Integrated Model for Hepatic Fat Fraction Assessment in MASLD: A QCT-Referenced Study

172 participants from a single-center Chinese health-examination cohort were categorized by quantitative computed tomography (QCT) to evaluate ultrasound attenuation and metabolic biomarkers for non-invasive assessment of hepatic steatosis. The dataset, authored by Zejun Ma and last updated in 2026, includes USAT values and serum levels of ALT, AST, and ferritin. Three predictive models were evaluated, with the USAT-only model achieving an AUC of 0.847 for detecting imaging-defined steatosis.

Use Cases

• Benchmarking ultrasound attenuation (USAT) models against QCT-derived hepatic fat fraction categories.
• Evaluating the additive value of metabolic biomarkers (ALT, AST, ferritin) to imaging data for liver disease classification.
• Training machine learning models for binary detection of imaging-defined hepatic steatosis based on clinical and imaging features.
• Conducting subgroup analyses to assess model performance across demographics like sex, age, and BMI.

Strengths

• Dataset includes 172 participants with QCT-referenced ground truth labels for hepatic fat fraction categories.
• Model performance was validated using fold-separated internal validation, including nested five-fold stratified cross-validation.
• Specific performance metrics are provided, such as an AUC of 0.847 for the USAT-only model.

Limitations

• Multiclass categorization, particularly for Category 2 (31 participants), showed weak discrimination (AUC=0.569).
• Data is from a single-center Chinese cohort, limiting generalizability to other populations.
• Column-level documentation is absent; field semantics must be inferred after download.

Provenance

Source

figshare, authored by Zejun Ma.

Collection Method

Clinical study enrolling participants and collecting USAT values and serum biomarker levels.

Freshness

Last updated 2026-05-29 04:53:51; freshness should be verified.

Geography

Single-center Chinese health-examination cohort.