Zero-shot audio prompting of GPT-4o & Sonnet 3.5 lets radiologists report 2× faster with 2× fewer errors vs Nuance Dragon !!! Soon, we won’t subscribe to traditional tools or our local RIS, we’ll go straight to OpenAI & friends. Times are changing
From dictation to diagnosis: enhancing radiology reporting with integrated speech recognition in multimodal large language models.
Mentions
Extended reach
Total likes
Video views
Mentions per month
Aug 2025 – Sep 2025Where it's discussed
-
X / Twitter10 posts
-
Bluesky3 posts
-
YouTube1 videos
AI synopsis
What people are saying
The discussions demonstrate strong support and fascination with the technological advances, such as mentions of GPT-5's architecture and the shift towards direct AI use in radiology, evident through enthusiastic words like 'wild,' 'potential,' and 'future.' The tone reflects optimism about AI's transformative role, emphasizing increased efficiency and future clinical implications.
Most discussions support the positive impact of multimodal large language models on radiology reporting, highlighting efficiency and accuracy improvements.
The posts reveal high interest by emphasizing emerging AI innovations, new architectures, and potential shifts in clinical workflows.
Posts reference specific technical developments like GPT-5's architecture and practical benefits such as faster reporting and error reduction, indicating active engagement.
The discussions suggest these advancements could significantly transform radiology and healthcare, with comments about moving away from traditional tools highlighting potential widespread influence.
Abstract in brief
- This study compares the effectiveness of multimodal large language models (LLMs) like GPT-4o and Claude Sonnet 35 in radiology reporting to traditional speech recognition software, finding that LLMs significantly reduce reporting time, corrections needed, and costs while producing higher-quality reports.
- Using only minimal audio input, these LLMs can generate detailed radiology reports faster and more accurately than conventional methods, which rely on speech recognition software.
- Despite promising results, the study notes potential future considerations such as licensing costs and the need for further research to confirm the models' applicability across different clinical settings.]
Top posts
Most liked on X and BlueskyGPT-5's hidden "router" architecture, Google's world-generating "Genie," and a model trained on 16 BILLION medical events... The last month in AI has been wild. ✅In our new podcast, @PRouzrokh & @MoeinShariatnia break it all down.
Large language model-based reporting from minimal audio has the potential to improve efficiency and report quality, supporting more streamlined workflows in clinical radiology (Roman Johannes Gertz et al.) #EuropeanRadiology https://t.co/SOi3diIC5K https://t.co/fNnbRy7Ccl
RT @readingroomai: GPT-5's hidden "router" architecture, Google's world-generating "Genie," and a model trained on 16 BILLION medical event…
RT @EurRadiology: Large language model-based reporting from minimal audio has the potential to improve efficiency and report quality, suppo…
RT @readingroomai: GPT-5's hidden "router" architecture, Google's world-generating "Genie," and a model trained on 16 BILLION medical event…
All mentions
AI Developments and Societal Impact of Latest Generative Models
This episode examines recent AI releases like GPT5, Google's Genie, and Nano Banana, discussing their societal impact, reliability issues, and future challenges. It also explores advancements in medical AI, including…
RT @EurRadiology: Large language model-based reporting from minimal audio has the potential to improve efficiency and report quality, suppo…
RT @EurRadiology: Large language model-based reporting from minimal audio has the potential to improve efficiency and report quality, suppo…
Large language model-based reporting from minimal audio has the potential to improve efficiency and report quality, supporting more streamlined workflows in clinical radiology (Roman Johannes Gertz et al.) #EuropeanRadiology https://t.co/SOi3diIC5K https://t.co/fNnbRy7Ccl
RT @readingroomai: GPT-5's hidden "router" architecture, Google's world-generating "Genie," and a model trained on 16 BILLION medical event…
RT @readingroomai: GPT-5's hidden "router" architecture, Google's world-generating "Genie," and a model trained on 16 BILLION medical event…
RT @readingroomai: GPT-5's hidden "router" architecture, Google's world-generating "Genie," and a model trained on 16 BILLION medical event…
GPT-5's hidden "router" architecture, Google's world-generating "Genie," and a model trained on 16 BILLION medical events... The last month in AI has been wild. ✅In our new podcast, @PRouzrokh & @MoeinShariatnia break it all down.
RT @AmineKorchiMD: Zero-shot audio prompting of GPT-4o & Sonnet 3.5 lets radiologists report 2× faster with 2× fewer errors vs Nuance Drago…
RT @AmineKorchiMD: Zero-shot audio prompting of GPT-4o & Sonnet 3.5 lets radiologists report 2× faster with 2× fewer errors vs Nuance Drago…
Zero-shot audio prompting of GPT-4o & Sonnet 3.5 lets radiologists report 2× faster with 2× fewer errors vs Nuance Dragon !!! Soon, we won’t subscribe to traditional tools or our local RIS, we’ll go straight to OpenAI & friends. Times are changing
From dictation to diagnosis: enhancing radiology reporting with integrated speech recognition in multimodal large language models https://pubmed.ncbi.nlm.nih.gov/40815310/
From dictation to diagnosis: enhancing radiology reporting with integrated speech recognition in multimodal large language models https://pubmed.ncbi.nlm.nih.gov/40815310/
From dictation to diagnosis: enhancing radiology reporting with integrated speech recognition in multimodal large language models https://pubmed.ncbi.nlm.nih.gov/40815310/
DK
ER
DS
BM
JD
MR