DTSR fMRI reconstruction

DTSR fMRI reconstruction reconstructs undersampled k-space fMRI data using double temporal sparsity to improve image fidelity and preserve Resting State Networks for accelerated fMRI acquisition.

Key Features:

• Double Temporal Sparsity: Imposes l1-l1 norm sparsity on voxel time series in a transformed domain and on their successive temporal differences to mitigate artifacts from accelerated acquisition.
• Undersampled k-space Measurements: Operates on undersampled k-space inputs to enable accelerated fMRI data collection.
• Performance Metrics: Demonstrates Peak Signal-to-Noise Ratio (PSNR) improvements of approximately 10–12 dB at acceleration factors up to 3.5.
• RSN Preservation and Reproducibility: Preserves and enables accurate detection and reproducibility of Resting State Networks (RSNs).

Scientific Applications:

• Resting State Networks (RSNs) Analysis: Supports preservation and detection of RSNs for studies of brain function and connectivity.
• Accelerated fMRI in Clinical and Experimental Neuroscience: Enables faster data acquisition while maintaining network integrity for clinical and experimental studies.
• Quantitative and Qualitative Evaluation: Provides quantitative PSNR metrics and qualitative reproducibility assessments for validating reconstruction quality.

Methodology:

The method reconstructs fMRI data from undersampled k-space by imposing l1-l1 norm double temporal sparsity constraints on voxel time series in a transformed domain and on their successive differences.