Defining non-contrast CT markers of the ischaemic penumbra in acute stroke

Background and purpose: Isodense swelling is a known early ischaemic change, and it is most likely to represent penumbra tissue on acute non-contrast computed tomography (NCCT) scan. However, in general, its detection by observers showed very poor interobserver agreement, which has been attributed to a lack of defining criteria that differentiate it from hypodensity. The aim of this study is to assess the reliability of detecting isodense swelling within the first six hours post stroke onset. Methods: A three-stage study was designed to test the effect of defining criteria and training on the reliability of detection of isodense swelling on acute NCCT. NCCT and perfusion computed tomography (PCT) scans of patients with acute stroke of less than six hours’ duration were reviewed retrospectively. A web-based tool was used to present NCCT scans to readers of different backgrounds (expert and trainee stroke neurologists and neuroradiologists), who evaluated the scans independently by the Alberta Stroke Program Early CT Score (ASPECTS). In the pre-training stage, nine readers assessed 19 repeated scans (a total of 40 scans) obtained at different slice thicknesses (0.9 mm and 5 mm) and both with and without clinical information. In the consensus stage, definitions for isodense swelling and hypodensity were extracted from simultaneous analysis of NCCT and PCT scans used in the pre-training stage, and potential definitions were circulated to participants for agreement. In the post-training stage, 11 readers assessed 32 scans (5 mm slice thickness) with clinical data after training using the consensus definitions and reviewing examples. Cerebral blood volume (CBV), cerebral blood flow (CBF) and mean transient time (MTT) in each ASPECTS region on all PCT scans were calculated and compared across the three NCCT appearances (normal, hypodense or isodense swelling, and the fate of each ASPECTS region was determined on follow-up NCCT scans. Results: Training increased detection of isodense swelling from 29.4% to 46.8%; significantly (P = 0.0001) improved interobserver agreement from very poor (k = 0.09) to fair (k = 0.30); and ameliorated the predictive power of isodense swelling for penumbra as classified by ASPECTS regions of interest (ROIs) from [sensitivity 9% (confidence interval (CI): 1.9%–24.3%); likelihood ratios positive and negative, 2.5 (CI: 0.6–10) and 0.95 (CI: 0.84–1.06), respectively] to [sensitivity 41% (CI: 30.3%–52.8%); likelihood ratios positive and negative, 5.5 (CI: 3.3–11) and 0.64 (CI: 0.53–0.77), respectively]. Detection of hypodensity did not change significantly with training. Exclusion of outliers improved interobserver agreement for isodense swelling to moderate (k = 0.50). Experience, speciality and clinical information had no significant effect on agreement; however, 5 mm slices increased interobserver agreement on hypodensity significantly (k = 0.34 to k = 0.46, P = 0.01). Intraobserver agreement on both hypodensity and isodense swelling was almost perfect. Hypodensity had low sensitivity for core (41.6% (CI: 33% - 50.7%)) but good likelihood ratios positive and negative (13.7 (CI: 7.5–24.8) and 0.6 (CI: 0.50–0.70), respectively). Both isodense swelling and hypodensity were highly specific (93.8% and 96.9% for penumbra and core, respectively). CBV was increased significantly in isodense swollen areas compared with hypodense areas (P < 0.05). Only 38.3% of isodense swollen ROIs ended in infarction and absence of recanalisation increased odds of infarction four-fold, whereas 90% of hypodense ROIs infarcted with odds of infarction 11 times greater than those of ROIs with isodense swelling. Conclusion: Penumbral tissue in acute stroke patients appears on NCCT as isodense swollen areas that are distinguishable from hypodense regions’ appearance. The detection of penumbral tissue can be improved significantly by training. Isodense swelling most frequently returns normal, particularly in evident recanalisation, and its prognostic value might differ from that of hypodensity.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.