54 mT超低场磁共振对脑血管疾病的初步临床研究

Preliminary Clinical Study of 54 mT Ultra-Low Field Magnetic Resonance Imaging in Cerebrovascular Diseases

摘要: 目的：通过分析正常志愿者及脑血管疾病患者的MR图像，评价ULF-MRI图像质量及其对脑血管疾病的诊断效能，探讨ULF-MRI的临床价值。方法：本研究纳入正常志愿者及脑血管疾病患者各20例。正常志愿组均行头颅HF-MRI及ULF-MRI扫描；脑血管疾病组均行头颅CT平扫或HF-MRI及ULF-MRI扫描。评价两位影像医生之间各项主观测量及客观评价指标的一致性，并对比两组受试者HF-MRI及ULF-MRI图像质量差异。评价医生基于ULF-MRI对脑血管疾病的诊断效能及诊断信心，并于对照学习CT及HF-MRI后再次评价医生基于ULF-MRI对脑血管疾病的诊断效能及诊断信心。结果：ULF-MRI T1WI、T2WI图像质量测量值的组间ICC系数分别为0.996、0.986；HF-MRI T1WI、T2WI图像质量测量值的组间ICC系数分别为0.998、0.977。两组受试者ULF-MRI各序列的图像质量均低于HF-MRI或CT。正常志愿组ULF-MRI与HF-MRI之间，T1WI及T2WI图像的灰质SNR、白质SNR、脑室SNR及三种组织的两两间CNR差异有统计学意义(P < 0.05)；T2-液体衰减反转恢复(T2-weighted fluid-attenuated inversion recovery, T2-FLAIR)图像的灰质SNR、白质SNR、灰质/脑脊液CNR及白质/脑脊液CNR差异有统计学意义(P < 0.05)，脑室SNR及灰质/白质的CNR差异无统计学意义(P = 0.088)。脑血管疾病组ULF-MRI与HF-MRI或CT之间的图像质量差异均有统计学意义(P < 0.05)。基于ULF-MRI，影像医师对脑血管疾病的初次诊断敏感度为60%，诊断信心评分为2.5 (1, 2.5)；在对照金标准图像学习后，诊断敏感度为85%，诊断信心评分为4 (3.625, 4)，学习前、后诊断信心评分差异有统计学意义(P < 0.05)，诊断敏感度差异无统计学意义(P = 0.077)。结论：目前ULF-MRI的图像质量及其对脑血管疾病的诊断效能尚有待提高，但能通过对照学习及经验总结显著提高脑血管疾病的诊断效能及医生诊断信心。

Abstract: Objective: To evaluate the image quality of ultra-low field magnetic resonance imaging (ULF-MRI) and its diagnostic efficacy for cerebrovascular diseases by analyzing MR images of healthy volunteers and patients with cerebrovascular diseases, and to explore the clinical value of ULF-MRI. Methods: A total of 20 normal volunteers and 20 patients with cerebrovascular diseases were enrolled in this study. All subjects in the normal volunteer group underwent cranial high field magnetic resonance imaging (HF-MRI) and ULF-MRI scans. For the cerebrovascular disease group, all patients underwent non-contrast cranial CT or HF-MRI (for patients with cerebral infarction) as well as ULF-MRI scan. The consistency of subjective measurements and objective evaluation indicators between the two observers was assessed, and the differences in image quality between HF-MRI and ULF-MRI were compared between the two groups. The diagnostic efficacy and diagnostic confidence of radiologists for cerebrovascular diseases based on ULF-MRI were evaluated. After the radiologists reviewed and learned from the gold standard images (CT and HF-MRI), their diagnostic efficacy and confidence based on ULF-MRI were re-evaluated. Results: The inter-observer reliability was good. The inter-observer intraclass correlation coefficients (ICC) for image quality measurements of ULF-MRI T1WI and T2WI were 0.996 and 0.986, respectively; while those for HF-MRI T1WI and T2WI were 0.998 and 0.977, respectively. The image quality of all ULF-MRI sequences in both groups was lower than that of HF-MRI or CT. In the normal volunteer group, there were statistically significant differences in the gray matter (GM) signal-to-noise ratio (SNR), white matter (WM) SNR, ventricular SNR, and pairwise contrast-to-noise ratios (CNR) among the three tissues on T1WI and T2WI images between ULF-MRI and HF-MRI (all P < 0.05); there were statistically significant differences in the gray matter SNR, white matter SNR, gray matter/cerebrospinal fluid (CSF) CNR, and white matter/cerebrospinal fluid CNR (GM SNR, WM SNR, GM/CSF CNR, and WM/CSF CNR) of T2-weighted fluid-attenuated inversion recovery (T2-FLAIR) (P < 0.05), while there were no statistically significant differences in the ventricular SNR and gray matter/white matter CNR (GM/WM CNR) (P = 0.088). In the group with cerebrovascular diseases, there were statistically significant differences in image quality between ULF-MRI and HF-MRI or CT (P < 0.05). Based on ULF-MRI, the initial diagnostic sensitivity of radiologists for patients with cerebrovascular diseases was 60%, and the diagnostic confidence score was 2.5 (1, 2.5). After learning with the gold standard images, the diagnostic sensitivity rate was 85%, and the diagnostic confidence score was 4 (3.625, 4). The difference in diagnostic confidence scores before and after learning was statistically significant (P < 0.05), while the difference in diagnostic sensitivity rates was not statistically significant (P = 0.077). Conclusion: At present, the image quality of ULF-MRI and its diagnostic performance for cerebrovascular diseases still need to be improved, but the diagnostic performance of cerebrovascular diseases and doctors’ diagnostic confidence can be significantly enhanced through comparative learning and experience summarization.