Cerebral diffuse glioma is a biological heterogeneous tumor [
1]. Patient clinical outcome was affected by many factors including age, anatomic location, tumor size, extent of resection, genetic alteration [
14]. Among them, anatomic location plays a crucial role not only for prognosis prediction but also for treatment strategy. It was widely acknowledged that prognosis is poor in midline glioma than non-midline tumor [
10]. Regarding to hemispheric glioma, patient with tumor located in frontal lobe tends to be younger, IDH1 mutation and longer survival time [
15]. This conclusion is accordance with our result. Furthermore, in our cohort, occipital lobe glioma implied negative impact on clinical outcome which is similar to Liu et al. [
16]. The reason might be larger tumor size commonly seen in this area. Meanwhile, anatomic location somehow determines the extent of resection, for example, midline or deep-seated glioma is hard to get gross total resection due to preservation of functional structure or complex surgical corridor [
17]. On the contrary, non-eloquent area tumor, especially superficial to the cortex is amendable to completely remove. For the same reason, longer survival time is strongly associated with gross total resection [
18]. On the other hand, based on huge amounts of exploration of glioma genetic alteration in the recent years, a panel of classic biomarkers begins to exert great impact on glioma precise diagnosis and prognostic assessment [
1]. IDH1/1p19q/H3F3A are the representatives introduced into newly revised 2016 WHO glioma guideline [
7]. In our previous study, 4 biomarkers were used to stratify lower grade glioma into 4 subgroups predicting better clinical outcome than the roles of histological diagnosis and WHO grade [
19]. The current findings firmly validate the great prognostic value of biomarkers in glioma. Similar to our findings, Jenkins et al. used a genetic combination of IDH1/1p19q/TERT to classify glioma into 5 subpopulations with unique clinical features and germline variants respectively which is highly recognized in the world [
8]. We referred to this 3-biomarkers scheme in our study, and drew the same conclusion Triple-positive and IDH1/TERT double-mutation cases are more likely to be oligo-lineage. IDH1 mutation only cases are astrocytoma with maximal possibility. IDH1 wild type and TERT mutation tumors are commonly seen in glioblastoma. According to this scheme, survival outcome in our patient cohort is distinctive among all the subgroups. All these data demonstrate integration of molecular and histology diagnosis being helpful in prognostic and predictive value for glioma patients. Nevertheless, the perfect integration of these two systems still calls for huge efforts like large cohort clinical validation on many aspects, such as image features. Thus, we hypothesized anatomic location, genetic biomarkers and histology diagnosis are highly correlated and intertwined.
In order to verify our hypothesis, we tried to study the interconnection between anatomic location and genetic biomarkers in our patient cohort. Beforehand, many papers published have put forward the cell origin theory underlying possible relationship between these two factors [
20,
21]. Many groups have successfully developed computational methods to predict glioma genetic alterations based on location features [
18,
22,
23]. For example, IDH1 mutation was commonly seen in left frontal lobe, where TERT mutation only exists as well[
2,
5,
23]. Such investigations were performed in the context of MGMT and TP53[
24,
25]. In our study, we used a panel composed of IDH1/1p19q/TERT which is worldwide recognized in the precise diagnosis of glioma to demonstrate the anatomic distribution of different molecular subsets. Our data showed similar results to previous research works such as IDH1 mutation prefers to localize in left frontal lobe[
2,
15]. Interestingly, we also found that triple-positive tumor located more superficial to cortex than TERT mutation only tumor. This finding may explain the differences in survival outcome and extent of resection. In spite of that,. highly consistency of location feature was observed between histological subpopulations and its corresponding molecular counterparts. For example, triple-positive tumor appears to have similar anatomic location with oligodendrogliomas. That means the axis of molecule-cell-tissue depicts the growth pattern of glioma which is additional evidence supporting the cell origin theory. Another interesting finding is that 1p19q codeletion oligoastrocytoma possessing different anatomic location with 1p19q intact oligoastrocytoma which is supportive to new 2016 WHO classification that the diagnosis of oligoastrocytoma is eliminated [
7]. Since now, the diagnosis of oligoastrocytoma converts to either oligodendroglioma or astrocytoma according to molecular biomarkers [
26]. These findings demonstrate definite molecular feature restricted to precise histological diagnosis. It strongly proved that molecular diagnosis can help clinicians make exactly right diagnostic decision facilitating to tailor personalized treatment.
On the other aspect, methods by using MR images to predict molecular biomarkers are popular recently, which was so-called Radiomics study. Ellingson et al. compared tumor volume ratio of T2 hyperintensity to contrast enhancement and central necrosis to differentiate mesechymal and non-mesenchymal molecular subtype in glioblastoma [
27]. His research team also used perfusion and diffusion MRI signatures to successfully stratify lower grade glioma into three subpopulations as IDH1 mut/1p19q codel, IDH1 mut/1p19q non-condel and IDH1 wt [
28]. MRS is another popular detectable technology to realize Radiomics study due to unique metabolic features inside glioma. It has been widely applied to predict IDH1 mutational status and medulloblastoma subgrouping [
29,
30]. Compared to these methods, our team used anatomic location as basic tumor feature to predict biomarkers like IDH1/1p19q/TERT, which is more simple, cost effective and visualized. The raw materials we need are only T2 flair and T1 contrast MR images without sophisticated computation process. However, our method has its own limitations, like rough estimation accuracy. In general, our team illustrated a simple method to predict molecular biomarkers and reveal anatomic location among different molecular subgroups which offered an alternative in Radiomics study.