Ordinary AI looks at medical images like puzzles made of unfamiliar pieces: it considers any two unmatched pictures completely different. It's like trying to tell twins apart without knowing they're related. In medicine, however, patients with the same condition look nearly identical on MRI, and the algorithm mistakes 'relatives'.
MseaCL adds doctors' text reports during training, creating a 'family album' for the AI. Now each pair of scans gets a caption: similar or not in essence, not just by pixels. Previously, hidden connections went unnoticed, like dark matter in astrophysics, but now the algorithm sees them.
Surprisingly, for pediatric scans, age had to be taken into account: a child's brain is actively developing, and without this, AI could mistake normality for pathology.
As a result, the accuracy of determining tumor subtypes improved by 22.6% compared to standard methods. Now diagnosis is less dependent on uncertainty, and AI has become a reliable assistant for pediatricians. This approach, like a spectroscopic prism, splits the disease into clear variants.
The MseaCL method was trained on data from children whose brains are still actively developing, so the AI must account for age-related changes to avoid mistaking normal development for pathology.