AI in medicine
5 important questions on AI in medicine
What are the two different kinds of models of AI? What is important to mention in this topic?
- Specific task -> specific outcome. Human training is needed.
- No specific task but a foundational model. Adapted downstream with little to no training.
We were unable to start with foundational models: there was too little information at that time.
AI is applied in radiology; what is the problem?
- products
- methods
- countries
AI is applied in drug discovery; how?
- New molecules that don’t exist in nature
- Target structure prediction
- Drug-drug interactio
- Drug- target binding affinity (I have a molecule -> what can I expect from it)
- Higher grades + faster learning
- Never study anything twice
- 100% sure, 100% understanding
AI in emergency room; how?
- Interpreting ECG > 24/7
- AI assisted consultation
- image assistance
- symptoms checker: not influenced with emotions
- Length of stay prediction
What are the challenges with AI in medicine?
- Data scarcity: privacy, not enough of the same cases, data not shared between organizations, rare diseases, resource limitations, lack of structure/annotation
- Explainability and trustworthiness: black box problem, what does the tool actually
- Bias: due to people training them and chosen endpoints, 44% gender bias (female voice works less well), 25% gender and race bias
The question on the page originate from the summary of the following study material:
- A unique study and practice tool
- Never study anything twice again
- Get the grades you hope for
- 100% sure, 100% understanding

















