Technical tools in data management and processing

• Head: Yang Bai, PhD, Docent, Assistant Professor (tenure-track)
• Faculty: Faculty of Information Technology and Electrical Engineering
• Group composition: One assistant professor, one postdoc, five doctoral researchers, and one project designer
• Websites: https://www.oulu.fi/en/researchers/yang-bai
• https://www.oulu.fi/en/university/faculties-and-units/faculty-information-technology-and-electrical-engineering/microelectronics
• Contact: yang.bai (at) oulu.fi

Physical body signals such as conscious and unconscious muscle movement, including change of muscle status, reflect early signs of potential human health issues. For instance, cough is the most common symptom for which individuals seek medical advice. In ICU (intensive care unit), especially for patients who have gone through tracheostomy, weak swallowing strengths may cause saliva going to lungs and thus inducing pneumonia. For spine trauma patients, the strength of breathing or coughing may indicate recovery state. Such types of human body signals need to be monitored closely and preferably autonomously. Beside acute and clinical patients, daily health status of outpatients should also be monitored wirelessly so that the number of follow-up visits to doctors can be reduced and thus to save healthcare resources. For instance, the number of patients recovering from anaesthesia is incredibly large, but their recovery requires careful monitoring. To fulfil the demand of acquiring massive health data from body signals, this research will explore feasible methods for data collection as well as analysis to deliver relevant critical health information using energy harvesting systems as energy-efficient, intelligent, and autonomous sensing systems. Energy harvesters convert ambient energy in the working environment, including optical, kinetical, and thermal energy, into electrical signals. Therefore, energy harvesters are naturally smart sensors which can dynamically respond to ambient stimuli. In this research, the stimuli will be representative muscle movements triggered by coughing, breathing, swallowing, etc. The methodology of this research is to collaborate with clinics and/or medical researchers to figure out a series of appropriate body signals that can be reliably detected by energy harvesting systems and thus can be potentially used as health data which help to predict and diagnose early diseases and health status, respectively.

Key words: Piezoelectric, signal processing, machine learning

• Head: Neslihan Bayramoglu, PhD, Docent(Associate Prof.), Senior Researcher
• Faculty: Faculty of Medicine, Research Unit of Health Sciences and Technology
• Group composition: 3 professors, 1 senior researcher, 2postdocs, 8 researchers
• Website: Unit: https://www.oulu.fi/en/university/faculties-and-units/faculty-medicine/research-unit-health-sciences-and-technology
• Research group: https://www.ai4health.fi/
• Lead researcher: https://www.neslihan.ai/
• Contact: Neslihan.Bayramoglu (at) oulu.fi

The Research Group of Applied AI for Health (AAI4H) at the University of Oulu develops advanced machine learning and AI methods to address complex challenges in the medical and healthcare fields. Our primary research areas include medical imaging analysis, clinical decision support, early disease detection, identification of novel biomarkers, and novel therapy applications for various conditions.

Project Themes

The group offers the opportunity to work on several cutting-edge research themes:

• Medical Imaging: AI-driven diagnostic and prognostic tools for diseases such as osteoarthritis and cancer.
• Personalized Therapy and Intervention: AI models that enable personalized treatment plans and real-time therapeutic support.
• Predictive Modeling and Disease Stratification: Development of predictive models for risk stratification and early intervention.
• Virtual Therapeutic Applications: Innovations such as AI-enhanced virtual speech therapy for children and psychiatric assessment tools.

In this early phase of the group, the Postdoctoral Researcher will also have a leadership presence, contributing to the group’s scientific direction, shaping research strategies, and mentoring junior researchers. The successful candidate will have the opportunity to foster interdisciplinary collaboration, and work closely with clinical experts. With access to diverse medical datasets and cutting-edge computational resources, you’ll play a central role in a collaborative environment dedicated to pioneering applied AI for health.

If you have a strong background in medical AI and are motivated by the chance to influence high-impact research, we encourage you to apply. Please contact us in advance to discuss your research interests and research plan.

Key words: AI, Machine Learning, Medical Imaging, Computer Vision, Natural Language Processing, Medical Image analysis

• Head: Aliaksandr Bykau, PhD, DrSc(Tech), Adj. Professor, Biophotonics group leader
• Faculty: Faculty of Information Technology and Electrical Engineering/Optoelectronic and Measurement Techniques Research Unit
• Group composition: 1 Professor, 2 postdocs, 3 PhD students
• Website: Unit: https://www.oulu.fi/en/research-groups/biophotonics
• Contact: alexander.bykov (at) oulu.fi

The project will introduce a new paradigm for collecting tissue-specific data, addressing key challenges in biotissue diagnostics, such as inter-observer variability and the limited sensitivity/specificity of current methods. This approach will enable label-free, non-destructive imaging of tissue samples for histopathological analysis, providing valuable structural and biochemical information relevant for supporting decisions during treatment procedures. Central to this innovation is the concept of multidimensional pathology, which combines polarization and spectrally resolved optical biopsy technique utilizing complex phase-tailored light. This type of light leverages both Orbital Angular Momentum (OAM) and Spin Angular Momentum (SAM), offering advanced capabilities for detailed tissue characterization and improved diagnostic precision.

OAM of light has emerged as a transformative concept in photonics, offering a unique capability to encode information within the phase structure of light beams. This property holds great potential for enabling future innovations in high-resolution imaging, secure data transmission, and precise sensing applications. In addition, by precisely controlling the phase and polarization of structured light beams, it is possible to generate strongly focused light patterns with enhanced capabilities for deep tissue penetration and high-resolution imaging. These advanced light structures enable detailed, non-destructive examination of cellular architecture and pathological changes, facilitating early disease detection and precise localization of tissue malformations.

Exploring the interaction of complex light with various media is showing great promise, as it opens new pathways to overcome limitations in traditional photonic systems. Despite their success, conventional imaging and sensing techniques face significant challenges in complex, scattering environments where light's phase and intensity are often distorted. Moreover, existing methods struggle to preserve the integrity of optical signals over extended distances and to detect subtle variations within heterogeneous or dynamically changing materials. Addressing these limitations could unlock significant capabilities in optical imaging and sensing.

This research represents a hot topic at the intersection of photonics and biomedical engineering, unlocking new opportunities for addressing some of the most pressing challenges in tissue diagnostics.

The group is seeking a highly motivated and skilled postdoctoral researcher with expertise in experimental work, data processing, and numerical simulations. The candidate will be responsible for designing and conducting advanced experiments, primarily focusing on optical imaging and characterization techniques, as well as processing and analyzing the resulting data to extract meaningful insights.

Key words: optical diagnostics of biotissues, complex light, digital histopathology

• Head: Lauri Eklund, Professor
• Faculty: Biocenter Oulu (BCO) and Research Unit of Health Sciences and Technology, Faculty of Medicine
• Group composition: 2 Prof., 1 senior researcher, 4 post docs, 5 PhD students, 1 engineer, 2 technicians
• Websites: https://www.oulu.fi/en/university/faculties-and-units/biocenter-oulu and
• https://www.oulu.fi/en/university/faculties-and-units/faculty-medicine/research-unit-health-sciences-and-technology
• Contact: lauri.eklund (at) oulu.fi

We seek a post-doctoral fellow to participate in the project aiming to develop image analysis tools for quantitative and qualitative analysis of extracellular matrix (ECM) content and organization. Exploded data quantities have generated a bottleneck for scientific breakthroughs and innovations in biomedical imaging. There is an urgent need to develop automated image analysis tools to fuse multimodal microscopic data and provide quantitative outcomes with supporting visualization. Here, we are seeking a post-doctoral fellow to participate the development for harmonized data storage integrated with state-of-the-art image analysis including image segmentation with neural networks and complex morphometric analyses. The overall aim is to provide accurate information on collagen content and organization in the various sample materials. Quantification of image data will provide more information from qualitative image information gathered from e.g. various diseased tissues, and allow more accurate statistical analysis of changes in different fibrotic diseases and treatment responses.

