1.

Introduction

In 1998, the fluorescence-guided resection of malignant gliomas using 5-aminolevulinic acid (5-ALA) was reported by Stummer et al. in Germany.¹ In 2006, Stummer et al. reported that fluorescence-guided resection using 5-ALA not only led to an increase in the complete resection rate but also improved progression-free survival.² In addition, the median overall from the time of surgery in patients without residual tumors as a result of complete resection achieved by fluorescence-guided resection with 5-ALA was longer than those in patients with residual contrast-enhancing tumor material.³ It was reported that even small volumes of residual tumor [radiation therapy oncology group recursive partitioning analysis (RTOG-RPA) classes III–V based on age] was associated with a worse prognosis compared with there being no visible and residual contrast-enhancing tumor.⁴ Differences in survival depending on resection status, especially in RPA classes IV and V, support a causal influence of resection on survival. However, fluorescence diagnosis with 5-ALA is very cost effective, whereas there have been problems with false negatives in fluorescence diagnosis; a false negative means that a brain tumor has “unobservable” fluorescence and is hence classified incorrectly as a nontumor.¹

Fluorescence intensity from PpIX observed through the surgical microscope has been graded according to intensity in two or three stages.⁵ On irradiation of violet–blue light to a brain tumor that included PpIX, necrotic tumors usually displayed no or only inhomogeneous red fluorescence. In addition, joining, perfused, and viable tumor tissue was distinguished by its deep red fluorescence, which we call “strong” in this report, whereas normal brain tissue was colored blue. Usually, a transition zone with clearly discernible, but lighter, pink fluorescence was encountered between solidly fluorescing tumor and nonfluorescing blue brain tissue.⁵ We term this fluorescence impression “vague.” “Strong” fluorescence from the region indicates that the region is highly likely to be tumor. In a primary malignant glioblastoma, the tumor probability (positive predictive value) for tumor tissue that shows strong fluorescence is 100%.¹ Moreover, in the detection of primary malignant glioma tissue using 5-ALA, sensitivity and specificity were 85% and 100%, respectively.¹ However, while diagnosis with 5-ALA by fluorescence provided an excellent ability to predict the presence of abnormal and tumor tissue, the mechanism of selective accumulation of PpIX in tumor tissue has not yet been understood. Pathological findings have revealed that strong correlation exists between tumor aggressiveness and corresponding intraoperative fluorescence.⁶ A previous study has indicated that solidly proliferating, central, viable tumors with vascular proliferation are associated with intense red fluorescence, and surrounding infiltrating tumor with moderate to high cell density is associated with pink fluorescence.⁵ Histologically, solidly fluorescing tissue generally consisted of coalescent tumor cells, whereas vaguely fluorescing tissue usually represented an infiltrating tumor of intermediate or low cellular density.⁵ Their analysis of tumor biopsy specimens revealed infiltrating tumor also to be present beyond vaguely fluorescing tissue portions. Takahashi et al. reported that the high mRNA level of CPOX expression was significantly correlated with the phenotype of strong 5-ALA-induced fluorescence.⁷ Elucidation of the cause of false negatives may ameliorate the sensitivity in fluorescence-guided resection for glioblastoma with 5-ALA.

To the best of our knowledge, no reports exist concerning whether the optical properties of false-negative brain tumor tissue differ from those of brain tumor tissue with strong fluorescence, when measured with a double integrating sphere and inverse Monte Carlo methods. If the cell density in the tumor tissue that contains the PpIX increases, the fluorescence intensity of the tumor tissue increases. The purpose of this study is to investigate the effect of optical properties of human brain tumor tissues on the fluorescence intensity. In addition, the cell density of the resected tumor tissue and the tissue that was classified by a neurosurgeon according to observable fluorescence are compared.

2.

Materials and Methods

3.

Results

3.1.

Optical Properties of Human Brain Tumor Tissues with “Strong,” “Vague,” and “Unobservable” Fluorescence, and Normal Human Brain Tissues

The absorption coefficient (${\mu }_{\mathrm{a}}$) spectra of the human brain tumor tissue from 350 to 1000 nm are shown in Fig. 1(a). There was no significant difference between the ${\mu }_{\mathrm{a}}$ values of brain tumor tissues in regions with “strong,” “vague,” and “unobservable” fluorescence, as shown in Fig. 1(a). The ${\mu }_{\mathrm{s}}^{\prime }$ spectra of the brain tumor tissues from 350 to 1000 nm are shown in Fig. 1(c). Again, no statistical difference was observed in the ${\mu }_{\mathrm{s}}^{\prime }$ of the tumor tissues with “strong,” “vague,” and “unobservable” fluorescence.

