Our AI Assess service increases trust for regulators like the FDA and your end-customers.
With AI Assess, we define and curate representative and edge-case patient data subgroups to assure the robustness of your AI application and identify potential biases encoded within.
Show that you’ve independently tested your model to gain the trust of regulatory bodies and end users.
AI Assess
How It Works
Register your AI Model into a privacy preserving container for assessment.
1.
2.
We use a federated Encrypted Model Registry to run an Inference Engine, pulling the appropriate reference dataset.
3.
We conduct reference dataset evaluation and reporting on your AI application.
4.
Receive an assessment of your performance. Results can be Certified and shared as part of an FDA submission or to end-user customers.
4.
Receive an assessment of your performance. Results can be Certified and shared as part of an FDA submission or to end-user customers.
3.
We conduct reference dataset evaluation and reporting on your AI application.
2.
We use a federated Encrypted Model Registry to run an Inference Engine, pulling the appropriate reference dataset.
Register your AI Model into a privacy preserving container for assessment.
1.
Service Options
Preliminary
• You want to run a full assessment again using real-world data to identify and measure several aspects prior to FDA submission or internal deployment.
• You may be at a beginning phase of the training of your AI model and want to run a narrow assessment for a specific “edge-case.”
• This is a service you may want to run at different points of AI model development.
Initial
• You want to validate your model performance metric and have the dataset preserved for life cycle management, certification of your performance directly to the FDA, or to share with end users.
Certify
We are privileged to be supporting and commercializing work outcomes that have been funded by the National Cancer Institute for their call for industry to develop tools for evaluating the safety and effectiveness of cancer-related devices that use artificial intelligence and machine learning.
Use Case - Oncology AI
Federal funding of $377,530.00 under Contract No. 75N91023C00021 from the National Cancer Institute, National Institutes of Health, Department of Health and Human Services, represents 100 percent of the total cost of this program.
"Supervised, Machine Learning-based oncologic algorithms degrade over time and often don't generalize well. The FDA requires the conduct of standalone testing of radiologic AI products to characterize their performance. Sourcing representative data with sufficient variability is time consuming, expensive, and often under -represent the variety of image quality experienced in clinical reality." - John F Kalafut, PhD, CSO of Asher Informatics PBC. "We're thrilled to be partnering with Gradient Health to further develop our solution for our response to this solicitation."
Phase I
We are currently in discussion with potential collaborators for our Phase II, two-year project application proposal. Oncology AI solution developers interested in independent validation of new models and enabling technology and data service companies interested in supporting this development are encouraged to reach out.
"We are excited about the ability to utilize the Rhino Health platform, an established and trusted federated learning and edge computing platform specially designed for medical imaging privacy preserving data collaborations in the Phase II application." state Charlotte Kalafut, CEO of Asher Informatics. "It allows us to validate and test performance with a large and diverse number of data sources while addressing privacy, security and cost concerns."
Phase II