ECM remodeling is essential for tissue homeostasis but when dysregulated, it leads to pathological processes including tissue fibrosis and cancer. Fibrosis and ECM research is one of the focus areas of the University of Oulu. FibroHealth is one of the UO interdisciplinary research fields comprised of several Faculties and covers two strategic research themes, fibrosis diseasome and fibrobesity, both supported by University proliferation funding from Research Council of Finland (11.5 M€).

Biocenter Oulu light and electron microscopy core facilities provide access to state-of-the-art confocal imaging techniques enabling real-time imaging of ECM in fixed and living cells, tissues and animals, and is selected as imaging service provider in European wide EuroBioimaging ERIC. 3D serial sectioning tomography of tissues, and light sheet and multiphoton microscopy of mesoscale optically cleared specimens provide additional imaging modalities to assess tissues and ECM macromolecular assemblies. In the project, a post-doctoral fellow will contribute for the development of image analysis tools for detecting abnormalities, changes in volume and quantity of ECM in various disease models including mouse and human samples.

The project is done in collaboration with Research unit of Health Science and Technology who are using x-ray microscopy, micro-computed tomography and spectroscopic imaging for in-depth understanding of ECM and fibrosis. Additionally, collaborations with various industry partners enhance our capabilities for research and development.

Key words: AI-assisted bioimage informatics, methodology development, fibrosis.

• Head: Caglar Elbuken, Professor
• Co-PI: Lauri Eklund, Professor
• Faculty: Faculty of Biochemistry and Molecular Medicine, Biocenter Oulu
• Group composition: 2 professors, 6 postdocs, 15 phd researchers
• Websites: https://www.oulu.fi/en/research-groups/elbuken-lab-microfluidics-and-biosensor-research-group
• https://www.oulu.fi/en/researchers/lauri-eklund
• Contact: caglar.elbuken (at) oulu.fi

Our groups work at the intersection of physics and biology focusing on cutting-edge research questions on vascularization, discrete particle distribution in networks, viscoelasticity of extracellular matrix, perivascular flow and disease modeling. We have developed imaging techniques for in vivo blood flow, biometrics and have acquired a wealth of digital data on vascular flow dynamics. We have on-going projects on dynamic vascular system modeling, in vitro vascularization, microfluidic hypoxia stations and organ-on-chip drug testing models. These projects produce huge amount of image data (e.g. video image sequences of 90 000 fps over several minutes, hours long 120 fps microcapillary flow at single cell resolution) requiring postprocessing steps of data analysis including mathematical approximation of autocorrelation function and fourier transform for spectral space information. In this project, we aim to develop a comprehensive model to study vascular architecture and flow characteristics by utilizing capillary flow data on brain and skin circulation from healthy mice, retinal network from databases and other vascular pathologies including amyloid angiopathy.

Advanced intravital vascular imaging technologies have generated vast amount of data at high resolution for extended durations in model organisms. These data have been used to study specific organ related diseases, such as retinopathies, as well as for systemic diseases caused by defective vascular functionality such as hypertension, diabetes or Alzheimer’s disease. Despite the existing data, our understanding of vascular architecture and flow functionality is limited due to complex network topology, tissue specific hemodynamics and physiological needs. The monitoring of hemodynamic changes and vascular remodeling happen at different time scales, hence a coarse time grain modeling is required to predict changes in long time scales while imaging at single cell spatial resolution. Additionally, the in vivo models allow limited controlled experiments in such a multiparameter space, hence there is significant need for in vitro experimental setups such as biomimicking microfluidic setups.

In this project, we aim to process big image datasets and develop a network-based model for fate mapping of a given vascular network with characteristic control of blood circulation. e.g cerebral vs. skin and retinal vasculature. Furthermore, we will utilize novel diffusion correlation spectroscopy (DCS) imaging technology for advanced blood flow visualization and comprehensive hemodynamic measurements. The in vivo data will be replicated by in vitro microfluidic experiments that allows fine tuning of geometry and flow parameters. Going beyond the existing hemodynamic models focusing on repartitioning due to plasma skimming and red blood cell discreteness effects, we will study real-time local hypoxic variations. This model will advance the existing vascular hemodynamics models with oxygen transport using data processing. In vitro microfluidic twin vascular systems will be used to correlate the function and the structure to study vascular pathologies. We foresee this project to deliver a generalized approach combining in vivo, in vitro and in silico data for various diseases where vascular pathologies have an important contribution.

Key words: Vascularization, vascular network modeling, vascular disease, big data analysis of video image data

• Head: Simo Hosio, Associate Professor
• Faculty: Faculty of Information Technology and Electrical Engineering, Center for Ubiquitous Computing
• Group composition: 1 Assoc. Prof., 1 postdoc, 4 doctoral researchers, 6 research assistants
• Website: https://ubicomp.oulu.fi/research/cc
• Contact: simo.hosio (at) oulu.fi

The Center for Ubiquitous Computing (UBICOMP) in the Faculty of Information Technology and Electrical Engineering welcomes a Postdoctoral Researcher to join the Crowd Computing (CC) research group (https://ubicomp.oulu.fi/research/cc), to conduct research on merging health informatics and data science to revolutionise the understanding and improvement of sleep quality and overall wellbeing.

UBICOMP is a vibrant research center with 65 employees, of which 70% are international research and teaching staff. In the most recent international assessment of the research conducted in the University of Oulu, UBICOMP was awarded the highest rating across all evaluation criteria. UBICOMP comprises of five independent research groups, of which the Crowd Computing (CC) group led by Associate Professor Simo Hosio investigates distributed forms of work, including crowd-powered decision support through modelling large-scale subjective data.

In this scope, you will have a unique opportunity to contribute to healthcare by employing the latest digital sleep-tracking technologies and data analysis techniques. The focus of the project is two-fold: 1) Prototyping new technologies for end-users, and 2) Discovering patterns and correlations between various lifestyle factors and sleep health. The objective is to engage with end-users who can directly benefit from your work.

Prototyping involves the development of mobile and web tools that integrate commercial API services from wearable vendors and combine this data with mobile usage data through visualisations and data analysis tools. The information is provided to end-users who can manage their personal data on the web platform.

Discovering patterns and correlations involves analysing wearable health data and mobile usage data to discover actionable data insights, patterns, and correlations between lifestyle and well-being. The project will use the latest machine learning and AI-based tools to enable users to engage through actionable data-driven insights - lifestyle changes that could have a measurable positive effect on their wellbeing and sleep.

Ultimately, you will have the freedom to choose your focus according to your personal interests.

In summary, your work will involve the collection and analysis of sleep data using state-of-the-art data science methodologies. This position promises professional growth through research and publication opportunities but also a chance to participate in the supervision of junior scholars. The project is distinctive in its approach to utilising real-world, user-generated data to drive healthcare interventions.

