Breaking the bottleneck: it’s time to scale up the analysis of muscle MRI
Decoding muscle health at AI speed
Traffic jam? AI is helping clear the data-processing bottlenecks behind large-scale MRI studies.
MRI data can now be processed in a fraction of the time once required, making it possible to compare muscle health across large groups of people.
Imagine trying to compare the muscle health of hundreds of people using MRI scans — manually identifying and outlining every individual muscle, image by image.
Until recently, that was the reality. In muscle MRI research, one of the biggest bottlenecks has been segmentation: the process of isolating specific muscles within scans so they can be measured and analysed. MRI can generate enormous amounts of detailed muscle data. But without the ability to process that data efficiently, large-scale comparisons between different populations, activity levels and age groups become extremely difficult.
Our new study explores how AI-assisted analysis could help change that. Using a deep learning method, we investigated whether muscle fat measurements could be generated automatically from Dixon MRI scans of the lumbar spine and pelvis. We measured fat fraction and lean muscle volume across several core muscles, including psoas major, quadratus lumborum and the gluteal muscles.
We then compared scans from highly active cyclists and physically inactive adults, helping build a broader picture of how muscle composition differs between groups with very different activity levels. Automation also enabled the creation of combined “core scores”, bringing together measurements across multiple muscles to provide an overall indicator of core muscle health.
Quantitative MRI allows us to move beyond simply looking at scans and begin extracting measurable data that can be compared across individuals and tracked over time.
We see this as part of a wider move towards meaningful imaging biomarkers for muscle health — helping clinicians understand muscles with the same precision that blood pressure and heart rate brought to cardiovascular medicine.
The potential is exciting. By dramatically reducing the time needed for segmentation and analysis, AI could help us analyse larger datasets, compare populations more effectively and accelerate progress towards meaningful biomarkers for muscle health.
