Clinical Validation
We are conducting prospective, real-time clinical validation studies of artificial intelligence models for medical imaging.
Research Questions
- How does an AI model influence the performance of a radiologist?
- How well does an AI model developed at one institution perform at another institution?
In the past, lack of due diligence in answering these questions stalled the adoption of computer-aided detection (CADe) algorithms for mammography. We want to prevent the same fate for AI in medical imaging by setting a higher standard of quality for AI models intended for use in clinical practice.
We are conducting prospective, real-time randomized controlled trials at multiple institutions to answer these questions. Conducting trials across multiple institutions affords a more robust clinical validation study across different patient populations and imaging devices.
We are starting with a Bone Age model; we will soon be adding new models for chest x-rays, head-CT and more.
Study Participation
We deeply integrate AI models into the clinical workflow to make participation in randomized controlled trials as easy as possible.
If you’re a radiologist ...
- Results from the AI model are directly filled into your report and integrated with your PACS viewer.
If you’re an AI researcher ...
- Model design and training are separated from clinical evaluation and use. After you train your deep learning model, you can initiate a validation study by uploading your model definition files and weights.
Let’s collaborate!
We are looking for new collaborations with hospital partners and AI researchers! Please reach out to info@aimi.stanford.edu if you’re interested!