quantum-brain-voxel-control

name	quantum-brain-voxel-control
description	Quantum-inspired neural network for vision-brain understanding using voxel controlling, phase shifting, and measurement-like projection in Hilbert space. Maps brain region connectivity via quantum-inspired modules for fMRI analysis. Use when: (1) analyzing fMRI voxel connectivity, (2) building vision-brain decoding models, (3) reconstructing images from brain signals, (4) designing quantum-inspired architectures for neuroimaging. Activation: quantum brain, vision-brain understanding, voxel controlling, phase shifting, measurement projection, fmri decoding, brain connectivity.
metadata	{"arxiv_id":"2411.13378","published":"2024-11-20","authors":"Hoang-Quan Nguyen, Xuan-Bac Nguyen, Hugh Churchill","tags":["quantum-inspired","fMRI","vision-brain","voxel-connectivity","neural-decoding"]}

Quantum-Brain: Quantum-Inspired Voxel Control for Vision-Brain

Core Concept

Uses quantum-inspired neural modules to model connectivity between brain regions (fMRI voxels) in Hilbert space, enabling effective vision-brain understanding tasks: image retrieval, brain signal retrieval, and fMRI-to-image reconstruction.

Architecture Modules

1. Quantum-Inspired Voxel-Controlling (QIVC)

• Models influence of one brain voxel on others
• Operates in Hilbert space representation
• Captures non-local voxel dependencies (entanglement-like)
• Replaces traditional attention for brain connectivity

2. Phase-Shifting Module (PSM)

• Calibrates brain signal values
• Inspired by quantum phase operations
• Adjusts signal amplitude and phase relationships
• Stabilizes learning across subjects

3. Measurement-like Projection (MLP)

• Projects connectivity information from Hilbert space to feature space
• Mimics quantum measurement collapse
• Extracts task-relevant features from quantum-inspired representation

Performance

• Natural Scene Dataset benchmarks:
  • Image retrieval: 95.1% Top-1 accuracy
  • Brain retrieval: 95.6% Top-1 accuracy
  • fMRI-to-image reconstruction: 95.3% Inception score

Methodology

Step 1: Encode fMRI to Hilbert Space

• Map voxel activations to quantum state representation
• Each voxel → amplitude in Hilbert space vector

Step 2: Apply Voxel-Controlling

• Compute voxel influence matrix
• Apply quantum-inspired transformation
• Capture inter-regional connectivity

Step 3: Phase Calibration

• Apply phase-shifting to stabilize representations
• Normalize across subjects and sessions

Step 4: Measurement Projection

• Project to task-specific feature space
• Use for downstream tasks (classification, reconstruction)

Implementation

• Can be implemented with standard deep learning frameworks
• Hilbert space = high-dimensional complex vector space
• Voxel-Controlling = parameterized unitary-like transformation
• Phase-Shifting = element-wise complex phase rotation
• Measurement = linear projection + nonlinearity

Pitfalls

• fMRI resolution: Spatial resolution limits voxel-level analysis
• Subject variability: Requires per-subject calibration or alignment
• Hilbert space dimensionality: Balance between expressivity and computational cost
• Training stability: Quantum-inspired modules can be sensitive to initialization

Related Work

• QEEGNet: Similar hybrid approach for EEG encoding (arXiv: 2407.19214)
• Quantum State Fidelity for functional networks (arXiv: 2508.16895)
• TRIBE v2: Multi-modal brain foundation model