MYO-Guide: a machine learning approach to the analysis of MRI
Research type
Research Study
Full title
Implementation of an artificial intelligence module on the online imaging portal MYO-Share for guiding the diagnosis of muscle diseases.
IRAS ID
313309
Contact name
Jordi Diaz Manera
Contact email
Sponsor organisation
Newcastle University
ISRCTN Number
ISRCTN14323809
Duration of Study in the UK
0 years, 8 months, 1 days
Research summary
Fat replacement in muscles is characteristic of genetic muscle diseases. Muscle Magnetic Resonance Imaging (MRI) can identify the magnitude of fat replacement for a clinician to quantify using the Lamminen-Mercuri scale. Patterns of fat replacement help guide the selection of genes tested using deoxyribonucleic acid (DNA) sequencing, currently the gold standard for diagnosing and categorizing muscle disease. Next generation sequencing (NGS) allows us to reach a genetic diagnosis earlier and easier thanks to the simultaneous analysis of hundreds of genes. An earlier diagnosis means patients benefits from genetic counselling, access to well defined and tailored care based in the diagnosis, clear definition of expectations about clinical progression and well-defined follow-up strategy and inclusion in natural history studies or clinical trials. However, most clinicians are not specialised in identifying muscle disease type from MRI image and NGS has several well-documented limitations. Therefore, a tool that can accelerate an accurate diagnosis will have a clear impact in the medical and patient communities. Specifically, a tool that uses machine learning to automatically segment muscle regions, analyse the amount of fat present on an MRI, and suggest a list of potential diagnoses could greatly facilitate the diagnosis process. In proof of concept, we created a machine-learning algorithm, MYO-Guide, which predicted muscle disease diagnosis with an accuracy of 95.7% based on Lamminen-Mercuri scores (Verdu-Diaz, 2020). Our present study aims to create an improved version of MYO-Guide to analyse a larger number of muscle diseases, which will include an automatic segmentation tool to identify and delineate all muscles and automatically quantify the muscle fat replacement using Lamminen-Mercuri scale.
Verdú-Díaz J, Alonso-Pérez J, Nuñez-Peralta C, Tasca G, Vissing J, Straub V, Fernández-Torrón R, Llauger J, Illa I, Díaz-Manera J. Accuracy of a machine learning muscle MRI-based tool for the diagnosis of muscular dystrophies. Neurology. 2020 Mar 10;94(10):e1094-e1102.
REC name
South West - Central Bristol Research Ethics Committee
REC reference
22/SW/0065
Date of REC Opinion
9 May 2022
REC opinion
Favourable Opinion