AIMIX, Inclusive Artificial Intelligence for Accessible Medical Imaging Across Resource-Limited Settings, will develop the first scientific framework for inclusive AI in medical imaging, and demonstrate its relevance for accessible and effective obstetric ultrasound screening in resource-limited rural settings. The project will greatly advance the state-of-the-art in AI for medical imaging, from the existing methods developed in high-income societies and mostly focused on performance as well as trustworthiness, towards new inclusive AI approaches that take into close consideration the local contextual factors and unmet clinical needs in resource-limited settings. To this end, a range of novel integrative-adaptive learning methods will be investigated to intelligently integrate existing large-scale, high-quality imaging cohorts with smaller, low-cost imaging datasets from resource-limited settings. AIMIX will enable the future development of imaging AI algorithms that are fundamentally inclusive, i.e:
Affordable for resource-limited healthcare centres;
Scalable to under-represented population groups;
Accessible to minimally trained clinical workers.
Importantly, AIMIX will investigate the socio-ethical principles and requirements that govern inclusive AI, and examine how they compare, conflict or complement those of trustworthy AI developed thus far in high-income settings.
How are we planning to achieve these objectives?
By combining big data from existing European repositories with small-size imaging studies from several Kenyan hospitals, AIMIX will address one of the main current obstacles, namely the lack of large retrospective datasets in resource-limited settings. The novel ground-breaking approach of AIMIX based on integrative-adaptive learning which combines small African samples with existing large-scale non-African databases, in such a way that the neural network is adjusted for optimal performance in African settings.
In AIMIX, we will develop an inclusive AI framework that will combine existing high-quality ultrasound clinical samples that can be obtained from standard ultrasound machines, with new low-cost ultrasound datasets that can be affordably used in resource-limited healthcare centres. The aim of this approach is to develop inclusive AI methods and diagnostic tools for obstetric ultrasound imaging that can be used by minimally trained clinicians in rural Africa such as midwives, nurses and technicians.
Our collaborating institution, the Centre of Excellence in Women and Child Health at the Aga Khan University (AKU), Nairobi, is a world-renowned research centre in maternal-fetal health that will ensure that the field research work is properly done in the hospitals in Kenya. Furthermore, AIMIX will count with the collaboration of experts from ISGlobal, BCNatal and the University of Copenhagen.