Paramedic Papers

Home Other GPT estimates pneumothorax size to guide care

Other

GPT estimates pneumothorax size to guide care

Chest X-Ray evaluation using GPT for tube thoracostomy or conservative care in non-tension spontaneous pneumothorax
Scandinavian Journal of Trauma, Resuscitation and Emergency Medicine · Added 22 April 2026

Tested whether a vision-enabled GPT model could estimate pneumothorax size and guide tube thoracostomy vs conservative care, and it matched radiologist measurements closely with good agreement to real-world ED decisions.

Study type
Cohort
Design
Single-centre retrospective observational study of vision-enabled GPT on chest radiographs
Participants
101 adults with confirmed spontaneous pneumothorax
Main finding
Sensitivity 86.8%, specificity 93.8%, accuracy 90.1%, AUC 0.90 for tube thoracostomy within 24h; depth agreement ICC 0.893, MAE 0.69 cm
Caveat
Single-centre retrospective design with confirmed pneumothorax only; no prehospital or prospective testing
Bottom line
Hypothesis-generating

Learning goal

Weigh how artificial intelligence imaging tools could support my judgement on conservative care versus intervention in spontaneous pneumothorax.

Open in dashboard Read the paper (DOI) PubMed

More

Related papers on Other