“Artificial intelligence in medicine: an ethical perspective”

Artificial intelligence holds great promise to transform how medicine is practiced with great implications for both medical professionals, patients and health systems more in general. A number of ethical challenges however may slow down or curtail progress in this rapidly evolving field. This keynote will discuss the main ones – namely, fairness and explainability – from an ethics point of view and offer some indications on how to overcome them.

Introductory keynote – Monday 3rd July – 9:30 – 10:30

“Digital health: policy and regulatory landscape in the European Union and globally”

Digital health is one of the most challenging and dynamic fields in health. The COVID-19 pandemic has accelerated development and deployment of digital health solutions, revitalised public debate, incentivised public and private investment, and resulted in regulatory responses in the EU, USA, Australia and globally, including WHO and World Bank. It became apparent that health systems need more innovative solutions, including reliable and affordable services and products; trusted, implementable, and agile regulatory framework; health data management and infrastructure; and finally, involvement of health professionals and patients and paradigm shift through “data-driven culture”. 

European Commission proposed the European Health Data Space Regulation (EHDS), which was welcomed by MS, industry, civil society and other stakeholders. EHDS sets up governance structures and rules for collecting, controlling, accessing, and sharing primary and secondary health data. It also holds a promise to boost the development of digital products and services, supporting data driven policy making, transformation of healthcare systems, support research and strengthen competitiveness of life science sector in the EU. Simultaneously, EC and MS are building MyHealth@EU and MyData@EU, EU-wide infrastructure supporting exchange of primary health data (e.g. e-prescriptions).

Simultaneously other regulatory frameworks are developed in the area Artificial Intelligence, medical devices, pharmaceuticals, cybersecurity and data governance. 

Several EU funding programs, including EU4Health, Digital Europe Program, Horizon Europe and RRF, support development and deployment of digital health solutions.

Introductory keynote – Monday 3rd July – 10:30 – 11:30

“The future of personalized medicine and its implications on the healthcare systems: A machine learning perspective.”

Medicine stands apart from other areas where machine learning can be applied. While we have seen advances in other fields with lots of data, it is not the volume of data that makes medicine so hard, it is the challenges arising from extracting actionable information from the complexity of clinical data. It is these challenges that make medicine the most exciting area for anyone who is really interested in the frontiers of machine learning – giving us real-world problems where the solutions are ones that are societally important and which potentially impact on us all. Think Cancer! Think Covid 19! In this talk I will show how, by working closely with clinicians, we are developing cutting-edge machine learning methods to create tangible clinical impact and how in turn, medicine is driving new advances in machine learning, including discovery of laws and equations from data, causal inference, time-series forecasting, reinforcement learning, generative models etc.

Introductory keynote – Monday 3rd July – 11:30 – 12:30

“Clinical Text Analysis: Methods and Applications.”

Rapid growth in accessing clinical text has led to an unprecedented expansion in the availability of large longitudinal clinical datasets. In these applications, natural language processing (NLP) technologies have played a crucial role as much of detailed patient information, symptoms and PICO elements is embedded in narrative clinical documents. Meanwhile, a number of clinical NLP systems have been developed and utilized to extract useful information from diverse types of clinical text, such as clinical notes, clinical trials, and clinical reports.

Through this lecture, I introduce NLP methods and tools developed in the clinical domain, and showcase the real-world NLP applications in clinical research, with a specific focus on argument mining and XAI methods for medicine.

Plenary Session – Monday 3rd July – 14:00 – 17:30

How complex simulations augment artificial intelligence

We will demonstrate how simulation with digital human brain twins augments AI based classification accuracy in dementia. We will describe novel digital twin technologies and cover related topics such as legal and ethical implications, reproducible research and data integration.

Use case – Monday 3rd July – 17:00 – 18:00

“Machine learning for time series with applications to healthcare.”

