Meningiomas are the most common primary central nervous system (CNS) tumors with an incidence of 8.6/100,000,1,2 accounting for 39.0% of all primary CNS tumors3 and 53.4% of all benign CNS tumors.4 The World Health Organization (WHO) classification of CNS tumors is a recognized tool for predicting prognosis of meningiomas, which are considered heterogeneous tumors and can be classified into 3 grades and 15 pathological subtypes.3 According to the European Association of Neuro-Oncology guidelines for the treatment of meningiomas, maximum safe resection of the tumor is the preferred treatment option for patients with meningiomas.5 Tumor heterogeneity determines the different biological behavior of meningiomas of different pathological grades or histological subtypes, leading to significant differences in their surgical treatment options and prognosis.6 Progression/recurrence (P/R) is defined as progressive enlargement of residual tumor with a 10% threshold increase in tumor volume compared with postoperative brain magnetic resonance imaging (MRI) (Figure 1),7,8 Otherwise, it is diagnosed as non-P/R (Figure 2). Although >80% of meningiomas are benign, they still exhibit aggressive behavior in some cases, with 7%–25% recurrence rate in WHO grade 1 tumors.4,9,10 A minority of tumors are atypical, with increased aggressiveness and recurrence rates of 29%–59% and 60%–94% for WHO grade 2/3 meningiomas, respectively.11, 12, 13 Several studies have found that the extent of surgical resection is a key factor influencing postoperative recurrence and overall survival of meningiomas.4,14, 15, 16 Therefore, preoperative noninvasive analysis of the biological behavior of meningiomas is beneficial for risk stratification and prediction P/R, providing a theoretical basis for the selection of clinical decision options.

At present, biopsy is the gold standard for determining meningioma P/R, but it is an invasive procedure that can be more stressful and painful for the patient. P/R can also be determined by other noninvasive methods, such as computed tomography and MRI.17,18 MRI provides essential information on tumor size, location, and adjacent brain tissue invasion.19 Contrast-enhanced T1-weighted imaging mainly shows the blood supply and blood-brain barrier disruption of meningiomas, whereas T2-weighted imaging is usually more sensitive to peritumoral edema and shows the tumor-brain interface. Functional MRI such as diffusion-weighted imaging (DWI) can provide information such as tumor cell density and physiological metabolism noninvasively.20,21 A previous study demonstrated that some MRI features are closely associated with meningioma recurrence, such as peritumoral edema, tumor diameter >5 cm, skull base meningioma, and irregular shape, which usually suggest aggressive tumor growth, and further suggested that MRI can be used as an alternative method to characterize tumor heterogeneity.22 Currently, in addition to these semantic features associated with meningioma recurrence, several machine learning (ML) methods have been shown to identify high-risk atypical meningiomas. ML constructs and validates the performance of the prediction model by extracting the phenotype of the tumor as well as the features of the intrinsic heterogeneity of the tumor, ultimately enabling individualized and precise patient treatment. In this review, we discuss the current state of research in meningioma P/R based on MRI, highlight the advantages and disadvantages of ML in meningioma P/R, and propose future research directions.