脳腫瘍における共焦点レーザー内視鏡検査の診断価値：盲検下・未研修の病理医による実施

Achieving maximal and safe tumour resection is a key goal in brain tumour surgery. Confocal laser endomicroscopy (CLE) enables real-time visualisation of the tissue microstructure at a cellular level, potentially helping neurosurgeons distinguish non-neoplastic from neoplastic tissue. The core aim of this study was to determine the baseline diagnostic accuracy that can be achieved with CLE alone, when assessed by neuropathologists without prior CLE training and without any additional clinical or contextual information, and to compare these findings to standard haematoxylin and eosin (H & E)-based histology in a blinded setting.

CLE images and corresponding H & E-stained slides from 100 brain tumour patients treated at the University Hospital Bern over a 22-month period were analysed. Five blinded neuropathologists with no prior CLE experience independently evaluated the data sets.

Based on CLE images, neuropathologists differentiated neoplastic from non-neoplastic tissue in 70.7%. The specific tumour type was correctly identified in 47.5%: gliomas in 59.8%, meningiomas in 43.8%, and metastases in 25.7%. In contrast, H & E slides were correctly classified as neoplastic in 87.6%, with 89.1% tumour-type-level accuracy (gliomas 85.6%, meningiomas 94.6%, metastases 88.2%). Confidence levels for CLE diagnoses were generally low, and no learning curve was observed.

CLE shows potential to distinguish between neoplastic and non-neoplastic tissue, but diagnostic accuracy remains lower than with H & E-stained slides. Training in CLE image interpretation is recommended to improve diagnostic accuracy. CLE imaging may have the potential to become a valuable tool for delineating brain tumour borders.