Key words: Wearable computing, human-computer interaction, mobile sensing

• Head: Jianan Huang, Assistant Professor
• Faculty: Faculty of Medicine, Faculty of Biochemistry and Molecular Medicine
• Group composition: Assistant Prof. Jianan Huang, Prof. Shuai Li, Dr. Kuo Zhan
• Website: https://www.oulu.fi/en/research-groups/jianan-lab-nanophotonic-biosensors
• Contact: jianan.huang (at) oulu.fi

Therapeutic drug monitoring (TDM) is a clinical practice that relies on the measurement of drug concentrations in blood or other body fluids. TDM adjusts the clinical use of drugs based on pharmacokinetic (PK) and pharmacodynamic (PD) principles with the aim of improving the efficacy and safety of drug therapy, up to and including the personalization of treatment. TDM is concerned with the determination of the optimal drug dose, which is crucial for the avoidance of sub- or supra-therapeutic concentrations and the prevention of ineffective treatment or toxicities, respectively. The objective is to facilitate the continuous monitoring of drug concentrations in order to optimise the adjustment of the drug dose at the point of care, thereby ensuring that the required concentrations are achieved in vivo and thus improving the treatment outcomes.

At present, conventional therapeutic drug monitoring (TDM) methodologies such as High-Performance Liquid Chromatography Mass Spectrometry (HPLC-MS) are too labor-intensive and time-consuming to meet the need of precision medicine, necessitating the exploration of novel technical paradigms. Surface-enhanced Raman spectroscopy (SERS), renowned for its capacity for molecular fingerprinting, presents distinct advantages in the fast detection and multiplex discrimination of drug molecular signals. However, challenges such as the complex SERS signals of drug molecules in blood, substantial background noise, and the inherent difficulties in signal acquisition using conventional enhancement substrates impede the application of this approach to TDM.

In this topic, we aim to develop a novel strategy for multimodal blood drug monitoring by integrating SERS technology with deep learning methodologies. Leveraging innovative gold nanoflower SERS substrates by modifications involving halide ions and aggregating agents, this initiative could overcome the technical hurdles associated with extracting drug molecular signals in blood via SERS, thereby realizing universality and high sensitivity in drug molecule detection. Inspired by recent advance in machine vision and medical image analysis, novel deep learning-based data processing pipeline such as convolutional neural networks (CNN) and recurrent neural networks (RNN) will be explored to address the computational challenges posed by conventional SERS methodologies in analyzing intricate datasets, thereby augmenting data interpretability and identification precision. This study would development an advanced real-time drug monitoring platform, dedicated not only to the meticulous regulation of therapeutic and toxic drug concentrations but also advancements in personalized precision pharmacotherapy within the domain of precision medicine.

We are now looking for a postdoctoral researcher to join the team and lead this research topic. It will be co-supervised by the AI and robotics expert, Prof. Shuai Li, from Faculty of Information Technology and Electronic Engineering in University of Oulu. This topic will contribute to the research cluster on wearable plasmonic sensing technologies that could benefit wireless sensors, digital health, early diagnosis of diseases, personalized healthcare, and precision medicine.

Key words: Surface enhanced Raman Spectroscopy, Deep learning algorithms, Personalized healthcare, Therapeutic Drug Monitoring, Precision Medicine

• Head: Miia Jansson, Adjunct Professor
• Faculty: Faculty of Medicine
• Group composition: Three postdocs (starting), six doctoral researchers, and two project researchers
• Website: https://www.oulu.fi/en/research-groups/digiimpact
• Contact: miia.jansson (at) oulu.fi

Research group is researching the impact of digital solutions in acute and critical care. Our goal is to improve short- and long-term outcomes of critical illness by developing a new real-time dynamic prediction models to tackle the timeliness and accuracy of the current static and intermittent risk assessment in a sustainable manner.

Intensive care (ICU) is a high-tech environment for intensive and invasive monitoring. ICU clinical information systems collect, store, process, and present high-volume, high-velocity, and high-variety information (e.g., structured and unstructured data, waveforms) from multiple systems (e.g., laboratory, pathology, and radiology systems) and solutions (e.g., monitors, ventilators, infusion pumps) simultaneously, continuously, and synchronously to support clinical decision-making. Predictive analytics can be used to estimate the probability of either presenting (diagnostic models) or developing a particular disease or outcome (prognostic models) whereas the best prediction performance can be detected by combining multimodal data (e.g., clinical, laboratory, and radiological variables, physiological signals). There is, however, a complete lack of time-dependent predictors to predict patient outcomes continuously.

The DigiImpact is an interdisciplinary research group that brings together a diverse team of experts (e.g., clinicians, epidemiologists, technology specialists, data scientists), who work in Finnish and international companies, hospitals, universities and research organizations to reduce the global burden of critical illness. We are also active in regional, cross-border and -boundary collaboration to scale our innovative solutions.

Our mission is to improve risk-based prevention, early detection and personalized care by utilizing AI, big data and machine learning. We use multimodal data from various sources to develop new machine learning-based prediction models to identify an individual who would benefit from a certain type of treatment based on the predicted disease course.

Key words: predictive analytics, ICU, prevention

• Head: Susanna Kosamo, Dr., Dos, Coordinator
• Faculty: Faculty of Biochemistry and Molecular Medicine, Research Unit of Disease Signaling Networks
• Group composition: 2 professors, 2 postdocs, 1 researcher. Secondments: Susanne Wiedmer professor (Department of Chemistry, University of Helsinki), Daniel Fürth (SciLife, Uppsala Univ.)
• Website: https://www.oulu.fi/en/research-groups/developmental-biology-laboratory-organogenesis-extracellular-vesicles
• Contact: susanna.kosamo (at) oulu.fi

Cell derived nanovesicles (NVs) provide a new mode of systemic signaling where they mediate local and remote metabolic control by delivering a panel of regulatory molecules to cells via endocytosis and endocytic escape. Given their nature it can expected that a highly large body of differentially molecularly assembled NVs can be generated, and these become altered in diseases. To advance the value of NVs diseases new ways to analyze them are essentially needed. Nano plasmonic sensors leverage the interaction of light with metallic nanoparticles to detect changes in the local refractive index. Nano plasmonic sensing, localized surface plasmon resonance, has great promises for the characterization and analysis of NVs. When NVs bind to the surface of these nanoparticles, they cause a shift in the resonance wavelength of surface plasmon resonances, which can be measured. In this post-doctoral project, we will carry out nanoparticle surface functionalization in collaboration with Insplorion AB (https://www.insplorion.com/en/; Sweden). By modifying the surface of the nanoparticles with ligands (e.g., phage display peptide libraries, selected antibodies, aptamers, or small molecular drugs) that specifically bind to markers or other components found on NVs, we ensure selective capture of specific target NVs. Since NPS relies on real-time monitoring, it will be possible to track the binding event over time to quantify the concentration of NVs and to analyze kinetics. Our instrumental NPS setup is designed for multiplex detection, meaning that we can functionalize different regions of the sensor surface with different ligands to detect multiple types of NVs or different NV markers in a single run from human blood samples in Finnish biobanks in refences to the associated clinical data and biobank linked GWAS data that is derived from the Finngen national Finnish genome project. In NPS we analyze the changes in the wavelength of the plasmonic signal to determine the amount of NVs bound to the sensor surface. By using appropriate statistical methods to interpret the data, we can validate the reliability and reproducibility of the measurements.

The NPS methodology can be readily applied to many applications including biomarker discovery (identification of NV-specific biomarkers relevant to diseases, including cancer, neurodegenerative disorders, and cardiovascular diseases) and diagnostic tools (development of diagnostic assays for early detection of diseases through profiling EV cargo (proteins, RNA, lipids).

The main challenges and considerations of NPS include the optimization of ligand density and surface properties to enhance detection limits and minimize nonspecific binding and in to address potential interference from other components in complex samples that could affect sensing performance or signal interpretation. By integrating these approaches, nano plasmonic sensing offers a versatile platform for the study of extracellular vesicles, helping to advance our understanding of their roles in cell communication and disease pathology associated to the selected ICD defined diseases derived from the Borealis/Arctic biobanks of Oulu university being part of the cluster of Finnish biobanks FinBB.

