Shaoshi Zhang

Spurious model comparisons are widespread in biomedical artificial intelligence

97% of 184 high-impact biomedical AI studies compare cross-validated models with tests assuming fold independence; in 69% of re-analyzable studies a significant comparison does not survive correction.

Longer scans boost prediction and cut costs in brain-wide association studies

The paper shows longer scans improve prediction accuracy in brain-wide association studies while reducing costs, with 30-minute scans being the most cost-effective.

Convergent and Divergent Brain-Cognition Development

The paper examines how stable and changing brain network features predict cognitive development in adolescents using rs-fMRI and cognitive data from the ABCD Study.

Optimizing Biophysical Large-Scale Brain Circuit Models With Deep Neural Networks

This paper introduces DELSSOME, a deep learning framework that accelerates biophysical brain circuit model optimization, enabling faster parameter estimation and lifespan trajectory analysis of cortical E/I ratios.

DeepResBat: deep residual batch harmonization accounting for covariate distribution differences

DeepResBat improves MRI data harmonization by addressing covariate distribution differences, outperforming existing methods while avoiding false associations.

Ventral attention network connectivity is linked to cortical maturation and cognitive ability in childhood

The paper links ventral attention network connectivity to cortical maturation and cognitive ability in children using independent datasets.

Multilayer meta-matching: translating phenotypic prediction models from multiple datasets to small data

The paper introduces multilayer meta-matching to improve phenotypic prediction in small datasets by translating models from multiple large datasets, outperforming classical methods and transfer learning.

In-vivo whole-cortex estimation of excitation-inhibition ratio indexes cortical maturation and cognitive ability in youth

The paper estimates the whole-cortex excitation-inhibition ratio using fMRI and shows it declines during youth, with lower ratios linked to better cognitive performance in children.

Relationship Between Prediction Accuracy and Feature Importance Reliability: an Empirical and Theoretical Study

This study investigates the relationship between prediction accuracy and feature importance reliability in neuroimaging data, showing that feature importance reliability can be high with sufficient sample size and is positively correlated with prediction accuracy.

Comparison of individualized behavioral predictions across anatomical, diffusion and functional connectivity MRI

This study compares the ability of anatomical, diffusion, and functional MRI to predict individual behavioral measures, finding that functional connectivity provides the best predictions, especially for cognition.

The Cost of Untracked Diversity in Brain-Imaging Prediction

This study quantifies how population diversity affects brain-imaging prediction accuracy and pattern stability using two clinical cohorts.

Anatomical and Functional Gradients Shape Dynamic Functional Connectivity in the Human Brain

This study develops a spatially-heterogeneous circuit model to simulate brain activity dynamics, showing that anatomical and functional gradients are essential for generating realistic static and dynamic functional connectivity patterns.