AutoTeaCH: Automated Teching Case Harvester

A typical radiology department is a radiological pathology banquet, with more food than the radiologist in training (ie all radiologists) can consume. The problem is that the cases are hidden - time constraints and diversions prevent the systematic logging of teaching cases and one remains unaware of great cases seen by colleagues.

The proportion of studies that are abnormal has fallen in advanced imaging modalities over recent decades, reflecting a more investigation-led style of medicine, and the wider availability of imaging. The trainee is faced with a double whammy of more studies and more refined diagnoses than ever before, more diluted with normals.

A decade ago a trainee could wait to catch pathology as it passed by. Today a more active search is needed. AutoTeaCH seeks to alleviate this problem. It takes as its input spreadsheets of reports that can be exported from radiology information (RIS) systems, and applies some natural language processing (NLP) to categorise and sort by diagnosis.

This yields lists of cases expressing the belief of the study reporter at the time, clearly not a gold standard diagnosis, and finds many more cases than are available by more traditional means. This is a good thing - it develops critical thinking in the ‘dirty’ real world and exposes the learner to the range of appearances for the diagnoses both typical and atypical - thereby avoiding the selection bias present in curated case archives.

We anticipate that the system will be of use at trainee and consultant level. A consultant who wants to broaden his range of reporting can see how relevant studies are reported by colleagues in his/her institution. An institution can gain insight into the pathology mix coming through the department. To facilitate these options we provide a user interface for the viewing of studies separately, but also the ability to download the extracted cases in bulk on a csv spreadsheet.

Recap: a leaflet elf

Recap was an idea generated in a workshop with clinicians, patients and carers, hosted by Nottinghamshire Healthcare. Clinicians were tired of having to give out paper leaflets, getting them printed and keeping them up to date. Patients and carers were finding them anachronistic in the age of the web. Why not give patients their own web-based account and into it put personalised videos, podcasts, information websites and, yes, leaflets. All that material could be managed online, and a lot of it could be given automatically when the patient reaches the right stage in the pathway. Unlike paper leaflets, patients could easily give feedback and ask for more, while clinicians get to see what is helpful and what is not.

Like all good microprojects we built the initial version of Recap in just a few weeks. It has gained many, many features since then. In fact the code repository today shows over 2,300 commits. Even with all that change Recap has run continuously for over ten years. In that time it has dispensed over 1.5 million information prescriptions. Over 44,000 Trust patients use it, in areas as varied as children and young people’s mental health, cardiac rehab and pre-admission services.

It is not just useful for replacing paper leaflets. In cardiac rehab a study showed patients using Recap were 42% less likely to be readmitted after six months, leading to savings of over £1.5 million per year in one Trust alone. Little elves can make a big difference.

Coming soon

Radiology Style Guide: an advisory elf

How good is your radiology report? The radiology style guide uses LLM technology to check for structural and wording errors which turn up again and again.

Digital Elf micro-project template: elves for elves

Great fleas have little fleas upon their backs to bite ‘em, And little fleas have lesser fleas, and so ad infinitum. And the great fleas themselves, in turn, have greater fleas to go on; While these again have greater still, and greater still, and so on.

We start a Digital Elf micro-project about once a week. They typically have a Jupyter notebook, some python and ruby code and tests, a page of documentation and a git repository in which to store it all. And of course, we have a digital elf to do all that for us.

Radiology Department Dashboard

Convert RIS statistical output files into a website of graphs and charts visible within the Trust.