The objective of the proposed research is to investigate the factors affecting the detectability and characterization of brain activation, and thus develop guidelines for designing activation studies and interpreting their results. The quantitative results of the proposed experiments will be plots of receiver operator characteristic (ROC) curves and related detectability parameters as well as the bias and variance measurements of activation site size under various experimental conditions. A software package will be developed for general distribution that will allow researchers to determine the operating characteristics obtainable under their unique experimental conditions. This will facilitate the design and interpretation of the results of their studies. The proposed studies will define the limits and probe the potential of what can realistically be measured in current brain-activation studies. The studies will indicate the experimental conditions for successful detection an the relative sensitivity of the results to the experimental parameters. The quantitative results produced will provide the data required for selecting decision thresholds, and for evaluating error probabilities in the detection tasks. Finally the proposed studies are expected to reveal weaknesses in the methodologies that may suggest fruitful research toward refining this potentially powerful approach to understanding the workings, and disease processes, of the human brain. The primary tool in the investigations will be computer simulation of brain-activation studies, which will enable various elements of the methodology to be studied in a controlled manner.
| Lukic, Ana S; Wernick, Miles N; Yang, Yongyi et al. (2007) Effect of spatial alignment transformations in PCA and ICA of functional neuroimages. IEEE Trans Med Imaging 26:1058-68 |
| Lukic, Ana S; Wernick, Miles N; Tzikas, Dimitris G et al. (2007) Bayesian kernel methods for analysis of functional neuroimages. IEEE Trans Med Imaging 26:1613-24 |
