Scholarlink

Background <p>Aneurysmal subarachnoid hemorrhage (aSAH) carries high morbidity and mortality. The Subarachnoid Hemorrhage Early Brain Edema Score (SEBES) and its extended version SEBES 6c are computed tomography (CT)-based markers of early brain edema, but their prognostic value remains uncertain. Radiomics enables quantitative characterization of imaging features beyond the visual assessment. Our objective was to compare the predictive performance of SEBES, SEBES 6c, a radiomic SEBES surrogate, and outcome-specific radiomic models for functional outcome, vasospasm, and hydrocephalus after aSAH.</p> Methods <p>We retrospectively analyzed 405 patients with aSAH (2007–2024). SEBES and SEBES 6c were visually scored on admission CT scans by anonymized observers. Radiomic features were extracted from gray and white matter, and models were trained either to reproduce SEBES (radiomic SEBES) or to directly predict outcomes. Multivariable analyses combined radiomic and clinical variables to assess prognostic performance. Model generalizability was additionally evaluated in an independent external cohort.</p> Results <p>SEBES and SEBES 6c alone showed poor discrimination for six-month functional outcome and lost significance after adjustment for World Federation of Neurological Surgeons and modified&#xa0;Fisher scores. The radiomic SEBES model accurately replicated the visual score but did not predict clinical outcomes. In contrast, outcome-specific radiomic models improved discrimination, particularly when combined with clinical variables, achieving the best predictive accuracy. When applied to the external cohort, the radiomics and clinical model preserved its discriminative ability, demonstrating robustness across data sets.</p> Conclusions <p>SEBES and SEBES 6c reflect visible CT edema but provide limited independent prognostic information. Radiomics offers a quantitative and reproducible alternative that complements, rather than replaces, clinical assessment. Outcome-specific radiomic models, especially when integrated with established clinical variables, show promise for improving prognostic stratification after aSAH, although external multicenter validation remains essential.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Radiomics Versus the Human Eye: Rethinking SEBES for Prognostic Stratification in Aneurysmal Subarachnoid Hemorrhage

  • 期刊论文

  • 发表日期：

    2026年1月15日

查看全文
  • Mónica Maldonado-Luna,
  • Gemma Urbanos,
  • Ana M. Castaño-León,
  • Andreea E. Baciu,
  • Luis Miguel Moreno-Gómez,
  • Guillermo García-Posadas,
  • Leandro Tosi,
  • Carlos Loynaz-Cardona,
  • Alfonso Lagares

摘要

Background

Aneurysmal subarachnoid hemorrhage (aSAH) carries high morbidity and mortality. The Subarachnoid Hemorrhage Early Brain Edema Score (SEBES) and its extended version SEBES 6c are computed tomography (CT)-based markers of early brain edema, but their prognostic value remains uncertain. Radiomics enables quantitative characterization of imaging features beyond the visual assessment. Our objective was to compare the predictive performance of SEBES, SEBES 6c, a radiomic SEBES surrogate, and outcome-specific radiomic models for functional outcome, vasospasm, and hydrocephalus after aSAH.

Methods

We retrospectively analyzed 405 patients with aSAH (2007–2024). SEBES and SEBES 6c were visually scored on admission CT scans by anonymized observers. Radiomic features were extracted from gray and white matter, and models were trained either to reproduce SEBES (radiomic SEBES) or to directly predict outcomes. Multivariable analyses combined radiomic and clinical variables to assess prognostic performance. Model generalizability was additionally evaluated in an independent external cohort.

Results

SEBES and SEBES 6c alone showed poor discrimination for six-month functional outcome and lost significance after adjustment for World Federation of Neurological Surgeons and modified Fisher scores. The radiomic SEBES model accurately replicated the visual score but did not predict clinical outcomes. In contrast, outcome-specific radiomic models improved discrimination, particularly when combined with clinical variables, achieving the best predictive accuracy. When applied to the external cohort, the radiomics and clinical model preserved its discriminative ability, demonstrating robustness across data sets.

Conclusions

SEBES and SEBES 6c reflect visible CT edema but provide limited independent prognostic information. Radiomics offers a quantitative and reproducible alternative that complements, rather than replaces, clinical assessment. Outcome-specific radiomic models, especially when integrated with established clinical variables, show promise for improving prognostic stratification after aSAH, although external multicenter validation remains essential.