Descrizione del progetto
Diagnostica per immagini alimentata dall’IA per contesti dalle risorse limitate
L’IA vanta applicazioni significative nel settore sanitario. La radiologia è una specialità medica che ha tratto grandi benefici dalle applicazioni dell’IA. Tuttavia, gli attuali sviluppi dell’IA nel campo della diagnostica per immagini si rivolgono principalmente ai paesi ad alto reddito. Se questa tendenza dovesse continuare, le già marcate disuguaglianze nella salute globale, in particolare nell’Africa rurale, aumenteranno. Il progetto AIMIX, finanziato dal CER, svilupperà un nuovo quadro scientifico per un’IA inclusiva nel campo dell’immaginografia in contesti dalle risorse limitate, scalabile a gruppi di popolazione sottorappresentati e accessibile a operatori clinici minimamente formati. Il progetto farà progredire l’attuale stato dell’arte, dai metodi di IA esistenti sviluppati per contesti clinici ad alte risorse verso nuovi algoritmi di IA per l’immaginografia sostanzialmente inclusivi. AIMIX indagherà anche sulle sfide socio-etiche e sui requisiti degli strumenti di IA inclusivi per promuovere future implementazioni in contesti locali.
Obiettivo
Artificial intelligence (AI) is widely regarded as one of the most promising and disruptive technologies for future healthcare. As AI algorithms such as deep neural networks are suited for the processing of large and complex datasets, radiology is the medical speciality that has seen some of the most important applications of AI in the recent years. However, despite these advances, a major limitation of current AI developments in medical imaging is that they have overwhelmingly, and almost entirely, targeted applications in high-income countries. There is a concern, if the current trend continues, that AI will increase the already pronounced inequalities in global health, in particular for resource-limited settings such as rural Africa, where the majority of the African population lives.
AIMIX will develop the first scientific framework for inclusive imaging AI in resource-limited settings. The project will greatly advance the current state-of-the-art, from existing AI methods mostly developed for high-income settings, towards new imaging AI algorithms that are fundamentally inclusive, i.e. (1) affordable for resource-limited clinical centres, (2) scalable to under-represented population groups, and (3) accessible to minimally trained clinical workers. Furthermore, 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. These innovations will be demonstrated for affordable and accessible AI-powered obstetric ultrasound screening by minimally trained clinicians such as midwives in rural Africa.
Ultimately, AIMIX’s scientific breakthroughs will enhance the democratisation of imaging AI in resource-limited settings, which will result in an important social impact, by empowering local communities, promoting inclusion, and reducing disparities between populations from low- and high-income societies.
Campo scientifico
Parole chiave
Programma(i)
- HORIZON.1.1 - European Research Council (ERC) Main Programme
Argomento(i)
Meccanismo di finanziamento
HORIZON-ERC - HORIZON ERC GrantsIstituzione ospitante
08007 Barcelona
Spagna