Machine Learning Outperforms Conventional Noninvasive Tests for MASH Fibrosis Detection: CGH | July 2026

Introduction:

Identifying patients with MASH and significant fibrosis (F2–F3) is critical because this group is eligible for emerging FDA-approved therapies. This study compared 28 existing noninvasive tests (NITs) with newly developed machine-learning (ML) models for detecting treatment-eligible fibrosis and cirrhosis.

Why was this study needed?

• Current noninvasive fibrosis tests have limited accuracy for identifying F2–F3 MASH, the key treatment population.
• Liver biopsy remains the reference standard but is invasive and impractical for widespread screening.
• More accurate tools are needed for patient selection and clinical trial enrollment.
• Machine-learning algorithms may integrate multiple routine clinical variables to improve diagnostic performance.
• Better identification of cirrhosis is also essential for surveillance and treatment planning.

Results:

• Existing noninvasive tests showed only modest performance for detecting F2–F3 fibrosis, with FAST being the best conventional test but demonstrating limited positive predictive value.
• Novel machine-learning models significantly outperformed all conventional noninvasive tests, achieving AUCs up to 0.86, higher diagnostic accuracy, and markedly improved positive predictive value for identifying treatment-eligible F2–F3 MASH.
• For cirrhosis (F4), AGILE4+ remained the best conventional noninvasive test, while the machine-learning models achieved even higher overall diagnostic performance, with AUCs around 0.93.

Clinical Impact:

This study highlights the potential of machine-learning–based fibrosis assessment to transform MASH care by accurately identifying patients eligible for emerging antifibrotic therapies. These models could reduce unnecessary liver biopsies, improve patient selection for treatment, and facilitate enrollment into clinical trials once externally validated.

Bottom Line:

Machine-learning models substantially outperform current noninvasive fibrosis tests for detecting both treatment-eligible MASH (F2–F3) and cirrhosis. If validated in routine clinical practice, these tools could become the next generation of precision diagnostics for MASLD/MASH.