When it comes to cfRNAs and their use in gliomas, Dong et al. discovered that the serum of GB patients exhibited a significant decrease in 24 miRNAs and a substantial increase in 115 miRNAs, which was not observed in the serum of healthy controls [93]. Furthermore, Wang et al. identified three downregulated miRNAs, namely miR-128, miR-485-3p, and miR-342-3p, in patients compared with controls. These miRNAs demonstrated a correlation with GB grades and served as biomarkers for assessing tumor grading and monitoring treatment response [94].

MicroRNA-21 is the most extensively studied miRNA in cancer and its overexpression has been consistently observed both in tissue and plasma of GB patients. This elevated expression is closely linked to lower overall survival rates and higher tumor grading [95]. miR-21 functions as an anti-apoptotic factor in glioblastoma cell lines by exerting its effects through caspase inhibition. The suppression of miR-21 activity consequently impedes cell growth, enhances apoptosis, and diminishes the proliferation of GB cancer cells [95,96].

It is worth noting a study that identified TP73-AS1 as a clinically relevant long non-coding RNA (lncRNA) in GB. The researchers observed a significant upregulation of TP73-AS1 in primary GB samples. They provided evidence that TP73-AS1 serves as a robust prognostic biomarker, as this lncRNA promotes tumor aggressiveness and confers resistance to TMZ treatment in GB cancer stem cells [97].

Given the remarkable patient and disease heterogeneity, as well as the broad protein involvement in multiple processes, it is difficult to find diagnostic plasma/serum markers for glioblastoma with sufficient clinical value. However, extensive research has identified the levels of several circulating proteins to be associated with the presence of GB. One example is the plasma glycoprotein, haptoglobin, which is an acute phase protein involved in the protection of tissue damage and oxidative stress [100]. Haptoglobin’s levels change during pathologies and, although its single estimation may not be specific and sensitive enough, several proteoforms of haptoglobin (α2 and β-chain) have been detected upregulated in GB patients’ plasma, indicating its potential use as a GB-specific blood biomarker [100].

In addition, a quantitative proteomic analysis showed that several plasma proteins, such as carnosinase 1 (CNDP1) that regulates carnosine levels, the inflammatory marker ferritin light chain (FTL), and the Ca²⁺ signaling protein S100A9, were found to be altered in GB tissues, with CNDP1 been suggested as a potential drug target. This confirms that plasma-based tests for initial diagnosis or recurrence monitoring can be highly useful for GB [101,102]. Moreover, the study of Pérez-Larraya et al. performed a comprehensive analysis of insulin-like growth factor binding protein 2 (IGFBP-2), the intermediate filament (IF) III protein GFAP, and the glycoprotein YKL-40 plasma levels as a supplementary diagnostic and prognostic tool [103]. This approach holds great promise, particularly for patients with inoperable brain lesions [103]. Interestingly, the presence of the glycoprotein’s fetuin-A auto-antibodies was also revealed as indicators of GB development [104]. Ectopic fetuin-A levels have been suggested to play a role in tumor progression, including GB. Therefore, evaluation of fetuin-A auto-antibodies in the serum of GB patients was proposed to potentially serve as a screening tool, presenting one of the earliest indicators for GB growth [104].

Numerous studies have confirmed the significant role of CTCs in CNS tumors. CTCs provide a less invasive way of obtaining tumor samples for cancer detection; certain CTC genotypes may indicate primary tumor progression and changes in malignant genetic information during disease relapse. GB has been reported to have a high prevalence of CTCs, which is estimated to be around 75% [108].

CTC concentrations have been utilized for the detection of various conditions. According to Santos et al., CTCs exhibit promising results in the early detection of colorectal cancer as they can be identified in the blood of patients who have recently developed the disease. As a result, CTC testing could be utilized to diagnose colorectal cancer [109].

