In this study, we analyzed the signal intensity outside residual cavity on FLAIR imaging of LGG patients with gross-total tumor resection. The present study indicates that the reproducibility for measurement of relative signal intensity on FLAIR images with the open-source software Image J. We showed that higher rFLAIR value (> 1.595) is an adverse prognostic factor for post-treatment progression and survival prognosis in LGG patients. Our data strongly supported the hypothesis that the ability of the quantitative metrics, the rFLAIR, in improving the survival prediction in combined prognosis model. Thus, besides the extent of hyperintensity disease, the relative gray intensity outside residual cavity on FLAIR images should be evaluated for post-treatment LGG patients.
The present study showed that the open-source software Image J could offer the quantitative metrics of FLAIR images with satisfactory reproducibility. As a widely used image-processing platform, Image J has already employed in biological image analysis for depicting weak signal variation beyond naked eye [16]. Image J had previously been used in the evaluation pontine glioma. In one series of 121 pediatric patients with post-treatment diffuse intrinsic pontine glioma (DIPG), Poussaint et al. generated quantitative metrics from FLAIR image and ADC map [17]. They demonstrated that pre-radiotherapy FLAIR skewness and standard deviation were associated with shorter PFS. In the new version of WHO classification of CNS tumors (5th edition, 2021) [17, 18], DIPG would mainly be diffuse midline glioma with H3K27-altered, which is a kind of high-grade glioma with poor prognosis. In comparison, we analyzed rFLAIR of LGG in cerebral parenchyma in the present study. However, both our results and the study of Poussaint et al. demonstrated the usefulness of Image J in quantitative analysis of FLAIR image in post-treatment gliomas with different grades.
There were several advantages of this method. Firstly, we calculated the relative signal gray value by comparing the gray intensity of hyperintensity lesions outside residual cavity with contralateral parenchyma as well as background of the image, instead of directly measuring signal intensity on certain MR equipment. Thus we deduce that the bias came from different parameters of FLAIR sequence of various scanners and diverse magnet field strength could be averted. Secondly, tumor progression after LGG treatment was evaluated RANO criteria, which mainly based on T2-FLAIR disease. LGG, especially WHO grade II gliomas often manifested as T2-FLAIR hyperintense lesions without enhancement after gadolinium contrast medium injection [4]. Therefore, evaluation of these non-enhancing FLAIR hyperintensity lesions is critical for the therapy regimens of WHO grade II and III gliomas [3]. Thirdly, there was significant difference of rFLAIR between the progression (1.80) and non-progression groups (1.55) in this study. The rFLAIR, and the combined model including rFLAIR, could effectively predict poor survival outcome. Therefore, rFLAIR may be an adequate surrogate metrics and even eliminate a long-term follow-up when a suspected non-enhancing FLAIR hyperintensity lesion is found. Finally, antiangiogenic agents, including bevacizumab and cediranib, had been extensively used in the treatment of gliomas. These agents may lead to pseudo-response for temporally decreasing the permeability of blood-brain barrier and consequently diminishing contrast enhancement [19]. However, T2-FLAIR imaging could be used to detect the increase of non-enhancing hyperintensity lesion and thus could be used to identify early tumor progression.
Higher rFLAIR outside the residual cavity is probably due to neoplastic cell infiltration or tumor remnant. In one study which included 10 patients with WHO grade II-IV gliomas, Amjad et al. investigated the so-called peri-tumoral high signal regions on FLAIR imaging with functional MR techniques and targeted biopsy [20]. They found tumor cell infiltration and tumor core in 75% samples in FLAIR hyperintense regions. Thus, tumor cells have a tendency to infiltrate and manifest as non-enhancing T2-FLAIR hyperintense lesions. Although LGG is less aggressive than glioblastoma, Amjad et al. still confirmed that a portion of tumor extending outside the gadolinium contrast enhancing border in 7 patients with WHO grade II and III gliomas. These non-contrast-enhancing lesions could be visualized well on T2-weighted FLAIR imaging [21]. On the other hand, Chang et al. investigated the signal intensity outside the residual cavity on T2-FLAIR imaging and found small but significant changes could be detected months before the development of abnormal contrast enhancing lesions [22]. We also confirmed that new enhancing-lesions developed on follow-up MRI within the earlier non-contrast-enhancing hyperintensity region on FLAIR imaging. For standardizing the intensity value of FLAIR images among patients, Chang et al. employed a histogram normalization algorithm [22]. Whereas, we normalized the measurement of signal intensity on FLAIR images with comparing the gray value of non-contrast-enhancing lesions with contralateral parenchyma as well as background. The calculation method of rFLAIR in this study may be in favor of eliminating the influence of diverse scan parameters and different magnet fields on FLAIR imaging. Therefore, rFLAIR in non-contrast-enhancing lesions outside the residual cavity of LGG can be used as an imaging marker for estimating the burden of microscopic non-enhancing tumor and predict the location of recurrent disease in post-treatment LGG patients.
