Abstract Body: Introduction: White matter hyperintensity (WMH) burden is associated with poor cognitive and functional outcomes after intracerebral hemorrhage (ICH). Grading WMH burden is commonly performed using morphometric scales like Fazeka’s scores, which require training and may be prone to subjectivity. Deep-learning models provide a potential alternative to quantify WMH burden. We compare WMH volumes calculated using the SPOTTY tool with categorical Fazekas scores.

Methods: An analytical group of adult (>18) ICH patients was created by randomly sampling from the Registry for Neurological Endpoint Assessment among patients with Ischemic and Hemorrhagic Stroke (REINAH) database, which includes Fazeka’s scores for ICH patients with available MRIs. Total Fazekas scores (Periventricular + Deep) were determined and patients were randomly selected at each level. T2 FLAIR images used for initial Fazekas scoring were retrieved and the previously trained SPOTTY tool was used mask WMH lesions and calculate total lesion volume in cm³. Calculated WMH volumes are reported as medians and interquartile ranges, with statistical differences assessed across Total Fazekas score (0-6), severe WMH burden in the periventricular (PV) (Fazeka’s score =3) and deep (Fazeka’s score ≥ 2) white matter, and in-hospital mortality using Mann Whitney Rank Sum and Kruskal Wallis tests of hypothesis.

Results: The final cohort included 476 total patients admitted between May 2016 and August 2024. Patients had a median age of 68 [55-77] years and included 236 (49.5%) female and 86 (18.0%) Hispanic patients, with a racial distribution of 289 (60.6%) White, 131 (27.5%) Black, 43 (9.0%) Asian/Native Hawaiian or Pacific Islander, and 14 (2.9%). 147 (30.9%) had severe PV WMH, 251 (52.7%) showed severe deep WMH, and 72 (15.1%) died in-hospital. Significant differences were found between SPOTTY-calculated volumes for non-severe and severe PV WMH (4.40 [1.47-10.15] vs 24.01 [14.99-35.69] p<0.001), non-severe and severe deep WMH (2.51 [0.98-5.72] vs 18.45 [9.47-31.06]; p<0.001) (Figure 1), and across Total Fazekas Score (Figure 2; p<0.001). Patients who died in-hospital similarly had higher SPOTTY WMH volumes than those who did not (12.51 [3.84-24.43] vs 7.85 [2.31-19.29]; p=0.032).

Discussion: The SPOTTY tool provides an avenue for automatic assessment of WMH burden in ICH patients.