Fig. 1

Optical properties of human brain tumor tissues obtained from glioblastoma, metastasis, meningioma, and oligodendroglioma (classified into three classes according to their fluorescence), normal brain gray matter ($n=2$), and white matter ($n=2$). (a) Absorption coefficient ${\mu }_{\mathrm{a}}$ spectra, (b) ${\mu }_{\mathrm{a}}$ spectra magnified scale between 650 and 1000 nm, and (c) reduced scattering coefficient ${\mu }_{\mathrm{s}}^{\prime }$ spectra.

3.2.

Optical Properties of Human Brain Tumor Tissues Classified by Histological Types, and Human Brain Normal Tissues

There was no significant difference between the ${\mu }_{\mathrm{a}}$ values of tissue with different tumor types (Figs. 2 and 3). The ${\mu }_{\mathrm{s}}^{\prime }$ values of human brain tumor tissues are shown in Figs. 2 and 3, respectively. The ${\mu }_{\mathrm{s}}^{\prime }$ values of the glioblastoma tissue were significantly higher than those of the meningioma tissue from 508 to 1000 nm. The ${\mu }_{\mathrm{s}}^{\prime }$ values of glioblastoma tissue were significantly higher than those of oligodendroglioma tissue at some wavelengths. There was no significant difference between the ${\mu }_{\mathrm{s}}^{\prime }$ values of glioblastoma and metastasis tissues. The ${\mu }_{\mathrm{s}}^{\prime }$ values of meningioma tissue at some wavelengths were significantly higher than those of oligodendroglioma tissue. Furthermore, there was no significant difference between the ${\mu }_{\mathrm{s}}^{\prime }$ values of meningioma and metastasis tissues.

Fig. 2

Optical properties of human glioblastoma ($n=14$), meningioma ($n=9$), normal brain gray matter ($n=2$), and white matter ($n=2$). (a) Absorption coefficient ${\mu }_{\mathrm{a}}$ spectra, (b) ${\mu }_{\mathrm{a}}$ spectra magnified scale between 650 and 1000 nm, and (c) reduced scattering coefficient ${\mu }_{\mathrm{s}}^{\prime }$ spectra.

Fig. 3

Optical properties of human oligodendroglioma ($n=3$) and metastasis ($n=6$). (a) Absorption coefficient ${\mu }_{\mathrm{a}}$ spectra, (b) ${\mu }_{\mathrm{a}}$ spectra magnified scale between 650 and 1000 nm, and (c) reduced scattering coefficient ${\mu }_{\mathrm{s}}^{\prime }$ spectra.

3.3.

Optical Properties of Human Brain Tumor Tissues Classified Following to Histological Types and Fluorescence Intensities

The optical properties of glioblastoma and meningioma tissues classified according to their fluorescence are shown in Figs. 4 and 5, respectively. There were no significant differences between the ${\mu }_{\mathrm{a}}$ values of glioblastoma tissue region with “strong” and “unobservable” fluorescence in Figs. 4(a) and 4(b). The maximum ${\mu }_{\mathrm{a}}$ values in the absorption band of hemoglobin appeared at 416 and 544 nm in glioblastoma tissue region with “strong” fluorescence. The maximum ${\mu }_{\mathrm{a}}$ values in the absorption band of hemoglobin appeared at 423 and 551 nm in glioblastoma tissue at the region with “unobservable” fluorescence. There were no significant differences between the ${\mu }_{\mathrm{s}}^{\prime }$ values of glioblastoma tissue in the regions with “strong” and “unobservable” fluorescence in Fig. 4(c). The maximum ${\mu }_{\mathrm{a}}$ values in the absorption band of hemoglobin appeared at 423 and 551 nm in meningioma tissue at the regions with “strong” fluorescence [Fig. 5(a)]. The dips observed for ${\mu }_{\mathrm{s}}^{\prime }$ in Figs. 1–5 at 420 nm are probably due to artifacts resulting from the hemoglobin absorption.

