Artificial intelligence applied to HEALTH

Personalised health uses biological, genetic and genomic data to provide the most effective and appropriate treatment or drug therapy for each individual. By applying artificial intelligence, the data is processed to enable personalised medical care and the development of digital diagnostic tools, ensuring patients receive the right treatment at the right time. 

Understanding diseases with AI

AI technologies enhance the ability to decipher the complex pathophysiology of diseases. Current medical advances foster a more proactive approach and a shift towards predictive, personalised, preventive, and participatory medicine (P4 Medicine) to improve people’s overall health. 

Bridging medicine and technology

Biomedicine combines experimental medicine with IT to improve patients' diagnosis, treatment, care and rehabilitation, while also improving the quality of life of healthy individuals. AI, big data and high-performance computing are the future of personalised health. Medical informatics plays a critical role in healthcare by supporting accurate diagnosis, informed therapy decisions and rapid and precise interpretation of medical data. 

By automating repetitive and time-intensive tasks, AI systems allow healthcare professionals to focus more on patient interaction, improving the quality of care and the overall patient experience.

Medical examinations such as electroencephalograms (EEG), electrocardiograms (ECG), positron emission tomography (PET), magnetic resonance imaging (MRI), histopathology and microscopic tissue analysis are essential for diagnosing diseases and offering treatment options. This data can be used in multimodal analyses. 

To improve care quality and patient centeredness, our researchers also use data from Patient Reported Outcome Measures (PROM) – questionnaires completed by patients to assess treatments. However, analysing the collected data is still time-consuming for healthcare staff. For this reason, the research teams at the Institute of Informatics are developing AI-driven analysis techniques. 

Biomedical informatics and predictive analytics

Biomedical informatics is developing technical solutions to synthesise complex phenotypic and molecular data. For example, it enables the prediction of tumour growth by analysing microscopic images. Computer-assisted deep learning approaches use digital images from scanners, optical coherence tomography, MRI or histopathological slides. These methods have shown that it is possible to train deep neural networks to differentiate between benign and malignant lesions or to predict the development of tumours or neurodegenerative diseases such as Parkinson's, Alzheimer's or multiple sclerosis. 

Cancer classification  

The TNM classification (Tumour, Nodes, Metastasis) is an international standard for assessing cancers according to their anatomical extent (primary tumour, regional lymph node involvement and distant metastasis). Work is also being carried out to classify cancer subtypes to predict therapeutic responses and clinical outcomes, taking into account tumour progression, tumour volume, lesion diameter, adhesion to neighbouring organs, lymph node swelling (adenopathies) or metastasis location. 

Decision support systems and data management

The institute is working on decision support systems (DSS) to support diagnostics and histological classification of diseases such as neurodegenerative disorders and cancer. Key areas include software architecture development, value and information management, decentralised data management, interoperability of heterogeneous information systems and semantic mapping. 

Multimodal analyses and intelligent agents

Multimodal analyses form the core of the research within the eHealth unit. Intelligent agents—autonomous programmes that perform tasks independently—are used to offer solutions for prevention and behavioural analysis. 

Learning agents and behavioural analysis

The intelligent learning agents developed at the Institute of Informatics go beyond rule-based responses by using machine learning techniques. These agents adapt and improve their performance by integrating new knowledge and adjusting their behaviour based on accumulated experience. This can be particularly useful for time-consuming tasks, performing behavioural analysis and exploring opportunities for behavioural change or persuasion. 

Innovative research methods and AI development

The Institute's projects also aim to develop innovative (deductive or abductive) research methods and prepare the necessary software for developing and implementing AI algorithms.

These innovative solutions for decision-making and early detection of diseases such as neurodegenerative conditions or autism spectrum disorders facilitate prevention while avoiding gender bias. 

Strengthening trust in medical AI tools

In sensitive fields such as medicine and healthcare, trust in decision-making tools is essential. The researchers of the eHealth unit work closely with healthcare professionals and experts in algorithm explainability to improve the transparency of their algorithms.

Clinical radiology professionals are involved in all applied research projects. Patients are often involved from the outset to improve trust, transparency, reliability and user-friendliness of the software solutions developed. The social acceptance of artificial intelligence in healthcare can only be achieved through an Open Science approach and trust. 

Reducing gender bias in diagnostics

Researchers in the eHealth unit aim to quantify the uncertainty of an AI model's predictions and interpret these values to reduce gender bias in diagnostics. Certain diseases, such as autism spectrum disorder, are diagnosed less frequently in girls because their symptoms are often less pronounced or differ from those of boys. This bias is reinforced by criteria traditionally focused on male test participants and gender-specific social expectations. When recruiting doctoral students, we pay attention to parity and diversity to minimise potential biases in the development of algorithms.

Artificial intelligence is reshaping the healthcare sector, particularly regarding the collection, management and analysis of large amounts of data. As part of a personalised healthcare approach, our institute leads research projects to support healthcare professionals adapting to new working methods as patients increasingly seek information online.  

Driving innovation with AI solutions

The eHealth unit deals with various areas of artificial intelligence such as medical knowledge representation or automated reasoning. It also carries out projects in sub-areas of artificial intelligence, including data scheduling, research and optimisation. The research teams also use cutting-edge technologies such as machine learning, computer vision and multi-agent systems, where several AI agents collaborate to solve complex tasks. 

Ethics and user-centered design

Our research also focuses on the intersection of AI with ethics and service design. Key areas include behavioural analyses, the explainability of algorithms and human-chatbot relationships. 

Our regional partners (Valais Hospital, Lausanne University Hospital CHUV, HUG), funding organisations (Swiss National Science Foundation SNSF, Innosuisse) and the EU funding programmes know and value the comprehensive specialist expertise of our institute in various sub-areas of artificial intelligence applied to the healthcare sector: 

1. Neural networks: Neural networks consist of interconnected artificial neurons, called nodes, organised into several layers that work in parallel. An artificial neuron receives a digital signal, processes it, and transmits it to other connected neurons. This improves the system's ability to recognise patterns, use specialist knowledge, and learn. 
2. Deep learning: Deep learning uses neural networks to analyse large, heterogeneous data sets.  
3. Machine learning: Machine learning uses algorithms that learn from data sets and recognise patterns. This algorithm progressively improves its performance as more data is processed. 
4. Artificial intelligence: AI refers to the ability of a machine to imitate human cognitive functions such as logical reasoning, learning, planning and creativity. AI systems interpret and analyse data to improve their learning process.  
5. Intelligent agents: Software that performs tasks. Unlike chatbots, they can take action independently. Intelligent agents can be installed on internet platforms or within smart devices.
6. Internet of Things (IoT): This technology connects physical objects to the Internet, transforming them into data-collecting and transmitting devices. Smartwatches or devices equipped with cameras or microphones can monitor or interact with patients.
7. Semantic data analysis: Analysing text data enables machines to understand the meaning of a sequence of words in context. The data is structured and analysed to understand human language better and optimise its processing.