Continuing our list of top AI terms medical communicators should know, we have entries 11-20. If you haven’t already, read 1-10 first.
The main difference between this glossary and content you will find at other resources (even ones for writers) is that I arrange the terms in an order of importance for medical writers and editors and write the related entries to be consistent with one another. For example, training data, validation data, and test data are related terms that are part of a process, so they are written together using the same example.
Without further ado, 11-20:
11. Training Data
Data used to teach machine learning models by exposing them to various scenarios so they can learn patterns and relationships, allowing it to learn from examples.
Example: 10,000 labeled chest X-rays used to teach an AI model to detect pneumonia.
12. Validation Data
A subset of data used to tune and improve the machine learning model, ensuring its performance generalizes well to unseen data.
Example: 2,000 different chest X-rays used during development to fine-tune the model’s sensitivity in detecting subtle pneumonia patterns
13. Test Data
A dataset used after training to evaluate how well a model performs on unseen data.
Example: 1,000 completely new chest X-rays, never seen by the model before, used only once at the end to evaluate real-world accuracy in pneumonia detection
14. Precision Medicine
The use of AI to analyze genetic, environmental, and lifestyle factors in order to tailor medical treatments to individual patients.
Example: By analyzing a patient’s genetic profile alongside their lifestyle habits and medical history, AI helped doctors select a specific chemotherapy drug that was most likely to be effective while causing fewer side effects for that individual breast cancer patient.
15. Predictive Analytics
The use of data, statistical algorithms, and machine learning techniques to predict future outcomes based on historical data.
Example: A hospital’s AI system analyzed five years of patient data to accurately forecast seasonal flu admission rates, allowing administrators to staff departments appropriately and maintain adequate medical supplies.
16. Chatbots
AI systems that use NLP to interact with users in real-time, often used in health care for patient engagement and administrative assistance.
Example: Memorial Hospital’s virtual assistant helps patients schedule appointments and refill prescriptions 24/7, responding naturally to messages like “I need to see my cardiologist next week” or “I’m running low on my blood pressure medication.”
17. Explainable AI
A set of processes and methods that allow human users to understand and trust the results and outputs created by machine learning algorithms.
Example: When the AI system flagged a chest X-ray as potentially showing pneumonia, it also highlighted the specific areas of opacity it detected and provided the statistical confidence level of its assessment, helping the radiologist understand and verify its reasoning.
18 Artificial General Intelligence (AGI)
A type of AI that aims to perform any intellectual task that a human can, as opposed to narrow AI, which is designed for specific tasks.
Example: While current AI systems can only perform specific tasks like analyzing medical images or processing health records, an AGI system would theoretically be able to conduct patient interviews, perform physical examinations, and make clinical decisions just like a human doctor.
19. Model
In AI, a model is a mathematical structure that represents real-world patterns, behaviors, or phenomena and can make predictions or decisions based on input data.
Example: The diabetes progression model analyzes patterns in thousands of patient records, incorporating factors like blood glucose readings, medication adherence, and lifestyle choices to predict how the disease might develop in similar patients.
20. Application
The practical use of AI algorithms or models to solve real-world problems, such as diagnosing diseases or predicting patient outcomes.
Example: A clinical software tool lets ophthalmologists upload retinal photographs and receive AI-powered diabetic retinopathy screening results within minutes. The application integrates with electronic health records to provide seamless documentation of the screening findings for clinical use.