Differentiating, processing and deciding in high-dimensional spaces related to health data

Comité de pilotage des Ateliers TAIMA

Summary:

One of the main issues with health data and images is related to the large size of the acquisition spaces and attributes generated. Several solutions have emerged thanks to advances in artificial intelligence to aid decision-making in these spaces. However, recent publications in radiology have highlighted the limitations in terms of uncertainty, reproducibility and generality.

The symposium proposed in the context of the ‘La Santé à 360°’ conference aims to bring together:

• researchers in mathematics and signal theory to establish appropriate distance/similarity measures,
• researchers in information and/or decision theory to implement protocols for analysing and paving these spaces,
• researchers in image analysis and processing to produce more relevant attribute spaces
• researchers in data science to find decision support solutions in these large spaces.

The proposed topic is also open to communication and exchange with clinical research concerning use cases and issues relating to data/measurement production or analysis support.

The issue of combining measurements with textual and/or contextual data also falls within the scope of this symposium.

Target audience: academic and clinical researchers

Keywords: high-dimensional space, multivalued image, Riemann, theoretical or learned distance, radiomics, metrology