Nimg-35. radio-pathomic maps of tumor probability identify glioblastoma invasion beyond 5-ala guided resection margins that eventually exhibit tumor recurrence

Current standard of care for glioblastoma includes surgery followed by chemotherapy and radiation, with extent of resection strongly correlated with survival. There is a need to improve the detection of tumor invasion beyond traditional imaging methods, as the full extent of tumor is known to be underestimated. Fluorescence-guided surgery with 5-Aminolevulinic acid (5-ALA) allows surgeons to visualize tumor cells beyond what is seen within traditional MR imaging margins. In this study, we compared tumor probability maps (TPMs) that detect tumor outside of contrast enhancement (CE) to the 5-ALA-guided resection cavity in glioblastoma patients. 3 pathologically confirmed glioblastoma patients that followed standard of care and underwent 5-ALA surgery were included. Patient 1 was a 56-year-old male with radiographic recurrence at 26 days, patient 2 was a 61-year-old male with radiographic recurrence at 130 days, and patient 3 was a 72-year-old male with cystic recurrence at 84 days, all post 5-ALA surgery. A previously published algorithm, using T1, T1C, FLAIR, and ADC as input and autopsy tissue as ground truth, was used to generate TPMs. Pre-surgical TPMs were qualitatively compared to pre-surgical CE, post-5-ALA surgical resection cavities, and radiographic recurrence seen as CE. In patients 1 and 2, TPMs identified areas suspicious of tumor that was not resected and later exhibited radiographic recurrence. Patient 3 had cystic recurrence only identified on post-resection T1C. In all patients, the extent of 5-ALA resection was greater than CE volume, but smaller than TPM identified tumor. These findings suggest that there was residual tumor that was unidentified via 5-ALA and traditional imaging but was identified by the TPMs. Future studies could utilize these maps in conjunction with 5-ALA to maximize surgical resections and prolong overall survival.