2.1.

Multiphoton Imaging System

The multiphoton imaging microscope used in this study is similar to one previously described.²² In short, the $780\text{-}\mathrm{nm}$ output of a femtosecond, titanium-sapphire laser (Tsunami, Spectra Physics, Mountain View, California) pumped by a diode-pumped, solid-state laser (Millennia X, Spectra Physics) was used for excitation. The excitation source was guided toward a modified upright microscope (E800, Nikon, Tokyo, Japan ) by a set of $x\text{-}y$ galvanometer-driven, scanning mirrors (model number 6210, Cambridge Technology, Cambridge, Massachusetts). Upon entering the microscope, the laser was beam-expanded to ensure overfilling of the objective’s back aperture. The expanded laser beam is reflected into the high numerical aperture focusing objective by the main dichroic mirror (435DCXR, Chroma Technology, Rockingham, Vermont). For high-resolution imaging, a high-numerical-aperture, oil-immersion objective (Nikon S Fluor $40\times ∕\mathrm{NA}$ 1.3) was used. Otherwise, an oil-immersion objective (Nikon Plan Fluor $20\times ∕\mathrm{NA}$ 0.75 MI) was used to acquire large-area images for SHG signal quantification. The MAF and SHG signals produced at the focal spot were collected in the epi-illuminated geometry and separated by a secondary dichroic and additional bandpass filters (E435LP 700SP, HQ 390/20, Chroma Technology) for the collection and isolation of broadband autofluorescence $\left(435\text{to}700\mathrm{nm}\right)$ and SHG $\left(380\text{to}400\mathrm{nm}\right)$ signals, respectively. Single-photon counting photomultiplier tubes (R7400P, Hamamatsu, Hamamatsu City, Japan) and home-built discriminators were used for the detection and processing of the signal photons. For the acquisition of large-area, multiphoton images, a sample translation stage (H101, Prior Scientific, Cambridge, UK ) was used to translate the specimen after each small-area optical scan. The individual small images $110\times 110\mu {\mathrm{m}}^2$ in area (S Fluor $40\times$ objective, Nikon) and $220\times 220\mu {\mathrm{m}}^2$ in area (Plan Fluor $20\times$ objective, Nikon) are then assembled to form a large-area image $1980\times 1540\mu {\mathrm{m}}^2$ (S Fluor $40\times$ objective) for qualitative multiphoton imaging and $3300\times 3300\mu {\mathrm{m}}^2$ (Plan Fluor $20\times$ objective) for quantitative SHG measurement. In this manner, high resolution, large-area view, and quantification of the severity of the tissue fibrosis can be achieved.

2.2.

Preparation of Frozen Specimens

In order to demonstrate the imaging capabilities of multiphoton microscopy in label-free imaging of liver fibrosis specimens, frozen tissues were used. One tissue block from each patient was frozen at $-80°\mathrm{C}$ after embedding with optical cutting temperature compound. Each frozen tissue block was sectioned into specimens approximately $10\mu \mathrm{m}$ in thickness. One of two adjacent sections was used for multiphoton imaging and the other sample was stained with Masson’s trichrome labeling for collagen-specific, histological comparison to the multiphoton results. In addition, since the intensity profile of the multiphoton signals was not uniform across the imaged area, multiphoton images of the liver fibrosis specimens were corrected using an image field flatness correction algorithm developed in our laboratory.²³

2.3.

Label-Free Optical Biopsy and Quantification of Liver Fibrosis

Following preparation of the liver fibrosis specimens as described earlier, we performed MAF and SHG imaging of liver fibrosis specimens that have been histologically graded using the METAVIR system (scores 0 to 4), and the results are compared to the histological images from the same patient.

The severity of liver fibrosis is highly correlated with the growth of collagen fibers. Therefore, to quantify liver fibrosis of the imaged specimens, we analyzed the collagen content in the fibrotic tissues. The previous work of Sun mentions that the intensity of SHG signal from aggregated and distributed collagen can be used to examine the extent of liver fibrosis in animal specimens.⁵ In this work, we apply the aggregated and total collagen (aggregated and dispersed collagen) fibers as indicators for grading human liver fibrosis. The ratio of the total collagen area to total specimen area (TC-ratio) and area of aggregated collagen to total specimen area (AC-ratio) are determined.

For the analysis of total collagen area, the pixel numbers of the SHG image [Fig. 1 ] having intensity above the threshold value are counted as the total collagen area. Since the intensity and distribution of aggregated collagen (black arrow) is much higher and denser than those of dispersed collagen (white arrow), we applied a $3\text{-pixel}$ -wide boxcar average to smooth out the dispersed collagen in the SHG images. The mean intensity of one random-selected region without aggregated collagen in the smoothed SHG image was measured as the threshold value. Then, the aggregated collagen area was isolated from the smoothed SHG image by setting this threshold value. The pixel numbers above the threshold value represent the aggregated collagen area.

Fig. 1

(a) SHG imaging shows the distinction between aggregated (black arrow) and dispersed collagen (white arrow). Scale bar is $100\mu \mathrm{m}$ . (b) Relationship between AC-ratios, TC-ratios, and METAVIR grades. The specimen numbers from METAVIR scores of 0 to 4 are 5, 3, 5, 7, and 5, respectively. ${}^{*}$ ; Both AC-ratios and TC-ratios of METAVIR score 0; 1 and 2 have significant difference from those of METAVIR score 4.

