Automated Segmentation of Muscle and Fat from MRI
Author Information
Author(s): Tam Kenneth, Liu Si Wen, Costa Sarah, Szabo Eva, Reitsma Shannon, Gillick Hana, Adachi Jonathan D., Wong Andy Kin
Primary Institution: University of California Davis
Hypothesis
Can a fully-automated algorithm accurately segment intermuscular and intramuscular fat from MRI images?
Conclusion
The fully-automated ITSA algorithm achieved high precision in segmenting muscle and fat from MRI scans, demonstrating its reliability for clinical use.
Supporting Evidence
- The algorithm demonstrated low short- and long-term precision error.
- Validation against semi-automated methods showed high correlation (R2 from 0.92 to 0.98).
- Quality control indicated that 7% of cases required manual correction.
Takeaway
This study created a computer program that can quickly and accurately measure fat in muscles using MRI scans, which helps doctors understand muscle health better.
Methodology
The study used a fully-automated iterative threshold-seeking algorithm (ITSA) to segment fat from MRI images across three cohorts.
Potential Biases
Potential bias exists in cases where muscle and fat boundaries are indistinct, leading to misclassification.
Limitations
The algorithm may struggle with cases of extreme fat distribution or muscle dystrophy, requiring manual corrections.
Participant Demographics
Participants were adults aged 45-85, predominantly female (89%), with an average age of 71.6 years and a BMI of 28.06 kg/m².
Statistical Information
P-Value
p<0.05
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website