MRI Diffusion-Weighted Imaging to Measure Infarct Volume: Assessment of Manual Segmentation Variability

Current State of Evidence for Neuroimaging Paradigms in Management of Acute Ischemic Stroke

Stroke is the chief differential diagnosis in patient presenting to the emergency room with abrupt onset focal neurological deficits. Neuroimaging, including non-contrast computed tomography (CT), magnetic resonance imaging (MRI), vascular and perfusion imaging, is a cornerstone in the diagnosis and treatment decision-making. This review examines the current state of evidence behind the different imaging paradigms for acute ischemic stroke diagnosis and treatment, including current recommendations from the guidelines. Non-contrast CT brain, or in some centers MRI, can help differentiate ischemic stroke and intracerebral hemorrhage (ICH), a pivotal juncture in stroke diagnosis and treatment algorithm, especially for early window thrombolytics. Advanced imaging such as MRI or perfusion imaging can also assist making a diagnosis of ischemic stroke versus mimics such as migraine, Todd's paresis, or functional disorders. Identification of medium-large vessel occlusions with CT or MR angiography triggers consideration of endovascular thrombectomy (EVT), with additional perfusion imaging help identify salvageable brain tissue in patients who are likely to benefit from reperfusion therapies, particularly in the ≥6 h window. We also review recent advances in neuroimaging and ongoing trials in key therapeutic areas and their imaging selection criteria to inform the readers on potential future transitions into use of neuroimaging for stroke diagnosis and treatment decision making. ANN NEUROL 2024;95:1017-1034.

Automatic Segmentation in Acute Ischemic Stroke: Prognostic Significance of Topological Stroke Volumes on Stroke Outcome

BACKGROUND

Stroke infarct volume predicts patient disability and has utility for clinical trial outcomes. Accurate infarct volume measurement requires manual segmentation of stroke boundaries in diffusion-weighted magnetic resonance imaging scans which is time-consuming and subject to variability. Automatic infarct segmentation should be robust to rotation and reflection; however, prior work has not encoded this property into deep learning architecture. Here, we use rotation-reflection equivariance and train a deep learning model to segment stroke volumes in a large cohort of well-characterized patients with acute ischemic stroke in different vascular territories.

METHODS

In this retrospective study, patients were selected from a stroke registry at Houston Methodist Hospital. Eight hundred seventy-five patients with acute ischemic stroke in any brain area who had magnetic resonance imaging with diffusion-weighted imaging were included for analysis and split 80/20 for training/testing. Infarct volumes were manually segmented by consensus of 3 independent clinical experts and cross-referenced against radiology reports. A rotation-reflection equivariant model was developed based on U-Net and grouped convolutions. Segmentation performance was evaluated using Dice score, precision, and recall. Ninety-day modified Rankin Scale outcome prediction was also evaluated using clinical variables and segmented stroke volumes in different brain regions.

RESULTS

Segmentation model Dice scores are 0.88 (95% CI, 0.87-0.89; training) and 0.85 (0.82-0.88; testing). The modified Rankin Scale outcome prediction AUC using stroke volume in 30 refined brain regions based upon modified Rankin Scale-relevance areas adjusted for clinical variables was 0.80 (0.76-0.83) with an accuracy of 0.75 (0.72-0.78).

CONCLUSIONS

We trained a deep learning model with encoded rotation-reflection equivariance to segment acute ischemic stroke lesions in diffusion- weighted imaging using a large data set from the Houston Methodist stroke center. The model achieved competitive performance in 175 well-balanced hold-out testing cases that include strokes from different vascular territories. Furthermore, the location specific stroke volume segmentations from the deep learning model combined with clinical factors demonstrated high AUC and accuracy for 90-day modified Rankin Scale in an outcome prediction model.

Variability assessment of manual segmentations of ischemic lesion volume on 24-h non-contrast CT

PURPOSE

Infarct lesion volume (ILV) may serve as an imaging biomarker for clinical outcomes in the early post-treatment stage in patients with acute ischemic stroke. The aim of this study was to evaluate the inter- and intra-rater reliability of manual segmentation of ILV on follow-up non-contrast CT (NCCT) scans.

METHODS

Fifty patients from the Prove-IT study were randomly selected for this analysis. Three raters manually segmented ILV on 24-h NCCT scans, slice by slice, three times. The reference standard for ILV was generated by the Simultaneous Truth And Performance Level estimation (STAPLE) algorithm. Intra- and inter-rater reliability was evaluated, using metrics of intraclass correlation coefficient (ICC) regarding lesion volume and the Dice similarity coefficient (DSC).

RESULTS

Median age of the 50 subjects included was 74.5 years (interquartile range [IQR] 67-80), 54% were women, median baseline National Institutes of Health Stroke Scale was 18 (IQR 11-22), median baseline ASPECTS was 9 (IQR 6-10). The mean reference standard ILV was 92.5 ml (standard deviation (SD) ± 100.9 ml). The manually segmented ILV ranged from 88.2 ± 91.5 to 135.5 ± 119.9 ml (means referring to the variation between readers, SD within readers). Inter-rater ICC was 0.83 (95%CI: 0.76-0.88); intra-rater ICC ranged from 0.85 (95%CI: 0.72-0.92) to 0.95 (95%CI: 0.91-0.97). The mean DSC among the three readers ranged from 65.5 ± 22.9 to 76.4 ± 17.1% and the mean overall DSC was 72.8 ± 23.0%.

CONCLUSION

Manual ILV measurements on follow-up CT scans are reliable to measure the radiological outcome despite some variability.