{
  "abstract": "Introduction Computational fluid dynamics (CFD)-derived hemodynamic parameters have been proposed as biomarkers for intracranial aneurysm (IA) rupture discrimination. However, most prior studies report inflated training-set performance without rigorous cross-validation or confounder-adjusted baselines. We sought to determine whether hemodynamic parameters provide statistically significant incremental discrimination of rupture status beyond morphological features and anatomical location alone using stringent cross-validation.Methods We analyzed 95 middle cerebral artery (MCA) bifurcation aneurysms (71 ruptured, 24 unruptured) from the Chinese Medical Hospital Aneurysm (CMHA) dataset (Anhui Medical University). Logistic regression models were constructed for morphology-only (5 features), hemodynamics-only (5 features), and combined (10 features) predictor sets. Primary outcome was 5-fold stratified cross-validated area under the receiver operating characteristic curve (CV-AUC). Negative controls included label permutation testing, size-only baseline, location-only baseline, and subsampling stability analysis. Patient-specific wall shear stress (WSS) distributions were computed using a pressure-Poisson flow solver applied to STL surface meshes and calibrated against published ANSYS Fluent values.Results Undulation index was the strongest individual discriminant (AUC = 0.639, p = 0.037). All hemodynamic parameters individually showed weaker discrimination (maximum AUC = 0.586 for minimum WSS). The morphology-only model achieved CV-AUC of 0.586; the hemodynamics-only model, 0.475; and the combined model, 0.503. A size-only baseline achieved CV-AUC of 0.649, outperforming the combined model by 0.146. Label permutation null mean training AUC (0.691) exceeded the observed training AUC (0.644), indicating severe overfitting.Conclusions Hemodynamic parameters derived from CFD do not provide incremental discrimination of rupture status beyond morphological features alone. A simple size-only baseline outperforms sophisticated multi-parameter models, raising fundamental questions about the methodological rigor of prior hemodynamic aneurysm rupture studies.Disclosures A. Gajjar: None. A. Custozzo: None. H. Greene: None.Abstract E-068 Figure 1Patient-specific wall shear stress (pressure-poisson CFD)Abstract E-068 Figure 2",
  "authors": [
    {
      "affiliations": [
        "Neurosurgery, Albany Medical College, Albany, NY"
      ],
      "name": "A Gajjar"
    },
    {
      "affiliations": [
        "Neurosurgery, Albany Medical Center, Albany, NY"
      ],
      "name": "A Custozzo"
    },
    {
      "affiliations": [
        "Neurosurgery, Albany Medical College, Albany, NY"
      ],
      "name": "H Greene"
    },
    {
      "affiliations": [
        "Albany, NY"
      ],
      "name": "A Paul"
    }
  ],
  "title": "E-068 Hemodynamic parameters do not improve discrimination beyond morphology: cross validated study of middle cerebral artery bifurcation aneurysms",
  "uid": "ec566199-1003-5062-b9e9-28d68fda3933"
}