Data Science Expert Center (DSEC)

The Data Science Expert Center aims to provide guidance and support in the field of data science in general and Health Data Science projects in particular within Amsterdam UMC.

About Data Science Expert Center

The Data Science Expert Center (DSEC) is a multidisciplinary help desk within Amsterdam UMC for all questions and support regarding data science, compute power and deployment within Amsterdam UMC.

DSEC consists of a multidisciplinary team of colleagues from EvA Service Center, Research Data Management (RDM) and the ICT department, established to support individuals who develop, purchase, or wish to deploy AI models within Amsterdam UMC.

DSEC provides support throughout the entire data science process, including advising: on the most suitable platform, how to access data, guiding through the entire ICT landscape of Amsterdam UMC, assisting with software requests, and mapping out the route to implementation together with the researchers. Above all, it is a partner who knows the way around Amsterdam UMC and can assist with the challenges of one's data science project.

DSEC always supports projects with a multidisciplinary group approach, involving data scientists, architects, analysts, and research consultants. For each project, the necessary expertise is identified and engaged. Utilizing DSEC's services is completely free of charge.

Implement your model in healthcare

When starting the process of creating an AI model, the initial thought often goes to data collection and which techniques to use. This is logical, as without a model, there is nothing to implement. However, it is wise to consider early on whether you want to implement the model (if it yields good results). The sooner this is considered, the faster actual implementation can take place.

Within Amsterdam UMC, there are various ways and platforms to implement a model. For each model DSEC helps to determine what and where the best place is. There are three types of AI implementations. DSEC can support all three types:

  • Implementation as part of a research project.
    It may be the case that the model shows good internal and external validation, but needs clinical validation, meaning in the actual workflow of the end user. For example, through an Randomized Controlled Trial (RCT), it is tested in a clinical setting (e.g. Epic) whether a model actually adds value.
  • Implementation of a self-developed model in healthcare.
    A model developed within Amsterdam UMC and used as a medical device can be implemented under Medical Device Regulation (MDR) article 5.5. This requires dossier formation, assessment by an expert group, and ultimately approval by the Medical Direction. A model that is not intended to be used as a medical device can be implemented without this dossier. However, within Amsterdam UMC we have internal guidelines and requirements how we handle these kinds of models.
  • Implementation in another hospital.
    As Amsterdam UMC cannot be seen as a manufacturer, the implementation of AI models is always done in consultation with IXA. Together, we choose the best route for the model to become available to other healthcare institutions.

A good start is to go through the slide deck and think about how the end users will ultimately use the model. You can also email DSEC on datascience@amsterdamumc.nl and we can go through the slide deck together.

Lacking necessary requirements for AI?

Every model and project is unique, but there are common characteristics among different models. To facilitate development, implementation, and management, Amsterdam UMC has acquired various scalable platforms and hardware.

However, it is possible that what your project requires may not yet be available. This could be related not only to computing power but also to connections, interfaces, access, integration with other platforms, or programming languages.

If this is the case, get in touch with DSEC, and together we will explore how you can be assisted. DSEC maintains an overview of all these aspects and knows what resources are available. Even if specific requirements are lacking, and new purchases or new connections/integrations are needed, we will guide you in the right direction.

Difference between Data Science and Health Data Science

Data science is a broad area, without an established definition. Depending on who you ask you will get a different anwser. Artificial intelligence, machine learning, data mining, data preparation and big data, DSEC will hang it all under the term data science. Data science is our umbrella term. ​

Health Data Science (HDS) refers to data-driven solutions that provide insight into healthcare issues and the related processes, using data science methods and techniques. HDS is about the implementation of data science research into healthcare (processes)​.

An HDS application is the created prediction model that has been/will be implemented in the healthcare domain (and the adjacent domains around it such as medical management), which is used by analyzing new data to gain new insights, make predictions or support (individual) decisions. We only talk about HDS applications when a model with data science techniques comes into production (within healthcare or adjacent domain).​

DSEC procedures en templates

  • Project document Projectdocument HDS - Sjabloon. This document (in Dutch) is a standardized/structured document describing aspects of the HDS project. Topics covered include: idea formation, plan formation, data specifications, DPIA, financing, the use case, model development, validation, implementation, management, phasing out and valorization of the model.
  • Policy framework Beleidskader Health Data Science Toepassingen. This document (in Dutch) describes frameworks and guidelines for the entire lifecycle of Health Data Science (HDS) applications, innovative applications in the Healthcare Domain, within Amsterdam UMC.
  • Software Management Plan

Contact

For advice and requests on other services, please contact DSEC using datascience@amsterdamumc.nl