ABSTRACT
INTRODUCTION: We aimed to determine the diagnostic performance of Hounsfield Units (HUs) in selected brain region using computed tomography for the clinical diagnosis of brain death (BD). METHODS: A retrospective, case-control study design. A total of 66 subjects (22 cases, 44 controls) underwent brain tomography between January 2011 and December 2016. Inclusion criteria for cases considered patients with a CT performed within the 24 first hours of a clinical diagnosis of brain death. Exclusion criteria applied to patients with no CT scan performed before BD diagnosis. Brain-healthy-control subjects were identified from the hospital's CT scan database. We selected 12 regions for each cerebral hemisphere (4 basal ganglia; 2 regions gray matter (GM) regions; 4 white matter (WM) regions; 2 brain stem regions); two GM and WM regions in each cerebellar hemisphere, and 4 GM/WM ratios. Measurements included analysis of variance, receiver operating characteristic (ROC) curve, and of pooled effect sizes. RESULTS: 72 measures per subject were recorded. Without contrast material, the best performance was the GM/WM ratio at the basal ganglia level (AUROC = 0.893, 95% C.I. = 0.83, 0.96; p-value <.001). After contrast enhancement, the greatest AUROC value corresponded to the thalamus (AUROC = .959, 95% C.I. = .93, .99; p-value < .001). CONCLUSIONS: There is not an absolute threshold of GM-WM differentiation below which all patients are diagnosed with BD, but a group of HUs in selected brain regions, some of them with very high sensitivity and specificity to be used as early predictors of BD.