Time series in healthcare are a fascinating, challenging and a much understudied area. There are many issues that arise from healthcare data, including missingness, non-stationarity, and in general violation of the commonly accepted assumptions that underlie most of the standard methods. Additionally, data leakages due to incorrectly designed experiments can result in inflated performance and a model that’s not practically applicable at deployment time. In this course we will highlight the challenges and mitigation strategies to help design practically effective machine learning methods for the high risk/high reward area of healthcare that are applicable to many real world scenarios. We will touch upon both pre-processing and ML methodologies in areas such as prediction and representation learning. We will conclude with potential considerations and strategies for updating the models once deployed in the real world. We hope that our participants will come out of this course comfortable using and critically assessing time series models for healthcare and beyond.

Plenary session – Tuesday 4th July – 9:00 – 12:30

Anna Goldenberg

Professor in the departments of Computer Science and the Laboratory Medicine and Pathobiology at the University of Toronto. 

TBD

Coming soon.

Use case – Tuesday 4th July – 14:00 – 15:00

“The basics of causal inference for health applications.”

Causal inference is a scientific paradigm which teaches us when and how can we use data to learn the effects of actions. For example, when and how can we use electronic health records to discover which of two medications is better for a certain group of patients?

The goal of the tutorial is to teach the participants the basic ideas of causal inference.

This includes:

1. Identifying problems and use cases which require causal inference
2. Understanding when and why can standard supervised machine learning fail when applied to causal problems
3. Learning some of the basic tools of inferring causal relationships from data, focusing on observational health data: outcome modeling, propensity score modeling and, if time permits, doubly-robust models

Plenary session – Tuesday 4th July – 15:00 – 18:00

“Deep learning for medical image processing.”

Medical image analysis has become a field of applied machine learning. In this lecture we will discuss machine learning algorithms for medical image classification, image segmentation, and image registration that are relying on recent deep learning techniques, such as convolutional neural networks.

Plenary session – Wednesday 5th July – 9:00 – 12:30

“Machine learning in the triage of major trauma , the Traumatrix project”

Despite the media hype about Artificial Intelligence and Machine Learning in medicine,  few projects explore the impact on patient outcome of predictive algorithms in real-life decisions. The project Traumatrix accepts this challenge. A multidisciplinary consortium developed three prediction models to generate a tailored and indiualized estimation the clinical needs of major trauma patients. The project will test these models in a real-life workflow with experienced clinicians.

Use case – Wednesday 5th July – 14:00 – 15:00

“Deep learning for pathology image analysis”

Advances in digital pathology and artificial intelligence have presented the potential to build assistive tools for objective diagnosis, prognosis and therapeutic-response and resistance prediction. In this talk we will discuss our work on: 1) Data-efficient methods for weakly-supervised whole slide classification with examples in cancer diagnosis and subtyping (Nature BME, 2021), and allograft rejection (Nature Medicine, 2022) 2) Harnessing weakly-supervised, fast and data-efficient WSI classification for identifying origins for cancers of unknown primary (Nature, 2021). 3) Discovering integrative histology-genomic prognostic markers via interpretable multimodal deep learning (Cancer Cell, 2022; IEEE TMI, 2020; ICCV, 2021). 4) Self-supervised deep learning for pathology and image retrieval (CVPR, 2022; Nature BME, 2022). 5) Deploying weakly supervised models in low resource settings without slide scanners, network connections, computational resources, and expensive microscopes. 6) Bias and fairness in computational pathology algorithms.

Use case – Wednesday 5th July – 14:30 – 16:00

TDB

Coming soon

Use case – Wednesday 5th July – 16:30 – 17:30

“Machine learning for time series with applications to healthcare”

This workshop will follow the general theme of the course presented by Dr. Anna Goldengerg on time series in healthcare. It will be divided into multiple sequential modules with increasing difficulty where attendees can follow based on their own pace. The morning session covers some essentials on data processing and basic time series modeling and will build the fundamentals required for the afternoon session; making sure everyone has the basic understanding of the tasks and dataset. The data used throughout the workshop is a Human Activity Recognition (HAR) dataset, collected from mobile devices and wearables with labeled activities.