Keywords: functionalization, surface array, reagent libraries, display peptides, biobank samples, nanovesicles, cell secreted extracellular vesicles, exosomes, cell secreted vesicles, biomarkers, diseases, data generation, clinical data

• Head: Shuai Li, Professor
• Faculty: Faculty of Information Technology and Electrical Engineering
• Group composition: 1Prof. Shuai Li (Data) and Prof. Jianan Huang (Healthcare)
• Websites: Shuai Li's website: https://www.oulu.fi/en/researchers/shuai-li
• Jianan Huang's website: https://www.oulu.fi/en/research-groups/jianan-lab-nanophotonic-biosensors
• Contact: shuai.li (at) oulu.fi

Proteins are the workhorses in biological systems, but the fact that proteins cannot be amplified results in a serious lag of proteomics behind genomics and transcriptomics. The big gap hampers not only mechanistic studies but also clinical applications. The latter includes digital health and personalized medicine that aims to provide customized treatment to individual patients based on dynamic personal health data collected by Point-of-care diagnostic devices.

Single-molecule protein sequencing technology based on plasmonic nanopore enhanced Raman spectroscopy is emegering to fill the gap. However, 2 obstacles must be solved before its practical applications in biomedical fields. The first one is the identification of single amino acids in single protein by the single-molecule Raman spectra. In strong contrast to the averaged Raman spectra with fixed peak positions of molecule solutions or solids, the single-molecule Raman data exhibit peak blinking and shifting due to the fact that the molecule diffuses in the plasmonic nanopore. Therefore, the fragments of individual protein can be monitored through the dynamic single-protein Raman data, which can be used to reconstruct the sequence of the whole protein. The second one is the fast translocation of the single protein through the plasmonic nanopore. The translocation speed is 1 amino acid per microsecond so that not enough Raman signals are collected to identify the amino acid by commercial Raman microscopes.

In this research topic, we aim to develop AI-assisted protein sequencing technology by combing machine learning algorithms in a FPGA device with the dual-nanopore Raman sensor device in Dr. Jianan Huang's group. The FPGA will analyze the dynamic single-molecule Raman data from the dual-nanopore sensor to firstly identify the amino acid residues and then use the amino acid signals for automatic movement of the protein through the two nanopores. It is synergetic with and supported by the Academy Research Fellow project (TwoPoreProSeq), BCO emerging project (DigiRaman) and BCO project (LiveProSeq). The whole team includes 2 postdoc fellows and 4 doctoral researchers for sensor fabrication and measurement, 2 doctoral researchers for detection automation and 2 doctoral researchers for data analysis. As a starting point, we have used the plasmonic nanopore biosensors to collect large amount of unprecedented single-molecule Raman spectra of 20 proteogenic amino acids, peptides and their post-translational modifications. The data-analysis team of 2 doctoral researchers is developing deep learning models to classify them.

We are now looking for a postdoctoral researcher to join the team and lead this research topic. The topic belongs to the Data4Healthcare Theme 3) Technical tools in data management and processing. It will be co-supervised by the biosensor expert, Assistant Prof. Jianan Huang, from Faculty of Medicine and Faculty of Biochemistry and Molecular Medicine in University of Oulu. This topic will contribute to the research cluster on single-protein Raman sequencing technologies that could open a new era for single-cell proteomics and diagnostic devices able to identify low-abundance protein, protein modifications and new biomarkers, benefiting early diagnosis of diseases, personalized healthcare, and precision medicine.

Keywords: Personalized healthcare, Single-molecule protein sequencing, Plasmonic nanopore Raman sensor, Amino acid identification, Machine learning algorithms

• Head: Miguel Bordallo López, Associate Professor
• Faculty: Faculty of ITEE (Research group: Multimodal Sensing Lab, Research Unit: Center for Machine Vision and Signal Analysis - CMVS)
• Group composition: At the Multimodal Sensing Lab research group: 1 professor, 4 post-docs, 4 phD students, 4 Master Students. At the Center for Machine Vision and Signal Analysis research unit: 6 Professors, 6 Senior Scientists, 25 post docs, 30 doctoral students, 10 Master Students, 10 visiting researchers, 2 support personnel.
• Websites: CMVS: https://www.oulu.fi/cmvs
• Multimodal Sensing Lab: https://sites.google.com/view/miguelbordallo/team
• Collaborator NE Device SW & Vitacam: https://www.nedevicesw.com/
• https://vitacam.health/
• Contact: miguel.bordallo (at) oulu.fi

The SViSenS research project, seeks to revolutionize healthcare through advanced technology. The project identifies a critical gap in current healthcare diagnostics, where physical examinations and doctors' oversights can lead to significant misdiagnoses. To address this, SViSenS proposes the innovative use of real-time video analysis, integrated with Artificial Intelligence (AI) and Computer Vision, to monitor human behavior and detect early signs of diseases. This approach aims to assist healthcare professionals in making more accurate diagnoses.

SViSenS intends to overcome the limitations of existing healthcare technologies that currently relies heavily on storing and processing raw image and video data on remote servers. This conventional method raises concerns about the security, safety, and privacy of sensitive healthcare information. The project's solution, to be developed in collaboration with the local industrial partner NE Device SW and its IIA Medical Device Vitacam, is to develop sensorized cameras and imaging devices that discard raw image and video footage, focusing instead on extracting and streaming only quantifiable, medically relevant features. This ensures data privacy and security while maintaining patient anonymity. The approach is grounded in four principles: real-time analysis for immediate information extraction, data minimization, privacy preservation, and scalable data storage and management.

The focuses on three interlinked research questions: 1) Explore the most effective methods for real-time extraction of medically relevant features from video data, 2) find ways to sensorize cameras to capture only essential information, and 3) applying the developed algorithms specific healthcare use cases with large amounts of available data, such as pediatric respiratory infections, stroke, Parkinson's disease, and psychiatric conditions related to depression. Additionally, the project investigates the potential challenges and limitations associated with sensorized camera systems in terms of data quality, feature selection, and algorithm performance, aiming to address these through iterative development, testing, and optimization.

The expected outcomes of the SViSenS project are groundbreaking in terms of methodological implementation and sensing capabilities. The project anticipates developing real-time feature extraction algorithms and sensorized camera systems that preserve patient anonymity. By applying these technologies to four use cases and addressing their inherent challenges, SViSenS aims to establish a new paradigm in healthcare data management. The project's potential for scientific breakthroughs lies in its approach to healthcare data acquisition and analysis, which places a strong emphasis on privacy, security, and quality healthcare outcomes.

SViSenS stands at the intersection of healthcare systems, computer vision, and embedded computing, aiming to provide a comprehensive understanding for future privacy-preserving healthcare applications, considering the constraints of real-world scenarios as underscored by the collaboration with NE Device SW Ltd. and the use of practical, data-driven health solutions. The project echoes Data4Health's emphasis on harnessing data for practical health solutions (Theme 1), the improvement of health data management practices (Theme 2), and the development of advanced technical tools that include predictive analytics, novel computational methods and AI (Theme 3).

The SviSenS project is led by Associate Professor Miguel Bordallo López at the Center for Machine Vision and Signal Analysis. SviSenS aligns closely with the Data4Health programme at the University of Oulu.

