Ischemic stroke, caused by blocked arteries in the brain, is one of the leading causes
of death and disability worldwide. Endovascular thrombectomy treatment (EVT) is one
of the best treatment strategies for restoring blood flow through blocked arteries, but its
success rate depends on a number of factors, including the extent of a patient’s collateral
circulation. Collateral circulation is a subsidiary vascular network that gets activated when
the main conduits fail due to ischemic stroke. It helps viable brain tissues to get oxygen
and nutrients temporarily.
Evaluation of collaterals by visual inspection of radiologists is time-consuming and
prone to inter and intra-rater variability. Thus, computer-aided systems can provide more
consistent and reliable assessments of collaterals. Four-dimensional computed tomography
angiography (4D CTA) is a reliable method for detailed cerebral vasculature imaging,
preventing inaccurate collateral estimation compared to single-phase CTA. Alongside 4D
CTA, readily available non-contrast computed tomography (NCCT) serves as a frontline
diagnostic tool, free from contrast agents’ potential adverse effects. Hence, we propose
computer-aided systems for automatic collateral evaluation in ischemic stroke using 4D
CTA and NCCT imaging.
We propose an automatic quantification method considering low-rank decomposition,
a classic machine learning (ML) method as well as deep learning (DL) methods for
the automatic evaluation of collaterals. DL models, while capable of automatic feature
extraction unlike classic ML models, face challenges due to limited ischemic stroke data. To
overcome data scarcity and class imbalance, we employ transfer learning with focal loss and
Siamese network. Furthermore, for efficient 3D vasculature segmentation without extensive
slice annotation, we introduce few-shot learning for cerebral blood vessel segmentation which
can be a preprocessing step to collateral evaluation.