使用疊代式自我一致性運算於核磁共振逆影像重建

Abstract

核磁共振逆影像(magnetic resonance inverse imaging, InI) 利用多組核磁共振射頻線圈同步接收核磁共振影像(magnetic resonance imaging, MRI) 信號以含括全腦的視區及100毫秒的時間解析度 。其基本原理是利用不同位置射頻線圈提供的空間敏感度重建在資料擷取中所忽略的空間編碼訊息。先前研究發現逆影像在K空間(k-space, K-InI)比起在影像空間(image space)使用最小範數估計解(minimum-norm estimate, MNE)的重建可有更高的空間解析度和對大腦活動訊號更高的靈敏度。最近的研究顯示使用多通道接受器的平行核磁共振影像在K空間中有一訊號的自我一致性(self-consistent property)。當運用這個特性在平行影像的重建時，可以提升重建影像的品質。根據這一特性，本研究提出運用自我一致性在核磁共振逆影像 的影像重建方法。其稱為自我一致性K空間逆影像(self-consistent K-InI)以及 範數自我一致性K空間逆影像( -self-consistent K-InI)。經由模擬，我們發現與K-InI 相比，self-consistent K-InI 和 -self-consistent K-InI可以提供更高的空間解析度。應用self-consistent K-InI 及 -self-consistent K-InI於真人視覺功能性核磁共振影像實驗時，這些影像重建方法可在100毫秒時間解析度下描繪出大腦視覺區域的血液動力學變化。Self-consistent K-InI與K-InI在偵測大腦活動訊號的BOLD對比靈敏度相近，但是 -self-consistent 則比K-InI提高約50%的偵測靈敏度。我們預期self-consistent K-InI 和 -self-consistent K-InI可以提供高時間和高空間解析度的大腦活動訊號來進一步了解人腦功能。

Magnetic resonance inverse imaging (InI) using multiple channel radio-frequency (RF) coil detection can achieve 100 ms temporal resolution with the whole brain coverage. InI reconstructions use the RF coil sensitivity information to reconstruct the omitted partition encoding data. Previously we proposed the k-space InI (K-InI) reconstruction to provide higher spatial resolution and higher sensitivity in detecting activated brain areas in BOLD fMRI experiment than the image domain minimum-norm estimate (MNE) InI reconstruction. Recently, the self-consistent property has been suggested as a useful property in k-space parallel MRI reconstruction because it improves the reconstruction image quality. Studying this study, we develop self-consistent K-InI and -self-consistent K-InI algorithms to use the self-consistent property to reconstruct highly accelerated InI acquisitions. Numerical simulations show that self-consistent K-InI and -self-consistent K-InI can provide higher spatial resolution than K-InI. Applying self-consistent K-InI and -self-consistent K-InI to BOLD contrast fMRI experiments, we found that all methods can reveal visual cortex activation at the 100 ms temporal resolution. Self-consistent K-InI has a comparable detection sensitivity to K-InI. -self-consistent K-InI the sensitivity of detecting brain activation is 50% higher than that of K-InI. Self-consistent K-InI and -self-consistent K-InI can be useful tools in fMRI data analysis to characterize brain activity with a high spatiotemporal resolution.