Key words: Real-Time Video Analysis, Sensorized Imaging Technology, AI-Driven Healthcare:

• Head: Dr. Lauri Lovén, Research Director,
• Faculty: Faculty of Information Technology and Electrical Engineering, Center for Ubiquitous Computing
• Group composition: 1 Research Director, 3 Professors, 1 Adjunct Professor, 1 University Teacher, 2 Postdoctoral researchers, 7 Doctoral researchers, 1 Project Engineer, 1 Project researcher
• Website: https://ubicomp.oulu.fi/research/iedge
• Contact: Lauri.Loven (at) oulu.fi

Citizens are the most valuable assets of the cities. Therefore, the quality of life of citizens, their health, and general well-being play the most important role in the prosperity of the city and the society overall. Understanding citizens’ quality of life and well-being in the city is challenging, as these constructs are composed of diverse subjective and objective parameters.

Nowadays, technology has become ubiquitous. Cities embed ICT to get insights and enhance their processes, citizens use connected devices to monitor their habits and health. By collecting measured data, available quantitative and qualitative information from Smart Cities and citizens, we assume that we should be able to provide insights into city performance and citizens’ life.

However, there are several challenges, ranging from data quality and availability, management and analysis, to providing actual insights and services for the users. Robust and efficient architectures are necessary for such data intensive systems, ensuring privacy and QoS. Proper analysis methods are essential to accommodate the variability in data availability and quality, privacy, and processing demands. Efficient solutions are required to communicate findings and link insights to actions for both decision-makers and citizens.

Edge computing distributes applications between the cloud, data producing devices, and devices at the edges of wireless and fixed networks. Edge computing brings AI methods closer to the physical environment. This enables deep integration of AI with applications supporting humans in their activities and with machines to realise autonomous systems. AI is integrated with the edge computing platform as well, for example, to decide optimal locations for computations.

We study situation-aware, self-organizing and self-managing software platforms that use efficiently edge computing infrastructures for 5G and beyond. We enable trust with distributed ledger technologies and privacy-by-design methods. We push edge computing to the limits by embedding computations and AI into physical objects that are digitally fabricated to fit to their purpose. We use distributed artificial intelligence, big data technologies, and machine learning methods to explore situation awareness for both end-
user applications and the edge computing platforms. We develop archtiectures and analysis for data-intensive systems. For example, in one of our current research projects, most closely related to the Data4Healthcare programme, we want to understand how available data from the city can be connected to citizens’ health and quality of urban life.

We implement AI methods in ethical manner and emphasise testing in real environments. We expect our research to shed light on how humans can benefit from edge computing, and distributed solutions in general, and how it can be used to implement resource-efficient systems.

We invite highly skilled and motivated postdoctoral candidates with an interest in smart cities, citizens' wellbeing, and expertise in AI, data management, and distributed architectures.

Key words: AI, machine learning, distributed architectures, urban computing

• Head: Jouko Miettunen, Professor
• Faculty: Faculty of Medicine
• Group composition: 1 professor, 6 postdocs, 5 doctoral researchers, 2 statisticians
• Website: https://www.oulu.fi/en/research-groups/psychiatric-epidemiology-northern-finland-birth-cohort-studies-1966-and-1986
• Contact: jouko.miettunen (at) oulu.fi

The research group investigates different topics relating to psychiatric epidemiology. The research topics include for instance risk and outtcome research on psychosis, depression, bipolar disorder and alcohol use. Schizophrenia and other psychotic disorders are a major public health problem and leading unsolved disease afflicting about 3-4 % of the population. In Finland, the prevalence of any psychotic disorder is even higher; 4.5 %. Schizophrenia is a complex genetic disorder manifesting combined environmental and genetic causation. To date, the candidate risk factors include e.g. advanced paternal age, obstetric complications, cannabis use, motor dysfunction, urbanicity, and low IQ. Data from Northern Finland birth cohorts have underpinned many of these key findings. Depression is major public health concern and affect about 10% of population. Rick factors include various biological and social factors, personality and life events. The team investigates also psychosocial factors on cancer and effect of temperament on physical and mental health. The team has long experience in epidemiology and access to unique data sets, the findings are likely to have both clinical and public health influence.

Key words: Psychiatry, epidemiology, cohort research

• Head: Konstantin Mikhaylov, Associate Professor
• Faculty: Faculty of Information Technology and Electrical Engineering, CWC-NS; Wireless Medical Communications (WiMeC) research group
• Group composition: 3 professors (one full, one associate, one assistant); two Docents and one postdoc; 8 researchers
• Website: https://www.oulu.fi/en/research-groups/wireless-medical-communications-wimec-research-group
• Contact: konstantin.mikhaylov (at) oulu.fi

With the ongoing digitalization and proliferation of the Internet-of-Things (IoT) health and medical devices, the importance of wireless communication (including integrated communication and sensing and simultaneous information and power transfer) technologies constantly increases. Due to the nature of healthcare data, wireless medical communication has to feature and balance reliability and security, robustness, low energy consumption and overall resource use efficiency, ubiquitous coverage, and communication availability. However, because of the diversity of the data and requirements, no single communication technology currently existing or being in development can address this request comprehensively.

Therefore, the research departs from the “one-size-fits-all” paradigm and focuses on conceptualizing, studying and developing a multi-connectivity-based wireless medical communication solution, seamlessly integrating multiple radio access technologies (multi-RAT), including, e.g., Smart Body Area Network (SmartBAN), Bluetooth Low Energy (BLE), Ultrawideband (UWB), 5G and beyond 5G terrestrial and non-terrestrial cellular. By this, we aim to create a single connectivity ecosystem for transferring healthcare data, accounting for their nature and the operation environment. The fundamental research questions we aim to approach are as follows: (i) Which existing and prospective wireless (radio and, optionally, optical) access technologies are the most promising in addressing the specific medical data traffic and scenarios, and what are the gaps for which new technologies are badly needed? (ii) How and for what should 5G and beyond 5G communication be employed for medical data transfer? (iii) How can one effectively enable multi-connectivity and seamlessly integrate different communication technologies, with a particular focus on integrating BAN and 5G/beyond 5G cellular technologies? (iv) How to optimize the selection and use of the different technologies at the device and network level? (v) How to design a practical wireless medical communication multiconnectivity system and what performance will it have? The research comprises the following phases: (i) state-of-the-art analysis, (ii) overall system architecture design, (iii) multi-RAT communication optimization (including RAT selection and handover, parameters of individual RATs especially w.r.t. prioritization of the emergency traffic, authentication and security, using analytical and semantic communication-based machine-learning methods), (iv) prototyping and experimental performance characterization in a laboratory and for selected illustrative medical and/or healthcare applications.

The proposals addressing the research topic above in full or in part focusing on the critical aspects (in that case, an explanation of why this aspect is considered by the applicant critical) are welcome.

Key words: wireless medical communication, 6G, 5G, radio, technology, IoT communication

• Head: Maisa Niemelä, PhD
• Faculty: Faculty of Medicine
• Group composition: 2 professors, 4 postdoctoral researchers, 6 doctoral researcher
• Website: www.oulu.fi/en/research-groups/wearables-for-population-health
• Contact: maisa.niemela (at) oulu.fi

Wearables for population health research group focuses on development of novel algorithms for wearable activity monitors tailored to assess 24-hour movement behaviors. The group is dedicated in revealing how the use of time and patterns of movement behaviors are related to several health outcomes. The research group utilizes data from large population-based cohort studies, integrating data from wearables, health records, and behavioral factors.

