Overview
Question:
Artificial intelligence technologies have the potential to transform healthcare by enabling rapid and reliable automated diagnoses on an unprecedented scale. This will allow clinicians to better address patient needs and, more ambitiously, provide telemedicine services in the developing world.
A major barrier to integrating machine learning into healthcare is the low availability of high-quality annotated medical data. Data of adequate quality and quantity are expensive and difficult to obtain in this area, and much of the innovation projects therefore inevitably yield limited benefits.
In our opinion, a deep exchange process between stakeholders from industry and technical researchers is the most important missing ingredient to effectively address the data scarcity problem and make machine learning truly productive in medical technology.
Methods:
With this project, we aim to create a platform to integrate the impressive work that has been done on both the academic and industrial sides to address the problem of data scarcity in medical technology.
The ingenious ideas, methods and techniques that have been developed in research in recent years will be related to the needs of Swiss companies and the data collection efforts they have started or are planning in the near future.
The exchange will take place both at the level of teaching and in the context of innovation projects with industrial partners, i.e. the unique position of a university of applied sciences will be exploited.
Output:
Guidelines and recommendations will be developed for the use of machine learning to solve real-world medical technology problems for which little data is available. By actively promoting discussion on this topic, we aim to create an engaged community composed of both academics and private sector representatives.
Specifically, we intend to organize one or more workshops on open problems, case studies, success and failure stories in working in a data-poor environment in medical technology. We will also develop a decision map to identify the best approaches for dealing with scarce data in healthcare, to be used as educational material and supporting documentation for project acquisition.
