CAMCA Investigator, Matthew Tivnan, got an early acceptance to MICCAI 2024 for his work titled “Hallucination Index: An Image Quality Metric for Generative Reconstruction Models”. The key innovation in this work is to define what hallucinations ARE by defining what they are NOT. The forward diffusion process has noise only, while the reverse diffusion has noise plus hallucinations. This work brings us one step closer to quantifying hallucinations to allow for ethical deployment of generative AI for medical imaging.
Matthew Tivnan, Early acceptance to MICCAI 2024 for “Hallucination Index: An Image Quality Metric for Generative Reconstruction Models”.