The recent European Congress of Radiology (ECR) featured the session “Intelligence at work: harnessing artificial intelligence (AI) for clinical decision-making”. This session included presentations from three different projects in the field of medical AI: ODELIA, AI-POD and RadioVal.
Alexandra Athanasiou (MHA) and Adrià Marcos Morales (VHIO) gave an insightful talk about AI-driven breast cancer diagnostics through swarm learning and summarised the current status of project activities.
Alexandra Athanasiou underlined the clinical need for AI tools in breast cancer MRI due to high variability in MRI to improve detection, e.g. of small lesions in dense breasts, and reduce the number of false positives. To achieve this, ODELIA will include 45k MRI cases, which will be the largest data set of its kind currently available for AI training, which are shared in a decentralized manner within the project.
Adrià Marcos Morales explained the deep learning approach for addressing the high variability of MRI images pointing out that deep learning outperforms classical statistical and machine learning methods. ODELIA avoids data privacy concerns by employing a federated learning and swarm learning approach, where the model travels but the data remains within the hospital. SL data has been proven to work equally well as centralized data for AI model training.
ODELIA is working to establish and implement an open access Swarm Learning network enabling collaborative AI training enabling advancement in MR breast imaging.
The session is available online for free.
To join the session online for free, simply log in via ESR Connect with your ESR MyUserArea credentials or create a free account. Then select the session you want to attend.
