eGPLearning Podblast: Artificial Intelligence in Primary care
Shownotes – Artificial Intelligence in Primary care
This is a themed episode where we talk about Artificial Intelligence (AI) in primary care – the applications, risks and benefits to patients and clinicians.
eGPlearning Podblast is a health tech talk by two Nottingham based GPs covering recent topics, useful clinical apps, and interviews with primary care health tech innovators.
Disclaimer – We are not experts, but we are interested GPs, we are merely discussing our impressions, ideas and concerns and optimism in the hope that you may find this interesting.
[maxbutton id=”1″ ] [maxbutton id=”4″ ] [maxbutton id=”5″ ] [maxbutton id=”2″ ]
@SonaliKinra @NottsLMC @NCGPANottm @jacey_melody @kalindikrishna @rcgpfaculties @2GPs_in_a_pod @mededbot @dme_health @karthikrishna86 @dr_zo
Andy: (1.15) – @ANorrisMP attended the practice
- Thank you for electing me as @RCGP National Council representative
- 200 Followers on eGPlearning Facebook page: https://www.facebook.com/Egplearning/
- #TipThursday – https://egplearning.co.uk/tag/tipthursday/
What is AI? (5.12)
Artificial intelligence (AI) in healthcare: is the use of algorithmsand software to approximate human cognition in the analysis of complex medical data. Specifically, AI is the ability for computer algorithms to approximate conclusions without direct human input.
When thinking about A lot of people think of Artificial General Intelligence, an artificial agent that thinks like a human, but this is thought to be a long way off.
HAL 9000 from 2001 A Space Odysey is a (rather sinister) example of Artificial General Intelligence.
AI is a combination of key technologies…
Key terms, definitions and technologies:
The 4th industrial revolution
Deep learning, algorithms, fast processing, storage capacity and ability to collect large amounts of data are the foundations of the 4th industrial revolution – AI.
Logic and rules based AI…
- “Top down approach”- system designers provide the rules for computer to follow
- Simpler than more modern approaches, but this can be very powerful
- Examples include: Automatic tax return software, Qrisk, FeverPain, FRAX (7.16)
- Already used for prescription safety like OptimizeRX (7.30) and advice, flagging investigation results
- Could be useful for automating simple tasks like processing some blood results
Pattern based AI…
Machine Learning (7.53) – see our episode with Jon Brassey
- Training a machine using data sets with known outcomes
- The algorithm analyses lots of data with known outcomes, makes connections – thus training the model which changes its own approach as it learns
- Eventually the software look at new novel data and reach reliable conclusion
Deep Learning (9.20)
- A key technology behind machine learning
- Uses “neural networks”, inspired by neurons, to make connections between data as it learns.
- Connections strengthen and weaken as it processes data and learns.
- E.g. Facial recognition – Facebook, google Deepmind trouble for breaching data handling regulations (10.15) – https://www.theguardian.com/technology/2017/jul/03/google-deepmind-16m-patient-royal-free-deal-data-protection-act
- IBM Watson tackling cancer – connects presentations with knowledge from research papers.
Natural Language Processing
- Key technology for understanding and communicating with humans naturally
- Human communication is HARD to understand
- Currently humans really have to adapt their behaviour to interact with computers, unnatural things – type on keyboard, use mouse, touch screen
- Once can understand human speech and communication much more data available for training algorithms – progress in AI will accelerate.
- Examples include: Dictation software – already here! – Dragon (13.30), Voice assistants (12.20) https://youtu.be/_ldoV5FQn0k , translation, learning from consultation transcripts
Similar to natural language processing in unlocking image and video data to train algorithms. Potential to initially impact, Radiology, dermatology, ENT(15.07), retinal imaging.
Implications for general practice…
Will it change things quickly or slowly?…
Some potential benefits of AI in General Practice:
- Triage and screening supporting access and effective use of resources
- Support medical practice, diagnosis and treatment. Big opportunity to improve quality, consistency, safety
- Supporting and caring for patients – supporting social care
- Accelerating medical research – data collection, processing, more data, quicker in new ways
- Drug/treatment/device development
Special mention for…
@BabylonHealth – They are sometimes controversial, but are a British company leading the world in Medical AI!
Babylon to be installed as standard on Samsung mobile phones https://www.ft.com/content/e7035e0c-634e-11e8-a39d-4df188287fff
Partnership in China https://www.digitalhealth.net/2018/04/babylon-ai-technology-china-tencent/
Potential problems with AI:
- Black box – does it matter if you don’t know how it works? – correlation is not causality.
- Bias in data sets:
- See Tay, Microsoft’s obnoxious chatbot https://en.wikipedia.org/wiki/Tay_(bot) (16.40)
- Normal laboratory ranges for blood tests based on young caucasian populations is a current example of potential bias due to the data used to calibrate a system.
- Risk management – who is responsible
- Security and privacy, data management – GDPR (https://egplearning.co.uk/ramblings/podblast/egplearning-podblast-episode-8-gdpr/)
Effect on Jobs?
- Mass unemployment or…
- New problems and needs arise as old ones are surmounted
- Potential new jobs in healthcare related to AI
- Facilitating and explaining engagement with AI – Healthcare workers are complex communication experts
- Training the AI
- Maintaining the AI
- Creativity in service design
- Original thought and innovation
- Focus on the right skills – creativity, innovation abilities, using and training technology
So… Are you feeling positive about AI in healthcare?
Thanks for listening and for your certificate of engagement click here
Subscribe, comment, share and keep eGPlearning.