Everyone's talking about AI in medicine. I want to talk about AI in POCUS. Here's a question that I am looking to answer clearly about AI in POCUS:
Can we actually trust it in clinical practice โ right now?
I'm reviewing the evidence, paper by paper, to find out ๐ฌ๐งต
Follow this series.
#POCUS #FOAMed #MedTwitter #AIinMedicine #EmergencyMedicine
Can AI spot a pneumothorax in real time from a bedside ultrasound โ no expert needed?
Short answer: pretty much yes. Here's the breakdown ๐
1/ The setup: 62 patients, 241 lung ultrasound scans. AI reads M-mode images in real time via a portable device connected to the probe. No cloud. No delay.
2/ The results: Sensitivity 92.1% ยท Specificity 80.2% ยท AUC 0.885. For a rule-out tool at 3am, that's impressive.
3/ The catch: COPD, smokers, subcutaneous emphysema, and apical views fool the model more. Know your false positive risk factors.
4/ The big picture: This opens POCUS to nurses, paramedics, and RTs. Expert not required. That's a game-changer for resource-limited settings.
๐ Fiedler et al. CHEST 2024 โ see infographic below โฌ๏ธ
#POCUS #FOAMed #MedTwitter #AIinMedicine #EmergencyMedicine
@journal_CHEST
We decided to do a left thoracentesis to evacuate effusion and start alveolar recruitment maneuvers. Patient was successfully extubated in the forthcoming hours (2/2)
1- Lidocaine injection
2- Pleural space entry
3- Guidewire insertion
4- Catheter insertion
5- Lung reexpanded
@lexfridman@m7mdkurd Thank you for these interviews! If you are still looking for people to interview on this matter, you should consider @UssamaMakdisi for a deep historical dive on the Palestinian - Israeli conflict