AI and Medicine
Integrating AI into existing medical infrastructure
Integrating AI systems into the existing medical infrastructure has the potential to be a complex process that requires careful planning and coordination between IT and healthcare professionals. Some of the key considerations include: Compatibility: AI systems must be compatible with existing IT systems, such as electronic health record (EHR) systems, medical imaging systems, and medical devices. This requires a thorough understanding of the data formats, protocols, and interfaces used by these systems. We are seeing this take place with the likes of the Prometheus trial, which has a computer interfacing with a standard echo machine to operate. The company EchoNous has been working on something similar in urology, again integrating AI into ultrasound machines in a manner that assists the sonographer in doing the scan, without requiring lots of new technology (EchoNous, 2018, 0:01:22). Data integration: AI systems require large amounts of data to learn and make accurate predictions. Healthcare organizations must ensure that data is integrated seamlessly from various sources and is of high quality. This requires vigorous data governance and management processes to ensure that data is accurate, complete, and consistent. An example of a simple and smooth way to do this in the UK would be to give the DL algorithms access to the data received from the National Data Disclosure scheme which the NHS already runs to aid research and planning. This would give the AI a large and varied data set that would supply information on an appropriate scale to ensure quality decision-making (NHS, 2022). Workflow integration: AI systems must be integrated into the existing clinical workflow to ensure that they are used effectively and do not disrupt patient care. This requires an understanding of the clinical processes and roles of healthcare professionals, such as physicians, nurses, and technicians. Examples of simpler forms of AI being integrated into regular workflows include the likes of CASNET in the UK (used to diagnose and prescribe drugs for glaucoma) (Weiss, 1978) and MYCIN in America (used to identify bacteria and recommend antibiotics for them) (Shortcliffe, 1974). Training and education: Healthcare professionals must be trained and educated on the use of AI systems to ensure that they are used effectively and safely. This includes an understanding of how to interpret AI-generated results and recommendations, as well as how to respond to errors and unexpected outcomes. This is something that could be integrated into the medical school curriculum alongside the likes of medical statistics and medical ethics. By addressing these considerations, healthcare organizations can successfully integrate AI systems into their existing medical infrastructure to improve patient care. AI as a technology has the potential to be very successfully integrated into existing healthcare as it is, at its core, a piece of software. Provided that the device you access it on is powerful enough, you simply need to download it and then you have access to its functionality. This makes it incredibly practical to use, as well as making it affordable. The need to ensure that the device you are using has the capacity to process all of the functions an AI needs to carry out potentially limits the ability for AI to have global use. However, the alternative of staffing and training workers to do these tasks in areas with a lack of expertise is potentially more expensive and so less feasible.
36
Made with FlippingBook - PDF hosting