Bastien LE GUELLEC

We're proud to welcome the journal's 2025-2027 Trainee Editorial Board! Learn more about these talented individuals: pubs.rsna.org/page/ai/teb

A New Era of Text Mining in Radiology with Privacy-Preserving LLMs! Open-source LLMs may be the best way to deploy LLMs in healthcare while ensuring patient privacy and data security. Study: doi.org/10.1148/ryai.2… Commentary: pubs.rsna.org/eprint/UVZP8JM… @cxbln @Radiology_AI

Using an open-source model off the shelf to extract information from real-life radiology reports without subsequent training! All you need are few examples

@BitValentine @Radiology_AI @AghilesHamroun @univ_lille Thank you! We hope so, and even more now that Llama 3 is out, with the biggest release tailgating GPT3.5 on numerous benchmarks

@Radiology_AI @BLeGuellec @AghilesHamroun @univ_lille @Radiology_AI Could this revolutionize radiological analysis and reporting? Impressive real-world application of open-source language models.

Local open-source #LLM extracted info from real-life free-text radiological reports with excellent accuracy without specific training doi.org/10.1148/ryai.2… @BleGuellec @AghilesHamroun @univ_lille #reporting #AI #MachineLearning

@FisherJuan76878 @Radiology_AI @AghilesHamroun @univ_lille Hi! Thank you for your interest! Not quite a record (pubs.rsna.org/doi/10.1148/ra… for example). However, we also reviewed those reports by hand, contrary to previous work on databases, which is by far more time-consuming than reviewing them by an AI ;)

#Vicuna, an open-source #LLM, reviewed 2398 emergency brain MRI free-text reports doi.org/10.1148/ryai.2… @BleGuellec @AghilesHamroun @univ_lille #reporting #AI #MachineLearning

Local open-source #LLM extracted info from real-life free-text radiological reports with excellent accuracy without specific training doi.org/10.1148/ryai.2… @BleGuellec @AghilesHamroun @univ_lille #AI #ML #MachineLearning

Proof-of-concept study shows the potential of open-source LLMs to perform information extraction tasks from real-life free-text radiology reports doi.org/10.1148/ryai.2… @AghilesHamroun @PasteurLille @univ_lille #NLP #Vicuna #AI

Searching for pertinent articles for your research? With a simple, transparent algorithm based on citation networks, BibliZap.org changes the game! And the best part: it is entirely free to use! Try it now!

Notre #EscapeGame pédagogique pour l'initiation à la #recherche et la lecture critique d'articles pour les étudiants de 3ème année de médecine mis à l'honneur sur le site de l'@UFR3S_ULille avec un reportage à la clé 😎🙌 : ufr3s.univ-lille.fr/detail-actuali…

Vicuna-13B, a large language model that can be run locally without deidentification of patient data, assigned labels to chest radiography reports, with moderate to substantial agreement with existing labelers. bit.ly/3ZNXoZy

My opening statement at the Senate Intelligence Committee yesterday.

Radiology-Llama2: Best-in-Class LLM for Radiology Presents an LLM based on Llama 2 tailored for radiology. It's tuned on a large dataset of radiology reports to generate coherent and clinically useful impressions from radiology findings. This an interesting use case of how to leverage instruction tuning for what I consider a hard and knowledge-intensive domain. A few models are assessed and compared including the proposed model, Llama 2, Dolly, ChatGPT, and Alpaca. It seems that Radiology-Llama2 produces the top performance on a couple of benchmarks. Great insights included! arxiv.org/abs/2309.06419

Neuroscience is slowly reaching a stage of advancement where it is going to become very useful for fundamental advances in AI. And this is partly thanks to AI, e.g. image segmentation.