Publications by authors named "Curt Langlotz"
Article Synopsis
- Large language models (LLMs) can perform various language tasks without specific training data, but their inaccurate and potentially harmful outputs limit their use in clinical settings.
- A study evaluated Almanac, an LLM framework designed for medical guidance, by comparing it with standard LLMs like ChatGPT-4, based on responses to 314 clinical questions from a panel of healthcare experts.
- Results indicated that Almanac significantly outperformed standard LLMs in factuality, completeness, user satisfaction, and safety, highlighting the need for thorough testing of these models before clinical application.
Article Synopsis
- Large-language models have shown strong performance in tasks like summarization and dialogue generation, but their use in clinical medicine is hindered by inaccuracies and harmful outputs.
- The study introduces Almanac, a model enhanced with retrieval features for providing medical guidelines and treatment recommendations, which was tested using clinical scenarios evaluated by physicians.
- Findings indicate that Almanac improved factuality by an average of 18% and also enhanced completeness and safety, highlighting the model's potential in clinical decision-making when properly tested and implemented.
This study describes the US geographic distribution of patient cohorts used to train deep learning algorithms in published radiology, ophthalmology, dermatology, pathology, gastroenterology, and cardiology machine learning articles published in 2015-2019.
View Article and Find Full Text PDF