Diagnostic Performance of Multimodal Large Language Models for Central Venous Catheter Assessment Chest Radiographs in the Intensive Care Unit
Medical Sciences · 14.06.2026 · 14 (2) · 315
DOI: 10.3390/medsci14020315 ↗Kurz gefasst
Fachartikel in der Zeitschrift Medical Sciences, online erschienen am 14.06.2026. Verfasst von Christina-Chrysanthi Theocharidou und 14 weiteren Autor:innen. Beteiligte Einrichtung: 1st Intensive Care Unit, General Hospital of Thessaloniki "G. Papanikolaou", 57010 Thessaloniki, Greece. Hauptthemen laut MEDLINE: Intensive Care Units, Central Venous Catheters, Radiography, Thoracic, Catheterization, Central Venous. Die Arbeit nennt 17 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).
Themen: Intensivpflege
Abstract (Originalsprache)
Background
Chest radiography remains central to post-procedural assessment of central venous catheter (CVC) placement in intensive care units. Multimodal large language models (MLLMs) can process medical images, but their reliability for practical radiography tasks remains uncertain. This study assessed the diagnostic performance of MLLMs and intensivists for CVC access classification, CVC tip assessment, and pneumothorax-related radiographic findings.
Methods
In this retrospective diagnostic performance study, consecutive portable anteroposterior chest radiographs obtained after CVC placement in adult critically ill patients were independently evaluated by four intensivists and five MLLMs. A radiologist consensus served as the reference standard. Interobserver agreement and diagnostic performance were assessed using Fleiss’ kappa, Gwet AC1, Cohen’s kappa, accuracy, sensitivity, specificity, precision, F1 score, balanced accuracy, and Matthews correlation coefficient.
Results
The final cohort included 183 unique radiographs. Intensivist reviewers showed high performance for CVC access classification but lower and more heterogeneous performance for CVC tip-position assessment. Among MLLMs, CVC access accuracy ranged from 0.339 to 0.874, whereas CVC tip assessment was dominated by almost universal classification of tips as appropriate, with near-zero specificity and chance-level balanced accuracy. For pneumothorax-related findings, all MLLMs classified every case as negative. Intensivist reviewers had higher balanced accuracy than MLLMs for CVC access classification (difference, 0.420; 95% CI, 0.349–0.490; p < 0.001) and CVC tip assessment (difference, 0.247; 95% CI, 0.205–0.290; p < 0.001). Pneumothorax analyses were exploratory because only five positive cases were present.
Conclusions
The evaluated MLLMs showed unreliable diagnostic performance compared with experienced intensivists. Apparent performance was influenced by class imbalance and dominant-response behavior, supporting cautious task-specific validation and complete diagnostic performance reporting.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- 1st Intensive Care Unit, General Hospital of Thessaloniki "G. Papanikolaou", 57010 Thessaloniki, Greece.
- Department of Microbiology, AHEPA University Hospital, School of Medicine, Aristotle University of Thessaloniki, 54636 Thessaloniki, Greece.
- Radiology Department, Asklipios Diagnostic Center, 54622 Thessaloniki, Greece.
- Interventional Radiology Unit, Clinical Radiology Department, AHEPA University Hospital, 54636 Thessaloniki, Greece.
- Radiology Department, Theagenion Anti-Cancer Hospital, 54639 Thessaloniki, Greece.
- Deparment of Medicine, Medical School, Democritus University of Thrace, 68100 Alexandroupoli, Greece.
ORCID-Profile: Christina-Chrysanthi Theocharidou, Zafeiris Tsinaris, Anastasia Theocharidou, Athanasia-Marina Peristeri, Myrto Tzimou, Evangelos Kaimakamis, Vasileios Voutsas
Zitieren
APA
Theocharidou, C. C., Tsinaris, Z., Karachristos, C., Theocharidou, A., Kourtidis, M., Papadopoulou, K., Peristeri, A. M., Astreinidis, A., Simichanidou, A., Giannaki, C., Tzimou, M., Kaimakamis, E., Voutsas, V., Soulountsi, V., & Lavrentieva, A. (2026). Diagnostic Performance of Multimodal Large Language Models for Central Venous Catheter Assessment Chest Radiographs in the Intensive Care Unit. Medical Sciences, 14(2), 315. https://doi.org/10.3390/medsci14020315
Vancouver
Theocharidou CC, Tsinaris Z, Karachristos C, Theocharidou A, Kourtidis M, Papadopoulou K, et al. Diagnostic Performance of Multimodal Large Language Models for Central Venous Catheter Assessment Chest Radiographs in the Intensive Care Unit. Medical Sciences. 2026;14(2):315. doi:10.3390/medsci14020315
BibTeX
@article{Theocharidou2026,
author = {Theocharidou, Christina-Chrysanthi and Tsinaris, Zafeiris and Karachristos, Christos and Theocharidou, Anastasia and Kourtidis, Michail and Papadopoulou, Kiriaki and Peristeri, Athanasia-Marina and Astreinidis, Athanasios and Simichanidou, Anna and Giannaki, Chrysavgi and Tzimou, Myrto and Kaimakamis, Evangelos and Voutsas, Vasileios and Soulountsi, Vasiliki and Lavrentieva, Athina},
title = {Diagnostic Performance of Multimodal Large Language Models for Central Venous Catheter Assessment Chest Radiographs in the Intensive Care Unit},
journal = {Medical Sciences},
year = {2026},
volume = {14},
number = {2},
pages = {315},
doi = {10.3390/medsci14020315}
}
EUVIMED