Bilteanu OneHealthCluj

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 30

Approaching complexity

in One Health
by mathematical modelling and machine
learning
Liviu Luca Bîlteanu, Andreea Dumachi, Radu Popa, Octavian
Buiu, Mihai-Florin Alexandrescu-Olteanu, Andreea Iren Șerban and
Corneliu Octavian Dumitru
Interdisciplinarity Background
• Physics (Engineer 2006 –
Bucharest, MSc 2007 –
Paris, PhD 2010 – PXI
Orsay @ CEA Saclay)
– Theoretical and solid
state physics
– Atomic scale simulations
– Multiscale calculations
• Medicine (Graduate -
2018, Specialist -2023)
• Mathematics (BSc 2021 -
Franch Comte, MSc)
Trandisciplinarity
Affiliations
• University of Agronomic Sciences and
Veterinary Medicine of Bucharest,
Faculty of Veterinary Medicine
– Associate Professor of
• Mathematics Applied to Life Sciences
• Biophysics
– Research in Cancer Molecular Biology
• Bucharest Oncological Institute
– Radiation Oncology
– Clinical Research, Screening, Health
Digital Technologies
• National Institute for Research and
Development in Microtechnologies
– Machine Learning for Medical
Applications
Contents
• Ecosystem models and their environmental
applications

• Interactions among multiple ecosystems

• Artificial Intelligence and One Health: a


building databases for modeling

• Concluding Remarks
ECOSYSTEM MODELS AND THEIR
ENVIRONMENTAL APPLICATIONS
Traditional vs. New Approaches
• Traditional Approach in Applied Ecology
– Initially focused on single-species management.
– Repeated failures led to paradigm shift.
• New Paradigm: Ecosystem-Level Management
– Involves complex interactions among organisms and
scientific disciplines.
– Accounts for interdependencies and feedback loops in
ecosystems.
– Key for understanding and managing complex
ecosystems.
New Ecosystem Modelling
• Objectives
– Describe and disentangle ecosystem components and
interactions.
– Predict future ecosystem states.
– Aid decision-making by evaluating strategies and uncertainties.
• Challenges in Ecosystem Modelling
– Balancing detailed representation with data limitations.
– Addressing the modelling objectives effectively.
• Dealing with Uncertainty
– A primary concern in modelling.
– Recommended strategies:
• Ensemble ecosystem models.
• Multi-model approaches.
Ecosystem modelling methods and their frequency of use for
specific purposes

A guide to ecosystem models and their environmental applications, by W. L. Geary et al., Nature Ecology &
Evolution volume 4, pages1459–1471 (2020)
Varying levels of ecosystem model complexity

A guide to ecosystem models and their environmental applications, by W. L. Geary et al., Nature Ecology &
Evolution volume 4, pages1459–1471 (2020)
Mathematics of applied ecosystem
modeling

A guide to ecosystem models and their environmental applications, by W. L. Geary et al., Nature Ecology &
Evolution volume 4, pages1459–1471 (2020)
End-to-end ecosystem models and their application in
fisheries management

A guide to ecosystem models and their environmental applications, by W. L. Geary et al., Nature Ecology &
Evolution volume 4, pages1459–1471 (2020)
Propagation of uncertainty through ecosystem
models
INTERACTIONS AMONG MULTIPLE
ECOSYSTEMS
Type of interaction among multiple ES
Interaction Type Definition Example
Positive interactions where the presence
Pollination of crops by bees not only
Synergy or functioning of one ecosystem service
increases food production but also
enhances another. supports biodiversity.
Increasing agricultural land may
Occur when the provision of one
increase food production but can
Trade-Offs ecosystem service is reduced as a
reduce water quality and habitat for
consequence of increasing another.
wildlife.
Movement of energy, materials, or Nutrients flowing from a forest to a
Flows
organisms among different ecosystems. stream can support aquatic life.
Wetlands provide a bundle of
Groups of ecosystem services that services: water purification, flood
Bundles/Clusters
commonly occur together. protection, and recreation
opportunities.

And their combinations, e.g.


• Trade-Offs + Synergy,
• Trade-Offs + Synergy + Bundles / clusters,
• Trade-Offs + Synergy + Bundles / clusters + flows
Modeling techniques applied to interactions among multiple ecosystem services
* Agudelo et al, 10.1016/j.ecolmodel.2020.109103.
Ecosystem Services
• Limited Consideration of Ecosystem Services:
– Researchers often focus on a narrow range of ecosystem services →
26 tools used across 209 studies (*)
– This limits exploration of interactions between different services.
– InVEST remains the most applied tool, despite limitations in
integrating multiple services.
• Modeling Techniques Used:
• Common methods: Logical and empirical models, extrapolations,
simulation processes.
• Less common but noteworthy: Data integration methods, direct-
mapping models.
• Trend in Modeling Techniques:
• Increasing use of two or more techniques in a complementary manner
(77% of studies)*.
• Need for broader adoption of multi-technique approaches.

* Agudelo et al, 10.1016/j.ecolmodel.2020.109103.


