Objective: We developed a detailed imaging phenotype of the cerebral complications in critically ill patients with infective endocarditis (IE) and determine whether any specific imaging pattern could impact prognostic information. Methods: One hundred ninety-two patients admitted to the intensive care units of seven tertiary centers with severe, definite left IE and neurological complications were included. All underwent cerebral imaging few days after admission to define the types of lesions, their volumes, and their locations using voxel-based lesion-symptom mapping (VLSM). We employed uni- and multi-variate logistic regression analyses to explore the associations among imaging features and other prognostic variables and the 6-month modified Rankin Scale (mRS) score. Results: Ischemic lesions were the most common lesions (75%; mean volume, 15.3 ± 33 mL) followed by microbleeds (50%; mean number, 4 ± 7.5), subarachnoidal hemorrhages (20%), hemorrhagic strokes (16%; mean volume, 14.6 ± 21 mL), and hemorrhagic transformations (10%; mean volume, 5.6 ± 11 mL). The volume of hemorrhagic transformations, the severity of leukopathy, and the compromises of certain locations on the motor pathway from the VLSM were associated with a poor 6-month mRS score on univariate analyses. However, upon multivariate analyses, no such specific imaging pattern independently predicted the mRS; this was instead influenced principally by age (OR = 1.03 [1.004–1.06]) and cardiac surgery status (OR = 0.06 [0.02–0.16]) in the entire cohort, and by age (OR = 1.04 [1.01–1.08]) and Staphylococcus aureus status (OR = 2.86 [1.19–6.89]) in operated patients. Conclusions: In a cohort of severely ill IE patients with neurological complications, no specific imaging pattern could be highlighted as a reliable predictor of prognosis. © 2021, Fondazione Società Italiana di Neurologia.< Réduire