ORIGINAL RESEARCH article
Front. Neurol.
Sec. Neuro-Oncology and Neurosurgical Oncology
Volume 16 - 2025 | doi: 10.3389/fneur.2025.1603536
This article is part of the Research TopicArtificial Intelligence in Neurosurgical Practices: Current Trends and Future OpportunitiesView all 4 articles
GPT prediction is associated with short-term survival outcomes after decompressive Hemicraniectomy in malignant media infarction
Provisionally accepted
- Leipzig University, Leipzig, Germany
Select one of your emails
You have multiple emails registered with Frontiers:
Notify me on publication
Please enter your email address:
If you already have an account, please login
You don't have a Frontiers account ? You can register here
Introduction: Analysis of the prognostic ability of the large language model (LLM) GPT to predict short-term survival and functional outcome of patients with malignant middle cerebral artery infarction (MCA) after undergoing decompressive hemicraniectomy.Methods: This retrospective study includes 100 patients with malignant MCA infarction after decompressive craniectomy (DC). GPT versions 4.0 and 4.Omni were used to assess the outcome of the patients. 20 patient-specific factors were provided. Each version of GPT was tested with and without context-enrichment (CE). CE versions were provided the the current AHA/ASA 2019 guideline and meta-analyzes of RCTs as context for decision-making. The real-life outcome of the patients, indicated by the modified Rankin Scale (mRS) served as reference. The following endpoints were evaluated: Survival of inpatient stay, achievement of functional status of mRS 0-4 at discharge and at 3, 6 and 12 months post discharge. We analyzed the prognostic prediction of GPT by calculating the area under the curve (AUC) and the Youden Index to determine the optimal cut-off in divergent answers predictions. After dichotomization according to the cut-off set, Chi-squared test (two-sided) was performed.Results:GPT versions 4.0 and 4.Omni were capable to estimate survival of the in-hospital stay. In both versions, CE GPT (GPT 4.Omni (CE) = AUC 0.67; 95% CI: 0.54-0.79; p=0.002; GPT 4.0 (CE) = AUC 0.70; 95% CI: 0.57-0.82; p=0.018) is superior to the non-CE version. GPT 4.0 achieves statistical significance even without CE (AUC 0.66; 95% CI: 0.53-0.78; p=0.018). Non-CE GPT 4.Omni does not reach significance in predicting survival of hospitalization (AUC 0.60; 95% CI: 0.48-0.73; p:0.07). For questions regarding the functional outcome of patients, GPT in both versions was not able to make a sufficient prognostic prediction, but given the pre-stroke mRS, GPT 4.Omni was able to predict the mRS at discharge (p=0.01, Pearson-correlation 0.696).Conclusions: The study shows the already existing high potential of AI in the calculation of short-term outcome. It also shows the existing limitations for the evaluation of more complex questions like functional outcome.
Keywords: decompressive hemicraniectomy, middle cerebral artery infarction, artificial intelligence, gpt, Survival, functional outcome, prediction
Received: 31 Mar 2025; Accepted: 27 Jun 2025.
Copyright: © 2025 Lehmann, Vychopen, Güresir and Wach. This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) or licensor are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
* Correspondence: Sebastian Lehmann, Leipzig University, Leipzig, Germany
Disclaimer: All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.