{
  "abstract": "Objective Mechanical thrombectomy (MT) is the standard of care for acute ischemic stroke resulting from large-vessel occlusion. However, predicting patient recovery and postprocedural care needs remains a significant challenge. This study evaluated the performance of TabPFN, a novel transformer-based machine learning model, in predicting two key endpoints: favorable 90-day functional outcomes and discharge disposition.Methods We conducted a retrospective analysis of 1,380 patients who underwent MT at a single comprehensive stroke center. TabPFN was compared with five established machine learning models: Random Forest, XGBoost, LightGBM, CatBoost, and Logistic Regression. The model performance was evaluated using the area under the receiver operating characteristic curve (AUROC), area under the precision-recall curve (AUPRC), and Brier score. Feature importance was determined using SHapley Additive exPlanations analysis.Results For predicting a favorable 90-day outcome (modified Rankin Scale [mRS] ≤2), TabPFN achieved an AUROC of 0.85 and an AUPRC of 0.72, with balanced sensitivity and specificity. For predicting discharge disposition, TabPFN demonstrated comparable performance, achieving an AUROC of 0.85 and an AUPRC of 0.92. Across both endpoints, TabPFN consistently matched or outperformed the other models, while maintaining reliable calibration. SHapley Additive exPlanations analysis identified the NIH Stroke Scale score at 24 hours as the most influential predictor of both outcomes.Conclusions TabPFN provided accurate and well-calibrated predictions of long-term functional outcomes and short-term discharge disposition after MT. Its performance, combined with its rapid implementation and interpretability, supports its potential for integration into clinical workflows to support patient counseling and care planning, pending necessary external validation.Disclosures H. Huang: None. J. Cheng: None. S. Huang: None. S. Collins: None. M. Jayaraman: None. D.N. Wolman: None. S. Yaghi: None. L. Shu: None. Z. Jiao: None.Abstract O-044 Figure 1Abstract O-044 Figure 2",
  "authors": [
    {
      "affiliations": [
        "Department of Neurology, Georgetown University School of Medicine, Washington, DC"
      ],
      "name": "H Huang"
    },
    {
      "affiliations": [
        "Department of Neurology, The Warren Alpert Medical School of Brown University, Providence, RI"
      ],
      "name": "J Cheng"
    },
    {
      "affiliations": [
        "Department of Radiology, The Warren Alpert Medical School of Brown University, Providence, RI"
      ],
      "name": "S Huang"
    },
    {
      "affiliations": [
        "Department of Radiology, The Warren Alpert Medical School of Brown University, Providence, RI"
      ],
      "name": "S Collins"
    },
    {
      "affiliations": [
        "Department of Radiology, The Warren Alpert Medical School of Brown University, Providence, RI"
      ],
      "name": "M Jayaraman"
    },
    {
      "affiliations": [
        "Department of Radiology, The Warren Alpert Medical School of Brown University, Providence, RI"
      ],
      "name": "DN Wolman"
    },
    {
      "affiliations": [
        "Department of Neurology, The Warren Alpert Medical School of Brown University, Providence, RI",
        "Department of Neurology, Jersey Shore University Medical Center at Hackensack Meridian Health, Neptune, NJ"
      ],
      "name": "S Yaghi"
    },
    {
      "affiliations": [
        "Department of Neurology, The Warren Alpert Medical School of Brown University, Providence, RI",
        "Department of Neurology, University of Pittsburgh Medical Center, Pittsburgh, PA"
      ],
      "name": "L Shu"
    },
    {
      "affiliations": [
        "Department of Radiology, The Warren Alpert Medical School of Brown University, Providence, RI"
      ],
      "name": "Z Jiao"
    }
  ],
  "title": "O-044 Early outcome prediction after mechanical thrombectomy: application of a tabular foundation (TabPFN) machine learning model",
  "uid": "43d8344d-4ee8-562e-9aae-24bb0abf3314"
}