Adopting AI: essential considerations for healthcare providers
Artificial intelligence (AI) is revolutionising the field of radiology, but making the leap into selecting, purchasing, and implementing an AI solution can be a daunting prospect for even the most sophisticated of clinical practices. There are many factors that can influence purchasing decisions; here we provide some crucial points to think about when comparing or adopting clinical AI solutions.
1. Understanding AI certifications and intended use
Understanding whether a medical device has been approved and meets regulatory standards in your region is key for clinical safety and effectiveness. Healthcare providers must ensure that AI vendors meet compliance obligations and align with clinical needs when evaluating AI solutions for use in clinical settings. The use of ‘CE’ under ‘EU MDR’ (Medical Device Regulations) indicates, in the context of medical devices, that the medical device complies with applicable EU regulations. This can provide comfort to clinicians when trialling and enabling a solution for clinical use.
- Annalise.ai solutions are available in 40+ countries and have received many regulatory clearances, including the highly recognised CE certification under the EU MDR 2017/745. The specified intended use under Annalise.ai’s CE certification is to support clinicians in reading imaging studies across a comprehensive set of 130 findings for non-contrast head CT studies and 124 findings for chest x-rays.
2. Validation and compatibility for your organisation’s need
Healthcare providers should examine the applicability of AI solutions to their intended clinical use case and their patient population. Identifying and avoiding geographic or demographic bias is critical to ensure the relevance and utility of AI to clinical practice. AI developers should be able to provide documentation describing how their AI tools are trained and validated across different populations (types and sizes), which will assist with making an informed selection.
- The Annalise Enterprise CXR and CTB tools were trained on large and heterogeneous databases using rigorous manual annotation by fully qualified radiologists. The solutions have been validated in large clinical studies in diverse patient populations. Annalise.ai’s seminal clinical studies are publicly available via peer-reviewed journal publications such as The Lancet Digital Health and European Radiology.
3. Integration and post-deployment support
Identifying and addressing integration challenges early on is critical to ensuring seamless deployment across client sites. IT departments should be engaged from the start of the procurement process to evaluate system interoperability and address potential deployment challenges.
Healthcare organisations must also consider bringing clinicians along for the ‘journey’ of AI adoption. Communication is key; an intentional and structured change management strategy will assist healthcare organisations to transition to new ways of working, ensuring the people who are tasked with implementing and using the software are informed, engaged, and prepared. The AI provider should be able to discuss how they ensure appropriate product maintenance and support, considering potential malfunctions. Many providers, such as teleradiology practices and network organisations, serve multiple clients with different PACS solutions, making customised integration complex. Developer’s transparency, integration agility, and customisability are essential factors to consider when selecting an AI partner.
Ongoing monitoring of the solution in clinical practice is key to identifying any opportunities for improvement and enhancement of the AI model performance. The ability to capture, proactively monitor and analyse feedback along with any performance changes is key to ensuring long-term adoption & accurate use of the solution in clinical practice.
- Annalise.ai has experience deploying its solutions into a range of clinical practice settings, from small private radiology practices to nationally distributed radiology practices with over 250 radiologists, to large public healthcare systems and large teleradiology companies. The Annalise.ai customer success team partners with our customers, providing guidance and support for both clinicians individually and the broader organisation, across the entire deployment journey and beyond. They focus on ensuring the solutions are tailored to your organisational needs and deliver the best possible outcomes.
4. Benefits and costs
Every organisation, team, or practice has its unique set of challenges and reasons for AI adoption. Defining meaningful, tailored success measures is crucial for calculating the benefits of AI solutions against investment costs. Organisations with an informed strategy and solution landscape should be able to identify appropriate outcome measures and beneficiaries.
The AI provider should be prepared to work together with the client to evaluate AI solutions using retrospective and prospective data and to perform formal ROI analyses to ensure the AI solution’s viability and utility.
- Annalise.ai strives to improve patient health outcomes by enabling healthcare institutions to work better. The Annalise team helps identify and measure the right performance indicators.
The table below covers essential factors to consider when you are evaluating or comparing AI tools.
Notes
- Seah, JCY et al. Effect of a comprehensive deep-learning model on the accuracy of chest x-ray interpretation by radiologists: a retrospective, multireader multicase study. The Lancet Digital Health, Volume 3, Issue 8, e496 – e506.
- Buchlak, Q et al. Effects of a comprehensive brain computed tomography deep learning model on radiologist detection accuracy. European Radiology, 2023 Aug 22. doi: 10.1007/s00330-023-10074-8. Epub ahead of print. PMID: 37606663.
The content on this page is provided by the individuals concerned and does not represent the views or opinions of RAD Magazine.
