Real-time Brain Tumor imaging with endogenous fluorophores: a diagnosis proof-of-concept study on fresh human samples

Abstract

The primary line of therapy for high-grade brain tumor is surgical resection, however, identifying tumor margins in vivo remains a major challenge. Despite the progress in computer-assisted imaging techniques, biopsy analysis remains the standard diagnostic tool when it comes to delineating tumor margins. Our group aims to answer this challenge by exploiting optical imaging of endogenous fluorescence in order to provide a reliable and reproducible diagnosis close to neuropathology. In this study, we first establish the ability of two-photon microscopy (TPM) to discriminate normal brain tissue from glioblastomas and brain metastasis using the endogenous fluorescence response of fresh human brain sample. Two-photon fluorescence images were compared to gold standard neuropathology. "Blind" diagnosis realized by a neuropathologist on a group of TPM images show a good sensitivity, 100%, and specificity, 50% to discriminate non tumoral brain tissue versus glioblastoma or brain metastasis. Quantitative analysis on spectral and fluorescence lifetime measurements resulted in building a scoring system to discriminate brain tissue samples.

Figure 1

Comparative H&E (a–c) and TPM images (d–f). (a,d) Control sample, stars: neurons, arrow: brain vessels in the SHG, scale bar 100 microns. (b,e) Brain carcinoma metastasis: scale bar 100 microns, stars: tumor cells, arrow: dense vascularization forming a dense network around the tumor cells, (c,f) GBM sample scale bar 40 microns: stars: proliferative endothelial cells, arrow: zoom on a proliferating vessel.

Figure 2

“Blind” histological analysis on TPM images. (a) Flowchart summarizing the results of the pathologists’ diagnosis of tumoral nature based on TPM images, sensitivity (Se), specificity (Sp) and accuracy (Acc) of this method were calculated. (b) Flowchart summarizing results of the pathologists’ discriminating GBM from brain metastasis. The sensitivity (Se), specificity (Sp) and accuracy (Acc) were also calculated.

Figure 3

Analysis of emitted fluorescence for the different groups. (a–c) Topological representation of the emitted fluorescence spectra at different excitation wavelengths for the control (a,d,g), the metastasis (b,e,h) and the GBM (c,f,i) groups. (d–f) Represent colormaps of the Emission-Excitation matrix for each type of tissue. The ideal excitation wavelength is highlighted by a red line and the part of the map corresponding to the SHG signal is identified by a green arrow. TPEF images (g–i) of the selected region is also shown with a scale bar of 100 microns.

Figure 4

Spectral decomposition and comparison of emitted spectra at 890 nm for each group. (a–c) Example of fitted spectra for control (a), metastasis (b), and GBM (c), under 890 nm excitation. (d) Mean spectra and standard deviations determined from 25 fresh human samples (10 metastasis samples, 8 GBM samples and 7 control samples) along with representative images of TPEF (red) and SHG (green) corresponding to each tissue group.

Figure 5

FLIM representation at 890 nm for each tissue type. (a–c) FLIM images of control sample (a), metastasis sample (b), and GBM (c). Scale bars of 100 μm, (d–f) give the color scale of FLIM imaging with the histogram of the average lifetime; shorter lifetimes (blue) are on the left side of the color scale, and longer lifetime (red) on the right side, two measurements were taken and the full width at half maximum of distribution and the average lifetime of the maximum. SHG is not a fluorescent but diffusing process, appearing as very short lifetime corresponding to the instrument response function (IRF), which is of the order of 0.06 ns and shown in blue in the FLIM images. The typical vascular structures of each tissue are consequently as recognizable as in the TPEF images.

Figure 6

(a) 3D scatter plot of three quantitative tissue indicators: the redox ratio, fluorescence lifetime, and SHG signal averaged for each tissue subgroup with a Gaussian ellipsoid fit, (b) box plot of the average lifetimes, (c) bar graph of the redox ratios for each tissue type with the errors corresponding to the standard deviations across all measurements, and (d) overlaid TPEF and SHG intensity images.