Prognostic Value of Non-Contrast CT Markers and Spot Sign for Outcome Prediction in Patients with Intracerebral Hemorrhage under Oral Anticoagulation

NCCT Markers of Intracerebral Hemorrhage Expansion Using Revised Criteria: An External Validation of Their Predictive Accuracy

BACKGROUND AND PURPOSE

Several NCCT expansion markers have been proposed to improve the prediction of hematoma expansion. We retrospectively evaluated the predictive accuracy of 9 expansion markers.

MATERIALS AND METHODS

Patients admitted for intracerebral hemorrhage within 24 hours of last seen well were retrospectively included from April 2016 to April 2020. The primary outcome was revised hematoma expansion, defined as any of a ≥6-mL or ≥33% increase in intracerebral hemorrhage volume, a ≥ 1-mL increase in intraventricular hemorrhage volume, or de novo intraventricular hemorrhage. We assessed the predictive accuracy of expansion markers and determined their association with revised hematoma expansion.

RESULTS

We included 124 patients, of whom 51 (41%) developed revised hematoma expansion. The sensitivity of each marker for the prediction of revised hematoma expansion ranged from 4% to 78%; the specificity, 37%-97%; the positive likelihood ratio, 0.41-7.16; and the negative likelihood ratio, 0.49-1.06. By means of univariable logistic regressions, 5 markers were significantly associated with revised hematoma expansion: black hole (OR = 8.66; 95% CI, 2.15-58.14; P = .007), hypodensity (OR = 3.18; 95% CI, 1.49-6.93; P = .003), blend (OR = 2.90; 95% CI, 1.08-8.38; P = .04), satellite (OR = 2.84; 95% CI, 1.29-6.61; P = .01), and Barras shape (OR = 2.41, 95% CI; 1.17-5.10; P = .02). In multivariable models, only the black hole marker remained independently associated with revised hematoma expansion (adjusted OR = 5.62; 95% CI, 1.23-40.23; P = .03).

CONCLUSIONS

No single NCCT expansion marker had both high sensitivity and specificity for the prediction of revised hematoma expansion. Improved image-based analysis is needed to tackle limitations associated with current NCCT-based expansion markers.

Machine Learning for Onset Prediction of Patients with Intracerebral Hemorrhage

Objective: Intracerebral hemorrhage (ICH) has a high mortality and long-term morbidity and thus has a significant overall health–economic impact. Outcomes are especially poor if the exact onset is unknown, but reliable imaging-based methods for onset estimation have not been established. We hypothesized that onset prediction of patients with ICH using artificial intelligence (AI) may be more accurate than human readers. Material and Methods: A total of 7421 computed tomography (CT) datasets between January 2007–July 2021 from the University Hospital Basel with confirmed ICH were extracted and an ICH-segmentation algorithm as well as two classifiers (one with radiomics, one with convolutional neural networks) for onset estimation were trained. The classifiers were trained based on the gold standard of 644 datasets with a known onset of >1 and <48 h. The results of the classifiers were compared to the ratings of two radiologists. Results: Both the AI-based classifiers and the radiologists had poor discrimination of the known onsets, with a mean absolute error (MAE) of 9.77 h (95% CI (confidence interval) = 8.52–11.03) for the convolutional neural network (CNN), 9.96 h (8.68–11.32) for the radiomics model, 13.38 h (11.21–15.74) for rater 1 and 11.21 h (9.61–12.90) for rater 2, respectively. The results of the CNN and radiomics model were both not significantly different to the mean of the known onsets (p = 0.705 and p = 0.423). Conclusions: In our study, the discriminatory power of AI-based classifiers and human readers for onset estimation of patients with ICH was poor. This indicates that accurate AI-based onset estimation of patients with ICH based only on CT-data may be unlikely to change clinical decision making in the near future. Perhaps multimodal AI-based approaches could improve ICH onset prediction and should be considered in future studies.

Non-contrast CT markers of intracerebral hematoma expansion: a reliability study

OBJECTIVES

We evaluated whether clinicians agree in the detection of non-contrast CT markers of intracerebral hemorrhage (ICH) expansion.

METHODS

From our local dataset, we randomly sampled 60 patients diagnosed with spontaneous ICH. Fifteen physicians and trainees (Stroke Neurology, Interventional and Diagnostic Neuroradiology) were trained to identify six density (Barras density, black hole, blend, hypodensity, fluid level, swirl) and three shape (Barras shape, island, satellite) expansion markers, using standardized definitions. Thirteen raters performed a second assessment. Inter- and intra-rater agreement were measured using Gwet's AC₁, with a coefficient > 0.60 indicating substantial to almost perfect agreement.

