Tenth MRMC Trial for CAD, AI and Deep Learning Software

Challenge

A developer of CAD, AI and deep-learning software required an MRMC study in support of an FDA submission, and needed the reader panel recruited, qualified and reading within weeks of contract. Reader recruitment is the most common cause of MRMC schedule slippage: panels of 15 to 30 readers must be board certified, representative of intended users, free of conflicts and prior exposure to the software, and available for multiple reading sessions separated by a washout period.

The sponsor also needed the study conducted by a core lab whose MRMC processes had already withstood FDA inspection, so that the design, reading environment and documentation would not become a point of regulatory discussion.

Intrinsic Imaging’s Solution

Intrinsic Imaging drew on its network of more than 200 qualified reviewers to recruit, screen and contract the reader panel within the sponsor's start-up window, documenting each reader's credentials, conflict-of-interest status and prior-exposure attestation. Reader training on the platform and case report form was completed before the first session.

The study followed the company's established MRMC methodology: a case set assembled against a documented reference standard, a fully crossed design with randomized case order, a washout period between unaided and aided sessions, and a validated Part 11 compliant reading environment capturing every interpretation and timestamp.

Project management ran parallel reader cohorts so the full case review volume was completed on schedule, and the dataset and documentation were delivered in the structure the sponsor's statisticians required for MRMC analysis and FDA submission.

Intrinsic Imaging’s Results

• Reader panel of 22 recruited, qualified and trained within four weeks of contract from a 200-plus reviewer network
• Approximately 9,500 case reviews completed on schedule across two reading sessions separated by a documented washout period
• Study conducted under the same MRMC methodology previously inspected by FDA with zero Form 483 observations
• Complete reader credential, conflict-of-interest and prior-exposure documentation delivered with the dataset for the sponsor's submission