Fig. 4

Optical properties of human glioblastoma [classified into three classes according to their fluorescence: strong ($n=8$), vague ($n=2$), and unobservable ($n=4$)]. (a) Absorption coefficient ${\mu }_{\mathrm{a}}$, (b) ${\mu }_{\mathrm{a}}$ spectra magnified scale between 650 and 1000 nm, and (c) reduced scattering coefficient ${\mu }_{\mathrm{s}}^{\prime }$ spectra.

Fig. 5

Optical properties of human meningioma [classified into three classes according to their fluorescence: strong ($n=5$), vague ($n=2$), and unobservable ($n=2$)]. (a) Absorption coefficient ${\mu }_{\mathrm{a}}$ spectra, (b) ${\mu }_{\mathrm{a}}$ spectra magnified scale between 650 and 1000 nm, and (c) reduced scattering coefficient ${\mu }_{\mathrm{s}}^{\prime }$ spectra.

3.4.

Correlation Between Visible Fluorescence and Histology

The sections of human brain tumor tissue, classified as meningioma, metastasis, and unclassified glioma with “strong” and “unobservable” fluorescence were stained with H&E, and histological images are shown in Fig. 6, from which the cell density was estimated. The cell densities [mean ± standard deviation (S.D.)] in regions with “strong” and “unobservable” fluorescence were $31\pm 9\times {10}^2\mathrm{per}{\mathrm{mm}}^2$ and $12\pm 4\times {10}^2\mathrm{per}{\mathrm{mm}}^2$, respectively. Thus, there is a statistically significant difference in the cell densities between the regions with “strong” and “unobservable” fluorescence. Meningioma showed typical whorl formation [Fig. 6(c)]. Some nuclei contained clear holes. Spindle-shaped tumor cells with elongated nuclei, forming parallel or storiform bundles, were observed.

Fig. 6

H&E staining of a tumor section. (a) Regions with “strong” fluorescence and (b) regions with “unobservable” fluorescence in unclassified glioma. (c) Regions of meningioma with “strong” fluorescence and (d) regions of metastasis with “strong” fluorescence. Scale bar, $100\mu \mathrm{m}$.

3.5.

Correlation Between Visible Fluorescence and the Optical Property of Tissue

To investigate whether optical properties affect emissions from glioblastoma tissues, the emission was calculated using Monte Carlo simulation (Fig. 7). The maximum fluorescence strength at the surface and the ratio of relative intensity at the surface are shown in Table 1. Although the probability distributions of emitted fluorescence in the three models were different, the intensities at the surface were almost identical.

Fig. 7

Upper row: (a) left, the distribution of the excitation light in the “strong” fluorescence human brain tumor model; (b) middle, in the “vague” fluorescence human brain tumor model; and (c) right, in the “unobservable” fluorescence human brain tumor model. Lower row: (d)–(f), corresponding images for the fluorescence emission light at 635 nm.

Table 1

Comparison of the maximum emission intensity at the surface of tissues estimated using fluorescence Monte Carlo simulation. The ratio of relative emission intensity is estimated by normalization with the value of the sample classified as “vague.” The ratio of emission to excitation at the tissue surface was estimated. In addition, the 1/(μ1+μ2) of tissues with strong and unobservable fluorescence was compared.

		Strong	Vague	Unobservable	P-value
Maximum emission intensity	[$1/{\mathrm{cm}}^2$]	0.16	0.21	0.16	—
Ratio of relative emission intensity	[—]	0.78	1.00	0.77	—
J_em/J_ex	[$\times {10}^{-4}$]	8.62	4.72	5.92	—
$1/({\mu }_1+{\mu }_2)$		$0.21\pm 0.06$		$0.18\pm 0.08$	0.18

3.6.

Sensitivity of the Optical Properties to Variations of Reflective Index and Anisotropy Factor

In Fig. 8(a), the absorption coefficient and the reduced scattering coefficient spectra are presented for different refractive indices. Figure 8(b) shows the absorption coefficient and the reduced scattering coefficient spectra when the anisotropy factor is changed. The absorption coefficient and the reduced scattering coefficient were not changed significantly by changing the refractive index and anisotropy factor.

