Why Accuracy Can Mislead Medical AI Decisions
Автор: Clinical AI Academy
Загружено: 2026-01-15
Просмотров: 9
Описание:
Most medical AI models are evaluated with metrics that do not correspond to clinical decision-making.
In this video, we critically examine Van Calster et al. (Lancet Digital Health) and explain why commonly used metrics such as AUC and accuracy can be misleading in medical AI, when they fail, and what should be used instead.
We discuss:
Which performance measures are appropriate for clinical decisions, not benchmarks
How poor metric choice leads to overconfident and unsafe models
Practical guidance for researchers reviewing or building medical AI systems
Paper: Evaluation of performance measures in predictive artificial intelligence models to support medical decisions (Van Calster et al., Lancet Digital Health).
📩 Join Clinical AI Academy: https://clinicalaiacademy.substack.com/
⚠️ Educational content only. Not medical advice.
#MedicalAI #ClinicalAI #MachineLearning #DigitalHealth
Повторяем попытку...
Доступные форматы для скачивания:
Скачать видео
-
Информация по загрузке:
![Generative AI Fully Explained by Daniel Wolf - Journal Club [13]](https://imager.clipsaver.ru/PziUDsbw5tk/max.jpg)
![Guidelines for Artificial Intelligence in Medicine - Journal Club [1]](https://imager.clipsaver.ru/DiYZ0KqJxqs/max.jpg)
![AI Fails Medical Image Spatial Reasoning - Daniel Wolf - Journal Club [10]](https://imager.clipsaver.ru/uTwr4i2XFb0/max.jpg)
![Generative AI for Detecting Drug-Drug interactions in Jordan - Abdullah Hamdan Journal Club [9]](https://imager.clipsaver.ru/UlBrUn6lEtY/max.jpg)
![AI chatbots versus human doctors: who shows more empathy? - Journal Club [6]](https://imager.clipsaver.ru/7vv-x5QCfuU/max.jpg)