3.1.

Optical Biopsy of Liver Fibrosis

The METAVIR scoring of liver fibrosis is determined by the fibrous expansion of the portal tract.¹ The portal tract is a component of the hepatic lobule and is composed of the hepatic artery, hepatic portal vein, bile duct, and lymphatic vessels. Shown in Fig. 2 and Fig. 3 are representative MAF and SHG images of liver fibrosis tissues that show morphological features corresponding to METAVIR grades 0 to 4, respectively. Histological images obtained using Masson’s trichrome stains are also shown for comparison. Since we used frozen sections for multiphoton imaging, histological comparison is made by identifying similar features in Masson’s trichrome–prepared specimens from an adjacent section of the tissue specimen.

Fig. 2

Imaging of METAVIR grade 0 tissue by (a) multiphoton microscopy and (b) histological comparison. The portal tract surrounded by SHG collagen fibers (enclosed region) can be delineated. Magnified images of the selected region of interest under (c) multiphoton and (d) histological examination. The yellow arrow indicates the normal fibrous tissue of the portal tract, and the white arrow indicates the hepatocyte nucleus. Scale bar is $200\mu \mathrm{m}$ . (Color online only.)

Fig. 3

(1a) to (4a) Multiphoton images of METAVIR grades 1 to 4 tissue, respectively. (1b) to (4b) The corresponding histological comparison for the images shown in (a). In the METAVIR grade 1 tissue, fibrotic collagen emanating from the portal tract (enclosed region) starts to appear, while in the METAVIR grade 2 tissue, mild bridging fibrosis is evident. In the METAVIR grade 3 tissue, marked bridging fibrosis (dashed lines) between adjacent portal tracks is visible, and in the METAVIR grade 4 tissue, formation of a cirrhotic nodule is evident. All scale bars are $200\mu \mathrm{m}$ .

Shown in Fig. 2 is the large-area, multiphoton image of a grade 0 METAVIR tissue, and the histological comparison is shown in Fig. 2. The selected region of interest (dashed frame) is shown in Fig. 2 as a magnified view of the portal tract surrounded by second-harmonic generating collagen fibers (yellow arrow). In addition, since hepatocyte nuclei (white arrow) lack autofluorescence, individual hepatocytes can be identified by round regions void of MAF surrounded by autofluorescent cytoplasm (color online only). Similar features are found in the histological image [Fig. 2]. Note that in the Masson’s trichrome–prepared specimen, the blue pseudo color represents the collagen fibers (yellow arrow), purple is the cytoplasm, and dark spots (white arrow) indicate positions of hepatocyte nuclei (color online only). Therefore, for collagen fibrosis identification, imaging with the spectrally separated MAF and SHG signals can provide high-contrast, label-free images comparable to that provided by conventional histological examination. In comparison, a multiphoton image of a METAVIR grade 1 specimen is shown in Fig. 3(1a). Note that the extension of liver fibrosis (yellow arrow) from a portal tract is clearly visible. Similar features can be observed in the histological comparison in Fig. 3(1b); (arrow). For imaging of a METAVIR grade 2 specimen [Fig. 3(2)], the formation of mild fibrotic bridging between two portal tracts, the hallmark of the METAVIR grade 2 pathology, can be easily identified [Fig. 3(2a)]. Once again, histological comparison of the METAVIR grade 2 tissue with the multiphoton results identifies similar morphological features. In multiphoton imaging of a METAVIR grade 3 specimen [Fig. 3(3a)], marked bridging fibrosis showing thick fibrotic bands is evident (dashed line). The same feature can be seen in the histological comparison [Fig. 3(3b)]. Last, for imaging of the representative grade 4 METAVIR tissue [Fig. 3(4)], the multiphoton imaging modality and the histological comparison show the enclosing of fibrotic collagen tissues and the formation of the cirrhotic nodule.

The results of liver fibrosis quantification analysis show that the TC-ratio and AC-ratio increase with the METAVIR scores [Fig. 1]. The relationship between METAVIR grades and the fibrotic collagen content is summarized in Table 1 . Both Fig. 1 and Table 1 show that the growth of fibrotic collagen in human liver increases with the severity of the METAVIR grades and that the increases of AC- and TC-ratios are nonlinear with METAVIR scores. For example, the AC-ratio increased threefold between METAVIR 3 and 0 specimens, while the same ratio increased by 5.71-fold between METAVIR 4 and 0 specimens. The statistical significance of the quantification of liver fibrosis was assessed by a one-way ANOVA analysis of variance and subsequent post hoc test using Bonferroni’s Multiple Comparison Test. With $P<0.05$ considered to be statistically significant, our analysis showed that both AC- and TC-ratios of METAVIR scores 0, 1, and 2 are significantly different from those of METAVIR score 4. Although the AC- and TC-ratios show increasing trends from METAVIR scores 0 to 3, the ratios cannot statistically differentiate between samples with those scores.

Table 1

Relationship between differently graded METAVIR liver fibrosis specimens, TC-ratios, and AC-ratios.