In this study, we examined the clinical utility of FLAIR imaging for the diagnosis of MELAS with recurrent SLEs compared to ASL imaging. The results showed that the rCBF values obtained from ASL images and the nSI obtained from FLAIR images were equivalent to their diagnostic performance in MELAS cases. We calculated Spearman’s rank correlation coefficient using the rCBF values and found no correlation between the normal and MELAS groups. We considered that the degree of increased blood flow in recurrent SLEs varied by case. However, as shown in Fig. 3, the overall rCBF values were higher in the MELAS cases than in the normal cases, especially in the cuneus. Therefore, only the degree of increased blood flow can easily indicate an abnormality of the brain; this is consistent with findings of previous reports [7, 8]. In contrast to the rCBF values, there was a strong positive correlation in the nSI results. The high correlation of the nSI obtained from FLAIR images is due to the smaller signal intensity in MELAS cases and normalization of the signal intensity by dividing it by the signal intensity of the brainstem. We considered that the correlation coefficient of nSI was also higher because normalization allowed the variation in signal values to be compared between cases. Based on the decision boundaries obtained from LDA, both the sensitivity and specificity for diagnosing MELAS using the rCBF value were higher than those of the nSI; however, there were few differences. Regarding sensitivity, in the positive cases diagnosed with MELAS on the border of the decision boundary obtained from the LDA of nSI, one MELAS-positive case could be distinguished by the ASL image, which was the cause for the slight difference. We considered that the difference in specificity was due to the larger rCBF values of cuneus in MELAS cases than in normal cases, and the smaller signal intensity in MELAS cases and the smaller difference between MELAS and normal cases. Care must be taken in cases of MELAS that cannot be distinguished using either rCBF values or nSI. In this study, MRI examination of recurrent SLEs was performed; however, there were no abnormalities on the ASL or FLAIR images (Fig. 5c). Li et al. reported that persistent cellular injury due to impaired adenosine triphosphate production may give rise to excessive loss of neurons, resulting in decreased rCBF in the affected brain regions during the chronic phase of SLEs [7]. Therefore, our method is not appropriate for these cases, and we need to perform a comprehensive diagnosis in conjunction with physical findings and other examination results. We calculated AUC values from ROC analysis using rCBF values and the nSI, and recognized that ASL images have better diagnostic performance than FLAIR images.

However, there was no statistically significant difference between the AUC values obtained from the ASL and FLAIR images. On the basis of the results, we concluded that the diagnostic performances of ASL and FLAIR images for MELAS with recurrent SLEs were equivalent using our method.

Edematous lesions with cortical predominance may persist long after symptom resolution, although FLAIR high-signal lesions may shrink in size. Using this method, it is possible to determine whether SLEs have recurred in residual FLAIR high-signal lesions that are difficult to assess visually. In addition, although the cases in this study were obtained at the time of recurrent SLEs, a case was reported in which ASL image over time captured increased blood flow in the preictal lesion, and it was also reported that ASL imaging was useful in predicting seizures [15, 16]. If our method has a diagnostic performance equivalent to that of ASL imaging, we suggest that FLAIR imaging may also be useful in predicting seizures.

A limitation of this study was the small number of MELAS cases; however, we considered that a population trend was obtained when considering the distribution of the scatter plots. In addition, both angioedema and cellular edema show high signal intensity on FLAIR images and may not be discriminated based solely on the signal intensity of the FLAIR image. Thus, we consider that the diagnosis of such cases should be proceeded with other sequences.