ABSTRACTObjectiveTo compare the diagnostic rate and accuracy of 3‐Tesla (T) postmortem magnetic resonance imaging (PM‐MRI) and postmortem ultrasound (PM‐US) in an unselected fetal population.MethodsWe performed prospectively, in a blinded manner, 3‐T PM‐MRI and PM‐US on 160 unselected fetuses at 13–41 weeks of gestation. All imaging was reported according to a prespecified template, for five anatomical regions: brain, thorax, heart, abdomen and spine. The rates of non‐diagnostic results for PM‐US and PM‐MRI were compared and, for results that were diagnostic, we calculated sensitivity, specificity and concordance rates for each anatomical region, using conventional autopsy as the reference standard.Results3‐T PM‐MRI performed significantly better than did PM‐US overall and in particular for fetuses ≥ 20 weeks' gestation. Specifically, the non‐diagnostic rates for PM‐MRI vs PM‐US were 4.4% vs 26.9% (7/160 vs 43/160; P < 0.001) for the brain, 5.2% vs 17.4% (8/155 vs 27/155; P < 0.001) for the thorax, 3.8% vs 30.6% (6/157 vs 48/157; P < 0.001) for the heart and 3.2% vs 23.6% (5/157 vs 37/157; P < 0.001) for the abdomen. For the spine, both techniques showed an equally low non‐diagnostic rate. When both postmortem imaging techniques were diagnostic, they had similar accuracy, with no difference in sensitivity or specificity, and similar concordance with autopsy (PM‐US, 79.5–96.5%; PM‐MRI, 81.6–99.1%).ConclusionsPM‐MRI performed significantly better than PM‐US in this unselected population, due mainly to a lower non‐diagnostic rate. PM‐MRI should remain the first‐line imaging investigation for perinatal autopsy, but PM‐US could be considered if MRI is not available, albeit with a higher non‐diagnostic rate. Copyright © 2019 ISUOG. Published by John Wiley & Sons Ltd.