Modular Clinical Decision Support Networks—Predicting What You Need with What You Have

The EDIC program is happy to invite you to a public talk by Mary-Anne Hartley, a post-doctoral scientist in Machine Learning and Optimization Laboratory at EPFL.
The aim of the talk is to present her achievements to a broad audience to prepare for hiring interviews coming up soon. You are warmly welcome to listen to the talk and participate in the Q&A session at the end of the presentation.

Abstract
Clinical Decision Support Systems (CDSS) are mobile applications that have the potential to improve and standardize care with probabilistic guidance. However, many CDSS deploy static, generic rule-based logic and require an inflexible combination of inputs that results in inconsistent performance in volatile or evolving clinical environments.
Paradoxically, it is these environments that have the most to gain from probabilistic guidance (as well as the most to lose when it goes wrong).
In this talk, I will describe the data-driven solution we developed to address this issue: MoDN (Modular Clinical Decision Support Networks).
MoDN is a novel, interpretable-by-design, modular decision tree network, comprising a flexible number and composition of question-specific encoder modules, which can be assembled in real-time to build tailored decision networks at the point-of-care. In short, it allows the clinician to combine the information that they have and inspect its ability to predict any information that they need.
I will first explain the motivation for MoDN and how it works, and then show the various advantages of its modular design like: 1) providing interpretable, continuous predictive feedback to the clinician, 2) being robust to the bias of systematic missingness, 3) enabling multi-modal modeling (i.e. combining text, images, sound, etc.) and 4) allowing privacy-preserving collaborative learning between imperfectly interoperable data sets.

Bio
Annie is an interdisciplinary biomedical data scientist and clinician with a focus on creating and implementing digital health solutions designed for low-resource and humanitarian settings.
She leads the Intelligent Global Health (iGH) research group in the computer science department of the Swiss Institute of Technology (EPFL): www.go.epfl.ch/igh iGH collaborates with international NGOs (Doctors Without Borders, the International Committee of the Red Cross, etc.) and several research/clinical institutions in Switzerland and Africa to collect large-scale, locally owned data/biobanks. The group has a special interest in ethical data management, privacy-preserving distributed collaborative learning, and the development of clinical decision-support tools that aim to improve care by filling resource gaps with probabilistic guidance.
As an MD, she regularly volunteers her medical skills in her home country of South Africa, where she also embeds her research and several capacity-building initiatives for clinical support (www.help-tintswalo.org) and bilateral academic exchange.
Annie obtained an undergraduate in biomedical sciences at the University of Cape Town, South Africa, followed by a PhD and MD at the University of Lausanne in Switzerland, and several further postgraduate specializations in public health and data science, including an MSc at the London School of Hygiene and Tropical Medicine (LSHTM).