Back to feed
arXiv cs.LG
arXiv cs.LG
7/1/2026
Accelerometry-Derived Digital Biomarkers for Cardiometabolic Risk: A Population-Representative Tabular Benchmark with Uncertainty Quantification

Accelerometry-Derived Digital Biomarkers for Cardiometabolic Risk: A Population-Representative Tabular Benchmark with Uncertainty Quantification

Short summary

New benchmark evaluates ML methods for predicting cardiometabolic risk from accelerometry and lifestyle data across 1,381 adults. TabPFN v2 performs best for HbA1c and CRP, but triglycerides remain largely unpredictable due to genetic dominance. Fairness analysis reveals demographic coverage gaps in marginalized groups despite marginal confidence targets; code and data are open-source.

  • •TabPFN v2 outperforms ridge regression and XGBoost for HbA1c and CRP prediction from activity phenotypes
  • •Triglycerides remain largely unpredictable (R² < 0.05), consistent with genetic dominance in the trait
  • •Demographic fairness analysis reveals localized undercoverage in marginalized groups despite marginal confidence targets

Generated with AI, which can make mistakes.

#research-breakthrough#ai-tools#open-source
Read full article at arXiv cs.LG

Is this a good recommendation for you?

Comments

Failed to load comments. Please try again.

Explore more

A Transportable Threshold-Based Framework for Interpretable Classification of Medical Data
arXiv cs.LG
Inertia-1: An Open Exploration of Wearable Motion Foundation Models
arXiv cs.LG
Who Gets Missed in the Tail? Thresholded Subgroup Underdiagnosis in Long-Tailed Chest X-ray Classification
arXiv cs.LG
Ask Before You Diagnose: Safe-Psych, a Sequential Evaluation Benchmark for LLMs in Psychiatry
arXiv cs.CL