Fig. 8

(a) Absorption coefficient spectra and (b) reduced scattering coefficient spectra of tumor tissue, when the refractive index or the anisotropy factor was changed.

4.

Discussion

The ${\mu }_{\mathrm{a}}$ values of glioblastoma tissue in the region with “strong” fluorescence and “unobservable” fluorescence were compared at 405 and 635 nm, and there was no significant difference as can be seen in Fig. 4(a). Our results have been compared with ones reported by other groups (Table 2).^17–19 The ${\mu }_{\mathrm{s}}^{\prime }$ values determined by Gebhart et al.¹⁹ are about 1.5 times lower than the values obtained in the present work. However, in that study, the tissues were stored at 193 K until in-vitro measurement of the optical properties. Roggan et al.²⁰ stated that the ${\mu }_{\mathrm{s}}$ of liver tissues is lowered by shock-freezing at 77 K by $-21.7\%$ in comparison with the untreated ones. Therefore, this might explain why the ${\mu }_{\mathrm{s}}^{\prime }$ values estimated in this study are higher than the values reported by Gebhart et al.¹⁹ Pitzschke et al.²¹ reported that tissue freezing reduces the ${\mu }_{\mathrm{s}}^{\prime }$.

Table 2

Comparison of (a) mean absorption coefficients, (b) mean reduced scattering coefficients, and (c) optical penetration depths of meningioma and brain tumor. Differences to the means of two measurements are calculated.

It is difficult to estimate accurately the experimental error in a study of this type in which many independent measurements are conducted. Experimental contributions to the error included the inaccuracy of the spectrometer, which we estimated to be $<1\%$ of the full-scale value. This error becomes more prominent as the measured values of $R_{\mathrm{d}}$ and $T_{\mathrm{t}}$ become smaller, such as in the wavelength range between 350 and 600 nm. Furthermore, there is a significant biological variability between the samples. For example, if the blood of the samples varied, this would be particularly noticeable between 350 and 600 nm. This may have introduced an error in the ${\mu }_{\mathrm{s}}^{\prime }$ spectra.

Tumors are associated with regions of hypoxia. Solid tumors contain regions at very low oxygen concentrations (hypoxia), often surrounding areas of necrosis.²² Perhaps, tumor tissues at very low oxygen concentration is account for certain curves in ${\mu }_{\mathrm{a}}$, mimicking the absorption curve of deoxyhemoglobin, whereas other curves correspond to the oxyhemoglobin form.

The higher fluorescence intensity is associated with higher cell density. However, while a higher cell density was found in “strong” than in “unobservable” fluorescence tissues, this is not reflected in the ${\mu }_{\mathrm{a}}$ and the ${\mu }_{\mathrm{s}}^{\prime }$ spectra. Perhaps this is caused by the higher ${\mu }_{\mathrm{a}}$ of blood than that of PpIX included in mitochondria; thus the ${\mu }_{\mathrm{a}}$ of blood is prominent in the ${\mu }_{\mathrm{a}}$ of the brain tumor tissue. In addition, the ${\mu }_{\mathrm{s}}^{\prime }$ is not linearly related with the concentration of scattering molecules for more strongly scattering materials, such as brain tissue. This leads to there being no difference between the ${\mu }_{\mathrm{s}}^{\prime }\mathrm{s}$ of “strong” and “unobservable” brain tumor tissue. Thus, the difference of cell density modified the scattering coefficient yet it does not lead to significant differences in the ${\mu }_{\mathrm{s}}^{\prime }$ and thus does not affect the propagation of the diffuse fluorescent light.

Diaz et al.²³ investigated whether fluorescein had the ability to specifically stain glioma cells in a study of fluorescein uptake by intracranial U87 malignant glioma xenografts in male nonobese diabetic/severe combined immunodeficient mice. In that report, false negatives could be caused by lack of fluorescence in areas of diffuse, low-density cellular infiltration. The results of our study are in agreement with this, as shown in Sec. 3.4. In addition, Bogaards et al.²⁴ reported that in the majority of animals, complete tumor resection was not achieved owing to small tumor cell nests in the surgical cavity, which were undetectable by fluorescence of PpIX. These could be considered as false negatives for the particular device, 5-ALA dose, and administration time. Moreover, in cases where the tumor invaded beyond the surgical cavity, this would not be detectable by fluorescence imaging of the cavity with fluorescent dye as the effective tissue sampling depth of the blue-light excitation below the exposed tissue surface is $<0.5\mathrm{mm}$.²⁵ These reports suggest the lack of fluorescent dye and/or the decline of the excitation energy in the tumor tissue may have the potential to cause false negatives in fluorescence-guided resection. The optical penetration depth ($\delta$) is one of the indices of light attenuation in biological tissue. Thus, the $\delta$ values of human brain tumor tissues were calculated from the optical properties shown in Figs. 1–5 and compared. Two equations were used to calculate the optical penetration depth ($\delta$). When ${\mu }_{\mathrm{a}}\ll 3{\mu }_{\mathrm{s}}^{\prime }$, the $\delta$ can be estimated as²⁶

