AI-Enabled Framework for Automated Pendelluft Detection in Critically Ill Patients Using Electrical Impedance Tomography
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Importance : While electrical impedance tomography (EIT) provides real-time monitoring of regional lung ventilation, its clinical implementation is hindered by complex data interpretation. We developed and validated a ChatGPT-generated system for automated pendelluft detection. Objective :To develop and validate an automated system for pendelluft detection using ChatGPT-generated architecture, with focus on real-time processing capability and clinical applicability. Design and Setting:A retrospective multicenter study conducted at three tertiary hospitals in China between January 2020 and December 2024.Participants:Among 458 screened patients, 278 mechanically ventilated patients met the inclusion criteria and were included for the final analysis. Main Outcomes and Measures : The proposed ChatGPT- generated system was developed using a three-component architecture: a modified ResNet-50 backbone for feature extraction, a bidirectional LSTM network (256 hidden units) with an 8-head self-attention layer for temporal analysis, and optimization for 32×32 EIT image processing. Primary outcomes included diagnostic accuracy (sensitivity, specificity, and AUC-ROC) and time efficiency. Secondary outcomes included correlation between pendelluft severity and clinical outcomes. Results : TheChatGPT-generated system achieved superior diagnostic performance (AUC 0.91, 95% CI: 0.87–0.95) compared to conventional machine learning approaches (CNN: 0.85, Random Forest: 0.82, SVM: 0.79). Analysis time decreased from 12.5 minutes (expert review) to 2.8 minutes (ChatGPT). Higher pendelluft grades correlated significantly with increased 28-day mortality (25.4% vs 9.0%, adjusted HR 2.8, 95% CI: 1.6–4.9, P = 0.001). Conclusions and Relevance : OurChatGPT-generated system provides an accurate, efficient solution for automated EIT analysis, potentially enabling real-time clinical decision support in mechanical ventilation management.
