AI Model Based on Diaphragm Ultrasound to Improve the Predictive Performance of Invasive Mechanical Ventilation Weaning: Prospective Cohort Study
KI-Modell auf Basis von Zwerchfell-Ultraschall zur Verbesserung der Vorhersage des Weaning-Erfolgs bei invasiver mechanischer Beatmung: prospektive Kohortenstudie
JMIR Formative Research · 08.09.2025 · 9 · e72482-e72482
DOI: 10.2196/72482 ↗Auf einen Blick
Bei 88 kritisch kranken Erwachsenen verbesserte ein multimodales KI-Modell aus klinischen Daten und Zwerchfell-Ultraschall die Vorhersage des Weaning-Erfolgs. Das Modell war den Einzelmodellen überlegen.
Fragestellung
Kann ein KI-basiertes multimodales Lernverfahren aus klinischen Daten und Zwerchfell-Ultraschall das Weaning-Versagen besser vorhersagen?
Methode
Prospektive Studie in 2 Intensivstationen in Guangzhou, China; 88 Patient:innen; vor dem Spontanatmungsversuch wurden Ultraschallvideos und klinische Daten erhoben; Vergleich mehrerer Modellkombinationen.
Ergebnisse
88 Patient:innen: 50 erfolgreich entwöhnt, 38 mit Weaning-Versagen. Das multimodale Modell erreichte die beste Leistung (Accuracy 0,8331; AUC 0,894; average precision 0,91) und war gut kalibriert.
Schlussfolgerung der Autor:innen
Die Kombination von Ultraschall- und klinischen Daten verbesserte die Vorhersage des Weaning-Ergebnisses. Das multimodale Lernen mit automatischer Messung erhöhte die praktische Anwendbarkeit.
Kernaussagen
- Multimodales Modell besser als Einzelmodelle
- Beste Leistung: Accuracy 0,8331; AUC 0,894
- Gut kalibriertes Modell mit average precision 0,91
Deutsche Zusammenfassung: KI-gestützt aus dem Original-Abstract erstellt; alle Zahlen wurden automatisch mit dem Original abgeglichen. Maßgeblich ist der Original-Abstract.
Kurz gefasst
Kohortenstudie in der Zeitschrift JMIR Formative Research, online erschienen am 08.09.2025. Verfasst von Feier Song und 7 weiteren Autor:innen. Beteiligte Einrichtung: Department of Emergency and Intensive Care Unit, Guangdong Provincial People's Hospital (Guangdong Academy of Medical Sciences), Southern Medical Uni…. Hauptthemen laut MEDLINE: Diaphragm, Ventilator Weaning, Artificial Intelligence, Respiration, Artificial. Die Arbeit wurde bisher 8-mal zitiert (Crossref) und nennt 33 Literaturangaben. Der Volltext ist frei in PubMed Central verfügbar.
Themen: Beatmung und Weaning
Abstract (Originalsprache)
Background
Point-of-care ultrasonography has become a valuable tool for assessing diaphragmatic function in critically ill patients receiving invasive mechanical ventilation. However, conventional diaphragm ultrasound assessment remains highly operator-dependent and subjective. Previous research introduced automatic measurement of diaphragmatic excursion and velocity using 2D speckle-tracking technology.
Objective
This study aimed to develop an artificial intelligence–multimodal learning framework to improve the prediction of weaning failure and guide individualized weaning strategies.
Methods
This prospective study enrolled critically ill patients older than 18 years who received mechanical ventilation for more than 48 hours and were eligible for a spontaneous breathing trial in 2 intensive care units in Guangzhou, China. Before the spontaneous breathing trial, diaphragm ultrasound videos were collected using a standardized protocol, and automatic measurements of excursion and velocity were obtained. A total of 88 patients were included, with 50 successfully weaned and 38 experiencing weaning failure. Each patient record included 27 clinical and 6 diaphragmatic indicators, selected based on previous literature and phenotyping studies. Clinical variables were preprocessed using OneHotEncoder, normalization, and scaling. Ultrasound videos were interpolated to a uniform resolution of 224×224×96. Artificial intelligence–multimodal learning based on clinical characteristics, laboratory parameters, and diaphragm ultrasonic videos was established. Four experiments were conducted in an ablation setting to evaluate model performance using different combinations of input data: (1) diaphragmatic excursion only, (2) clinical and diaphragmatic indicators, (3) ultrasound videos only, and (4) all modalities combined (multimodal). Metrics for evaluation included classification accuracy, area under the receiver operating characteristic curve (AUC), average precision in the precision-recall curve, and calibration curve. Variable importance was assessed using SHAP (Shapley Additive Explanation) to interpret feature contributions and understand model predictions.
