Project

Discipline

multivariate analysis; Brain connectivity; Random field theory; False discovery rate FDR; Diffusion imaging; Magnetic resonance imaging (MRI); Bonferroni; Multiple testing; Family-wise error rate (FWER); Multiple comparisons

Lead
Lay summary
The recent developments in medical imaging and image analysis allowed the determination of the interregional brain connectivity trough Diffusion Magnetic Resonance Imaging tractography. The brain connectivity is then represented by a connection matrix where each cell represents a certain measure of connectivity between two regions of interests of the brain. Our group is one of the pioneers in the domain of diffusion medical image processing and is being continuously improving the processing of such images to construct more robust connection matrices. An important part of the project is the ability of inferring from the connection matrices and the ability to perform rigorous statistical analysis to derive new significant medical results. In particular, we focus on the brain regions of interest level of inference. Performing inference at the level of regions of interest involves the problem of multiplicity correction or the so-called multiple testing or multiple comparisons problem. The problem of multiple comparisons in brain connectivity analysis is the subject of a current PhD thesis started in 05.09.2009, directed by Prof. Jean Philippe Thiran (EPFL – Signal Processing Laboratory 5) and co-directed by Prof. Stephan Morgenthaler (EPFL – Chair of Applied Statistics). The program of this PhD thesis can be divided into three essential parts. 1.The general problem of multiple comparisons. The objective of this part is to develop new multiple comparisons procedure that has a certain property of optimality. 2.The multiple comparisons for positively dependent data. By supposing that the brain regions of interest which are in the same vicinity or which belong to the same functional network are positively correlated, we develop new multiple comparison strategies that exploit this additional information in the aim of increasing the sensitivity of detecting real effects. We call this strategy the two-step methods. 3. The application of multiple comparisons on brain connectivity matrices. This part is not only an application of the two precedent parts. In fact, we need to adapt it by estimating the distributions of the statistical tests when network measures are used and by estimating the influence of the MRI images noise on these measures. In addition, we need to estimate the structure of positive dependence to define brain sub-networks to use the two-step methods.  This part is crucial for the validation of the developed adaptive strategy. The work done so far has been focused on the two first parts, which are more related to the problem of multiple comparisons. The third part, which is an adaptation of the previous work to the brain connectivity analysis, will be the main subject of this 9 months project, in order to complete the PhD thesis. Considering the results of the previous project, at the end of this 9 months project we will have developed a framework of an adaptive strategy for the statistical analysis of both functional and structural connection matrices ready to be used in single subject analysis or group comparison. The framework will be general enough to be used in many neuroscientific research domains and will include a variety of network measures that permit an efficient investigation of a large range of pathologies where changes in brain connectivity is a key ingredient of medical interpretation, such as Alzheimer or Schizophrenia.

Direct link to Lay Summary

Last update: 21.02.2013

Number	Title	Start	Funding scheme

The recent developments in medical imaging and image analysis allowed the determination of the interregional brain connectivity trough Diffusion Magnetic Resonance Imaging tractography. The brain connectivity is then represented by a connection matrix where each cell represents a certain measure of connectivity between two regions of interests of the brain. Our group is one of the pioneers in the domain of diffusion medical image processing and is being continuously improving the processing of such images to construct more robust connection matrices.An important part of the project is the ability of inferring from the connection matrices and the ability to perform rigorous statistical analysis to derive new significant medical results. In particular, we focus on the brain regions of interest level of inference. Performing inference at the level of regions of interest involves the problem of multiplicity correction or the so-called multiple testing or multiple comparisons problem.The problem of multiple comparisons in brain connectivity analysis is the subject of a current PhD thesis started in 05.09.2009, directed by Prof. Jean Philippe Thiran (EPFL - Signal Processing Laboratory 5) and co-directed by Prof. Stephan Morgenthaler (EPFL - Chair of Applied Statistics). The program of this PhD thesis can be divided into three essential parts.1.The general problem of multiple comparisons. The objective of this part is to develop new multiple comparisons procedure that has a certain property of optimality. 2.The multiple comparisons for positively dependent data.By supposing that the brain regions of interest which are in the same vicinity or which belong to the same functional network are positively correlated, we develop new multiple comparison strategies that exploit this additional information in the aim of increasing the sensitivity of detecting real effects. We call this strategy the two-step methods.3.The application of multiple comparisons on brain connectivity matrices.This part is not only an application of the two precedent parts. In fact, we need to adapt it by estimating the distributions of the statistical tests when network measures are used and by estimating the influence of the MRI images noise on these measures. In addition, we need to estimate the structure of positive dependence to define brain sub-networks to use the two-step methods. This part is crucial for the validation of the developed adaptive strategy.The work done so far has been focused on the two first parts, which are more related to the problem of multiple comparisons. The third part, which is an adaptation of the previous work to the brain connectivity analysis, will be the main subject of this 9 months project, in order to complete the PhD thesis. Considering the results of the previous project, at the end of this 9 months project we will have developed a framework of an adaptive strategy for the statistical analysis of both functional and structural connection matrices ready to be used in single subject analysis or group comparison. The framework will be general enough to be used in many neuroscientific research domains and will include a variety of network measures that permit an efficient investigation of a large range of pathologies where changes in brain connectivity is a key ingredient of medical interpretation, such as Alzheimer or Schizophrenia.