Industrial Maths study group for AI and Health: can machine learning be used in rare disease prediction?

The first in a series of problems for KTN’s Industrial Maths in AI and Health event has been announced and researchers are encouraged to apply.

The first problem that will be tackled at the AI in Health study group has been announced.  The three-day study groups will be run in Manchester from 26-28 June.

If you are a researcher working in a UK university who would like to work on this problem, please register here.

 

The organisers, KTN, alongside the Universities of Manchester, are looking for researchers to work on the following conundrum:

Secure Machine Learning for Rare Disease Prediction presented by Mendelian

In the UK there is an estimated 3.5 million rare disease patients Рthat’s more than ​all cancer patients.  On average, these patients wait 5.6 years for a diagnosis and are referred to 7.3 different doctors.  It’s a significant burden on our healthcare system.  The problem is that rare diseases are difficult to diagnose.  They are numerous and varied, making it difficult for doctors to recognise them on demand.  

Mendelian has spent several years creating tools for clinicians, geneticists and healthcare systems who manage rare diseases.  Its clinical search engine has had much success with doctors seeking the latest relevant knowledge for their troublesome diagnoses.  It has ongoing studies showing how its technology can improve rare disease screening and clinical decision making.

In academia, there has been machine learning research on predicting rare disease in electronic health records.  These studies have shown that machine learning methods are able to predict diagnoses that expert clinicians would make.  Unfortunately, it is often infeasible to find large enough datasets to sufficiently train classifiers for diseases with such a low prevalence.  In practice, it would require access to data held by several institutions that is very sensitive in nature.

Mendelian is interested in exploring machine learning methods that are effective and secure on siloed data.  It can provide a patient data for ideation and to validate proofs of concept.  These datasets are either limited, anonymised, public domain patient records or synthetic data produced from the latest rare disease statistics. For well-formed solutions, it would be able to test them on larger, secured patient record databases to feedback summary statistics and performance metrics.

 

In recent years, there has been a huge surge in interest in artificial intelligence (AI) for health and care.  There are huge commercial opportunities in public health, vaccination, medicine discovery and manufacture, mental health services, medical technology for monitoring lifestyle and behaviour.

A key component to developing a successful and booming AI in health and care eco-system is providing tangible examples of the successful application of AI detailing the techniques, successes, and limitations on real data.

This study group will bring researchers from across the UK together with a number of industrial organisations (large, governmental, SMEs) to tackle some of the most interesting challenges in this sector.

This is the first of several problems that will be discussed at the three-day study group in Manchester from 26-28 June.

If you are a researcher working in a UK university who would like to work on this problem, please register here.

For a longer version of this problem with full referencing, please click here.