When it comes to gliomas, the blood brain barrier (BBB) may restrict the quantity of CTCs which are found in the bloodstream. Despite this, circulating cells have been identified in glioma patients’ blood and several studies have reported the detection of CTCs in the bloodstream of glioma patients through the use of various isolation techniques. After using the technology of CTC-iChip to enrich CTCs and staining them with antibodies, such as SOX2, EGFR, tubulin b-3, c-MET, and A2B5, Sullivan et al. identified CTCs in GB patients’ blood with a frequency of 39%. They also discovered that the CTCs exhibited a mesenchymal molecular signature [48]. Muller et al. conducted a study where they employed Ficoll–Paque gradients through differential centrifugation, followed by GFAP staining, to identify 20.6% of CTCs [108]. In addition, Gao et al. demonstrated that CTCs were present in 77% of blood samples from patients with brain cancer and in 82% of patients with GB [52].

An interesting study by MacArthur et al. suggested that CTCs could also serve as a prognostic factor. Using density gradient centrifugation, they isolated CTCs and analyzed the activity of the telomerase enzyme in brain cancer patients. The study revealed that before radiotherapy treatment, CTCs were isolated in 72% of patients versus only in 8% after treatment [70]. Moreover, the anti-glioma aptamers Gli-233 and Gli-55 have been used to detect CTCs in LBs in order to increase diagnostic specificity [110]. Aptamers are nucleic acid molecules that bind exclusively to their targets, such as cancer-related proteins, with high affinity and selectivity [111]. They have therefore already been used in biosensing, bioregulation, and bioimaging as reliable recognition ligands [112]. Kichkailo et al. also demonstrated that aptamers may specifically bind to glial tumor cells for CNS tumor detection [110]. Although more research is necessary to uncover the full potential of CTCs in diagnosing gliomas and GB, their use as prognostic factors for gliomas is very promising.

Another important characteristic of GB cells is their ability to use invadopodia, through which they are capable of invading adjacent cells, specifically astrocytes. Invadopodia are GB cell membrane-derived extensions that adhere to neighboring tissues and promote the proteolytical degradation of the matrix, thus allowing invasion [113,114]. Several proteins excreted from GB-derived exosomes are linked to invadopodia biogenesis and subsequently GB-invasive potential, among which are annexin A1 (ANXA1), actin-related protein 3 (ACTR3), integrin β1 (ITGB1), calreticulin (CALR), and programmed cell death 6-interacting protein (PDCD6IP) [114]. Hallal et al. also observed the formation of podosomes in astrocytes and degradation of the matrix after their interaction with exosomes derived from GB cells, a process that appears to be favored by decreased p53 levels. Exosomes thereby exhibit carcinogenic potential and promote the neighboring astrocytes to become tumorigenic [115].

In this context, GB exosomes can be detected both in the blood and the CSF by using the minimally invasive technique of liquid biopsy. A robust number of tumor-derived exosomes may be found in the CSF. It is important to note that CSF is not contaminated with blood-related EVs, for example, platelet-derived exosomes. However, on the other hand, blood samples are more conveniently and less invasively collected compared with the CSF [113,116].

Furthermore, Shao et al. described how a nuclear magnetic resonance system can detect GB-expelled microvesicles in blood samples [117]. Studies have shown that exosome-carried molecules may promote tumor development and therapy resistance through the formation of a tumor-friendly microenvironment. Therefore, exosomes have been suggested as promising tools in GB diagnosis and prognosis, enabling a more specific characterization of the tumor [118,119].

Despite being carriers of many miRNA types, including miR-21/-29a/-221/-222, etc., as demonstrated in several studies (in vitro and microarray analyses), exosomes also play a key role in boosting proliferation and inhibiting GB cell apoptosis [58,120]. They could therefore be utilized as a promising delivery vehicle for tumor-suppressive miRNAs that target gliomas [121]. Approximately 1000 proteins have been identified in GB exosomes using mass spectrometry analysis; they mostly act as pro-angiogenic factors in normal endothelial cells of the brain, such as angiogenin, IL-6/-8, and tissue inhibitor of metalloproteinase-1 and 2 (TIMP-1, and TIMP-2), capable of promoting malignancy by causing hypoxia [122,123].

