Mediwhale

Not always in a technical sense, but transparency about the product is increasingly expected. Many machine learning models cannot give a simple step-by-step reason for each output, so regulators focus on clear information for users: intended use, inputs, how the model was developed and tested, performance, and known limitations.

In the US, one of the criteria for clinical decision support to fall outside device regulation is that a healthcare professional can independently review the basis for its recommendations. FDA has also published transparency principles for machine learning devices with Health Canada and MHRA.

For buyers, a vendor's ability to document how its model works and where it fails is often more useful than heat maps or confidence scores alone.

AI is used to screen large volumes of ECG data from patches, wearables and remote monitors, flagging possible arrhythmias or recordings that need clinician review. This helps manage the large amount of data produced by continuous monitoring.

Important distinctions for buyers:

• A consumer feature on a wearable and a clinical-grade analysis service can have very different intended uses and authorizations.
• Authorization often covers specific recording hardware, leads and populations.
• Signal quality and motion artifacts affect results.

In the US, software that analyzes physiological signals does not meet the criteria for non-device CDS. Confirm each product's status, and check how alerts are routed to clinicians and documented.

AI image reconstruction and denoising use deep learning to produce clearer images from raw scanner data or to reduce noise and artifacts, for example in CT or MRI. Vendors often position these tools as supporting lower-dose or faster scanning protocols.

Unlike detection software, these tools change the image the clinician reads rather than adding a finding. Questions to ask include:

• Is the function built into the scanner or a separate software product?
• Which scanners, protocols and body regions is it authorized for?
• How was diagnostic image quality evaluated, and could fine detail be lost or altered?

These functions are typically regulated as part of the imaging device or as image processing software, depending on the product and market.