When ${\mu }_{\mathrm{a}}$ is comparable with ${\mu }_{\mathrm{s}}^{\prime }$ ($10{\mu }_{\mathrm{a}}\ge 3{\mu }_{\mathrm{s}}^{\prime }$), $\delta$ was estimated as²⁷

The $\delta$ depends on the wavelength, as shown in Figs. 9Fig. 10–11. The $\delta \mathrm{s}$ at 635 nm in brain tumor tissues at the regions with “strong,” “vague,” and “unobservable” fluorescence were about 0.98, 0.97, and 0.78 mm, respectively, as shown in Fig. 9. Moreover, to confirm whether there is a difference in $\delta$ in glioblastoma tissues between the regions with “strong” and “unobservable” fluorescence, the $\delta$ values of glioblastoma tissues were compared [Fig. 11(a)]. The $\delta \mathrm{s}$ at 405 nm in glioblastoma tissues with “strong” and “unobservable” fluorescence were about $0.28\pm 0.08\mathrm{mm}$ and $0.20\pm 0.05\mathrm{mm}$, respectively, and thus are comparable within the error limits. Assuming the attenuation coefficient to be ${\mu }_1=1/{\delta }_1$ at 405 nm and ${\mu }_2=1/{\delta }_2$ at 635 nm, the detected fluorescence intensity under a uniform distribution of PpIX of density [PpIX] under pencil illumination is proportional to: $[\mathrm{PpIX}]/({\mu }_1+{\mu }_2)$. “$1/({\mu }_1+{\mu }_2)$” and $p$-values comparing “strong” and “unobservable” fluorescence tissue are shown in Table 1. They were comparable within the error limits. From the result of the fluorescence Monte Carlo simulation (Fig. 7), the ratio of the relative intensity of emission at the surface was almost identical (Table 1). The measured approximate fluorescence intensity ratio of “strong,” “vague,” and “unobservable” was 2.51:0.49:0.1 by spectroscopy;⁸ this explains that this difference is caused by the difference of the concentration of fluorescence, or PpIX.

Fig. 9

Optical penetration depths $\delta$ of human brain tumor tissues obtained from glioblastoma, metastasis, meningioma, and oligodendroglioma (classified into three classes according to their fluorescence), normal brain gray matter ($n=2$), and white matter ($n=2$).

Fig. 10

Optical penetration depth $\delta$ spectra of human glioblastoma ($n=14$), meningioma ($n=9$), oligodendroglioma ($n=3$), metastasis ($n=6$), normal brain gray matter ($n=2$), and white matter ($n=2$).

Fig. 11

Optical penetration depth $\delta$ spectra of (a) human glioblastoma [classified into three classes according to their fluorescence: strong ($n=8$), vague ($n=2$), and unobservable ($n=4$)], and (b) human meningioma [classified into three classes according to their fluorescence: strong ($n=5$), vague ($n=2$), and unobservable ($n=2$)].

If in-vivo measurement shows that tissue has higher ${\mu }_{\mathrm{a}}$ than in this ex-vivo study, the $\delta$ would become lower and the detectable fluorescence would be mainly from PpIX included in the tissue in the submillimeter range from the surface. Thus, the effect of optical properties on the false negatives would become smaller or zero as ${\mu }_{\mathrm{a}}$ increases.