Movement behaviors occurring in 24-hour days are associated with many major health outcomes and risk of non-communicable diseases such as cardiometabolic diseases. Previous research has relied on self-reported measures of physical activity or measured waking hour activity behaviors with accelerometers enabling more precise estimates on sedentary time, light intensity activity and moderate-to-vigorous intensity physical activity. However, recently the codependent relationship within sleep and waking behaviors and the association of this composition to health has been recognized as well the versatile health effects of sleep duration and quality. Yet, accelerometer-based activity monitoring has failed to precisely measure sleep because accelerometers measure movements and their intensity and lack of motions makes it difficult to distinguish between sedentary time on waking hours and sleep. With multimodal wearable measurement setups, recognizing sleep periods from the 24-hour activity patterns has shown promising results compared to purely self-reporting times in bed or accelerometer only measurements. When continuous measurements of heart rate, heart rate variability, respiratory rate, temperature, and movements are combined, other metrics than just sleep duration like sleep efficiency, quality, restorativeness, and sleep stages can be assessed. These metrics are of importance in recognizing healthy sleep patterns and determining how interrelationship between sleep and waking hours activity behaviors affect these behaviors and their health effects.

In the Northern Finland Birth cohorts (https://www.oulu.fi/en/university/faculties-and-units/faculty-medicine/…), our research group has measured from 2012 onwards from more than 7000 individuals habitual 24-hour movement and sleep behaviors with different wearable sensors. Up to 14 days measurement duration during daily living has enabled recognition of various movement patterns and profiles predicting future disease risk and healthy aging. We have also measured sleep patterns with Oura rings from more than 1600 cohort members. Our group focuses on innovative analytical methods to reveal the associations between different movement patterns and life-long health outcomes. The aim of this project is recognizing novel patterns and features from the 24-hour movement behavior data especially focusing on sleep and recovery. The knowledge produced enables us to target evidence-based, effective preventive interventions to those at risk in future.

Key words: activity behaviors, wearable devices, disease risk, sleep

• Head: Ilkka Nissinen, Professor
• Factulty: Faculty of Biochemistry and Molecular Medicine
• Group composition: Profs. 4 and PhDs 2
• Website: https://www.oulu.fi/fi/yliopisto/tiedekunnat-ja-yksikot/infotech-oulu
• Contact: ilkka.nissinen (at) oulu.fi

To create optimal medical diagnostic value the analyte measures should allow identification of its detailed molecular compositions in association to a disease. Such measures can be provided by the electromagnetic spectrum (EMS) based applications. Use of the EMS diagnostics offers molecular classification on the grounds of emission and absorption relationships at certain frequencies serving to biochemically characterize the samples. The more in depth spectroscopic analytic results can be obtained if machine learning strategies will be integrated. The other line of diagnostic research relies on specific molecular depicting reagents. These are offered by recombinant and synthetic antigens and antibodies against them, nanobodies, aptamers and DNA probes and molecularly imprinted polymers. The reagents were used to enable massive need for citizen science-based SARS-CoV2 diagnostics in COVID-19.

However, there is now a great need to be able to diagnose single cells, single molecules and single cell secreted extracellular vesicles (EV) with its multiple molecular analytes. EVs represent a rather recently identified complex and a robust secretory information system. In EVs are assembled in cells in their multivesicular bodies as nano- and micro size secreted droplets. Human health status involves in theory millions of specifically molecularly assembled EVs whose composition depends on exposomic/environmental conditions.

This EU CODUND Data4Healthcare post-doctoral project focuses to advance technology to enable EVs molecular composition at single and multiple, an EV array level. Moreover, given the large degree of EV molecular variation a challenge is to translate EV analysis to the real world, situation. Such capacities can be offered by the citizen science concept that represents a co-creation strategy where citizens help to collect the data. The aim here is in this Data4Health project to advance the capacities to diagnose the excreted EV based analytes in sweat, saliva or urine. For this purpose, we propose to take use of the Time-Gated/Raman technologies and machine learning to obtain measures for specific biological sport parameters. Better analytic powers to target sports medicine related needs such as aerobic/anaerobic exercise load, recovery from intense exercise period and strategies to offer more precision sports values and efficient training. These proposed measures will be conducted in the ODL sports clinic and VTT in collaborating sports network partner premises. We will capitalize to take use of our novel biological wearable solutions and aim to advance establishment of a bio- and databank based on sweat excreted EV analytes derived from wearable blasters. Putative secondments: VTT, ODL Sports Clinic/KIHU and Time Gated Instruments Oy.

Key words: Raman, wearable devices, wireless medical communications, biosensors

• Head: Juha Röning, Professor
• Faculty: Faculty of Information technology and electrical engineering
• Group composition: Seppo Vainio, Valerio Izzi and Vincent Huang
• Website: https://www.oulu.fi/en/research-groups/developmental-biology-laboratory-organogenesis-extracellular-vesicles
• Contact: juha.roning (at) oulu.fi

The rapid evolution in understanding individual variability in health and disease has driven aims toward preventive and personalized medicine. This field aims to optimize healthcare by predicting, preventing, and personalizing treatment options based on each individual’s unique molecular and genetic profile. A promising pathway in this pursuit involves the analysis of cell-secreted nanovesicles, such as extracellular vesicles (EVs), which carry a cargo of diverse molecular classes that reflect the physiological and pathological states of their source cells. These nanovesicles offer rich insights into the early detection and classification of diseases, allowing for minimally invasive diagnostics. Nanovesicle-Based Diagnostics Using Raman Spectroscopy and Machine Learning: Our project will take use of the Raman spectroscopy and machine learning to develop an innovative, cost-effective approach for identifying disease signatures from EVs. The process begins with the collection of excreted body fluids, utilizing clinically accessible blasters, to gather secreted EVs and their molecular components. We will enrich the samples by an expanded microscope applied hydrogel polymer, which facilitates the isolation and stability of EVs, optimizing them for Raman analysis. Integration of Biobank Resources and Clinical Data: We will utilize samples from the Borealis and Arctic biobanks, whose participants have granted broad consent for research use. By coupling the Raman-based molecular signatures with Genome-Wide Association Study (GWAS) data and Finland’s FinRegistry clinical data, we aim to gain a better view of each patient’s health profile. This integrative approach will enable us to correlate specific molecular markers identified via Raman spectroscopy with clinically relevant genomic data and historical health records, thereby improving the precision of our disease models. Advancements in Raman Technology: Time Gated Instruments have introduced a new Raman device capable of high-throughput analysis across 96 well plates, supporting the concurrent processing of multiple samples. This device, combined with machine learning, will enable us to diagnose and classify samples based on distinct Raman signal signatures. Machine learning algorithms will be employed to analyze the complex spectral data, distinguishing healthy and diseased states with high sensitivity and specificity. These data can be then verified in with the NanoPore Raman applications advanced by Dr. Vincent Huang in Biocenter Oulu. The proposed research approach has the potential to revolutionize non-invasive diagnostics and advance cost effective personalized care, paving the way for early disease detection and tailored treatment strategies. The project is based on our key original perr reviewed research papers depicting also that the human excreted body fluidics including the sweat indeed appears to obtain exsosome EVs from the internal organs. This is one essential element that we aim to prodice solid evidence in this Post Doc project.

Key words: Time-Gated Raman, NanoPore Raman, high through put screen, nanovesicle exosomes, transcytosis, biobanks, recall, AI, Machine learning, ICD codes, disease signatures

• Head: Simo Saarakkala, Professor / Head of Research Unit
• Faculty: Faculty of Medicine
• Group composition: The research team in this proposal includes the following persons: University of Oulu: 1 x Professor (Simo Saarakkala), 1 x Assistant Professor (Aleksei Tiulpin), 1 x Doctoral Researcher (Leon Robertz) Polar Electro: 1 x Senior Researcher (Lassi Rieppo), 1 x Principal Researcher (Topi Korhonen)
• Website: https://www.oulu.fi/en/research-groups/diagnostics-osteoarthritis-dios
• Contact: simo.saarakkala (at) oulu.fi

This research topic is carried out as a joint effort with our industrial partner, Polar Electro, which has a strong research heritage and is a world-wide known manufacturer of wearable devices for health and wellbeing technology and guidance. We are looking for candidates who are interested in wearable technology and are skilled in programming (Python) and machine learning.

