Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

TV-based Deep 3D Self Super-Resolution for fMRI

Fernando Pérez-Bueno, Hongwei Bran Li, Matthew S. Rosen, Shahin Nasr, Cesar Caballero-Gaudes, Juan Eugenio Iglesias

TL;DR

The paper tackles the challenge of limited fMRI spatial resolution by introducing a self-supervised 3D DL SR method that integrates a CNN with an analytical degradation model and a Total Variation prior, enabling high-resolution reconstruction without HR ground-truth data. The approach optimizes a MAP-like loss $\mathcal L = \lVert \mathbf x - \mathbf B f^{\boldsymbol\theta}(\mathbf x) \rVert_2^2 + \alpha \mathrm{TV}(f^{\boldsymbol\theta}(\mathbf x))$, leveraging the degradation operator $\mathbf B = \mathbf D\mathbf H$ and training on LR observations only. Empirical results on 7T rs-fMRI data show that the method outperforms interpolation baselines and is competitive with supervised DL SR, while preserving functional maps (seed-based connectivity) with high accuracy and low FDR. This yields a practical pathway to enhance fMRI spatial detail without demanding GT HR data, potentially improving mesoscale brain analyses and clinical/experimental workflows.

Abstract

While functional Magnetic Resonance Imaging (fMRI) offers valuable insights into cognitive processes, its inherent spatial limitations pose challenges for detailed analysis of the fine-grained functional architecture of the brain. More specifically, MRI scanner and sequence specifications impose a trade-off between temporal resolution, spatial resolution, signal-to-noise ratio, and scan time. Deep Learning (DL) Super-Resolution (SR) methods have emerged as a promising solution to enhance fMRI resolution, generating high-resolution (HR) images from low-resolution (LR) images typically acquired with lower scanning times. However, most existing SR approaches depend on supervised DL techniques, which require training ground truth (GT) HR data, which is often difficult to acquire and simultaneously sets a bound for how far SR can go. In this paper, we introduce a novel self-supervised DL SR model that combines a DL network with an analytical approach and Total Variation (TV) regularization. Our method eliminates the need for external GT images, achieving competitive performance compared to supervised DL techniques and preserving the functional maps.

TV-based Deep 3D Self Super-Resolution for fMRI

TL;DR

The paper tackles the challenge of limited fMRI spatial resolution by introducing a self-supervised 3D DL SR method that integrates a CNN with an analytical degradation model and a Total Variation prior, enabling high-resolution reconstruction without HR ground-truth data. The approach optimizes a MAP-like loss , leveraging the degradation operator and training on LR observations only. Empirical results on 7T rs-fMRI data show that the method outperforms interpolation baselines and is competitive with supervised DL SR, while preserving functional maps (seed-based connectivity) with high accuracy and low FDR. This yields a practical pathway to enhance fMRI spatial detail without demanding GT HR data, potentially improving mesoscale brain analyses and clinical/experimental workflows.

Abstract

While functional Magnetic Resonance Imaging (fMRI) offers valuable insights into cognitive processes, its inherent spatial limitations pose challenges for detailed analysis of the fine-grained functional architecture of the brain. More specifically, MRI scanner and sequence specifications impose a trade-off between temporal resolution, spatial resolution, signal-to-noise ratio, and scan time. Deep Learning (DL) Super-Resolution (SR) methods have emerged as a promising solution to enhance fMRI resolution, generating high-resolution (HR) images from low-resolution (LR) images typically acquired with lower scanning times. However, most existing SR approaches depend on supervised DL techniques, which require training ground truth (GT) HR data, which is often difficult to acquire and simultaneously sets a bound for how far SR can go. In this paper, we introduce a novel self-supervised DL SR model that combines a DL network with an analytical approach and Total Variation (TV) regularization. Our method eliminates the need for external GT images, achieving competitive performance compared to supervised DL techniques and preserving the functional maps.
Paper Structure (12 sections, 2 equations, 2 figures, 3 tables)

This paper contains 12 sections, 2 equations, 2 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related work
  3. Method
  4. Deep inference
  5. Network architecture and training details
  6. Experiments and Results
  7. Data: Gorgolewski Resting State dataset
  8. Image quality comparison
  9. Functional analysis
  10. Discussion and Conclusion
  11. Compliance with ethical standards
  12. Acknowledgments

Figures (2)

  • Figure 1: Overview of the proposed self SR model. The observed LR image is fed to the SR Network to produce an HR estimation. During training, the HR output is used to calculate the TV regularization and downsampled to be compared to the observed input.
  • Figure 2: Single-subject seed correlation maps. Top: Thresholded maps ($r \geq 0.5$) showing a clear pattern of the sensorimotor network with clusters in bilateral motor and somatosensory cortices, and supplementary motor areas. Bottom: Unthresholded maps. a) Real observed image $1.5mm$. Super-resolved image at $1.5mm$ from an input of isotropic voxels size b) $1.875mm$$(f=1.25)$ and c) $3mm$$(f=2)$.