In the present study, we also confirmed prognostic prediction value of other previously described MRI features [9, 14, 23], including the enhancement types of residual wall and new distal enhancement and new SVZ involvement. However, the prediction performance of these feature was relative lower (AUC of 0.672 for PFS, AUC of735 for OS) and would be improved when combined with rFLAIR (AUC of 0.799 for PFS, AUC of 0.827 for OS). This phenomenon may be explained by the gliomas enrolled with lower grade in this study. The above mentioned MRI features could be detected more often in those post-treatment glioblastoma patients [12, 13]. The incidences of thick-linear and nodular enhancement (34.87%), new distal enhancement (5.26%), new SVZ involvement (20.39%) were lower than those of glioma (51.52%, 25.43%, and 49.14% separately) [12, 13]. On MRI, LGG often manifested as ill-defined hyperintensity T2-FLAIR lesions and without post-contrast enhancement because of less invasive, less angiogenesis and minimal disruption of blood-brain barrier [4]. New involvement of SVZ could be manifested as both new enhancement and non-enhancing FLAIR hyperintensity lesions in SVZ region. In this study, new SVZ involvement was detected as non-enhancing hyperintensity T2-FLAIR lesions in 80.65% patients (25/31). SVZ could increase invasiveness and migratory potential because it is the source of tumor precursor stem cells [11, 23]. Thus, higher incidence of SVZ involvement (83.69%) in progression group in the present study, even without post-contrast enhancement, had adverse impaction on survival outcome of LGG patients.
Our data also confirmed that previously reported clinical factors, including age, post-operative KPS score, Ki-67 scores, WHO tumor grades, were associated with survival outcome of LGG [8, 10, 24, 25]. Previous studies had already identified that age was an important prognostic factor of gliomas. In an analysis of 113 grade III glioma patients, Hong et al. found 51 and 55 years old were the cutoff values of PFS and OS separately [8]. In the present study, we confirmed that the patients in progression group (48.00 year-old) were elder than those in non-progression group (43.13 year-old). The prognostic value of Ki-67 index for LGG was similar to that for glioblastoma [25] since higher Ki-67 index in LGG was associated with malignant transformation and poor survival outcome. We also found the Ki-67 index in progression group (0.20) was higher than that of non-progression group (0.10) ( p = 0.014). However, our study revealed that the superiority of the outcome diagnostic performance with the combination of conventional MRI and rFLAIR (AUC of 0.799 for PFS, and AUC of 0.827 for OS) to those of clinical factors (AUC of 0.726 for PFS, and AUC of 0.799 for OS). Thus, we further combined conventional imaging findings with the quantitative metrics of FLAIR images, rFLAIR, which can be more reliable in differentiating tumor progression from non-progression patients with non-enhancing hyperintensity lesions outside residual cavity. This combination improved the prognostic prediction performance effectively. Therefore, we recommend that LGG patients with suspicious non-enhancing hyperintensity lesions on T2-FLAIR images should additionally calculate the quantitative metrics from conventional MRI sequence, such as rFLAIR.
Several limitations should be mentioned to the present study. First, although we revealed that the rFLAIR was independent prognostic factor of post-treatment LGG,the sample in this retrospective study was relatively small. We enrolled the consecutive LGG patients who had operated in a period of five years. But, because of relatively lower incidence (43.2% of all gliomas) [26] and comparative more benign course, LGG was less often encountered and treated aggressively in clinical practice. Thus, the results of this study warrant further validation with larger multicenter investigation. Second, as a retrospective analysis, there may be a selection bias of the patients. A few cases were excluded because of lost follow-up, without hyperintensity of FLAIR images, without standard treatment and follow-up measurement and so on. Third, the discrimination of progression from non-progression lesions was based on follow-up data except 9 patients who were confirmed as progression disease by re-operation. Fourth, we measured T2-weighted hyperintensity lesions in single ROI without discrimination tumor remnants from post-treatment cerebral edema, ischemic change. We confirmed that some FLAIR hyperintesity lesions were due to cerebral edema and ischemia based on the decrease in size of lesions with a long-term follow-up. Whereas, as a consequence of operation, ischemia plays an important role in inducing hyperintensity on T2-FLAIR imaging and probably leads to overestimation of tumor remnants [27]. Finally, although we collected molecular pathological data from some patients, including mutation of IDH, MGMT, and 1p19q co-deletion, molecular pathology examination has not been widely included in routine clinical examination in author’s institution, especially in the era before 2016. Recently, Consortium to Inform Molecular and Practical Approaches to CNS Tumor Taxonomy working committee considered that histologic grade II and III IDH wild-type astrocytic glioma should be referred as diffuse astrocytic glioma, IDH-wild-type, for these gliomas containing high-level EGFR amplification or TERT promoter mutations [26]. Further analysis on the outcome evaluation of LGG in the light of FLAIR hyperintensity lesions should be based on genetics of LGG in the future.
In conclusion, we found that higher rFALIR (> 1.595) of non-contrast-enhancing lesions outside the residual cavity was a useful predictor of poor survival of LGG. As one reproducible, accessible quantitative metrics based on conventional sequence, rFLAIR was helpful to improve the survival prediction of post-treatment LGG patients in clinical practice. An early post-treatment MRI performed after the completion of radiotherapy might be more appropriate for the delineation of tumor remnants in the region with non-enhancing hyperintensity on FLAIR imaging. The combination of rFLAIR, clinical factors and conventional MRI features may even eliminate a long-term follow-up for LGG patients when a suspected non-enhancing hyperintensity lesion is found.