LSMA

LONDON SCHOOL OF MULTIOMICS & AI

Free course

Free course

0 of 11 lessons

Optional

Further reading

Further reading

Further reading

Optional reading for the free course Introduction to AI in Healthcare. It follows the course's three modules: foundations, judging a model, and AI in practice. No coding is needed for any of it. Read at your own pace. Nothing here is submitted or marked.

⏱ 7 min readFree
Lesson at a glance5 steps · about 7 min
  1. Module by module
  2. Papers and guides
  3. Reporting standards
  4. Glossary
  5. Going further
Download the PDF PDF · 8 pages · 772 KB · updated November 2026
  • Start with the module cards. Each module has two or three key readings. If you read only those, you have the core of the subject.
  • Most of it is free to read. Papers with a free full text link to PubMed Central or the publisher; the few that may need a subscription are marked.
  • The questions have no answer key. Use them to check your understanding after a module, or to discuss with other learners.

Module by module

The key readings for each module, and questions to think about once you have finished it. Full references are under Papers and guides.

Module 1 · FoundationsFoundations

Medical AI learns patterns from health data. What it can do depends on the data, the labels and how carefully it was tested.

Read
Think about
  • A model learned from one hospital's records. What could change when it is used in another?
  • Which labels in your own work would be hardest to get right for training a model?
Module 2 · IntermediateJudging a model

Accuracy alone says little. Sensitivity, specificity and predictive values depend on prevalence, and good evidence moves from retrospective tests to trials.

Read
Think about
  • A test has 95% sensitivity and 95% specificity. Why might most of its positive results be wrong?
  • What is the difference between external validation and a randomised trial?
Module 3 · AdvancedAI in practice

Tools succeed or fail in real workflows. Bias, drift and over-trust are the usual problems; language models add fluent errors and privacy risks.

Read
Think about
  • Who in your organisation would need to agree before an AI tool is used with patients?
  • Which of your daily tasks could an approved assistant help with safely, and how would you check its output?

Papers and guides

The papers, books and guidance behind the course, grouped by module.

Module 1 · Foundations

If you read one of these, start with Rajkomar and colleagues, which covers the ground of Module 1 in more depth; Topol is wider in scope and is best read after the course, as a survey of where the field has gone.

  1. Rajkomar A, Dean J, Kohane I (2019). Machine learning in medicine. New England Journal of Medicine 380:1347–1358. A review of machine learning in medicine. May need a subscription
  2. Topol EJ (2019). High-performance medicine: the convergence of human and artificial intelligence. Nature Medicine 25:44–56. A broad overview of medical AI. May need a subscription

Module 2 · Judging a model

Read Trevethan alongside Measuring performance, and Liu and colleagues (2019) alongside Evidence: from validation to trials; the three reporting standards are reference documents, best consulted when you are reading or planning a particular study rather than read from start to finish.

  1. Trevethan R (2017). Sensitivity, specificity, and predictive values: foundations, pliabilities, and pitfalls in research and practice. Frontiers in Public Health 5:307. The measures used in Measuring performance, explained.
  2. Liu X et al. (2019). A comparison of deep learning performance against health-care professionals in detecting diseases from medical imaging: a systematic review and meta-analysis. The Lancet Digital Health 1:e271–e297. A systematic review of AI imaging studies.
  3. Collins GS et al. (2024). TRIPOD+AI statement: updated guidance for reporting clinical prediction models that use regression or machine learning methods. BMJ 385:e078378. Reporting standard for prediction models.
  4. Liu X et al. (2020). Reporting guidelines for clinical trial reports for interventions involving artificial intelligence: the CONSORT-AI extension. Nature Medicine 26:1364–1374. Reporting standard for trials of AI interventions.
  5. Vasey B et al. (2022). Reporting guideline for the early-stage clinical evaluation of decision support systems driven by artificial intelligence: DECIDE-AI. Nature Medicine 28:924–933. Reporting standard for early clinical evaluation of AI tools. May need a subscription

Module 3 · AI in practice

Wong and colleagues and Obermeyer and colleagues are short and readable, and each shows one of the module's main warnings with real data; the regulatory and journal guidance changes from time to time, so check the current version when you need it for a decision.

  1. Gulshan V et al. (2016). Development and validation of a deep learning algorithm for detection of diabetic retinopathy in retinal fundus photographs. JAMA 316:2402–2410. An early, influential AI screening study (diabetic retinopathy). May need a subscription
  2. Wong A et al. (2021). External validation of a widely implemented proprietary sepsis prediction model in hospitalized patients. JAMA Internal Medicine 181:1065–1070. External validation of a sepsis model, as in the case study.
  3. Obermeyer Z et al. (2019). Dissecting racial bias in an algorithm used to manage the health of populations. Science 366:447–453. Bias from the choice of label. May need a subscription
  4. Thirunavukarasu AJ et al. (2023). Large language models in medicine. Nature Medicine 29:1930–1940. Language models in medicine (Large language models in medicine and Using AI assistants safely at work). May need a subscription
  5. Medicines and Healthcare products Regulatory Agency. Software and artificial intelligence (AI) as a medical device. GOV.UK guidance. How the UK regulates software and AI as a medical device (Safety, bias and regulation).
  6. National Institute for Health and Care Excellence. Evidence standards framework for digital health technologies. NICE guidance. The evidence NICE expects for a digital tool, by level of risk (Evidence: from validation to trials).
  7. International Committee of Medical Journal Editors. Recommendations for the conduct, reporting, editing, and publication of scholarly work in medical journals. see the section on artificial intelligence. Journal guidance on disclosing AI-assisted work.

Reporting standards

When you read or write about a medical AI tool, these standards list what a good report should contain.

StandardUse it forModuleWhere
TRIPOD+AIDeveloping or validating a prediction model2Collins et al. 2024
DECIDE-AIEarly clinical evaluation of an AI decision-support tool2Vasey et al. 2022
CONSORT-AIRandomised trials of AI interventions2Liu et al. 2020

Glossary

Algorithm
A set of steps a computer follows; in AI, often a model learned from data.
AUC
How well a model ranks positives above negatives; 0.5 is chance.
Bias
Systematic error that affects some groups more than others.
Calibration
Whether predicted risks match what really happens.
Drift
A fall in performance as patients, practice or equipment change.
External validation
Testing a model on data from a different place or time.
Hallucination
Plausible output from a language model that is not supported by facts.
Label
The answer a model learns to predict, such as a diagnosis.
Large language model
A model trained on text to predict the next word or part of a word.
Positive predictive value
Of those the test calls positive, the share who truly are.
Prevalence
How common a condition is in the people tested.
Sensitivity
Of those with the condition, the share the test finds.
Specificity
Of those without the condition, the share the test correctly clears.
Special category data
Under UK data protection law, sensitive data such as health information.

Going further

  • Build a model yourself. Learning a programming language used for data analysis, such as Python or R, lets you train and evaluate a simple classifier and see the measures from Module 2 for yourself.
  • Understand the biology. Precision medicine rests on molecular data such as DNA sequencing and gene expression, which is worth learning in its own right from a good introductory text.
  • Keep up to date. Regulation and evidence in this area change quickly. The regulator guidance and reporting standards above are the first places to check.
Back to the overviewPDF

Up next

Finish

Course complete

Ask BioTutor

Create a free website with Framer, the website builder loved by startups, designers and agencies.