Biophysics

B.D. Ward

Senior Biostatistician
Department of Biophysics

B.D. Ward was originally trained as an electrical engineer, and worked seven years as an aeronautical engineer, responsible for development and evaluation of missile guidance logic. Later, he pursued graduate training in mathematics and statistics, receiving an M.S. in Mathematics from the University of Wisconsin – Milwaukee.

In 1996, Ward was hired to work in the Biophysics Research Institute of the Medical College of Wisconsin. His responsibilities in the fMRI research program have included: AFNI software development for image processing, signal processing, statistical analysis, and display of functional MRI (fMRI) data; documentation and maintenance of AFNI software; and statistical consulting for the fMRI research community.

In the area of image processing, Ward developed software for automatic segmentation of the intracranial region (program 3dIntracranial), image intensity uniformization (3dUniformize), and automatic segmentation of subcortical structures (3dSegment).

Ward has done extensive work in the area of software development for fMRI signal detection and analysis. This includes the widely-used program 3dDeconvolve for multiple regression and deconvolution analysis of fMRI data. For nonlinear regression analysis, he wrote program 3dNLfim, which has been particularly useful for modeling of the pharmacokinetic fMRI signal. Additionally, he wrote program 3dWavelet, for wavelet analysis of fMRI time series data. He developed each of these fMRI time series analysis programs in two separate versions: a non-interactive batch processing version, and an interactive version for graphical display of the fitted waveforms and statistical results.

Ward developed several programs for group statistical analysis of fMRI data; i.e., combining data from multiple subjects, either as a single group, or for across-group comparisons, on a voxel-by-voxel basis. For classical statistical analysis applications (which assume normality of the data), he wrote programs 3dANOVA, 3dANOVA2, and 3dANOVA3 for one-, two-, and three-factor analysis of variance, including tests for fixed and random effects, interactions, and contrasts in factor level means. Also, he wrote program 3dRegAna for multiple linear regression analysis across fMRI datasets.

For nonparametric statistical analysis of group fMRI data, Ward wrote programs 3dMannWhitney (Wilcoxon rank-sum test of two groups), 3dWilcoxon (Wilcoxon signed-rank test for paired fMRI data), 3dKruskalWallis (Kruskal-Wallis test for comparing multiple treatments), and program 3dFriedman (comparison of blocked multiple treatments).

To help the user determine the appropriate statistical threshold level for fMRI datasets, Ward wrote program AlphaSim, which calculates the tradeoff between the individual voxel probability threshold and minimum cluster size threshold to achieve the desired overall statistical significance level. He also wrote program 3dFDR, which implements the false discovery rate method as an alternative thresholding criterion. In addition, he wrote program 3dStatClust for agglomerative hierarchical clustering of multiple parameters.

Ward's current research activities and interests include: Kalman filtering for analysis of nonstationary fMRI time series, real-time fMRI data analysis for clinical applications (such as real-time mapping of the visual field), mathematical modeling of pharmacokinetic fMRI data, automatic image segmentation, and numerical calculation of the MR signal due to magnetic field perturbations arising from inhomogeneities in magnetic susceptibility at the vascular level.