RESULTS

Almost perfect inter-rater agreement was observed for the swirl (0.85, 95% CI: 0.78-0.90) and fluid level (0.84, 95% CI: 0.76-0.90) markers, while the hypodensity (0.67, 95% CI: 0.56-0.76) and blend (0.62, 95% CI: 0.51-0.71) markers showed substantial agreement. Inter-rater agreement was otherwise moderate, and comparable between density and shape markers. Inter-rater agreement was lower for the three markers that require the rater to identify one specific axial slice (Barras density, Barras shape, island: 0.46, 95% CI: 0.40-0.52 versus others: 0.60, 95% CI: 0.56-0.63). Inter-observer agreement did not differ when stratified for raters' experience, hematoma location, volume, or anticoagulation status. Intra-rater agreement was substantial to almost perfect for all but the black hole marker.

CONCLUSION

In a large sample of raters with different backgrounds and expertise levels, only four of nine non-contrast CT markers of ICH expansion showed substantial to almost perfect inter-rater agreement.

KEY POINTS

• In a sample of 15 raters and 60 patients, only four of nine non-contrast CT markers of ICH expansion showed substantial to almost perfect inter-rater agreement (Gwet's AC₁> 0.60). • Intra-rater agreement was substantial to almost perfect for eight of nine hematoma expansion markers. • Only the blend, fluid level, and swirl markers achieved substantial to almost perfect agreement across all three measures of reliability (inter-rater agreement, intra-rater agreement, agreement with the results of a reference reading).

Location-Specific Radiomics Score: Novel Imaging Marker for Predicting Poor Outcome of Deep and Lobar Spontaneous Intracerebral Hemorrhage

Objective: To derive and validate a location-specific radiomics score (Rad-score) based on noncontrast computed tomography for predicting poor deep and lobar spontaneous intracerebral hemorrhage (SICH) outcome.

Methods: In total, 494 SICH patients from multiple centers were retrospectively reviewed. Poor outcome was considered mRS 3–6 at 6 months. The Rad-score was derived using optimal radiomics features. The optimal location-specific Rad-score cut-offs for poor deep and lobar SICH outcomes were identified using receiver operating characteristic curve analysis. Univariable and multivariable analyses were used to determine independent poor outcome predictors. The combined models for deep and lobar SICH were constructed using independent predictors of poor outcomes, including dichotomized Rad-score in the derivation cohort, which was validated in the validation cohort.

Results: Of 494 SICH patients, 392 (79%) had deep SICH, and 373 (76%) had poor outcomes. The Glasgow Coma Scale score, haematoma enlargement, haematoma location, haematoma volume and Rad-score were independent predictors of poor outcomes (all P < 0.05). Cut-offs of Rad-score, 82.90 (AUC = 0.794) in deep SICH and 80.77 (AUC = 0.823) in lobar SICH, were identified for predicting poor outcomes. For deep SICH, the AUCs of the combined model were 0.856 and 0.831 in the derivation and validation cohorts, respectively. For lobar SICH, the combined model AUCs were 0.866 and 0.843 in the derivation and validation cohorts, respectively.

Conclusion: Location-specific Rad-scores and combined models can identify subjects at high risk of poor deep and lobar SICH outcomes, which could improve clinical trial design by screening target patients.

New Prediction Models of Functional Outcome in Acute Intracerebral Hemorrhage: The dICH Score and uICH Score

Objectives: The original intracerebral hemorrhage (oICH) score is the severity score most commonly used in clinical intracerebral hemorrhage (ICH) research but may be influenced by hematoma expansion or intraventricular hemorrhage (IVH) growth in acute ICH. Here, we aimed to develop new clinical scores to improve the prediction of functional outcomes in patients with ICH.

Methods: Patients admitted to the First Affiliated Hospital of Chongqing Medical University with primary ICH were prospectively enrolled in this study. Hematoma volume was measured using a semiautomated, computer-assisted technique. The dynamic ICH (dICH) score was developed by incorporating hematoma expansion and IVH growth into the oICH score. The ultra-early ICH (uICH) score was developed by adding the independent non-contrast CT markers to the oICH score. Receiver operating characteristic curve analysis was used to compare performance among the oICH score, dICH score, and uICH score.

Results: There were 310 patients in this study which included 72 patients (23.2%) with hematoma expansion and 58 patients (18.7%) with IVH growth. Of 31 patients with two or more non-contrast computed tomography markers, 61.3% died, and 96.8% had poor outcomes at 90 days. After adjustment for potential confounding variables, we found that age, baseline Glasgow Coma Scale score, presence of IVH on initial CT, baseline ICH volume, infratentorial hemorrhage, hematoma expansion, IVH growth, blend sign, black hole sign, and island sign could independently predict poor outcomes in multivariate analysis. In comparison with the oICH score, the dICH score and uICH score exhibited better performance in the prediction of poor functional outcomes.

Conclusions: The dICH score and uICH score were useful clinical assessment tools that could be used for risk stratification concerning functional outcomes and provide guidance in clinical decision-making in acute ICH.