A PROSPECTIVE MULTI-CENTERED, ASSESSOR BLINDED INTERIM STUDY FOR BRAIN TUMOR DIAGNOSIS BY USING CONFOCAL FLUORESCENCE ENDOMICROSCOPY (CCELL - EX VIVO)

초록

BACKGROUND To address an intraoperative brain tumor diagnosis by frozen section(FS) which is time-consuming and traumatic, confocal laser endomicroscopy (CLE) has actively been explored as a promising alternative tool. In this study, we aimed to demonstrate that novel CLE image evaluation by a neuropathologist is as effective as frozen section diagnosis for histopathological assessment during brain tumor surgery. MATERIAL AND METHODS Utilizing a novel confocal endomicroscopy (cCeLL) wherein tissues were pretreated with indocyanine green, a prospective multicenter, assessor-blinded interim analysis was conducted at three tertiary medical institutions in the Republic of Korea. Patients newly diagnosed with brain tumors between May 2023 and April 2024 were included. Each tissue sample was divided into three smaller pieces for permanent section analysis, frozen section analysis and cCeLL-Ex vivo imaging. A designated neuropathologist interpreted all cCeLL-Ex vivo images. Statistic evaluation was performed by comparing both frozen section analysis and cCeLL-Ex vivo with permanent H&E diagnosis. The total time from sample preparation to diagnosis was also calculated. RESULTS A total of 274 samples were acquired from 242 patients. The predominant tumor types were meningioma (32.1%), glioma (24.8%) and pituitary adenoma (9.49%). The majority of tissue samples (86.9%) were obtained from the tumor core. Non-diagnostic tissues constituted 4.38% in frozen section and 6.57% in cCeLL-Ex vivo imaging. The diagnostic accuracy of cCeLL-Ex vivo was on par with frozen section analysis, with 96.1% and 96.9%, respectively. Both methods demonstrated comparable diagnostic sensitivity (FS vs cCeLL-Ex vivo: 99.2% vs 98.3%) and specificity(FS vs cCeLL-Ex vivo: 76.0% vs 77.8%). There were no statistically significant differences in diagnostic accuracy between frozen section (FS) and cCeLL imaging for each tumor type. The average number of cCeLL images required for neuropathologic assessment was 15.84, varing based on the tumor type, with a range of 5.05 to 27.8. The mean duration from sample preparation to diagnosis was 8 minutes and 40 seconds in cCeLL-Ex vivo and 24 minutes and 35 seconds in frozen section (p-value < 0.001). CONCLUSION cCeLL-Ex vivo has shown promising tool as an alternative intraoperative brain tumor diagnosis compared to frozen section analysis. Future studies are anticipated to further validate its clinical utility and provide additional evidence supporting the potential of cCeLL imaging in clinical practice.