Imaging biomarkers of intracranial meningioma consistency: a scoping review of conventional MRI, advanced MRI, radiomics and artificial intelligence techniques

Preoperative prediction of intracranial meningioma consistency may improve surgical planning by anticipating the ease of internal debulking, operative difficulty, and the feasibility of minimally invasive approaches. Numerous imaging biomarkers have been proposed, ranging from routine MRI measurements to advanced quantitative techniques and artificial intelligence-based models. This scoping review aimed to critically evaluate their diagnostic performance, methodological quality, biological rationale, and potential for clinical implementation. PubMed, Embase, Web of Science, and the Cochrane Library were searched from database inception to July 2026 in accordance with PRISMA-ScR guidelines. Original human studies assessing MRI-derived biomarkers for the preoperative prediction of meningioma consistency were eligible. Studies were grouped into routine MRI, advanced MRI, and artificial intelligence-based approaches. Methodological quality was assessed using QUADAS-2, the Newcastle–Ottawa Scale, and the Radiomics Quality Score, as appropriate. Owing to substantial methodological heterogeneity, findings were synthesized narratively. Thirty-five publications were included: 14 evaluated routine MRI biomarkers, 13 advanced MRI techniques, and 8 radiomics or artificial intelligence approaches. Among routine sequences, T2-weighted signal intensity ratios, particularly the tumor-to-cerebellar peduncle T2-weighted intensity ratio, showed the most consistent association with softer tumors and was the most readily deployable in routine practice. Diffusion-derived parameters provided complementary but less reproducible information. Magnetic resonance elastography demonstrated the strongest evidence among advanced techniques by directly quantifying tissue stiffness and correlating with intraoperative consistency and histopathological features. Radiomics and deep-learning models reported high apparent predictive performance, frequently with AUCs above 0.90, but most studies were retrospective, single-center, internally validated, and in some cases based on overlapping patient cohorts. Routine T2-based biomarkers currently offer the best balance between accessibility and predictive value. Magnetic resonance elastography provides greater biomechanical specificity, whereas artificial intelligence approaches remain promising but insufficiently validated for routine use. Standardized reference standards, multicenter external validation, and stronger biological correlation are required before widespread clinical implementation.