以磁振造影術定位單向傳導白質纖維束之邊界：理論探討

Abstract

簡介 神經科多項臨床決策之擬定皆須考慮動作及感覺神經纖維束在病患體內的確切位置，然而目前尚無非侵襲性影像技術可精確辨識此類單向傳導纖維束在白質中的邊界。先前研究指出：纖維束傳導複合神經動作電位時會產生足以用腦磁圖與脊髓磁圖偵測的微弱磁場訊號，且神經軸突磁場會調控核磁共振訊號強度。本篇研究以一組織型態及訊號特性皆近似真實白質纖維束的數值仿體模擬纖維束邊界處產生的核磁共振訊號，藉以探討以磁振造影術定位神經纖維束邊界位置之可行性。 方法 本研究參考大鼠頸段皮質脊髓徑之組織構造，以AxonPacking軟體與自編程式建立一包含兩條相鄰髓鞘化纖維束之數值仿體，並以霍奇金-赫胥黎模型及纜線模型模擬動作電位傳導。依據歐姆定律與必歐-沙伐定律，我們將模擬所得之動態膜電位分佈轉換為仿體中的動態磁場分佈，並以布洛赫方程式及拉莫爾方程式計算該動態磁場分佈對核磁共振訊號的影響。 結果 動作電位沿單一髓鞘化軸突傳導所產生的典型磁場變化為雙向波，由一寬度小於0.5毫秒的前導波瓣及一寬度大於1毫秒的尾隨波瓣所組成。我們操弄兩條纖維束的傳導方向以組合模擬出四種靜息狀態訊號；在四種靜息狀態中，兩條纖維束皆以平均約7.5赫茲之頻率隨機傳導動作電位。當主磁場平行於纖維束交界面並垂直於纖維束長軸時，異向傳導靜息狀態所引致之交界處磁場調控可在90毫秒內造成約0.0002度的核磁共振訊號相角偏移。我們操弄交界處體素與外來刺激之相對距離以模擬五種激活狀態訊號；五種激活狀態皆僅有左側纖維束接受刺激，而刺激處分別位於體素上游0毫米，2.5毫米，5毫米，7.5毫米與10毫米之位置。模擬所得之複合神經磁場波形為雙向波，前導波瓣幅度於刺激處可高達5毫微特斯拉，並隨訊號往下游移動而漸有波形離散現象。五種激活狀態之相角累積曲線圖皆顯示單一波峰；隨著刺激處往上游移動，波峰幅度由0.014度漸減為0.004度，而波鋒寬度則由約2毫秒漸增為7毫秒。神經磁場之相關係數及標準差分析顯示：與纖維束截面之磁場變化相較，沿著纖維束長軸的磁場變化較為一致。在超低磁場核磁共振之實驗設定下，以激活狀態之振盪磁場作為翻轉脈衝所能產生之最大翻轉角度為0.021度。 結論 模擬結果顯示：由於纖維束邊界處磁場變化對核磁共振訊號的調控效應極其微弱，以磁振造影術定位神經纖維束邊界位置將是極度困難。將沿著纖維束長軸方向的解析度降低，並將垂直於纖維束交界面方向的解析度提高，可能有助於提高訊雜比。激活狀態複合神經磁場所造成的核磁共振訊號相位變化或許可用回音間距較小的多回音梯度回音序列進行追蹤。

Introduction Accurately localizing unidirectional motor and sensory pathways assists in clinical and operative decision making. Currently there are no imaging techniques that reliably and non-invasively delineate the boundaries of these neural pathways. Previous studies suggest that signal propagation through neural pathways generates weak magnetic signals that can be detected by magnetoencephalography or magnetospinography systems, and detectable modulation of MR signal magnitude by axonal magnetic fields has been reported. In the present study, we assessed the potential of MRI as a method to delineate the boundaries of neural pathways by simulating MR signals generated at the tract boundary using a histologically realistic numerical phantom propagating physiologically realistic signals. Methods A numerical phantom containing two neighboring myelinated white matter tracts with microstructure resembling rat cervical corticospinal tracts was built using the AxonPacking software and an in-house software. Action potential propagation was simulated using a combination of Hodgkin-Huxley model and cable model. Dynamic transmembrane voltage distribution was converted to dynamic magnetic field distribution using Ohm’s law and Biot-Savart law. The degree of MR signal modulation by axonal magnetic fields was computed using Bloch equation and Larmor equation. Results The typical waveform of axonal magnetic field produced by signal conduction through a myelinated axon comprised a leading high-amplitude component lasting less than 0.5 millisecond and a trailing low-amplitude component lasting more than 1 millisecond. Four resting-state conditions were simulated in which both tracts propagated random action potentials in either one or the other direction with a mean firing frequency of 7.5 Hz. With a static field B0 oriented perpendicular to the tracts, the net phase shift of MR signal from the boundary voxel accumulated to roughly 0.0002 degrees in 90 milliseconds when two tracts propagated signals toward opposite directions. Five activated-state conditions were simulated in which a stimulus was applied to one of the two tracts at 0 mm, 2.5 mm, 5 mm, 7.5 mm and 10 mm upstream of the boundary voxel. The compound magnetic field signal had a bipolar waveform, with the leading component reaching 5 nano-Tesla at the location of stimulus and gradually dispersed as it propagated downstream. The phase accumulation trace showed a single peak that gradually decreased in amplitude from 0.014 to 0.004 degrees and increased in duration from about 2 milliseconds to 7 milliseconds as the location of stimulus was moved upstream. Analysis of correlation coefficients and standard deviations of magnetic fields indicated that magnetic fields were more coherent along the long-axis of the tract than across the cross-section of the tract. The maximal flip angle generated by the compound magnetic field, which served as a tipping pulse in the ultra-low-field MR setting, was about 0.021 degrees. Conclusion Simulation results indicate that delineating the tract boundaries using MRI is extremely challenging because axonal magnetic fields exert weak modulation on MR signals. Using anisotropic voxels with the resolution being higher along the direction perpendicular to the boundary plane and lower along the long-axis of the tract might increase signal-to-noise ratio. Phase changes induced by the propagation of compound magnetic fields could be tracked using multi-echo gradient echo sequences with sufficiently short echo spacing.