Today our early career reporters cover sessions on machine learning in diagnostics, large vessel vasculitis and more.
Today our early career reporters cover sessions on machine learning in diagnostics, large vessel vasculitis and more.
Dr Bonnia Liu
Day 3 suggests that we can be better at diagnosing large vessel vasculitis (LVV) and perhaps classification goes beyond just giant cell arteritis (GCA) and Takayasu’s arteritis.
Identifying LVV is still tricky. Delayed time to diagnosis may be attributed to the variable clinical manifestations of LVV – and at 20 weeks, patients presenting with isolated extracranial LVV have the longest symptom duration prior to diagnosis (Poster #1409).
But we are improving! The ARTESER study explores observational retrospective data collected from multiple centres in Spain. The identification of extracranial involvement has increased in recent years, coinciding with the increase of the use of imaging tests in diagnosis (Poster #1403).
However, we still appear to be struggling with treatment, with 19.9% patients having relapses, despite high mean cumulative prednisone doses of 8514mg during a mean treatment duration of 22.65 months (Poster #1408).
Can we do even better with some extra help? It might be so! Our very own Dr Chris McMaster utilised machine learning techniques to differentiate GCA from mimics based on rapid assessment of a range of variables including symptoms and biochemical markers (Poster #1417).
Furthermore, machine learning algorithms may be able to accurately identify histopathological features on temporal artery biopsies in order to identify GCA (Poster #1415).
It is indeed an exciting time to be in rheumatology! With the collaboration of man (or woman) and machine, we can be better at identifying subtypes of LVV and work towards improved treatment outcomes.
- 1409: Clinical Features at Disease Onset of Different Subsets of Large-vessel-giant Cell Arteritis in a Monocentric Cohort of 100 Patients
- 1403: Giant Cell Arteritis Subtypes: Data from the ARTESER Registry
- 1408: Treatment of Giant Cell Arteritis in the ARTESER Multicenter Study of 1675 Patients
- 1417: Machine Learning Enhances the Identification of GCA from Its Mimics Based on Clinical Factors
- 1415: Temporal Artery Biopsy Reports Can Be Accurately Classified by Artificial Intelligence
Dr Bonnia Liu is a final year rheumatology trainee pursuing dual training in rheumatology and nuclear medicine at Austin Hospital in Melbourne.
Dr Chris McMaster
I want to highlight just one abstract from day 3 that may have been missed but has very important implications.
Yiwei Yuan, from Dartmouth College in New Hampshire, presented her work, bringing the power of complex gene expression analysis to your neighbourhood histopathology lab (#1374).
In work likely inspired by a recent spate of oncology papers in this area, researchers looked at standard skin immunohistochemistry (IHC) slides form scleroderma patients who had paired tissue gene expression testing to determine molecular subtype.
The original gene expression tests split them into four subtypes: inflammatory, fibroproliferative, limited and normal-like. By training a machine learning model on the IHC slides, they were able to accurately predict the subtype.
This area of research has already caused a stir in oncology, and it’s sure to do the same for us. This brings the power of scleroderma sub-classification into the hands of community pathology laboratories – a simple skin biopsy is all that will be needed to determine the patient subtype.
Let’s hope this opens the door for greater precision in selecting therapy for our scleroderma patients.
- 1374: A Deep Neural Network Classifier to Identify Inflammatory Systemic Sclerosis Patients from Histological Images
Dr Chris McMaster is a Melbourne-based final-year rheumatology and clinical pharmacology trainee.