Wearable technology has gained popularity among large populations for its ability to continuously track multiple health-related parameters, such as sleep, physical activity, stress, physical motion, and cardiac function. Wearable devices produce datasets that characterize human health in a longitudinal setting and can be used to monitor long-term changes in health.

Data modelling methods enable the development of a Digital Twin, a digital representation of a person, based on the wearable data. The Digital Twin can be used to understand the health status of the person and to make predictive simulations of how the person’s health would change in different interventions. Consequently, the Digital Twin modelling and associated predictions can be used to design interventions that would lead to desired outcomes in health. In this research proposal, the primary interest is to study and develop Digital Twin modelling technologies and methodologies based on wearable data in a health context.

The specific research topic can relate to, for example:

a) Developing Digital Twin methodologies that employ large datasets of wearable measurements and computational physiology,

b) Computational model-based design of interventions to improve health and well-being, or

c) Digital Twin models integrating wearable data to other data sources such as environmental data, clinical data, behavioral data, etc.

Key words: wearable technology, digital twin, computational modeling and AI

• Head: Dr Pekka Siirtola
• Faculty: Faculty of Information Technology and Electrical Engineering
• Group composition: 2 professors (Seppo Vainio & Juha Röning); Secondment contacts: Titti Kallio, Terhi Paananen, Kristiina Fekete
• Website: https://www.oulu.fi/fi/yliopisto/tiedekunnat-ja-yksikot/infotech-oulu
• Contact: pekka.siirtola (at) oulu.fi

Traditionally human relations were based on direct interactions between individuals and nature. Today social media offers alternatives for this and impacts our actions. Digital communication offers benefits such as non-locality and easy engaging with people. Online communities provide a space for individuals with shared interests, emotional support and personal well-being. Digitalization brings together individuals who might never have crossed otherwise. In case of forming partnership, digital tools offer a spectrum of potential partners. Digitalization shapes the traditional rituals, the cornerstone of our cultural identity. A supportive social network offers psychological comfort, promotes mental well-being and deeper meanings reinforcing of belonging. Irrespective of the depicted values of the social media feeling of loneness has increased. Moreover, mental diseases and psychiatric medication prescriptions have increases especially in the younger citizens.

We hypothesize that the digital communication lacks the critical transmitted true embodiment biosignalling that are transmitted when physically present. This in turn may contribute to sensory imbalance reflected at one end as sensation of lack of meaning among the younger generations.

In relevance to this problematic is the discovered mechanism that humans are more directly connected to others and nature as considered. This notion is based on the findings that all living organisms use the same mechanism as their overall systemic communication mechanism. Such communication is composed of the cell secreted molecularly loaded extracellular nanosized lipid membrane encircled vesicles (EVs). EVs are similar to those of the synaptic vesicles in the brain. In addition, most of the brain cells communicate with EVs. Importantly EVs can cross the biological barriers and may be well presented in excreted body fluids such as sweat. Transmitted individual EV molecular constituents can be analyzed to date via clinical grade skin plasters and offers effective bio- and data banking from cohorts.

In this project we will collaborate the Evangelical Lutheran Church (ELC) of Finland Evangelical Lutheran, the largest church in Finland with 3,6 million members. The Christian dating App VALO will participate in the research by activating users to provide their expertise by experience on searching for a biologically and mentally compatible partner. The App will offer eventually the knowledge that will be derived from the nano level bioparticles to be integrated into the App to better the performance of its social algorithm. VALO has around 250 000 registered users. We aim to couple digitally collected data sets from the connected community to those of the molecular genetic ones. This may provide value and motivation that may reinforce social gathering and mental wellbeing promotion via such a novel citizen science program. It is speculated that EV exchanges during social interactions provide a novel physical sociobiological control mechanism and reasoning of our social and self-assembly. Realization of the hypothesis is based on analysis of questionnaires, wearable data and social narratives from an app, EV OMICS and spectral data derived from targeted cohorts. The results are expected to provide novelties towards science depicted value of social bonding and mental wellbeing. Putative Secondments: Valo and ELC collaborators.

Key words: Wearable data, social events, dating applications, network stress control, body-mind nanocommunication, synaptic vesicles, soma-brain interactions, mindfulness, biosensors.

• Head: Ilya Skovorodkin, Adjunct Professor
• Faculty: Kvantum Institute, Disease Networks Research Unit, Oulu University
• Group composition: 3 professors, 2 postdoctoral researchers, 2 researchers, Seppo Vainio, Daniel Fürth, Caglar Elbuken, Pekka Varmanen
• Website: https://www.oulu.fi/en/research-groups/developmental-biology-laboratory-organogenesis-extracellular-vesicles
• Contact: ilya.skovorodkin (at) oulu.fi

Personalized, non-invasive diagnostics that can be performed by patients at home, particularly for therapeutic interventions in remote clinical trials, are needed. Such diagnostics could alleviate strain on healthcare systems by enabling patients to monitor their own health, reducing unnecessary, and allowing resources to focus on high-needs patients, who account for most medical expenses. This BioSense project aims to address this challenge by developing a massively multiplexed wearable lab-on-a-chip with a personalized diagnostic cell library, which adheres to the patient’s skin like a band-aid and provides high throughput diagnostic capability. The strategy is based on our data depicting that the same analytes that serve to diagnose diseases in the clinical laboratory setting are presented in the excreted body fluids. The BioSense project leverages expertise from Oulu Univ. (OU) renowned for the work in organogenesis, extracellular vesicles and microfluidic, alongside Daniel Fürth’s lab at SciLifeLab/Uppsala Univ., specializing in molecular tool development, and Pekka Varmanen that targets in Helsinki Univ. the Propionibacterium freudenreichii, used in food fermentation proving a valuable bioreactor applicable model for the BioSense. Such food associated bacteria can be subjected to bioreactors and serves as one key model to develop the BioSense solutions. We will integrate cutting-edge biological insights with innovative molecular technologies, providing a strong foundation to advance the next generation of personalized diagnostics and therapeutics that stems from the biobanks, registry and cohort/GWAS data.

The Aim 1 is to assemble a Live Cell Based Patch-Biosensor. Here the post doc will use commercial edited live reporters or in house assembled library composed of non-pathogenic skin flora bacterial cells that thrive in a hydrogel patch. Upon analyte-induced systemic activation, each bacterial cell will express a specific signal pathway reporter labeled with bio-orthogonal fluorescent click chemistry. This allows live on lone detection of an array of diffusible analytes on the skin. Thus, the applied strategy provides a versatile, non-invasive real time biosensor platform for personalized health monitoring.

In the Aim 2 the post doc will advance the BioSense towards a Citizen Science Reporter platform, together with patients that had given their biobank consent allowing the recall. Biobank-consented patients will use hydrogel patches equipped with engineered biosensor cells targeting disease pathways. Patients will transfer data via image capture of the expressed fluorescent reporters via their smartphone adapter. Such a solution enables real-time live cell sensor monitoring, data collection and data transfer. Such a scaled-up technology offers a user-friendly system for widespread health data collection via a citizen science opening.

