MUNICH, BOSTON, and STANFORD, Calif., Nov. 30, 2022 /PRNewswire/ -- With AI becoming an integral part of clinical radiology workflows, the data used to develop AI algorithms is ever essential. An identified industry challenge is ensuring post-market surveillance of AI at scale, providing a foundation for the continued clinical safety and improvement of algorithms before and after clinical deployment. AI algorithms already need to meet high-performance standards for clearance by notified bodies such as the FDA.
However, in some clinical scenarios, the performance of an AI algorithm may not meet expectations or degrade over time, e.g. due to different target patient populations or new scanner types. To identify such effects immediately, it is crucial to assess the performance of AI before and during its deployment in routine clinical use. In such cases, AI algorithms can then be re-trained on broader datasets, reflecting the target population, and making them more robust and safe. To address this need, deepc, Centaur Labs, and Segmed have joined forces and started a Radiology AI Safety Initiative to strengthen the continued AI performance in clinical workflows.
Three leaders in the field of Radiology AI have formed a novel AI Safety initiative to make AI algorithm use in clinical routines safer, as well as management and retraining easier.
MedTech company deepc with its leading radiology AI platform deepcOS offers radiologists access to a myriad of regulatory-cleared AI solutions from their globally leading partner network.
Segmed provides secured, de-identified medical data, and Centaur Labs a labeling solution for such data.Collaborating with its AI Partners, deepc helps AI applications fulfill the highest safety and performance standards.
Customers that use deepcOS commonly seek validation and benchmarking assistance to identify the optimal performance of algorithms and pre-launch safety checks, offering surety of selection meeting the specific needs of the clinical environment.
Segmed's self-serve data platform Insight allows AI developers to build cohorts of de-identified and structured data that are readily available for training, testing, and validation of AI algorithms.
Together with Centaur Labs, medical labeling services can be provided for given clinical indication fields to ensure safe operations of AI engines on specific customer data prior to deployment.
The companies are developing a joint offering. AI vendors will be able to perform real-time post-market surveillance via the deepcOS platform and identify when an AI model is performing suboptimally over time, a phenomenon that is often regarded as 'AI drift'. This may be caused by a shifting patient population or new scanner types, for instance. deepc will avail this precise information to AI vendor partners and offer on-demand access to data that is anonymized and cleared for AI R&D through Segmed's data platform, and annotated by Centaur Lab's scalable data annotation platform. This data can be used to retrain the AI algorithm, improve performance, robustness, and clinical AI safety. Obtaining diverse, relevant data and labeling this data are two of the biggest bottlenecks in AI pipelines, solved by deepc's partners Segmed and Centaur Labs.
"An outstanding effort of our companies leads to improved clinical AI safety, eventually leading to better clinical outcomes and increased patient safety. We are very proud to be leading this very important initiative in the industry", says deepc's Co-Founder and CEO Dr. Franz Pfister.