Study finds smart glasses boost AI scribe accuracy and more briefs
Flinders study tests vision-enabled AI scribe Researchers at Flinders University have tested a vision-enabled AI clinical scribe combining Google Gemini and Ray-Ban Meta smart glasses and found improved accuracy in documenting pharmacist-patient consultations.
Photo: Pattanaphong Khuankaew/EyeEm via Getty Images
Flinders study tests vision-enabled AI scribe
Researchers at Flinders University have tested a vision-enabled AI clinical scribe combining Google Gemini and Ray-Ban Meta smart glasses and found improved accuracy in documenting pharmacist-patient consultations.
The system processes both audio and video from consultations to capture clinical details such as medication type and strength, achieving 98% documentation accuracy compared with 81% for audio-only systems. It also integrates features such as transcript generation and visual verification within clinician workflows.
The study, based on 110 mock interviews by 10 clinical pharmacists, suggests multimodal AI scribes could reduce documentation burden further, while highlighting the need for human oversight and addressing privacy, consent, and data integration before wider clinical adoption.
4DMedical raises $57M, receives EU CE mark
ASX-listed 4DMedical has secured AU$83 million ($57 million) in a private placement and received CE Mark certification in the European Union for its non-contrast ventilation-perfusion imaging software.
The company's CT:VQ software generates ventilation and perfusion maps from routine non-contrast CT scans without radiotracers. It provides quantitative lung function insights for radiology and respiratory care.
The EU approval and funding follow recent deployments at major academic centres in the United States. Based on a media release, 4DMedical is now preparing to expand its commercial rollout in Europe, where 400,000 nuclear VQ scans are estimated to be performed each year.
Orchestral to pilot medication safety AI in New Zealand
Auckland-based Orchestral, a venture under McCrae Tech, is in discussions to pilot its new medication safety tool, which checks prescriptions in real time to flag dosing and drug interaction risks.
Based on a media release, the tool, delivered through its AI orchestration platform, integrates into existing dispensing platforms to automatically assess prescriptions against clinical guidelines and patient medication histories and raise alerts to support decision-making under time pressure.
The pilot follows evidence of prescribing errors in New Zealand, including a one-week audit last year across 68 Midland pharmacies that identified 1,257 prescription errors, with over a quarter assessed as high risk for patient harm.
Telstra Health's e-requesting goes live
Telstra Health has launched Smart Connect, a pathology electronic requesting capability within MedicalDirector Clinical, with Healius Pathology Network as its first integrated partner.
According to a media release, the feature enables GPs to generate, edit, electronically sign, and securely send pathology requests directly within their existing workflow, using a FHIR-native health information exchange to share data in near real time between clinics and laboratories. Future releases will introduce radiology e-requesting.
Smart Connect adds to the MedicalDirector Clinical suite, which includes an AI scribe and clinician training tools.
Tags:
MobiHealthNews
https://www.mobihealthnews.com/news/anz/study-finds-smart-glasses-boost-ai-scribe-accuracy-and-more-briefsSign in to highlight and annotate this article
Conversation starters
Daily AI Digest
Get the top 5 AI stories delivered to your inbox every morning.
More about
geministudyresearch
Salt: Self-Consistent Distribution Matching with Cache-Aware Training for Fast Video Generation
Video generation models are distilled using self-consistent distribution matching to improve quality under extreme inference constraints, with cache-aware training enhancing real-time autoregressive generation. (1 upvotes on HuggingFace)
Knowledge Map
Connected Articles — Knowledge Graph
This article is connected to other articles through shared AI topics and tags.
More in Research Papers
Salt: Self-Consistent Distribution Matching with Cache-Aware Training for Fast Video Generation
Video generation models are distilled using self-consistent distribution matching to improve quality under extreme inference constraints, with cache-aware training enhancing real-time autoregressive generation. (1 upvotes on HuggingFace)
Swift-SVD: Theoretical Optimality Meets Practical Efficiency in Low-Rank LLM Compression
Swift-SVD is a compression framework that achieves optimal low-rank approximations for large language models through efficient covariance aggregation and eigenvalue decomposition, enabling faster and more accurate model compression. (3 upvotes on HuggingFace)
Communicating about Space: Language-Mediated Spatial Integration Across Partial Views
MLLMs demonstrate limited capability in collaborative spatial communication tasks, achieving only 72% accuracy compared to humans' 95%, with models struggling to build consistent shared mental models unlike human dialogues that become more specific during convergence. (9 upvotes on HuggingFace)
Test-Time Scaling Makes Overtraining Compute-Optimal
Train-to-Test scaling laws jointly optimize model size, training tokens, and inference samples under fixed budgets, revealing that optimal pretraining decisions shift into overtraining regimes when inference costs are considered. (11 upvotes on HuggingFace)
Discussion
Sign in to join the discussion
No comments yet — be the first to share your thoughts!