Methods for measuring uncertainty used in the
studies of interactions among multiple ecosystem

Uncertainty type Uncertainty methods


Classification error Confusion Matrix
Natural variability Probability distribution
Process-based model Monte Carlo simulations
Empirical model Probabilistic equations
Measurement error Normal distribution
Expert knowledge Four-point estimation method*
Linguistic uncertainty Diffuse logic

*distribution based on the lowest and highest expected value


obtained, the best estimate and confidence
ARTIFICIAL INTELLIGENCE AND ONE
HEALTH: A BUILDING DATABASES FOR
MODELING
One Digital Health
A Unified Framework for Future Health Ecosystems

• Aims: Digitally Transform Future Health Ecosystems To Address COMPLEXITY In Health And
Care Scenarios
• Approach: Systemic Health and Life Sciences
• Focus:
– Human Health
– Animal Health
– Environmental Management
• Hybrid Landscape:
– Central Role of Citizens and Health Data
– Individual-level and Population-level Data Management
• Main Challenges:
– Facilitating Interactions Between Communities
– Delivering Data-Driven Contributions in Systems Medicine and Ecology
• Necessities:
– Digital Health Literacy
– Engagement in Health Prevention and Self-Management
– Collaboration in Health Crisis Management
• Outcome: Active Approach to Prevent and Manage Health Crises (e.g., COVID-19 Pandemic)
One Digital Health Steering Wheel
• Two Keys:
• One Health
• Digital Health
• Three Perspectives:
• Individual Health and Well-
being
• Population and Society
• Ecosystem
• Five Dimensions:
• Citizens’ Engagement
• Education
• Environment

Benis et al. J Med Internet Res. 2021 Feb; 23(2): e22189.


Extended Dimensions
• When focusing on specific topics
> 5 dimension emerge
• Case Study: Antimicrobial
Resistance (AMR) in Germany
• An increasing global threat in One
Health context
• AI-based solutions for:
• Drug development
• Personalized therapy
• Monitoring antibiotic use in animal
husbandry and agriculture
• Comprehensive environmental
surveillance for future AMR risk
assessments

Anwendungsbereiche von künstlicher Intelligenz im Kontext von One


Health mit Fokus auf antimikrobielle Resistenzen, Irrgang et al.
Bundesgesundheitsblatt Gesundheitsforschung
Gesundheitsschutz. 2023; 66(6): 652–659
Artificial Intelligence (AI) Overview
• Definition: AI is a broad term for learning-capable algorithmic
structures that mimic and automate intelligent behavior for
problem-solving.
• Scope: AI includes various machine learning (ML) methods and is
integral in foundational and applied research, as well as in the
industry.
• Machine Learning (ML) vs. Traditional Data Analysis:
• Key Difference: ML can identify structures, patterns, and correlations
in problem-specific data without prior specific feature assumptions.
• Universal Property: Allows ML methods to be equally applied in
health, environmental/climate, and population research.
• Quality and Bias in ML Models:
• Dependency: The quality of an ML model largely depends on the
underlying training data.
• Challenge: Imperfections in data can lead to biased model predictions,
highlighting the importance of data generation and curation.
ML in Global and Planetary Health
• One Health Approach: ML
contributes significantly due to
its universal property.
• Data Streams: In every sector of
One Health (humans,
environment, wildlife), there are
data streams recording variables,
processes, and correlations.
• Opportunities: Careful data
collection and curation improve
ML methods for process
mapping, diagnostic capabilities,
and predictive powers.
• Challenge: Moving from sector-
specific to cross-sector solutions.

Anwendungsbereiche von künstlicher Intelligenz im Kontext von One


Health mit Fokus auf antimikrobielle Resistenzen, Irrgang et al.
Bundesgesundheitsblatt Gesundheitsforschung
Gesundheitsschutz. 2023; 66(6): 652–659
Operationalizing “One Digital Health” through the Use
of AI and Related Digital Technologies (1)
• Advancements in AI and Big Data: Leveraging cloud and supercomputing
technologies for large-scale studies in genomics and metagenomics.
• Computational Power and Statistical Methods: Utilizing Bayesian
methods to analyze complex systems in animal epidemiology and socio-
ecological systems.
• Technology Integration: High-speed telecommunications, Internet-of-
Things, GPS, and sensors enhancing understanding of health effects
related to urbanization and social networks.
• Urban Environment and Health: Applying statistical methods and multi-
agent simulation for urban planning and public health interventions.
• Real-Time Data Processing: Enhanced capability to collect and process
data through sensors and advanced analytics for early detection of
infections and rationalizing treatments.
Operationalizing “One Digital Health” through the Use
of AI and Related Digital Technologies
• Holistic Health Approaches: Embracing EcoHealth and Planetary
Health for a multidisciplinary approach to complex health issues.
• Adaptive Monitoring: Using digital tools for assessing exposure and
organism response at individual and population levels across
various scales.
• AI in Pathogen Control: Machine learning and deep learning AI for
identifying genotypes of animal pathogens and understanding
pathogen transmission.
• Antimicrobial Resistance (AMR) Research: AI-based software
predicting antibiotic production and resistance in microorganisms.
• AI in Animal Health: Potential applications in evaluating animal
epidemiological systems, studying environmental patterns, and
supporting decision-making systems.
CONCLUDING REMARKS
Summary
• Ecosystem Modelling and Management:
– Central to understanding and managing ecosystems.
– Requires appropriate models considering decision/problem
constraints.
– Hybrid models gaining prominence due to uncertainty
integration.
• Uncertainty in Ecological Systems:
– Inherent in complex ecological systems.
– Strategies like model skill assessment and multiple model usage
aid in managing uncertainty.
• Interactions Among Multiple Ecosystem Services (ESs)
– Recent shift towards exploring multiple ES interactions.
– Common modeling tools: logical, empirical models, simulation,
and data integration.
– Challenges: Incorporating diverse ecosystem services and
predicting inflection points.
Perspectives
• COVID-19 Impact and ODH Framework:
• Highlighted interconnectivity of environment, animals, and
humans.
• Emphasizes systematic collaboration among various
stakeholders.
• ODH integrates digital and One Health, focusing on
interconnected health dimensions.
• Operationalization of One Health (OH) as ODH:
• Uses AI and big data for improved data collection and
insights.
• Faces challenges due to fragmented data landscape and
restrictive data access.
• ABS framework's limitations in addressing non-human
biological materials and data.
Thank you!

Questions?

You might also like