In many cases, optical properties change the wavelength-dependent manner and the corresponding excitation and emission intensity changes. However, this situation is restricted to when $\delta$ is high, or incident light can easily penetrate tissues and the intensity of excitation light becomes high and results in the emission light becoming easy to detect. In 2017, Wirth et al.²⁸ reported that the depth of fluorescence was estimated under a bulk phantom, having ${\mu }_{\mathrm{s}}^{\prime }$ of $1.0{\mathrm{mm}}^{-1}$ and ${\mu }_{\mathrm{a}}$ of $0.001{\mathrm{mm}}^{-1}$. Although the results by Wirth et al. suggest that the depth of fluorescence was correctly estimated for shallower inclusions, the error of the predicted depth drastically increased for inclusions at depths greater than $\delta$, even when using a correction based on optical properties. With regard to the optical properties of the human glioblastoma tissues, the $\delta$ was 0.2 or 0.3 mm at 405 nm. Whether we can recognize, or detect, the PpIX fluorescence emitted from the glioblastoma tissue does not mainly depend on the optical properties but is more strongly affected by the quantity of PpIX present within 0.2- to 0.3-mm depth, or the $\delta$, from the surface when PpIX is excited by the light at 405 nm. Hence, the small amount of PpIX that is contained in the tissue within the depth of optical penetration more strongly affects the false negatives under the conditions that the tissue has low $\delta$.

Technical advances^29–31 have increased the sensitivity of tumor tissue detection using 5-ALA-PpIX, but so far, the sensitivity has not reached 100%.^32,33 The correction of optical properties for the detection of PpIX validated with a phantom was found to be suitable for quantitative fluorescence spectroscopy and imaging.^28,34 If there is a region with high $\delta$ in brain tumor tissue caused by local variation, the method of correction using the optical properties would work effectively in quantitative fluorescence spectroscopy and imaging. In vivo measurement of PpIX content in false-negative tissue, such as reported by Valdés et al.,³⁵ it is also important to reveal the correlation in each patient between the presence of tumor tissue and the corresponding local in-vivo optical properties.

Stummer et al.³⁶ reported that “strong” visual fluorescence in individual samples was related to high cell densities and strong fluorescence, whereas samples designated as “weak” by study surgeons had low cell densities and low spectrometric fluorescence. The mRNA level of ABC transporter ABCG2 was lower in malignant glioma cells in the brain tumor that exhibited strong fluorescence of PpIX after 5-ALA treatment, whereas the surrounding cells emitted “vague” fluorescence.³⁷ ABCG2, a porphyrin efflux pump, is downregulated in tumors. These reports and our results suggest that the primary cause of false negatives may be a lack of PpIX or a low accumulation of PpIX in malignant glioblastoma.

5.

Conclusion

In human glioblastoma tissue obtained from fluorescence-guided resection with 5-ALA for malignant tumors, a significant difference was not observed between the optical properties of the glioblastoma regions with “strong” and “unobservable” fluorescence. However, the variation in hemoglobin contents caused by the use of ex-vivo measurement would affect these results. The ${\mu }_{\mathrm{a}}$ values of glioblastoma tissue in the region with “strong” fluorescence and “unobservable” fluorescence were compared at 405 and 635 nm, and there was no significant difference. There were no significant differences between the ${\mu }_{\mathrm{s}}^{\prime }$ values of glioblastoma tissues in “strong” and “unobservable” fluorescence at 405 and 635 nm. The fluorescence signal strongly correlated with the tumor cell density. Although a higher cell density was found in “strong” and “unobservable” in fluorescence tissues, this is not reflected in the ${\mu }_{\mathrm{a}}$ and the ${\mu }_{\mathrm{s}}^{\prime }$ spectra. In addition, from the result of the fluorescence Monte Carlo simulation, the ratio of the relative intensity of emission at the surface was almost identical. The $\delta \mathrm{s}$ at 405 nm in glioblastoma tissues with “strong” and “unobservable” fluorescence were about $0.28\pm 0.08\mathrm{mm}$ and $0.20\pm 0.05\mathrm{mm}$, respectively. The small amount of PpIX that is contained in the tissue within the depth of optical penetration more strongly affects the false negatives under the conditions that the tissue has low $\delta$. Hence, it was considered that there was no correlation between fluorescence signals and the patient-averaged ex-vivo optical properties of the malignant glioma tissues. Thus, a false negative in using 5-ALA for detection of human glioblastoma does not result from the differences in optical properties of human brain glioblastoma tissues.