van der Schaar Lab

Unlocking the True Potential of Machine Learning for Healthcare

There are few fields expanding as rapidly in scope and impact as machine learning. However, its tangible impacts on healthcare remain largely unrealised. In an attempt to assess why this is the case, a research team comprising experts from leading academic and industrial heavyweights in the field, including the van der Schaar lab, MIT, Microsoft, and IBM, critically examines the limiting factors in the structure of machine learning for healthcare (MLHC) in a newly published paper.

The work illustrates how the community currently prioritises scientifically novel endeavours over initiatives that directly benefit patients – highlighting the need to refocus on a truly reality-centric AI.

Key issues plaguing the MLHC landscape include a dearth of diverse data, an overemphasis on easily measurable metrics that may lack clinical significance, and inadequate funding for deployment-focused projects. These structural barriers inhibit the translation of cutting-edge research into tangible improvements in patient care.

To address these challenges head-on, the experts advocate for a fundamental shift in approach. They propose a clear distinction between “machine research performed ON healthcare data” and true “machine FOR healthcare.” The latter paradigm places patient impact at the forefront, prioritising solutions that directly address clinical needs from inception.

A series of recommendations are put forth to tackle these systemic issues, targeting stakeholders ranging from machine learning researchers and healthcare practitioners to institutions and government bodies. The necessary direction is clear: prioritise initiatives that prioritise patient outcomes and equity in healthcare delivery.

Furthermore, the concept of “impact challenges” emerges as a call to action for the MLHC community. These challenges represent specific, measurable goals with a strong focus on health equity and community impact. By rallying behind such objectives, the field can strive towards meaningful advancements that benefit patients on a global scale using reality-centric AI.

Through collaborative action and a commitment to prioritising patient welfare, the MLHC community can pave the way for a future where reality-centric AI and refocused machine learning revolutionise healthcare delivery for the better.

Andreas Bedorf