Clinical Scorecard: Utilizing Advanced Machine Learning to Forecast Metabolic Dysfunction–Associated Steatotic Liver Disease in the Han Chinese Population

At a Glance

Key Highlights

• MASLD affects approximately 25% of the global population and requires early diagnosis to prevent progression.
• Traditional diagnostic methods (ultrasound, liver function tests, biopsy) have limitations including invasiveness, subjectivity, and accessibility.
• Machine learning algorithms can analyze complex clinical and anthropometric data to improve MASLD prediction accuracy.

Guideline-Based Recommendations

Diagnosis

• Use Philips IU22 Color ultrasound system for MASLD diagnosis as per study protocol.
• Interpret liver function tests in context with clinical history due to indirect measurement and confounding factors.
• Consider machine learning models incorporating clinical, anthropometric, and biochemical variables for enhanced diagnostic accuracy.

Management

• Early detection of MASLD is critical to implement effective treatment and prevent disease progression.
• Personalized treatment approaches may be guided by risk stratification from machine learning predictions.

Monitoring & Follow-up

• Regular anthropometric and biochemical assessments including BMI, blood pressure, fasting glucose, liver enzymes, lipid profile, and tumor markers.
• Monitor changes in selected biomarkers identified by machine learning for disease progression.

Risks

• Invasive biopsy carries risks such as infection and limited accessibility.
• Ultrasound accuracy may be affected by patient factors like weight and bowel gas, leading to misdiagnosis.
• Sole reliance on liver function tests can be misleading due to influence from medications and other conditions.

Patient & Prescribing Data

Older Han Chinese adults (aged >60) undergoing routine health checks in Shanghai

Machine learning models can assist clinicians in early MASLD detection, enabling timely intervention and personalized management.

Clinical Best Practices

• Collect comprehensive clinical, anthropometric, and biochemical data for accurate MASLD risk assessment.
• Apply variable selection methods to identify key biomarkers for simplified and practical diagnostic models.
• Incorporate machine learning algorithms such as KNN, SVM, logistic regression, and ANN for improved prediction and classification.
• Ensure ethical approval and informed consent when implementing new diagnostic technologies.
• Interpret diagnostic results within the clinical context to avoid misdiagnosis and overdiagnosis.

References

• Global prevalence of MASLD
• Traditional MASLD diagnostic methods
• Limitations of liver function tests and ultrasound
• Machine learning in MASLD prediction
• Types of machine learning algorithms
• Common supervised learning algorithms
• Variable selection method in protein selection
• Greedy feature-selection algorithm