AUTOMATED ALBERTA STROKE PROGRAM EARLY CT SCORE ASSESSMENT USING A MACHINE LEARNING BASED ALGORITHM (ID 1362)
Abstract
Background And Aims
The Alberta Stroke Program Early CT Score (ASPECTS) quantifies the extent of early ischemic changes in the middle cerebral artery (MCA) territory on baseline non-contrast CT (NCCT). The StrokeSENS ASPECTS v1.2 software utilizes machine learning to automatically assess ASPECTS on NCCT. The aim of this study was to validate the software’s performance.
Methods
ASPECTS v1.2 was evaluated on a dataset of 200 NCCT studies with ICA, M1 or distal MCA occlusions in patients presenting within 12h from the stroke onset. Expert consensus reading served as reference standard. Software performance was evaluated at an ASPECTS region level using ROC analysis, sensitivity, and specificity. In addition, total ASPECTS was dichotomized at ASPECTS£4 vs >4 and ASPECTS£7 vs > 7 for similar analysis.
Results
AUC for region level ASPECTS agreement vs. reference standard was 0.80 (0.81 for ASPECTS cortical regions; and 0.78 for deep regions). Results for each ASPECTS region are shown in Table 1. When assessing dichotomous ASPECTS, sensitivity/specificity of 0.83/ 0.77 were noted for ASPECTS£4 vs. > 4 and 0.85/0.77 for ASPECTS£7 vs. > 7 dichotomy, respectively.
Table 1. AUC, sensitivity, and specificity for automated ASPECTS at regional-level vs. expert reads.
| M1 | M2 | M3 | M4 | M5 | M6 | I | L | C | IC | |
| AUC | 0.83 | 0.77 | 0.84 | 0.82 | 0.82 | 0.79 | 0.87 | 0.76 | 0.70 | 0.59 |
| Sensitivity | 0.77 | 0.78 | 0.74 | 0.77 | 0.77 | 0.73 | 0.91 | 0.73 | 0.45 | 0.38 |
| Specificity | 0.78 | 0.59 | 0.80 | 0.82 | 0.77 | 0.80 | 0.63 | 0.66 | 0.87 | 0.91 |
Conclusions
When compared to expert reading, StrokeSENS automated algorithms assess ASPECTS at a regional level and globally with good accuracy.
Trial Registration Number
Not applicable
