Related Experiment Video
Updated: Apr 26, 2026
13:51
Cross-Modal Multivariate Pattern Analysis
Published on: November 9, 2011
21.0K
An empirical comparison of different approaches for combining multimodal neuroimaging data with support vector
William Pettersson-Yeo1, Stefania Benetti2, Andre F Marquand3
1Department of Psychosis Studies, Institute of Psychiatry, King's College London London, UK.
Frontiers in Neuroscience
|August 1, 2014
Summary
Integrating multiple neuroimaging data types can improve classification accuracy for psychiatric and neurological disorders. However, simple integration methods may be most effective, and benefits depend on the specific diagnostic comparison.
Area of Science:
- Neuroimaging
- Psychiatric Research
- Neurological Illness
Background:
- Clinical utility of neuroimaging for psychiatric and neurological disorders requires improved single-subject classification accuracy.
- Current single-modality approaches show limitations in achieving desired classification accuracies.
- Data integration offers a potential solution to enhance classification performance.
Purpose of the Study:
- To compare four data integration approaches for enhancing classification accuracy in neuroimaging.
- To evaluate the effectiveness of integrating structural, functional, and diffusion tensor magnetic resonance imaging data.
- To assess classification enhancement across ultra-high risk, first episode psychosis, and healthy control groups.
Main Methods:
- Described and compared four integrative approaches: un-weighted sum of kernels, multi-kernel learning, prediction averaging, and majority voting.
- Integrated structural, functional, and diffusion tensor magnetic resonance imaging data.
- Compared classification accuracy against best single-modality performance in three subject groups (n=19, 19, 23).
Main Results:
- Data integration enhanced classification accuracy by up to 13% in some instances, but not universally.
- Simpler integration methods (e.g., un-weighted sum, prediction averaging) often yielded greater accuracy increases than complex methods.
- The potential for classification enhancement was significantly influenced by the specific diagnostic comparison being made.
Conclusions:
- Combining different neuroimaging modalities using data-driven methods is not a guaranteed solution for increasing classification accuracy in moderately sized clinical datasets.
- The limited complementary information between the studied modalities may have restricted enhancement.
- Future research should explore integrating a wider variety of data types (e.g., genetic, cognitive) for optimal classification of early-stage psychosis.
