Abstract Glioblastomas are heterogeneous, primary brain tumors hiding several sub-populations. Patient-derived xenografts are considered gold-standards to study glioblastoma invasion. However, they present many disadvantages, including time consumption, complex standardization, high cost. To counteract these issues and rapidly identify the most invasive sub-populations, we developed an in vivo mimicry platform named SP2G (SPheroid SPreading on Grids). Live imaging of tumor-derived spheroids spreading on gridded micro patterns mimicking the brain vasculature recapitulated 3D motility features observed in brain or 3D matrices. Using patient-derived samples coupled with a semi-automated macro suite, SP2G easily characterized and sorted differences in cell migration and motility modes. Moreover, SP2G exposed the hidden intra-patient heterogeneity in cell motility that correlated molecularly to specific integrins. Thus, SP2G constitute a versatile and potentially pan-cancer workflow to identify the diverse invasive tumor sub-populations in patient-derived specimens. SP2G includes an integrative tool, available as open-source Fiji macro suite, for therapeutic evaluations at single patient level. Teaser Cracking the inter and intra-patient diversity in Glioblastoma migration profiles