Practical session – Thursday 6th and Friday 7th of July

“Information Extraction and Argument Mining for Medicine”

Argumentation mining aims at automatically extracting arguments from textual corpora, to provide structured data for computational models of argument and reasoning engines. In a typical argumentation mining pipeline, sentences recognized as argumentative are extracted from the input document, and argument components (claims and supporting evidences) are identified within such sentences. Then, links between argument components are predicted to construct complete arguments. Finally, the connections between arguments are inferred (e.g., support and attack relations), so as to produce a complete argument graph. Recent advances in computational linguistics and machine learning promise to enable breakthrough applications to this research area.
In this practical session, we will introduce information extraction and argument mining methods applied to clinical trial text, and discuss challenges and perspectives of this new research area with respect to its applications in medicine.

Practical session – Thursday 6th and Friday 7th of July

“Fundamentals of Deep Learning for Pathology Image Analysis”

This session will focus on fundamentals of pathology image analysis using deep learning and hands on training for: (a) preprocessing histology whole slides images for deep learning-based analysis (b) training deep models using whole slide images and slide level labels (c) evaluation and assessment of pathology classification models (d) visualization and interpretability. We will provide a well curated pathology dataset form the cancer genome atlas for this training session, which will be hosted over google co-lab.

Pre-requisites: Basic familiarity with Python is required, familiarity with Pytorch and Google co-lab is preferred but not required.

Practical session – Thursday 6th and Friday 7th of July

“Deep learning for medical image processing”

We will explore recent techniques for medical image segmentation using convolutional neural networks. Specifically, we will train – and test – algorithms that automatically find and delineate organs in 3D whole body image volumes, tumor structures in multiparametric brain scans, or cells in microscopy image data. Prerequisites are knowledge in Python, additional prior exposure to biomedical image data, image processing, computer vision, machine learning with Pytorch are helpful.

Practical session – Thursday 6th and Friday 7th of July

“Disease course mapping with longitudinal data”

Longitudinal data consist of the repeated observations of subjects or objects over time. They are ubiquitous in biology and medicine as they inform about the progression of a biological phenomenon such as growth or the progression of a chronic disease. Analysing longitudinal data requires a careful attention because repeated data of the same subjects are not independent. Linear mixed-effect models have long been a piece of choice to address this problem. They model the progression of the underlying phenomenon and how it manifests itself in variable forms across subjects. Recent developments made it possible to address some limitations of these methods. They account for the non-linear dynamics of progression. They do not use age as a regressor, so that they can compare subjects data even if the subjects differ in their age at onset or pace of progression. In this workshop, we will cover the theory of non linear mixed effect models and their corresponding inference algorithm: maximum likelihood estimation and expectation-maximisation (EM) algorithms. We will present first these tools in the framework of regression models, and then a particular class of disease progression modeling called “disease course mapping”. You will practice using several data longitudinal sets from patients developing neurodegenerative diseases: Alzheimer and Parkinson disease. You will learn how disease course mapping allows you to characterise the variability of the progression profiles across subjects, impute missing data, resample data sets at intermediate time-points, predict the future progression of new patients, and even simulate cohorts of virtual patients.

Practical session – Thursday 6th and Friday 7th of July

“Fed-BioMed, an open source framework for federated learning in real world healthcare applications”

This practical session focuses on federated learning (FL) for healthcare applications, and is based on Fed-BioMed, an open source framework for deploying FL in real world use-cases. Throughout the session the participants will get introduced to the basics of federated learning, and will learn to deploy a federated training in a network of clients by using the Fed-BioMed software components. We will focus on the federation of general machine learning approaches for the analysis of medical data (such as tabular or medical images), using a variety of AI frameworks, from Pytorch to scikit-learn. Most advanced topics include the use of privacy-preserving techniques in FL, and the definition of custom data types, models and optimisation routines.

Practical session – Thursday 6th and Friday 7th of July

“Omics data analysis for precision medicine”

The training will cover the following aspects:
* introduction to molecular biology data (sequencing data, single cell, spatial transcriptomics and multi-omics)
* Bioinformatics methods and tools for high-throughput data analysis and integrative bioinformatics
* Network-based approaches for the analysis of complex diseases

Practical session – Thursday 6th and Friday 7th of July