Results
The multimodal co-learning model outperformed all single-modal approaches. The accuracy improved when predicted through diaphragm ultrasound video data using Video Vision Transformer (accuracy=0.8095, AUC=0.852), clinical or ultrasound indicators (accuracy=0.7381, AUC=0.746), and the multimodal co-learning (accuracy=0.8331, AUC=0.894). The proposed co-learning model achieved the highest score (average precision=0.91) among the 4 experiments. Furthermore, calibration curve analysis demonstrated that the proposed colearning model was well calibrated, as the curve was closest to the perfectly calibrated line.
Conclusions
Combining ultrasound and clinical data for colearning improved the accuracy of the weaning outcome prediction. Multimodal learning based on automatic measurement of point-of-care ultrasonography and automated collection of objective clinical indicators greatly enhanced the practical operability and user-friendliness of the system. The proposed model offered promising potential for widespread clinical application in intensive care settings.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- Department of Emergency and Intensive Care Unit, Guangdong Provincial People's Hospital (Guangdong Academy of Medical Sciences), Southern Medical University, Guangzhou, 510080, China.
- Medical Big Data Center, Guangdong Provincial People's Hospital (Guangdong Academy of Medical Sciences), Southern Medical University, Number 106, Zhongshaner Road, Guangzhou, 510080, China, 86 13416404410.
- Department of Cardiology, Guangdong Provincial People's Hospital (Guangdong Academy of Medical Sciences), Southern Medical University, Guangdong Provincial Cardiovascular Institute, Guangzhou, 510080, China.
- Department of Critical Care Medicine, Guangdong Provincial People's Hospital (Guangdong Academy of Medical Sciences), Southern Medical University, Guangdong Provincial Geriatrics Institute, No. 106, Zhongshaner Rd, Guan…
- Office of Organ Procurement Organizations, Medical Department, Guangdong Provincial People's Hospital (Guangdong Academy of Medical Sciences), Southern Medical University, Guangzhou, 510080, China.
ORCID-Profile: Feier Song, Huazhang Liu, Huan Ma, Xuanhui Chen, Shouhong Wang, Tiehe Qin, Huiying Liang, Daozheng Huang
Zitieren
APA
Song, F., Liu, H., Ma, H., Chen, X., Wang, S., Qin, T., Liang, H., & Huang, D. (2025). AI Model Based on Diaphragm Ultrasound to Improve the Predictive Performance of Invasive Mechanical Ventilation Weaning: Prospective Cohort Study. JMIR Formative Research, 9, e72482-e72482. https://doi.org/10.2196/72482
Vancouver
Song F, Liu H, Ma H, Chen X, Wang S, Qin T, et al. AI Model Based on Diaphragm Ultrasound to Improve the Predictive Performance of Invasive Mechanical Ventilation Weaning: Prospective Cohort Study. JMIR Formative Research. 2025;9:e72482-e72482. doi:10.2196/72482
BibTeX
@article{Song2025,
author = {Song, Feier and Liu, Huazhang and Ma, Huan and Chen, Xuanhui and Wang, Shouhong and Qin, Tiehe and Liang, Huiying and Huang, Daozheng},
title = {AI Model Based on Diaphragm Ultrasound to Improve the Predictive Performance of Invasive Mechanical Ventilation Weaning: Prospective Cohort Study},
journal = {JMIR Formative Research},
year = {2025},
volume = {9},
pages = {e72482--e72482},
doi = {10.2196/72482}
}
EUVIMED