NexOP: Deep Learning Optimizes Low-Field MRI Sampling and Reconstruction

Researchers have introduced NexOP, a deep-learning framework that jointly optimizes k-space sampling and image reconstruction for low-field MRI. Low-field MRI offers portable, low-cost systems but suffers from low Signal-to-Noise Ratio (SNR), which limits diagnostic quality. A common SNR-boosting method is repetitive signal acquisitions (NEX), but this prolongs scan times. While prior work optimized k-space sampling, the NEX dimension was neglected—typically using a single sampling mask across repetitions. NexOP optimizes sampling density probabilities across the extended k-space-NEX domain under a fixed sampling-budget constraint. The framework is tailored for low-SNR settings and aims to improve image quality without increasing scan duration. The work is detailed in a preprint on arXiv (ID: 2605.11583).