Deep reinforcement learning-designed radiofrequency waveform in MRI

Abstract

A carefully engineered radiofrequency (RF) pulse plays a key role in a number of systems such as mobile phone, radar, and magnetic resonance imaging (MRI). The design of an RF waveform, however, is often posed as an inverse problem that has no general solution. As a result, various design methods each with a specific purpose have been developed based on the intuition of human experts. In this work, we propose an artificial intelligence-powered RF pulse design framework, DeepRF, which utilizes the self-learning characteristics of deep reinforcement learning (DRL) to generate a novel RF beyond human intuition. Additionally, the method can design various types of RF pulses via customized reward functions. The algorithm of DeepRF consists of two modules: the RF generation module, which utilizes DRL to explore new RF pulses, and the RF refinement module, which optimizes the seed RF pulses from the generation module via gradient ascent. The effectiveness of DeepRF is demonstrated using four exemplary RF pulses, slice-selective excitation pulse, slice-selective inversion pulse, B1-insensitive volume inversion pulse, and B1-insensitive selective inversion pulse, that are commonly used 2 in MRI. The results show that the DeepRF-designed pulses successfully satisfy the design criteria while improving specific absorption rates when compared to those of the conventional RF pulses. Further analyses suggest that the DeepRF-designed pulses utilize new mechanisms of magnetization manipulation that are difficult to be explained by conventional theory, suggesting the potentials of DeepRF in discovering unseen design dimensions beyond human intuition. This work may lay the foundation for an emerging field of AI-driven RF waveform design.

Applications of Artificial Intelligence in MR Image Acquisition and Reconstruction

“…최근에는 불충분하게 얻어진 데이터로부터 영상을 복원하는 과정뿐만 아니라, 영상복원을 위한 생체신호를 생성하고 기록하는 단계, 즉 경사자장(Gx, Gy, Gz)과 전자기파(RF)의 세기를 시간에 따라 어떤 방식으로 가하여 줄 것인지를 결정하는 MRI 펄스시퀀스 디자인과정에서도 인공지능기술이 적용되고 있다( 43 44 45 ). 펄스시퀀스의 디자인은 인체 내부의 어떤 물성을 강조하게 만들 것인가를 결정할 수도 있고, 특정 물성을 강조하고자 할 때에 하드웨어의 사양 및 인체에 대한 안전성 등 주어진 제약조건 내에서 보다 효율적인 영상정보의 부호화 방법을 제안해 줄 수도 있다.…”

Autonomous Single-Molecule Manipulation Based on Reinforcement Learning

“…Machine learning algorithms provide a promising solution to this kind of problem, and they have already demonstrated their capabilities for a variety of tasks: they can solve highly complex computer games at super-human performance, − even when the rules of the game are a priori unknown. In more science-related context, machine learning has been integrated to simulate environments, − autonomous data acquisition in SPM experiments, , and the detection and movement of nanowires using an atomic force microscope . The notable capabilities of STMs to autonomously assemble atoms into atomically perfect nanostructures have been demonstrated .…”

Method to Minimize the Errors of AI: Quantifying and Exploiting Uncertainty of Deep Learning in Brain Tumor Segmentation

“…Therefore, AI has not been widely applied in the clinical field yet [ 14 , 15 ]. There have been many studies that seek to overcome this dilemma [ 16 , 17 , 18 , 19 , 20 ]. To fully trust the prediction of AI, other metrics may be needed in addition to accuracy only [ 14 , 21 ].…”