A quality-control driven framework using multiple AI algorithms for robust decision making in MRI and mission-critical AI
A method using multiple AI algorithms to enhance the robustness of data processing.
Current automatic AI methods may fail, especially in a typical case which calls for visual inspection, thus eradicating any improvements in efficiency. This technology addresses the increasing need to enhance the robustness of data processing pipeline with automatic quality scoring; demonstrated in T1 mapping segmentation, yet potentially applicable to all AI decision making.
| Features | Benefits |
|---|---|
| Estimates uncertainty in selected naturally denominated units, e.g. dice coefficients (F, G, I) | Ensures inter-rater reliability between results; Automated quality control in selected units |
| Detects and localises anomalies (E, H) | Aids the identification of abnormal patterns in medical data with highlighted areas of uncertainty |
| Reduces laborious manual review in datasets above the pre-selected confidence levels | Saves both time and costs by reducing the number of decisions to be reviewed by human observers; Increases productivity |
| Ability to be applied to general AI applications | Potential application to decision making beyond clinical imaging |
Awaiting patent and Available For
- Co-development
- Consulting
- Licensing