Highly Scalable Image Reconstruction using Deep Neural Networks with Bandpass Filtering

TL;DR

This paper introduces a scalable deep neural network framework for image reconstruction that incorporates bandpass filtering to leverage imaging physics, demonstrated on MRI data to improve speed and accuracy.

Contribution

The novel framework combines bandpass filtering with deep learning to enhance scalability and physics-based consistency in image reconstruction tasks.

Findings

Enables high subsampling rates in MRI reconstruction

Speeds up MRI acquisitions significantly

Maintains high reconstruction accuracy

Abstract

To increase the flexibility and scalability of deep neural networks for image reconstruction, a framework is proposed based on bandpass filtering. For many applications, sensing measurements are performed indirectly. For example, in magnetic resonance imaging, data are sampled in the frequency domain. The introduction of bandpass filtering enables leveraging known imaging physics while ensuring that the final reconstruction is consistent with actual measurements to maintain reconstruction accuracy. We demonstrate this flexible architecture for reconstructing subsampled datasets of MRI scans. The resulting high subsampling rates increase the speed of MRI acquisitions and enable the visualization rapid hemodynamics.