Accelerate clinical trials using CARPL

What is a clinical trial for an AI model?

Almost all Radiology AI algorithms fall under the broad category of SaMD (Software as Medical Device). Every such algorithm needs local/ regional regulatory clearance for clinical deployment. The scope and ambit of these approvals vary from region to region. For example, US-FDA is needed for deployments in the US, CE is needed for European deployments, and TGA for Australia. But the common need for all such regulatory approvals is a demonstration of clinical evidence and safety by a clinical trial.

Challenges in conducting a clinical trial

The execution of a clinical trial for AI algorithms appears simple and straightforward in theory, but in reality, involves several challenges:

1.?Finding appropriate data – Most clinical trials that provide evidence for a SaMD are done in a retrospective manner. It is crucial to design a protocol that simulates the real-world deployment environment. The right type of clinical data needs to be defined and sourced for the execution of the study. Putting together diverse datasets in a legal, ethically compliant manner matching the predefined inclusion and exclusion criteria is a critical part of the study.

2.?Finding skilled radiologists – Experienced, board-certified radiologists, are needed to read images in a manner defined in the study protocol (with or without AI outputs, as the case may be), to establish ground truth or to realize the intended end-points for the trial.

3.?Finding the right execution platform – In order efficiently run a clinical trial, the data mentioned above has to be loaded onto a reading platform that can be easily accessed by radiologists and has the ability to integrate AI outputs appropriately. Such uniform and user-friendly platforms are difficult to come by.

How can CARPL execute a successful clinical trial?

CARPL’s three-way network easily helps AI providers identify the right healthcare provider as well as radiologists for reading studies. More importantly, CARPL’s feature-rich modules can help the execution of the different phases of the study – screening, ground-truthing, and reading efficiently. Also, CARPL logs the entire course of the clinical study and enables the extraction of these logs in a regulatory-ready fashion. Many radiologists on the CARPL network have previously participated in studies and are familiar with modules on the platform – hence the learning curve, the lead time to initiate a study, and the reading time are significantly lesser.

So if you feel overwhelmed by the challenges of conducting a clinical trial for your AI algorithm. Please feel free to reach out to me.

*LUNIT used CARPL for conducting one of its clinical trials, you can learn more about it through this short video.?

- Dr. Bhanushree Bahl, Sr. Analyst, Ops & Strategy, CARPL.ai