DSA-NRP: No-Reflow Prediction from Angiographic Perfusion Dynamics in Stroke EVT
Shreeram Athreya, Carlos Olivares, Ameera Ismail, Kambiz Nael, William Speier, Corey W. Arnold · 2025
Read DSA-NRP: No-Reflow Prediction from Angiographic Perfusion Dynamics in Stroke EVT by Shreeram Athreya, Carlos Olivares, Ameera Ismail, Kambiz Nael, William Speier, Corey W. Arnold (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2506-17501.txt.
Introduction
Athreya and co-authors predict a hidden stroke complication straight from procedure images in this 2025 paper. After a clot is pulled from a brain artery, some patients suffer no-reflow: tiny vessels stay starved despite the open highway. Clinicians usually learn this only from an MRI within a day.
The team trained classifiers on X-ray angiography sequences recorded during the procedure itself. They extracted flow statistics from the tissue downstream of the clot. Their model scored 0.77 in area under the curve, against 0.57 for a clinical-variables model. Accuracy reached about 0.81 against 0.63. Earlier warning could let doctors act while the patient is still on the table.
Catalog record
| CREATOR | Shreeram Athreya, Carlos Olivares, Ameera Ismail, Kambiz Nael, William Speier, Corey W. Arnold |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 35 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2506-17501
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
DSA-NRP: No-Reflow Prediction from Angiographic Perfusion Dynamics in Stroke EVT Shreeram Athreya, Carlos Olivares, Ameera Ismail, Kambiz Nael, William Speier, Corey W. Arnold Abstract Following successful large-vessel recanalization via endovascular thrombectomy (EVT) for acute ischemic stroke (AIS), some patients experience a complication known as no-reflow , defined by persistent microvascular hypoperfusion that undermines tissue recovery and worsens clinical outcomes. Although prompt identification is crucial, standard clinical practice relies on perfusion magnetic resonance imaging (MRI) within 24 hours post-procedure, delaying intervention. In this work, we introduce the first-ever machine learning (ML) framework to predict no-reflow immediately after EVT by leveraging previously unexplored intra-procedural digital subtraction angiography (DSA) sequences and clinical variables. Our retrospective analysis included AIS patients treated at UCLA Medical Center (2011–2024) who achieved favorable mTICI scores (2b–3) and underwent pre- and post-procedure MRI. No-reflow was defined as persistent hypoperfusion (Tmax > > 6 s) on post-procedural imaging. From DSA sequences (AP and lateral views), we extracted statistical and temporal perfusion features from the target downstream territory to train ML classifiers for predicting no-reflow. Our novel method significantly outperformed a clinical-features baseline (AUC: 0.7703 ± \pm 0.12 vs. 0.5728 ± \pm 0.12; accuracy: 0.8125 ± \pm 0.10 vs. 0.6331 ± \pm 0.09), demonstrating that real-time DSA perfusion dynamics encode critical insights into microvascular integrity. This approach establishes a foundation for immediate, accurate no-reflow prediction, enabling clinicians to proactively manage high-risk patients without reliance on delayed imaging. Keywords Acute Ischemic Stroke, Digital Subtraction Angiography, Endovascular Thrombectomy, MRI, mTICI, No-reflow. 1 Introduction Stroke is the third most prevalent cause of death worldwide, with acute ischemic stroke (AIS) being the most common type [ 1 , 2 ] . AIS occurs when blood flow to the brain is blocked by an occlusion, usually blood clots, in a cerebral artery, leading to a rapid deterioration of brain tissue from ischemia (tissue degeneration due to lack of oxygen) to infarction (tissue death). It is critical to restore blood flow as quickly as possible to prevent permanent brain damage [ 3 ] . Thrombolysis and endovascular thrombectomy (EVT) are the primary treatments for AIS, with EVT becoming more widely adopted due to its higher efficacy [ 4 , 5 ] . Upon admission, stroke patients receive urgent neuroimaging evaluation, typically non-contrast CT/magnetic resonance imaging (MRI) along with CT/MR perfusion and angiography of the head and neck to confirm large-vessel occlusion and assess infarct core and salvageable tissue ( Figure 1 ( left )). Eligible patients undergo EVT, guided by digital subtraction angiography (DSA), and reperfusion success is graded using the modified Thrombolysis in Cerebral Infarction (mTICI) score [ 6 ] . The mTICI scale ranges from 0 (no perfusion) to 3 (complete perfusion), with 2b or higher considered successful macrovascular reperfusion. However, mTICI \geq 2b does not ensure restored microvascular perfusion. The no-reflow phenomenon, characterized by persistent tissue hypoperfusion despite macrovascular recanalization, is well documented [ 7 , 8 , 9 ] . Currently, no-reflow is identified only through follow-up perfusion