Human Bytes
Implementing AI
- Implementing AI
Deploying AI in a safe and controlled manner is critical to create confidence in the organisation and to harvest the full value. We have extensive experience with implementing AI, and can support with the most commonly asked questions and concerns, whether solutions are deployed on-prem or cloud.
We offer regulatory approved Software-as-a-Medical Device (SaMD). All our solutions are backed by strong scientific evidence and in clinical use in multiple countries, supported by strong value cases for patient outcomes and health economics.
The AI solutions are standard software, hence using standard data formats and integrates as part of existing work flows. That makes implementation of AI a relatively simple task, especially when deployed via cloud. We have successfully completed end-2-end large-scale installations within a few weeks delivering real-world value to patients, HCPs and health systems.
Here are some of the myths, barriers and misunderstandings we often hear.
AI software is a static clinical software developed with AI, but not learning from any local data. This means, that the AI software does not change, and any result can be reproduced and quality controlled. AI software is a regulatory approved medical device, like an MRI scanner and a patient monitor, or any other medical device or software used in healthcare.
AI software has a normal product liability responsibility, as all other medical devices. AI software is most often decision support for the clinical staff, but autonomous AI software is now seen approved as well. Here AI as has responsibility for the final result or score.
For any medical device, also clinical AI software, there is an intended use that defines what the solution can be used for, or not used for, and on whom. As example, most AI solutions would in their intended use have a definition for which type of patients can be analyzed (e.g. pediatrics), or which type of equipment and exams (e.g. scanners or protocols). We recommend only to procure well tested and documented AI solutions, and do extra quality insurance in the first period.
AI is often compared to the gold standard, of the best radiologist or most difficult patient case. However, AI value creation should be compared to representative clinical data and the actual use case. We see that AI can support in handling or triaging normal or routine cases freeing up time for the clinical staff.
AI is standard software, regulatory approved as a Medical Device. Our solutions are vendor agnostic and use standard data exchange format to ensure fast and seamless integration within scanners, devices, PACS, LIS and other clinical systems. Too often we see Procurement and IT departments thinking it is complex and a large change management exercise, but in fact you should consider it more like buying and implementing a MRI license. Keep simpler things, simple.
Change management and onboarding of staff, is easy and often only requires few hours of training. Reason is that most of our AI solutions are deeply integrated in your normal workflows or operate in the background and only flagging if anything not normal.
Cloud or Local On-Prem
Our solutions can be installed either as Cloud or as a local On-Prem. The benefit of the cloud is often more secure, scalable, future proof and much faster implementation. We have significant experience within cloud for the Nordic healthcare, including the many questions and concerns about privacy and data protection. Before you decide on one or the other, then we recommend you to take an initial discussion with us.
More than 1,000 sites
Our solutions are running in more than 1,000 sites across the world, including several in the Nordics. We have a vast experience of handling large scale implementation, across multiple sites and various legacy systems. We recommend involving your IT department very early in your procurement considerations, to de-mystify the task and create an aligned project plan.
We often come across university hospitals who develops own AI solutions, which is great to build knowledge about how AI works and the opportunities it can bring. However, with the new Medical Device Regulation (MDR), all such home-grown solutions must be fully regulatory certified or alternative get a section 5.5 exceptions. Such an exception will only be valid for own legal entity and can only be granted if no commercial alternative is available. Failure to comply places clear responsibility for any patient harm with the local user and management.