Dataset Information

Diagnostic performance of an algorithm for automated collateral scoring on computed tomography angiography.

ABSTRACT:

Objectives

Outcome of endovascular treatment in acute ischemic stroke patients depends on collateral circulation to provide blood supply to the ischemic territory. We evaluated the performance of a commercially available algorithm for assessing the collateral score (CS) in acute ischemic stroke patients.

Methods

Retrospectively, baseline CTA scans (≤ 3-mm slice thickness) with an intracranial carotid artery (ICA), middle cerebral artery segment M1 or M2 occlusion, from the MR CLEAN Registry (n = 1627) were evaluated. All CTA scans were evaluated for visual CS (0-3) by eight expert radiologists (reference standard). A Web-based AI algorithm quantified the collateral circulation (0-100%) for correctly detected occlusion sides. Agreement between visual CS and categorized automated CS (0: 0%, 1: > 0- ≤ 50%, 2: > 50- < 100%, 3: 100%) was assessed. Area under the curve (AUC) values for classifying patients in having good (CS: 2-3) versus poor (CS: 0-1) collaterals and for predicting functional independence (90-day modified Rankin Scale 0-2) were computed. Influence of CTA acquisition timing after contrast material administration was reported.

Results

In the analyzed scans (n = 1024), 59% agreement was found between visual CS and automated CS. An AUC of 0.87 (95% CI: 0.85-0.90) was found for discriminating good versus poor CS. Timing of CTA acquisition did not influence discriminatory performance. AUC for predicting functional independence was 0.66 (95% CI 0.62-0.69) for automated CS, similar to visual CS 0.64 (95% CI 0.61-0.68).

Conclusions

The automated CS performs similar to radiologists in determining a good versus poor collateral score and predicting functional independence in acute ischemic stroke patients with a large vessel occlusion.

Key points

• Software for automated quantification of intracerebral collateral circulation on computed tomography angiography performs similar to expert radiologists in determining a good versus poor collateral score. • Software for automated quantification of intracerebral collateral circulation on computed tomography angiography performs similar to expert radiologists in predicting functional independence in acute ischemic stroke patients with a large vessel occlusion. • The timing of computed tomography angiography acquisition after contrast material administration did not influence the performance of automated quantification of the collateral status.

SUBMITTER: Wolff L

PROVIDER: S-EPMC9279191 | biostudies-literature |

REPOSITORIES: biostudies-literature

ACCESS DATA

Similar Datasets

Project description:OBJECTIVE:To compare assessment of collaterals by single-phase computed tomography (CT) angiography (CTA) and CT perfusion-derived 3-phase CTA, multiphase CTA and temporal maximum-intensity projection (tMIP) images to digital subtraction angiography (DSA), and relate collateral assessments to clinical outcome in patients with acute ischemic stroke. METHODS:Consecutive acute ischemic stroke patients who underwent CT perfusion, CTA, and DSA before thrombectomy with occlusion of the internal carotid artery, the M1 or the M2 segments were included. Two observers assessed all CT images and one separate observer assessed DSA (reference standard) with static and dynamic (modified American Society of Interventional and Therapeutic Neuroradiology) collateral grading methods. Interobserver agreement and concordance were quantified with Cohen-weighted ? and concordance correlation coefficient, respectively. Imaging assessments were related to clinical outcome (modified Rankin Scale, ? 2). RESULTS:Interobserver agreement (n = 101) was 0.46 (tMIP), 0.58 (3-phase CTA), 0.67 (multiphase CTA), and 0.69 (single-phase CTA) for static assessments and 0.52 (3-phase CTA) and 0.54 (multiphase CTA) for dynamic assessments. Concordance correlation coefficient (n = 80) was 0.08 (3-phase CTA), 0.09 (single-phase CTA), and 0.23 (multiphase CTA) for static assessments and 0.10 (3-phase CTA) and 0.27 (multiphase CTA) for dynamic assessments. Higher static collateral scores on multiphase CTA (odds ratio [OR], 1.7; 95% confidence interval [CI], 1.1-2.7) and tMIP images (OR, 2.0; 95% CI, 1.1-3.4) were associated with modified Rankin Scale of 2 or less as were higher modified American Society of Interventional and Therapeutic Neuroradiology scores on 3-phase CTA (OR, 1.5; 95% CI, 1.1-2.2) and multiphase CTA (OR, 1.7; 95% CI, 1.1-2.6). CONCLUSIONS:Concordance between assessments on CT and DSA was poor. Collateral status evaluated on 3-phase CTA and multiphase CTA, but not on DSA, was associated with clinical outcome.

Project description:Background and purpose: Anatomical labeling of the cerebral vasculature is a crucial topic in determining the morphological nature and characterizing the vital variations of vessels, yet precise labeling of the intracranial arteries is time-consuming and challenging, given anatomical structural variability and surging imaging data. We present a U-Net-based deep learning (DL) model to automatically label detailed anatomical segments in computed tomography angiography (CTA) for the first time. The trained DL algorithm was further tested on a clinically relevant set for the localization of intracranial aneurysms (IAs). Methods: 457 examinations with varying degrees of arterial stenosis were used to train, validate, and test the model, aiming to automatically label 42 segments of the intracranial arteries [e.g., 7 segments of the internal carotid artery (ICA)]. Evaluation metrics included Dice similarity coefficient (DSC), mean surface distance (MSD), and Hausdorff distance (HD). Additionally, 96 examinations containing at least one IA were enrolled to assess the model's potential in enhancing clinicians' precision in IA localization. A total of 5 clinicians with different experience levels participated as readers in the clinical experiment and identified the precise location of IA without and with algorithm assistance, where there was a washout period of 14 days between two interpretations. The diagnostic accuracy, time, and mean interrater agreement (Fleiss' Kappa) were calculated to assess the differences in clinical performance of clinicians. Results: The proposed model exhibited notable labeling performance on 42 segments that included 7 anatomical segments of ICA, with the mean DSC of 0.88, MSD of 0.82 mm and HD of 6.59 mm. Furthermore, the model demonstrated superior labeling performance in healthy subjects compared to patients with stenosis (DSC: 0.91 vs. 0.89, p < 0.05; HD: 4.75 vs. 6.19, p < 0.05). Concurrently, clinicians with model predictions achieved significant improvements when interpreting the precise location of IA. The clinicians' mean accuracy increased by 0.04 (p = 0.003), mean time to diagnosis reduced by 9.76 s (p < 0.001), and mean interrater agreement (Fleiss' Kappa) increased by 0.07 (p = 0.029). Conclusion: Our model stands proficient for labeling intracranial arteries using the largest CTA dataset. Crucially, it demonstrates clinical utility, helping prioritize the patients with high risks and ease clinical workload.