Related Experiment Videos
Rapid generation of biexponential and diffusional kurtosis maps using multi-layer perceptrons: a preliminary
1Fondazione Istituto Nazionale Neurologico Carlo Besta, via Celoria 11, Milano, Italy. lminati@istituto-besta.it
Magma (New York, N.Y.)
|July 30, 2008
Summary
Multi-layer perceptrons (MLPs) show promise for analyzing diffusion-weighted images. These artificial neural networks can determine biexponential and diffusional kurtosis model parameters, offering a potentially useful curve fitting method.
Area of Science:
- Medical Imaging
- Computational Neuroscience
- Biophysics
Background:
- Diffusion-weighted imaging (DWI) is crucial for assessing tissue microstructure.
- Accurate modeling of diffusion processes is essential for quantitative analysis.
- Traditional methods for parameter estimation can be computationally intensive.
Purpose of the Study:
- To evaluate the efficacy of multi-layer perceptrons (MLPs) for estimating model parameters directly from DWI data.
- To compare MLP-derived parameters with those obtained from conventional least-squares fitting.
Main Methods:
- Model parameters were estimated using both least-squares fitting and MLPs.
- Statistical comparisons included linear regressions, t tests, and Levene's tests.
- Residual analysis was performed to assess model fit.
Main Results:
- MLPs demonstrated strong linear correlation for all estimated parameters.
- MLP estimates were unbiased for the biexponential model but showed bias for the kurtosis model.
- MLP-derived parameters generally exhibited smaller variance and residuals compared to least-squares fitting.
- Visual comparison of parameter maps generated by both methods showed high similarity.
Conclusions:
- Multi-layer perceptrons show potential as an effective curve fitting technique for biexponential and diffusional kurtosis models in DWI.
- MLPs offer a viable alternative for quantitative analysis of diffusion MRI data.
