NativeT1图像纹理分析作为透析患者心肌纤维化无创评估的新方法（社论）。Editorial on “Texture Analysis of Native T1 Images as a Novel Method for Non-Invasive Assessment of Myocardial Fibrosis in Dialysis Patients”
The most common cause of death in patients with renal failure is cardiovascular disease. Myocardial fibrosis has been shown to correlate with increased cardiovascular mortality. Cardiac MRI with gadolinium-based contrast agent (GBCA) is often used to identify myocardial fibrosis. Late gadolinium enhancement techniques are useful in identifying dense, regional areas of fibrosis. In the assessment of diffuse myocardial fibrosis, different tissue characterization techniques are needed. The one technique that has been most validated is extracellular volume fraction (ECV), which requires acquisition of T1 map before and after contrast injection, and has been shown to correlate with poor prognosis.
In patients with renal failure, GBCA is rarely used due to the risk of nephrogenic systemic fibrosis. Thus, non-contrast techniques such as native T1 map, T1rho, and magnetization transfer techniques have been explored. Native T1 values have been found to be statistically elevated in myocardial fibrosis compared to normal myocardium; however, there is significant overlap in the T1 values. This limits the techniques diagnostic abilities to discriminate between normal myocardium and myocardial fibrosis in the clinical setting.
In this issue of JMRI, the authors explored machine learning techniques involving texture analysis to distinguish normal myocardium from diffuse myocardial fibrosis without the use of GBCA. The authors performed myocardial native T1 mapping on 66 dialysis patients and 51 healthy controls with a 3 T MRI scanner. Texture analysis techniques were employed on the native T1 maps: sum entropy, vertical run-length non-uniformity (VRLN), 45° short run emphasis, wavelet energy LL, and wavelet energy HH. Of the five techniques, wavelet energy LL did not demonstrate a statistical different between control and dialysis patients. Whereas, VRLN was not only statistically different between the two groups, but also had a 92% accuracy (88% sensitivity, 97% specificity) to distinguish between them. VRLN also negatively correlate to ejection fraction and global longitudinal strain (GLS), indicating the possible correlation to myocardial fibrosis.
The findings are quite impressive in the ability to differentiate the myocardium of healthy controls from dialysis patients without the use of GBCA. The study is consistent with other texture analysis studies performed in patients with hypertrophic cardiomyopathy in the identification of abnormal myocardium. However, the texture analysis findings need to be validated against biopsy data or ECV mapping, which are the accepted techniques in demonstrating myocardial fibrosis. Although GLS does correlate with prognosis, it does not correlate well with myocardial fibrosis as detected by ECV.
To determine the specificity of the findings to myocardial fibrosis, these techniques should be applied to patients who have abnormal native T1 values that are not related to myocardial fibrosis. These may include patients who have low myocardial native T1: cardiac siderosis and Fabry's, as well as patients with elevated myocardial native T1: acute myocardial infarction, acute myocarditis, and cardiac amyloidosis. Regardless of whether texture analysis is truly identifying myocardial fibrosis or just myocardial pathology, large population prognostic studies need to be performed to ensure it is clinically relevant, and ultimately lead to development of therapies that can improve patient outcomes.
Machine learning has become an important area of study in clinical imaging. There has been on its increasing role in identification of pathology that is not possible with the human eye. Further studies will be necessary as it continues to transform how we practice, increase our diagnostic accuracy, and ultimately improve the lives of our patients.
Leung SW. Editorial on "Texture Analysis of Native T1 Images as a Novel Method for Non-Invasive Assessment of Myocardial Fibrosis in Dialysis Patients". J Magn Reson Imaging. 2021 Jul;54(1):301-302. doi: 10.1002/jmri.27523. Epub 2021 Feb 26. PMID: 33634909.
Zhou et al. Texture analysis of native T1 images as a novel method for non-invasive assessment of myocardial fibrosis in dialysis patients. J Magn Reson Imaging 2021.