Fully Automated Thrombus Segmentation on CT Images of Patients with Acute Ischemic Stroke

Abstract

Thrombus imaging characteristics are associated with treatment success and functional outcomes in stroke patients. However, assessing these characteristics based on manual annotations is labor intensive and subject to observer bias. Therefore, we aimed to create an automated pipeline for consistent and fast full thrombus segmentation. We used multi-center, multi-scanner datasets of anterior circulation stroke patients with baseline NCCT and CTA for training (n = 228) and testing (n = 100). We first found the occlusion location using StrokeViewer LVO and created a bounding box around it. Subsequently, we trained dual modality U-Net based convolutional neural networks (CNNs) to segment the thrombus inside this bounding box. We experimented with: (1) U-Net with two input channels for NCCT and CTA, and U-Nets with two encoders where (2) concatenate, (3) add, and (4) weighted-sum operators were used for feature fusion. Furthermore, we proposed a dynamic bounding box algorithm to adjust the bounding box. The dynamic bounding box algorithm reduces the missed cases but does not improve Dice. The two-encoder U-Net with a weighted-sum feature fusion shows the best performance (surface Dice 0.78, Dice 0.62, and 4% missed cases). Final segmentation results have high spatial accuracies and can therefore be used to determine thrombus characteristics and potentially benefit radiologists in clinical practice.

Agreement between estimated computed tomography perfusion ischemic core and follow-up infarct on diffusion-weighted imaging

“…The StrokeViewer platform is a CE marked platform which comprises various—FDA-cleared—stroke technology applications including arterial occlusion location detection, collateral status quantification, ASPECTS scoring, hemorrhage segmentation, and analysis of the perfusion status in acute ischemic stroke. Although the diagnostic performance of several applications available on the StrokeViewer platform have been studied and published before [ 17 , 18 ], the automated perfusion software available in StrokeViewer has not been volumetrically and spatially compared with 24 h FU DWI in a clinical validation set. Given the increased use of the StrokeViewer platform and the fact that treatment decisions may be based on CTP-estimated results—especially in the 6-24 h time window–, it is of great importance to determine the diagnostic performance of the perfusion algorithm available on the StrokeViewer platform and compare this to the performance of other commercially available software packages.…”

Thrombus radiomics in patients with anterior circulation acute ischemic stroke undergoing endovascular treatment

“…Second, full thrombus segmentations are time consuming, preventing clinical implementation of TR. However, recently, an automated thrombus segmentation has been proposed 29. The segmentations obtained by this method had high spatial overlap with expert based ground truth annotations.…”

Thrombus imaging characteristics within acute ischemic stroke: similarities and interdependence

“…With the enhanced development of automated measurements and thrombus segmentations, quantitative characterization of the thrombus might be optimized. 28 Finally, in-silico models that simulate stroke onset, treatment, and outcome require, among others, an accurate and quantitative representation of the thrombus. 29 Combining thrombus imaging, histological and biomechanical properties can help to capture the whole spectrum of thrombus characteristics needed for such in-silico modeling.…”

“…Correlations between occlusion-location/CBS and DT/length/density/perviousness offer the possibility of uncomplicated characterization of the thrombus, which could ultimately lead to a reduction of time-consuming assessments. With the enhanced development of automated measurements and thrombus segmentations, quantitative characterization of the thrombus might be optimized 28…”