State-of-the-art generative imaging models were trained to generate synthetic data for CT, MRIs, Ultrasound, and Endoscopic surgery. These models can generate over 160 pathologic classifications, as well as create synthetic segmentations on top of the synthetic image frames. This data can then be used to train or augment downstream AI model training. Segmed is making said data available for licensing to researchers and companies doing medical research.

In addition, Segmed is developing generative AI models to create high-quality synthetic images. These images will also be made available via their Insight platform in the coming months.

By generating large quantities of synthetic images that closely mimic real-world data, this partnership will help to expand the availability of training data, while also augmenting the scope and variability of patient datasets. Potential use cases of the generated data include classification of modality, body part, and reconstruction plane. Synthetic data has the added benefit of protecting patient privacy, as synthetic records cannot be linked back to real patients.

"We're thrilled to be working with NVIDIA and RadImageNet on this initiative, as this collaboration is a great step towards enhancing datasets used for research" said Adam Koszek, CTO & Co-founder of Segmed. "Supplementing the real-world data Segmed already provides with synthetic data can further increase the robustness and adaptability of our customers' AI algorithms and models."

"Generative AI for imaging is at an inflection point, and has the capability to truly democratize healthcare imaging data," said an NVIDIA representative. "We're excited to work with partners like Segmed and RadImageNet to make this a reality."

The goal of this partnership is to accelerate the refinement of medical AI algorithms to improve the accuracy and consistency of medical diagnoses, ultimately leading to better patient outcomes.