The Aim 3 focuses on the Personalized Medicine BioSense solution. The post doc will use of the hydrogel patch to enable short-term live mammalian cell-based monitoring. Thus, the BioSense hydrogel strategy offers to take wide use of the human cellular models and offer an avenue for true personalized diagnostics, resulting in a precise, patient-specific tools to monitor health and therapy efficacy.

Keywords: Live biosensors, probiotics, wearables, citizen science, biobanks, cohorts, GWAS, recall option, orthogonal chemistry, decentralized clinical trials therapy efficacy, gene editing, genomic knock in

• Head: Jaakko Suutala, Asst. Prof
• Factulty: Faculty of Information Technology and Electrical Engineering (ITEE)
• Group composition: 1 Professor, 1 Adjunct professor, 1 senior researcher, 2 postdocs, 3 doctoral researchers, 4 research assistant
• Website: https://www.oulu.fi/en/research-groups/data-analysis-and-inference-group
• Contact: jaakko.suutala (at) oulu.fi

As a part of Biomimetics and Intelligent Systems research unit at ITEE, Data Analysis and Inference Group (DataAI) conducts high-quality research in the field of computational data analysis and develops methods for diverse real-world modeling problems. We are experts in applied AI and ML, especially in the areas of health and wellbeing and sustainability in industry and environment. Our long-term goal is to enable Human-AI collaboration in next generation decision support to make your work and everyday life easier.

In this project, we are developing novel machine learning approaches that are able to perform robust predictive uncertainty quantification and explaining the model's decisions. Furthermore, data-efficient ML models that are capable of learning from sparse data in small data regimes and adapt to distributional shifts as a part of big multimodal health datasets, are considered with the recent advances of deep learning, non-parametric Bayesian methods, and ensemble learning. Developed methodologies are applied and their performance are evaluated on large-scale medical records and wearable sensor data with the goal of building personalized models to help clinical decision making and health monitoring systems.

Key words: explainable AI, robust predictive models, multimodal data

• Head: Aleksei Tiulpin, PhD, Assistant Professor
• Faculty: Faculty of Medicine
• Group composition: 7 professors, 1 senior researcher, 4 postdocs, 6 PhD students
• Website: https://aleksei.tiulpin.ai
• Contact: aleksei.tiulpin (at) oulu.fi

The consortium led by the Intelligent Medical Systems research group of the University of Oulu offers projects, which are done in collaboration with Oulu University Hospital. We are looking candidates interested in different subfields of AI, namely medical imaging and computer vision, natural language processing, and machine learning in general.

The consortium is co-led by clinical collaborators Prof. Tero Rautio, MD PhD, Prof. Juha Saarnio, MD PhD, Prof. Nina Hautala, MD PhD, Prof. Mika Martikainen, MD PhD, Adj. Prof Niina Salokorpi, MD PhD, and Adj. Prof Mika Nevalainen, MD PhD. The team is strongly reinforced by several senior researchers and existing PhD students.

Our group is focused on the development of new methods, and we are looking for candidates with excellent skills in mathematics, medical image analysis, and programming. Please apply if you are looking to work on: Reinforcement Learning, Bayesian Optimization, Uncertainty, Computer Vision, and Foundation Models. The offered projects are mainly under the themes (1) Data as a new resource in the society and (3) Technical tools in data management and processing.

-Sub-theme: Machine Learning:

(1) Differentially-private disease trajectory modelling

(2) Hospital-scale foundation models for radiology and ophthalmology

(3) Medical foundation models for time-varying multi-modal data

-Sub-theme: Medical imaging:

(1) AI-driven care pathways in glioma

(2) Pre-operative prognostic biomarkers of colorectal cancer

(3) Personalized diabetic retinopathy screening

(4) Population-scale body composition analysis with Deep Learning.

-Sub-theme: Natural language processing:

(1) Care need assessment in neurology

(2) Pseudonymization of medical text

(3) Medical data structuring

-Sub-theme: Surgical data science:

(1) Automatic analytics of surgical processes, and AI-driven procedure scheduling

(2) Surgical room monitoring with computer-vision

(3) Virtual reality for surgical training

In your application, specify which theme and which project within it you would like to work on. We also have a very extensive collaboration network beyond the partner organizations in the program. Contact us in advance to discuss your interests and the research plan.      

Key words: Multimodal Deep Learning for Medicine, Bayesian Inference, Hospital-scale AI

• Head: Guoying Zhao, Academy Professor
• Faculty: Faculty of Information Technology and Electrical Engineering
• Group composition: 3 professors, 1 academy research fellow, 7 postdocs, 23 researchers
• Website: https://gyzhao-nm.github.io/Guoying/
• Contact: guoying.zhao (at) oulu.fi

The group led by Academy Professor Guoying Zhao is affiliated with the Center for Machine Vision and Signal Analysis (CMVS), Faculty of Information Technology and Electrical Engineering. It is a creative, open, and internationally attractive research center, combining the expertise of University of Oulu computer vision, machine learning, biosignal analysis and biomedical image analysis scientists to explore novel computer vision methods and harness the power of big data to improve healthcare services.

The group is committed to revolutionizing healthcare through mining the knowledge embedded in the big data produced by both the sensors in our daily life such as normal cameras, and the medical imaging equipment such as fundus cameras. Our mission is to improve healthcare for every patient, particularly those who have difficulty expressing their feelings and enabling remote healthcare.

The research “Non-intrusive computer vision-based big data processing and multi-modal healthcare” aims to develop novel learning methods, particularly multi-modal learning, domain adaption, domain generalization, deep generative models etc. to analyze the large-scale multi-modal health data, and apply them to the tasks including but not limited to:

1) Multi-modal pain analysis: The goal is to collect and integrate large-scale, diverse multi-modal datasets, including video and audio data captured by daily life cameras and information from wearable devices. By leveraging advanced big data processing techniques, this research aims to explore multi-modal learning methods capable of estimating pain levels with high accuracy. The focus is on analyzing vast amounts of data in real-time, employing deep learning and data fusion methods to enhance the reliability of pain assessment for the applications, ranging from monitoring patients in intensive care units to assessment of chronic pains, across different environments and contexts.

2) Non-contact remote heart rate for emotional well-being and heart diseases: This research aims to develop novel big data-driven learning methods for predicting heart rate using facial videos and images. By processing large datasets of facial imagery, the goal is to assess emotional well-being and detect heart diseases non-invasively. Advanced machine learning and big data analytics will be applied to handle the massive volume of information, enabling real-time predictions and improving the accuracy of emotional wellbeing and heart health assessments.

3) Multi-modal ophthalmic data driven eye disease detection: aiming to develop novel multi-modal learning methods particularly foundation model based multi-modal learning methods to learn with multi-modal ophthalmic data with missing modalities to detect eye diseases such as glaucoma, macular edema etc.

4) Doctor and patients’ interaction: aiming to develop novel paradigms of doctor-patient interaction such as (a) explainable AI methods to explain how AI health systems make decisions to doctors and patients in an easy way and increase the trustworthiness of AI health systems, and (b) learning-based interactive behavior analysis and visualization methods to better visualize the big health data, and make the medical expertise be easily and well understood by patients, further facilitate the interaction of doctor and patients.

5) Cancer image analysis: novel deep learning techniques are developed for cancer image analysis for computer-aided diagnosis in digital pathology.

The co-supervisors are Professor Tapio Seppänen from Biomedical Engineering group and Academy Fellow Adj. Prof. Qing Liu.

Key words: Multimodal learning, pain analysis, remote heart rate measure, medical image analysis, disease detection, doctor-patient interaction, cancer image analysis