Characterization of Clot Composition Through Radiomic Features
Andres Gudino1, Carlos Dier1, Martin Cabarique1, Elena Sagues1, Gustavo Chiriboga1, Alexander Van Dam1, Shubhangi Setia1, Sebastian Sanchez Herrera2, Navami Shenoy1, Edgar Samaniego1
1University of Iowa, 2Yale University
Objective:
Characterize clots biological composition in non-contrast computed tomography (NCCT) images.
Background:
Clots composition could provide insights of the etiology in acute ischemic stroke (AIS). Radiomics is a non-invasive imaging tool that can characterize images at a voxel-by-voxel level. We correlated the histology of clots from patients with large vessel occlusion (LVO) with specific radiomic features (RFs)to characterize clot composition. 
Design/Methods:

Ten clots retrieved from mechanical thrombectomy were imaged using micro-computed tomography (micro-CT) and histologically analyzed. RFs from each clot were extracted with 3D-Slicer and were correlated with different clot components determined by histology.  NCCT and computed tomography angiography images were co-registered with the corresponding slices obtained from histology.  A large cohort of 426 patient with NCCT obtained at the time of presentation were then analyzed.  

Results:
In micro-CT, total energy (TE), joint average (JA) and large dependence high gray level emphasis (LDHGLE) were associated to RBCs (p<0.001, p 0.003, and p<0.002, respectively). Fibrin's RFs included minimum (p .005). TE, JA and LDHGLE were correlated with clots composed by 70% RBCs (Rho 0.654 and 0.652, respectively). Minimum was strongly correlated (Rho 0.795) with clots with >80% of fibrin. TE (Rho 0.687) JA (0.809) and LDHGLE (rho 0.657) were correlated between micro-CT and NCCT. Minimum values were negatively correlated (Rho -0.851). TE (sensitivity 0.750, specificity 0.800) and LDHGLE (sensitivity 0.750, specificity 0.800) were accurate to determine RBCs composition in NCCT. Mnimum (sensitivity: 0.500, specificity: 0.400) was not accurate for fibrin determination. 145 of 426 patients NCCT images were finally analyzed. 50 patients have a stroke of cardioembolic origin, 45 due to large artery atherosclerosis (LAA) and 50 were cryptogenic. Radiomics analysis showed that, 35 (75%) cardioembolic, 12 (27%) of LAA and 25 (50%) of cryptogenic AIS were mainly composed of RBCs.
Conclusions:

RFs are sensitive and specific to determine clots rich in RBCs composition in NCCT.

10.1212/WNL.0000000000211787
Disclaimer: Abstracts were not reviewed by Neurology® and do not reflect the views of Neurology® editors or staff.