Sequence Search: Automated Sequence Design using Neural Architecture Search

arXiv cs.AI / 4/17/2026

📰 NewsIdeas & Deep AnalysisModels & Research

共有:

Key Points

The paper introduces “Sequence Search,” a neural architecture search–based framework to automate MR (magnetic resonance) pulse sequence design without relying on prior knowledge of conventional sequence structures.
Instead of requiring an initial hand-crafted sequence or large training datasets, the method uses tissue properties, imaging parameters, and explicit design objectives as inputs to generate sequences that meet those objectives.
It iteratively proposes candidate sequences via NAS and then optimizes them using a differentiable Bloch simulator combined with objective-specific loss functions trained with gradient-based learning.
The framework can reproduce established sequences (e.g., spin-echo, T2-weighted spin-echo, inversion recovery) and also discovers less intuitive alternatives that reduce RF energy or deviate from standard refocusing phases.
Overall, the work presents a generalizable approach for exploring MR sequence configurations beyond human-intuition-driven designs.

Abstract

Developing an MR sequence is challenging and remains largely constrained by human intuition. Recently, AI-driven approaches have been proposed; however, most require an initial sequence for parameter optimization or extensive training datasets, limiting their general applicability. In this study, we propose "Sequence Search," an automated sequence design framework based on neural architecture search. The method takes tissue properties, imaging parameters, and design objectives as inputs and generates pulse sequences satisfying the design objectives, without requiring prior knowledge of conventional sequence structures. Sequence Search iteratively generates candidate sequences through neural architecture search and optimizes them via a differentiable Bloch simulator and objective-specific loss functions using gradient-based learning. The framework successfully replicated conventional spin-echo, T2-weighted spin-echo, and inversion recovery sequences. Less intuitive solutions were also discovered, such as three-RF spin-echo-like sequences with reduced RF energy and refocusing phases deviating from the conventional Hahn-echo. This work establishes a generalizable framework for automated MR sequence design, highlighting the potential to explore configurations beyond conventional designs based on human intuition.