Development and validation of a prognostic nomogram incorporating automated collateral score and computed tomography perfusion parameters for patients with acute ischemic stroke undergoing endovascular thrombectomy

Abstract

Background: Prognostic evaluation of patients with acute ischemic stroke (AIS) after endovascular thrombectomy (EVT) remains challenging. Traditional computed tomography angiography (CTA) and computed tomography perfusion (CTP) have limitations, whereas multiphase CTA (mCTA) allows for more accurate collateral circulation assessment. However, the predictive utility of automated collateral scoring remains underexplored. This study aimed to develop and validate a nomogram combining automated collateral scores and CTP parameters to enhance the prognostic prediction in patients with AIS undergoing EVT.

Methods: This retrospective study enrolled patients with AIS due to large-vessel occlusion (LVO) who underwent multimodal computed tomography (CT) and EVT between January 2017 and December 2022. The inclusion criteria were as follows: (I) age ≥18 years; (II) symptom onset within 24 hours; (III) available non-contrast CT (NCCT) and CTP; and (IV) 90-day modified Rankin Scale (mRS) score. Meanwhile, the exclusion criteria included posterior circulation stroke, intracranial hemorrhage, extensive prior infarction, prestroke mRS ≥2, poor image quality, and missing outcome data. A total of 111 patients were included and randomly assigned to development (n=77) and validation (n=34) cohorts. Clinical data, National Institutes of Health Stroke Scale (NIHSS) features, and imaging features [including automated collateral scores from NeuBrainCARE (NBC) software and relative cerebral blood flow (rCBF) <30% volume] were analyzed. Agreement between automated and manual Menon collateral scores was assessed via Cohen's kappa. A nomogram was constructed through multivariable logistic regression and validated with bootstrapping (500 iterations). Clinical outcomes were assessed at 90 days through the mRS, with outcomes categorized as good (mRS 0-2) or poor (mRS 3-6).

Results: Manual collateral scoring between physicians showed substantial agreement (κ=0.667). For the automated scoring system, the consistency with manual scoring performed by a junior physician resulted in a kappa value of 0.597, while comparison with senior physicians yielded a higher kappa value of 0.872. The median age was 67 years [interquartile range (IQR), 54-74] years, and 60.3% were male. At 90 days, 54.1% had good outcomes (mRS 0-2). Key independent predictors included age, admission NIHSS, automated collateral score, and rCBF <30% volume (P<0.05). The nomogram achieved strong discrimination in both the development [concordance index (C-index) 0.872, sensitivity 82.9%, and specificity 80.6%] and validation (C-index 0.807, sensitivity 68.4%, and specificity 93.3%) cohorts.

Conclusions: This study supports the use of automated collateral scoring as a reliable and unbiased method in AIS prognosis. The proposed nomogram, integrating clinical and imaging parameters, provides an effective tool for outcome prediction after EVT, facilitating individualized treatment planning and optimizing patient selection.

Keywords: Acute ischemic stroke (AIS); collateral status; computed tomography perfusion (CTP); endovascular thrombectomy (EVT); multiphase computed tomography angiography (mCTA).

Figure 1

Flowchart of patient inclusion in the study. AIS, acute ischemic stroke; CT, computed tomography; EVMT, endovascular mechanical thrombectomy; mRS, modified Rankin Scale; NCCT, non-contrast computed tomography.

Figure 2

Schematic representation of the automated algorithm for assessing patients’ collateral status during different CTA phases. (A,B) The filling time phase of the left middle cerebral artery blood-supplying area during the arterial phase, demonstrating a filling range of 94.18% and a collateral status score of 5. (C,D) The filling time phase during the venous phase, specifically in the left middle cerebral artery blood-supplying area, with a filling range of 83.08% and a collateral status score of 3. The color scheme indicates distinct phases: red for the arterial phase, green for the venous phase, and blue for the late venous phase. CTA, computed tomography angiography.

Figure 3

Percentage bar chart of automated collateral scoring. The automated collateral scores range from 1 to 5 and are represented using five different colors (from light gray to black). The horizontal axis indicates the percentage (ranging from 0% to 100%). In the good outcome group, the percentages for scores 1 to 5 are 1.7%, 3.3%, 15.0%, 58.3%, and 16.7%, respectively. In the poor outcome group, the percentages for scores 1 to 5 are 5.9%, 11.8%, 43.1%, 37.3%, and 1.9%, respectively.

Figure 4

Nomogram for predicting individual outcomes based on the developed model, along with its discrimination and calibration performance. (A) Constructed nomogram for individualized risk estimation. (B) The C-index values of the nomogram in the development and validation cohorts were 0.872 and 0.807, respectively, indicating good discriminative ability. (C) Calibration curve in the development cohort. (D) Calibration curve in the validation cohort, showing good agreement between the predicted and observed outcomes. CI, confidence interval; C-index, concordance index; NIHSS, National Institutes of Health Stroke Scale; rCBF, relative cerebral blood flow.

Figure 5

The nomogram applied in a 47-year-old male patient with left limb symptoms and a baseline NIHSS score of 12. (A) NCCT did not show significant low-density area and bleeding. (B) CTA showed right middle cerebral artery occlusion. (C) CTP showed right temporal lobe hypoperfusion area (color). (D) Multiphase CTA showed peripheral vessels of the right middle cerebral artery supplying area delayed by 1 phase, with normal filling degree, and the automated collateral score was 4. (E) The predicted probability of poor outcome is 0.258 (less than 0.5). The clinical outcome is 90 days, and the mRS score is 0. CTA, computed tomography angiography; CTP, computed tomography perfusion; mRS, modified Rankin Scale; NCCT, non-contrast computed tomography; NIHSS, National Institutes of Health Stroke Scale; rCBF, relative cerebral blood flow.

Figure 6

The nomogram applied in a 78-year-old male patient with left limb symptoms and a baseline NIHSS score of 14. (A) NCCT showed slightly low-density area in the right temporal lobe without bleeding. (B) CTA showed right middle cerebral artery occlusion. (C) CTP showed right temporal lobe hypoperfusion area (color). The red and blue arrows were used in postprocessing to indicate the anterior cerebral artery and venous sinus, respectively. (D) Multiphase CTA showed delayed peripheral vessels of the right middle cerebral artery feeder area for two phases, with a reduced filled vessel count, and the automated collateral score was 2. (E) The predicted probability of poor outcome was 0.853 (greater than 0.5). The clinical outcome was 90 days, and the mRS score was 4. CTA, computed tomography angiography; CTP, computed tomography perfusion; mRS, modified Rankin Scale; NCCT, non-contrast computed tomography; NIHSS, National Institutes of Health Stroke Scale; rCBF, relative cerebral blood flow.