Evaluating artificial intelligence software for delineating hemorrhage extent on CT brain imaging in stroke: AI delineation of ICH on CT.
Artificial intelligence
CT
Hemorrhage
Stroke
Journal
Journal of stroke and cerebrovascular diseases : the official journal of National Stroke Association
ISSN: 1532-8511
Titre abrégé: J Stroke Cerebrovasc Dis
Pays: United States
ID NLM: 9111633
Informations de publication
Date de publication:
25 Nov 2023
25 Nov 2023
Historique:
received:
07
06
2023
revised:
25
10
2023
accepted:
21
11
2023
medline:
27
11
2023
pubmed:
27
11
2023
entrez:
26
11
2023
Statut:
aheadofprint
Résumé
The extent and distribution of intracranial hemorrhage (ICH) directly affects clinical management. Artificial intelligence (AI) software can detect and may delineate ICH extent on brain CT. We evaluated e-ASPECTS software (Brainomix Ltd.) performance for ICH delineation. We qualitatively assessed software delineation of ICH on CT using patients from six stroke trials. We assessed hemorrhage delineation in five compartments: lobar, deep, posterior fossa, intraventricular, extra-axial. We categorized delineation as excellent, good, moderate, or poor. We assessed quality of software delineation with number of affected compartments in univariate analysis (Kruskall-Wallis test) and ICH location using logistic regression (dependent variable: dichotomous delineation categories 'excellent-good' versus 'moderate-poor'), and report odds ratios (OR) and 95 % confidence intervals (95 %CI). From 651 patients with ICH (median age 75 years, 53 % male), we included 628 with assessable CTs. Software delineation of ICH extent was 'excellent' in 189/628 (30 %), 'good' in 255/628 (41 %), 'moderate' in 127/628 (20 %), and 'poor' in 57/628 cases (9 %). The quality of software delineation of ICH was better when fewer compartments were affected (Z = 3.61-6.27; p = 0.0063). Software delineation of ICH extent was more likely to be 'excellent-good' quality when lobar alone (OR = 1.56, 95 %CI = 0.97-2.53) but 'moderate-poor' with any intraventricular (OR = 0.56, 95 %CI = 0.39-0.81, p = 0.002) or any extra-axial (OR = 0.41, 95 %CI = 0.27-0.62, p<0.001) extension. Delineation of ICH extent on stroke CT scans by AI software was excellent or good in 71 % of cases but was more likely to over- or under-estimate extent when ICH was either more extensive, intraventricular, or extra-axial.
Sections du résumé
BACKGROUND
BACKGROUND
The extent and distribution of intracranial hemorrhage (ICH) directly affects clinical management. Artificial intelligence (AI) software can detect and may delineate ICH extent on brain CT. We evaluated e-ASPECTS software (Brainomix Ltd.) performance for ICH delineation.
METHODS
METHODS
We qualitatively assessed software delineation of ICH on CT using patients from six stroke trials. We assessed hemorrhage delineation in five compartments: lobar, deep, posterior fossa, intraventricular, extra-axial. We categorized delineation as excellent, good, moderate, or poor. We assessed quality of software delineation with number of affected compartments in univariate analysis (Kruskall-Wallis test) and ICH location using logistic regression (dependent variable: dichotomous delineation categories 'excellent-good' versus 'moderate-poor'), and report odds ratios (OR) and 95 % confidence intervals (95 %CI).
RESULTS
RESULTS
From 651 patients with ICH (median age 75 years, 53 % male), we included 628 with assessable CTs. Software delineation of ICH extent was 'excellent' in 189/628 (30 %), 'good' in 255/628 (41 %), 'moderate' in 127/628 (20 %), and 'poor' in 57/628 cases (9 %). The quality of software delineation of ICH was better when fewer compartments were affected (Z = 3.61-6.27; p = 0.0063). Software delineation of ICH extent was more likely to be 'excellent-good' quality when lobar alone (OR = 1.56, 95 %CI = 0.97-2.53) but 'moderate-poor' with any intraventricular (OR = 0.56, 95 %CI = 0.39-0.81, p = 0.002) or any extra-axial (OR = 0.41, 95 %CI = 0.27-0.62, p<0.001) extension.
CONCLUSIONS
CONCLUSIONS
Delineation of ICH extent on stroke CT scans by AI software was excellent or good in 71 % of cases but was more likely to over- or under-estimate extent when ICH was either more extensive, intraventricular, or extra-axial.
Identifiants
pubmed: 38007987
pii: S1052-3057(23)00533-5
doi: 10.1016/j.jstrokecerebrovasdis.2023.107512
pii:
doi:
Types de publication
Journal Article
Langues
eng
Sous-ensembles de citation
IM
Pagination
107512Informations de copyright
Copyright © 2023. Published by Elsevier Inc.
Déclaration de conflit d'intérêts
Declaration of Competing Interest GM: Consultancy fees from Canon Medical for stroke imaging software development. KWM: Institution has research agreement with Brainomix. PMB is Stroke Association Professor of Stroke Medicine and an emeritus NIHR Senior Investigator. PMW has no relevant disclosures. He reports institutional educational grants from Medtronic, Stryker and Penumbra in the last 3 years.