Researchers at the University of Colorado Anschutz School of Medicine have developed a vision-language AI model to assist radiologists by interpreting right upper quadrant abdominal ultrasounds. Published in Nature Communications, the AI agent evaluated approximately 11,000 ultrasounds and achieved a 99.6% diagnostic accuracy in manual spot checks. Designed to draft preliminary imaging reports for conditions like acute cholecystitis, the tool aims to reduce diagnostic delays for emergency surgical patients while mitigating severe radiologist workforce shortages.
Capabilities of the CU ultrasound AI tool
- ▪The University of Colorado AI agent combines visual abdominal ultrasound data with written clinical laboratory data to diagnose common health conditions affecting the liver, gallbladder, right kidney, and pancreas.
- ▪The University of Colorado vision-language AI model rapidly identifies severe abdominal conditions like acute gallbladder inflammation (cholecystitis) to accelerate emergency surgery decisions.
- ▪Researchers at the University of Colorado Anschutz School of Medicine created a vision-language AI agent that reads right upper quadrant (RUQ) abdominal ultrasounds to assist emergency care diagnoses.
Role in writing preliminary imaging reports
- ▪Senior author Harrison Bai stated the University of Colorado AI tool is designed to work alongside trained radiologists by writing preliminary imaging reports rather than replacing physicians.
- ▪The University of Colorado AI tool generates preliminary right upper quadrant ultrasound reports to save time for radiologists facing increasing emergency department workloads.
Performance evaluation of CU AI ultrasound tool
- ▪University of Colorado researchers evaluated their vision-language AI model on a dataset of approximately 11,000 abdominal ultrasounds from Stanford University, University of Colorado, and Johns Hopkins University.
- ▪In a manual review of 100 AI-generated ultrasound reports, the vision-language AI model correctly identified the target diagnoses 99.6% of the time, according to a study in Nature Communications.
- ▪In a blinded evaluation, expert reviewers could generally not distinguish radiologist-edited AI-written ultrasound reports from reports written entirely by human radiologists.
Dubai Medical University triage chair
- ▪Hessa Yahya Lootah stated that Dubai Medical University is adjusting its triage chair technology to handle challenges related to facial light reflection, glare, varying skin tones, and skin conditions.
- ▪Dubai Medical University is developing the AI Diagnostics Vital Chair, a prototype designed to reduce hospital triage time from 5–7 minutes to approximately 45 seconds using cameras and sensors.
Dubai Medical University first-aid drone
- ▪Dubai Medical University's drone first-aid kit utilizes an internal AI system programmed around six compartments to assess injuries and guide emergency responses without needing internet connectivity.
- ▪Dubai Medical University is developing a medical assistance drone equipped with an offline AI system designed to guide individuals through first-aid kits in remote or network-unstable locations.
Story comments
Loading comments…