Hello everyone!
I am supporting health care managers in the identification and classification of all AI-based health care technologies present in their hospitals (technologies for clinical, managerial or patient uses). By AI, I mean a broad set of computation techniques from advanced expert systems to machine-learning algorithms to NLP-driven techniques, and much more.
Due to my lack of knowledge, I am having trouble finding reliable references to set up categories to help managers conduct this survey. How should they classify the AI technologies? Can they rely on clear criteria to tell what technologies are AI-driven and what are not?
I am looking especially for public agency reports (from the FDA or NHS for instance) that expose criteria used to classify information technologies before market approval. But any other literature advice would be really appreciated!
Thanks in advance to the community!
Cécile
Cecile,
You have at least two different goals that merit two different strategies
a) catalogue the existing tools
b) criteria used to evaluate IT before market approval
Further, if I get your question's underlying objective correctly, the focus is more on business operations rather than academic research purposes. The objectives are very different and merit different approaches.
At least start with dividing them into:
- Decision support systems for business operations
- Clinical decision support systems.
Once you have this you can then tackle further dive into the AI problem. What is under consideration is whether a technology is AI driven or not and that is where consensus breaks down and a subjective criteria comes into play. since what you will find is mostly subjective I will at least provide you with one possible scheme(based on [1]) which you can use as a matrix for evaluation:
- Image processing
- Natural language processing
- Robotics
- Planning
- Knowledge and reasoning (Expert systems, probabilistic systems)
- Machine learning
You can then make subcategories within these categories. Notice, that almost all of them when taken in isolation are hard to categorize as AI, it is when taken in combinations that the AI-ness comes to light. This will happen with all the taxonomies that you encounter in the literature.
should you need further assistance, let me know
Regards
[1] Artificial Intelligence A Modern Approach by Stuart J. Russell and Peter Norvig
There is going on a lot in the area you are asking about. AI, ML, DL, DM areas are being more applied in biomedical praxis and we need tools to address their safety, accuracy, predictability, and so on. I
For the beginning, I would recommend the following review paper
https://www.researchgate.net/publication/329945443_Machine_learning_and_AI_research_for_Patient_Benefit_20_Critical_Questions_on_Transparency_Replicability_Ethics_and_Effectiveness
It is written by experts in the field and is of great value to anyone who wants to penetrate into the core of this problematic.
One of the authors, Ioannidis, wrote another excellent review on the validity of research results. I recommend you to read this one too (there is free access)! It is a bestseller!
https://www.researchgate.net/publication/7686290_Why_Most_Published_Research_Findings_Are_False
Lastly, there is a necessity to know about complex systems and how they can be applied to describe and model medical applications. A review, authored by a team of a mathematician and medical doctors, can be found here
https://www.researchgate.net/publication/330546521_Complex_Systems_and_Their_Use_in_Medicine_Concepts_Methods_and_Bio-Medical_Applications
Only after getting all there views on the topic of your question, you will start to get the global overview of seriousness, complications, and bottlenecks of all current research in biomedicine.
All the best in your important work. It is very important as it can help save many lives of patients.f
Cecile,
You have at least two different goals that merit two different strategies
a) catalogue the existing tools
b) criteria used to evaluate IT before market approval
Further, if I get your question's underlying objective correctly, the focus is more on business operations rather than academic research purposes. The objectives are very different and merit different approaches.
At least start with dividing them into:
- Decision support systems for business operations
- Clinical decision support systems.
Once you have this you can then tackle further dive into the AI problem. What is under consideration is whether a technology is AI driven or not and that is where consensus breaks down and a subjective criteria comes into play. since what you will find is mostly subjective I will at least provide you with one possible scheme(based on [1]) which you can use as a matrix for evaluation:
- Image processing
- Natural language processing
- Robotics
- Planning
- Knowledge and reasoning (Expert systems, probabilistic systems)
- Machine learning
You can then make subcategories within these categories. Notice, that almost all of them when taken in isolation are hard to categorize as AI, it is when taken in combinations that the AI-ness comes to light. This will happen with all the taxonomies that you encounter in the literature.
should you need further assistance, let me know
Regards
[1] Artificial Intelligence A Modern Approach by Stuart J. Russell and Peter Norvig
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Artificial Intelligence