An introduction to artificial intelligence
If you’re interested in artificial intelligence (AI) but don’t fully understand it, we provide some explanations and context about how AI can be used in healthcare.
Wednesday, 14 August 2024
Artificial intelligence (AI) is rapidly changing many aspects of clinical practice. As AI becomes an increasingly significant feature in the delivery of healthcare, it is important to understand the fundamentals.
Start your AI journey by understanding some terminology and how AI is being used in clinical care. This information is introductory, not comprehensive, so we have also provided links for further reading.
What is AI?
At its simplest, artificial intelligence is a branch of computer programming that focuses on creating machines or software that can perform tasks typically requiring human intelligence. AI uses ‘algorithms’ which are step-by-step instructions or rules that use mathematic and statistical techniques to make predictions, provide recommendations and/or generate new content (outputs) based on the information provided (inputs). Inputs can be audio (such as a conversation between people), visual (such as images) or text (via keyboard and mouse). AI can operate at various levels of automation, and includes systems that:
- automate decision-making
- categorise complex data
- generate new content.
How AI is used in healthcare
AI is increasingly used in healthcare, including for:
- system efficiency, such as
- waiting list management, triage, appointment scheduling
- medical documentation, clinical notes
- synthesising data
- prediction – length of stay, unplanned readmission, staffing requirements
- clinical decision support to make treatment recommendations
- diagnostics, such as
- radiology scan review and interpretation
- IVF embryo selection
- breast screening
- dermatology screening
- pathology
- diabetic retinopathy and glaucoma screening in ophthalmology
- knowledge generation, such as
- discovering new drugs
- keeping up to date with scientific papers
- patient information
- preventive medicine, such as
- risk profiling for early intervention
- sepsis prediction.
Types of AI
Generative AI
Generative AI generates new content in response to a prompt. It can answer questions, create images (including deep fake images of real people) and audio, code software and write songs, poems, essays or any piece of content.
Large language model (LLM)
Large language models is a type of generative AI that works with text. It can interpret, summarise, generate and predict new content. LLMs predict text and generate language that is grammatically and semantically correct within the context of the prompt and the information the LLM has been trained on. Chat GPT is an example of a LLM generative AI.
LLMs can be used to:
- generate text on any topic the LLM has been trained on
- translate from one language to another
- summarise blocks or multiple pages of text
- rewrite content
- classify and categorise content
- analyse the sentiment of a piece of content
- converse with people via a chatbot.
Multimodal foundation model (MfM)
Multimodal foundational model is a type of generative AI that can process and produce multiple data types such as text, images, audio and video.
Generally, AI scribes use generative AI to ‘listen’ to a patient consultation and generate clinical notes, referral letters, patient information sheets and treatment plans. AI scribes process sound (audio data) into text to generate the content.
Hallucination
Hallucination is a risk of generative AI and occurs when the AI produces false information (text, image, audio or video) in response to a prompt. Another term for an AI hallucination is a confabulation.
Hallucinations often appear plausible because LLMs are designed to produce fluent, coherent text. They occur because LLMs have no understanding of the underlying reality that language describes. Hallucinations could inadvertently spread misinformation, fabricate references or misrepresent a clinical observation or diagnosis.
This is one of the reasons why the output from any generative AI system needs to be carefully checked for accuracy. In a clinical context, responsibility rests with the doctor to ensure the output is accurate.
Machine learning
Machine learning (ML) is a type of artificial intelligence (AI) that enables computers to learn and make decisions on their own. Instead of being explicitly programmed, these systems use mathematical models and algorithms to recognise patterns, make predictions, and offer recommendations based on data they've been trained on.
ML systems improve over time by learning from new data, similar to how humans learn from experience. There are different types of machine learning, including methods that mimic the human brain's structure, like artificial neural networks. Examples of ML include tools in healthcare, where AI helps analyse medical images to identify potential pathology.
It's crucial to remember that while machine learning can reduce errors caused by human fatigue or bias, it can still make mistakes if the data it's trained on is flawed. Therefore, human expertise is always needed to confirm AI's suggestions or diagnoses.
Natural language processing
Natural language processing (NLP) is the ability of computers to understand and process human language. It is used in translation tools, chatbots and virtual assistants. NLP recognises and interprets speech, mimicking human hearing (machine hearing).
A chatbot is a computer program that uses NLP to simulate and process human conversation (either written or spoken), allowing humans to interact with digital devices as if they were communicating with a real person. They can be simple, taking the user through pre-defined questions, or more complex, using LLMs and ML to produce more sophisticated answers to questions.
Virtual assistants such as Siri and Alexa are designed to assist users with tasks and retrieving information, using machine hearing and natural language processing.
During the COVID pandemic, a chatbot, Billie the Bot, was used in Queensland to triage patients and provide care options based on information provided by the patient about their symptoms.
Computer vision
Computer vision (CV) AI recognises and analyses images and videos, mimicking human sight. CV is used in facial recognition, object detection and self-driving cars.
CV, together with ML, is used in radiology and medical imaging to interpret images and detect potential abnormalities.
Resources
Avant factsheet - Artificial Intelligence for medical documentation
Digital NSW - A common understanding: simplified AI definitions from leading standards
RANZCR - Artificial Intelligence
Australian Government Department of Industry, Science and Resources - Supporting responsible AI: discussion paper and Safe and responsible AI in Australia
Australian Government Department of Industry, Science and Resources - Artificial intelligence
Australian Government Department of Industry, Science and Resources - Australia’s Artificial Intelligence Ethics Framework
More information
For medico-legal advice, please contact us here, or call 1800 128 268, 24/7 in emergencies.
Disclaimers
This publication is not comprehensive and does not constitute legal or medical advice. You should seek legal or other professional advice before relying on any content, and practise proper clinical decision making with regard to the individual circumstances. Persons implementing any recommendations contained in this publication must exercise their own independent skill or judgement or seek appropriate professional advice relevant to their own particular practice. Compliance with any recommendations will not in any way guarantee discharge of the duty of care owed to patients and others coming into contact with the health professional or practice. Avant is not responsible to you or anyone else for any loss suffered in connection with the use of this information. Information is only current at the date initially published.
More ways we can help you
Our CPD courses for Avant members
Tick off some CPD hours and learn more with our in-depth eLearning courses, free for Avant members. Our courses include education activities, reviewing performance and measuring outcomes.