studies several hours after thrombectomy, which limits opportunities for immediate intervention [ 10 , 11 , 12 ] . Figure 1: Overview of clinical workflow and the proposed DSA-NRP pipeline for no-reflow prediction. The proposed approach enables immediate, intra-procedural prediction of no-reflow using features extracted from DSA sequences and machine learning, reducing reliance on delayed follow-up MRI and supporting earlier, more targeted intervention. No-reflow is particularly relevant for proximal large-vessel occlusions involving the internal carotid artery (ICA) and the M1 segment of the middle cerebral artery (MCA), as these are the primary EVT targets with robust support from randomized trials like SWIFT PRIME, REVASCAT, and ESCAPE [ 13 , 14 ] . Conversely, recent randomized clinical trials showed no significant EVT benefit for distal medium-vessel (M2) occlusions, reducing the clinical relevance of no-reflow in these cases [ 15 , 16 ] . Accurate detection of no-reflow depends on perfusion imaging sensitive to subtle microvascular hemodynamic disturbances, typically represented by Tmax delay mapping ( > 6 seconds) generated from MR or CT perfusion imaging [ 7 , 17 , 18 ] . MRI, particularly within the first three hours post-onset, provides superior sensitivity to ischemia compared to CT: DWI, FLAIR, and GRE/SWI sequences effectively capture cytotoxic edema, vasogenic edema, and hemorrhagic transformation [ 19 , 17 , 20 ] . To address the limitations of delayed MRI assessment, we propose DSA-NRP , a machine learning solution that predicts no-reflow immediately after EVT by integrating clinical features with intra-procedural DSA sequences ( Figure 1 ). Early identification of patients at risk enables timely interventions such as neurocritical care [ 21 , 22 ] , individualized blood pressure management [ 23 , 24 ] , thrombus aspiration [ 25 ] , or intra-arterial thrombolysis [ 26 , 27 ] . We hypothesize that DSA sequences obtained during EVT contain perfusion patterns in the target downstream territory (TDT) [ 28 ] that are indicative of microcirculatory status, and these patterns can be extracted to predict no-reflow before follow-up imaging. Our contributions: 1. We present the first ML framework to predict post-thrombectomy no-reflow using structured, interpretable features derived from intra-procedural DSA sequences. 2. We demonstrate that DSA-derived features: specifically, statistical features and time-series signal descriptors, used with simple, interpretable classifiers outperform traditional clinical baselines, enabling early identification of patients at risk of no-reflow immediately following EVT. 3. We conduct a detailed feature importance analysis, revealing physiologically meaningful predictors such as peak tracer intensity, temporal variance, and flow decay, which are associated with downstream perfusion failure. These results offer novel insights into DSA-based hemodynamic signatures of microvascular dysfunction. 2 Related Work 2.1 Predicting No-Reflow after EVT in AIS Early prediction of no-reflow is vital for managing acute ischemic stroke (AIS). Clinical factors such as age, hypertension, diabetes, high NIHSS, and large infarct core are strongly linked to no-reflow and futile recanalization [ 29 ] . Imaging biomarkers, like the modified Capillary Index Score (mCIS), quantify delayed microvascular filling and correlate with neurological decline and hemorrhagic complications [ 30 ] . MRI and CT perfusion imaging within 24 hours post-procedure reveal persistent hypoperfusion, predicting worse outcomes [ 10 ] . Recent reviews highlight advanced modalities (arterial spin labeling MRI, transcranial Doppler, AI-driven qTICI) for objective no-reflow quantification [ 17 , 31 ] . Our study builds on these multimodal indicators, uniquely combining intraprocedural DSA videos and clinical variables to extract angiographic features for accurate no-reflow prediction [ 32 ] . 2.2 DSA Biomarkers for Stroke Intervention Outcomes DSA sequences acquired during thrombectomy provide biomarkers for stroke outcome assessment, including reperfusion quality, microvascular perfusion, and residual thrombus. Angiographic indices like the modified Capillary Index Score (mCIS) quantify delayed or absent capillary filling despite vessel reopening, identifying microvascular no-reflow and correlating with early neurological decline and hemorrhagic transformation [ 30 ] . Automated and semi-automated techniques such as quantitative TICI (qTICI) use biplane DSA processing to objectively measure reperfusion, achieving predictive accuracy comparable to experts [ 31 , 33 ] . Deep learning models analyzing DSA temporal flow patterns reliably detect residual thrombi, allowing real-time incomplete recanalization identification [ 34 ] . Other biomarkers, such as parenchymal blush and collateral filling, also predict complications [ 17 ] . Our work builds on these advances,