Handling missing data: AI approach for survival prediction in lung cancer despite missing data.

Optimal cancer prognostication combines multimodal data including biopsy results, CT scan, patient characteristics, and clinical trajectory thus far. However, many patients do not have all modalities of data available, so requiring physicians to have all patient data modalities to use an AI prediction algorithm limits the tool’s clinical utility. To increase the clinical potential of these AI algorithms, Ruffini et al. developed a “missing data aware” survival prediction approach that is able to handle inputs with missing data for risk stratification in non-small cell lung cancer patients.