MIDWAY ++

The project MIDWAY ++ was a 2022 winner of the PREMATURATION call for projects, led by the Innovation Lab of IP Paris.

Medical images (CT, MRI, Xray) are not normalized nor calibrated. Therefore, pixel intensity values cannot be compared directly between two scans.

Today, clinicians adjust image contrast interactively on their display console. This is subjective, not reproducible, and not practical for large image cohorts.

MIDWAY can calibrate contrast between sets of images (2 to thousands). 
Method initially patented for automated quantification of tumor growth on longitudinal pairs of brain MRIs (with CNRS and Univ. Paris Cité).

MIDWAY++ is a specialized algorithm based on an explicit mathematical model to match image histograms to a “Midway” target, defined from a population of scans.

Midway-mapped images enable:

• Direct comparison of pixel intensity values;

• Computation of informative “Difference maps”

• Automated  characterization of lesion growth (e.g. tumors)

• Training more robust Deep-Learning segmentation networks.

• Transferring Deep-learning models to new medicl image cohorts more easily. 

• Enhanced radiological interpretation of images by healthcare professionals.

• More robust transfer of Deep-Learning architectures to new image cohorts.

• Explainable and tunable for new use-cases or new cohorts.

• Light-computational overhead (no cloud computing)

• Unique solution of its kind.