The histology slide is analyzed by an open-source artificial intelligence model called LLaVA 1.5, which identifies the cells as being from the cheek. Another version of the model, LLaVA-Med, trained on medical data, correctly identifies the cells as being from breast tissue. The color and pattern of the cells on the slide are described in a whimsical manner, but ultimately the AI provides valuable insights for researchers. GPT4 provides a technical but unhelpful response to the image. To read the full story, one must subscribe to STAT+.
Source link
STAT+: New AI for biopsies can analyze microscopic images, talk to pathologists about your tumor
