possibility of monitoring minor differences of the haemodynamic regulation and controlling of the major components of the measurement uncertainty so as to support decision making during aviation physiological and aerospace medical examinations.
Download full paper:
conference-paper
Authors: Dr. Janos Palhalmi (PhD)
Journal: Computing in Cardiology 2018, Vol 45
DOI: 10.22489/CinC.2018.080
Abstract
Repolarization heterogeneity expressed by QT interval prolongation and abnormal temporal dynamics of the QT interval time series is an important factor in relation to coronary heart disease and lethal arrhythmias.
Based on our observations, the calculation of window correlation between the mean and variance of features extracted from QT interval time series can reveal natural and disease specific fluctuation patterns. Our algorithm is potentially a sensitive biometric measure to quantify personalized differences and the properties of repolarization heterogeneity, and also a potential biomarker to characterize disease specific QT interval temporal dynamics.
Download full paper:
Environmental Monitoring & Bio-threat Detection
Publications related to air samples, microbial composition, digital holographic microscopy, AI-supported microscopy, and CBRN-related detection methodology.
peer-reviewed
Authors: János Pálhalmi, Marcin Niemcewicz, Łukasz Krzowski, Anna Mező, Rafał Szelenberger, Marcin Podogrocki and Michal Bijak
Journal: Applied Sciences 2025 15(4) MDPI
DOI: https://doi.org/10.3390/app15041778
Abstract
This study examines the differences in particulate matter (PM) properties and microbial compositions between natural and urban environments, providing foundational data for environmental monitoring and biothreat detection.
Air samples were collected during the spring and early summer from two distinct locations: a forest/lake area, and an urban parking lot adjacent to a high-traffic roadway. Quantitative phase imaging microscopy and genomic sequencing were employed to characterize particle size distributions, statistical properties, and microbial community structures in these environments.
The results revealed significant differences in PM properties between the two locations. Urban air exhibited higher particle concentrations that reflect pollution sources, whereas the natural environment displayed greater variability in particle size and distribution, correlating with diverse biological content. Genomic sequencing showed a lower diversity of microbial communities compared to the forest/lake area but with greater uniformity.
To sum up, by integrating optical microscopy and genomic sequencing, this research demonstrates the feasibility of establishing environmental baselines for PM characteristics and bio-component diversity. The findings underscore the potential of combining real-time optical sensing with genomic tools for early biothreat detection and improved environmental monitoring in diverse settings.
Download full paper:
peer-reviewed / book chapter
Authors: Janos Palhalmi and Anna Mezo
Journal: Security Informatics and Law Enforcement Series – Paradigms on Technology Development for Security Practitioners (pp 117-127) – Springer Nature Link
DOI: https://doi.org/10.1007/978-3-031-62083-6_10#DOI
Abstract
Because Bacillus anthracis is one of the most lethal bioweapons, it is critical to create rapid, label-free screening and early warning systems to detect and classify anomalies in bacillus form vegetative cell and spore concentrations in the air.
Even though significant effort has been invested in the development of various sensor solutions to detect, monitor, and identify airborne biological agents, no standard, interoperable, real-time or near-real-time optical sensor-based biothreat monitoring solution exists. Aside from the numerous advantages of genomic methods in microbe identification, optical sensors and microscopy-based technologies provide advantages in terms of rapid detection and classification capabilities.
The AI-powered biothreat detection software platform from DataSenseLabs can perform intermethod comparison to cross-validate the results acquired by various quantitative phase imaging (QPI) measurement methodologies. This platform feature—support for multisensory data input—is not merely the foundation of the R&D level cross-validation approach, but also the key component of interoperable verification of air sample content in the case of airborne biothreat.
Depending on the study design, sample type, and light microscopic or QPI measurement method, the platform’s algorithm system can detect and monitor abnormalities in the concentration of bacillus form objects taken from the air with greater than 80–95% accuracy. Another goal of the platform is to serve as a standardized tool for biomedical, environmental, and CBRN scientists to train and validate their concepts in pathogen detection and classification use cases, allowing them to better understand the gaps and challenges associated with artificial intelligence-powered optical sensor systems.
Download full paper:
conference paper
Authors: Molani, A.; Pennati, F.; Aliverti, A.; Pálhalmi, J.
Journal: 2023 IEEE International Conference on Metrology for eXtended Reality, Artificial Intelligence and Neural Engineering (MetroXRAINE)
DOI: 10.1109/MetroXRAINE58569.2023.10405639
Abstract
Bioaerosols are an important component of air quality and their detection has emerged as a central concern. Current bio-detection tools do not meet the need to be fast, autonomous and ready for use in the field. Artificial intelligence-powered digital holographic microscopy (DHM) offers a possible solution for real-time, cost-effective and field-deployable bioaerosol analysis.
This study investigates the characteristics of DHM using a simulation-based approach to define the theoretical limit of detection for submicron objects under different experimental conditions. Spherical particles with different diameters were simulated according to Mie scattering theory, optimising their size and refractive index to be comparable to bacteria. Statistical analysis was used to analyse particles’ holograms, both individually and across a range of concentrations.
The evidence from this study indicates that air rather than water medium allows more reliable identification of smaller particles. In addition, basic statistical indices enable the differentiation between particles and the background, as well as discrimination among particles based on their diameter, even with higher particle concentrations.
Overall, this study suggests that DHM allows for detection and distinction of airborne micro- and even nano-particles. The theoretical detection limit with a lensless setup is around 500 nm, while with a lens-based setup it can reach at least 30 nm under ideal conditions.
Download full paper:
peer-reviewed
Authors: Molani, A.; Mihalik, B.; Pennati, F.; Rahi, P.; Mező, A.; Pálhalmi, J.; Aliverti, A.; Bela, G
Journal: The European Physical Journal Plus. Springer Nature Switzerland
DOI: https://doi.org/10.1140/epjp/s13360-024-05672-4
Abstract
There is a global need to advance bio-aerosol sensing for CBRN (Chemical, Biological, Radiological, and Nuclear) applications by compact and cost-effective devices. Employing digital holographic microscopy (DHM) and deep learning, we developed a system called HoloZcan to automate the analysis of airborne microbial pathogens and particles.
DHM provides valuable information, but obtaining data from biological specimens for robust investigations is challenging. This paper introduces a custom simulation approach using the open-source software Meep and the finite-difference time-domain (FDTD) method to overcome limitations of existing Mie-based simulators, especially when dealing with complex microbial shapes. The simulation tool enables the modelling of specific microorganisms, offering a safer and more flexible alternative for CBRN research by bypassing ethical and logistical constraints associated with live pathogens.
The study details the simulation workflow, built upon the construction of a database of optical properties of biological materials, for realistic simulations of light-microbe interactions. Evaluations on homogeneous and non-homogeneous objects demonstrate the tool’s limited intrinsic errors and superior sensitivity to refractive index changes compared to traditional Mie-based simulations.
This work significantly advances our capability to accurately simulate and analyse CBRN-related scenarios, enhancing comprehensive research in bio-aerosol sensing.
Download full paper:
Scientific background, connected to practical work
If your next research, validation, or technology decision depends on reliable scientific evidence, we can help you interpret the methods, data quality, and uncertainty behind it. We’ll look at your situation with you and help clarify what is already strong, what may need further validation, and what the most practical next step could be.