fMRI Analysis · Sensor Fusion · Image Segmentation
Functional MRI (fMRI), BOLD signal decomposition, sensor fusion, image segmentation, tumor detection.
This approach performs whole-brain estimation of the haemodynamic response function in fMRI using convolutional sparse coding between the hemodynamic system and temporal atoms coding for neural activity. The method addresses resting-state fMRI analysis without requiring experimental paradigms, demonstrates statistical significance in discriminating stroke patients from healthy controls, and shows predictive accuracy of 74% for individual age classification across 459 subjects.
The proposed unsupervised Bayesian method for segmenting multifractal textures jointly estimates multifractal parameters and pixel-level labels using a multiscale Potts Markov random field to model inherent spatial and scale correlations, with Gibbs sampling for posterior distribution analysis.
A guided filtering-based fusion method produces weighted averages accounting for ultrasound speckle noise while enhancing image contrast for surgical visualization, aiding intraoperative guidance.