Exosomes may also transfer receptors with tumorigenic features, such as epidermal growth factor receptor vIII (EGFRvIII), human epidermal growth factor receptor 2 (HER2), and platelet-derived growth factor (PDGFR), which promote GB proliferation to healthy stromal cells [123]. Furthermore, exosomes deliver phosphatase and tensin homolog (PTEN), which is present in the nucleus or cytoplasm and whose absence has been associated with tumorigenesis [124]. Ndfifip1 protein plays a major role in exosome internalization. Interestingly, the Ndfifip1 protein is repressed in GB; subsequently, the intranuclear concentration of PTEN is also suppressed, allowing tumor cells to proliferate and survive [124].

Radiogenomics is a field that integrates large amounts of quantitative data obtained from medical images with individual genomic phenotypes and uses deep learning techniques to develop a predictive model. This model is useful for stratifying patients, guiding therapeutic strategies, and assessing clinical outcomes [140].

The combination of liquid biopsy and radiomics/radiogenomics can offer a promising avenue for non-invasive disease diagnosis and provide valuable insights for treatment planning (Figure 2). Both approaches have individually shown great potential, but their combination can lead to even more accurate and comprehensive diagnoses. The study of images using artificial intelligence (AI) and machine and deep learning techniques has progressed significantly in recent years. These methods involve an increasingly complex analysis of computational processes and imaging features, with the potential to accurately predict the molecular alterations necessary for correct disease diagnosis. As a result, they represent a promising approach for improving the accuracy and precision of glioma diagnosis [141,142]. In more detail, the field of radiogenomics offers a unique opportunity to predict genetic mutations, the status of molecular markers, and chromosomal aberrations through the analysis of imaging features. This approach utilizes imaging data as a substitute for the presence of genetic alterations, providing a non-invasive and efficient way of diagnosing and monitoring gliomas [143].

Radiogenomics has already been employed in other forms of cancer, such as breast [144] and lung cancer [145]. These studies have demonstrated the potential of radiogenomics to accurately predict the genetic characteristics of tumors through the analysis of imaging data. The extension of this approach to gliomas represents a significant opportunity to improve the accuracy and efficiency of non-invasive diagnosis and monitoring.

In the study of Li Y et al., for example, that used MRI radiomics (machine learning), the authors were able to predict ATRX mutations in low-grade gliomas [146]. In another study, the same mutations were identified in gliomas with deep learning, with 94% sensitivity and 92% specificity [147]. Diffuse midline gliomas with H3F3A histone mutations have greater enhancement in the T2 sequence, while tumors without the mutation have a poor identification of the non-contrast-enhancing tumor (NCET) margin. Furthermore, non-mutated tumors present with sound edema and cortical invasion [148]. A study of radiomics analysis (multiparameter MRI) managed to identify that tumors harboring TERT mutations have more necrotic areas, since TERT mutations suggest a high-grade profile [149].

Moon et al. also identified several characteristics of high-grade gliomas relative to MGMT promoter methylation [150]. When it comes to other brain tumors and, particularly, oligodendrogliomas, which are subcategorized to IDH-mutant and 1p/19q co-deleted [151], studies have been able to predict the 1p/19q codeletion with high specificity and sensitivity using, for example, a textural analysis of the T2 sequence [152,153,154]. It therefore becomes evident that the combination of liquid biopsy with the genomics component of radiogenomics profiling represents a promising approach to revolutionize cancer treatment. This combination may have the potential to facilitate early diagnosis, provide more accurate prognostic assessment, and enable real-time disease monitoring, all while minimizing invasiveness. Furthermore, this personalized approach to non-invasive brain tumor diagnosis could pave the way for precision medicine, tailored to the individual needs of each patient.