2.1.

Study Design

The goals of our study were to expand the applicability of MPM as a diagnostic tool through development of a new methodological approach and to apply that approach in the optical diagnosis of pancreatic neoplasms. To accomplish the first goal, we determined whether the structure of normal pancreatic tissue can be identified in MPM images of unstained histological sections as clearly as in the corresponding H&E-stained samples. The second goal of the study was to test the applicability of the new methodology in a diagnostic setting, probing whether multiphoton imaging of unstained samples can optically distinguish between three types of pancreatic neoplasms as well as H&E-stained tissue samples and lead to an accurate diagnosis.

2.2.

Sample Preparation

In our study, pancreatic neoplasms embedded in paraffin were provided by the First Affiliated Hospital of Fujian Medical University (Fuzhou, China). Samples included 25 cases of pancreatic neoplasm (9 pancreatic mucinous cystic neoplasm, 10 pancreatic ductal adenocarcinoma, and 6 pancreatic colloid carcinoma), as well as 10 normal pancreatic tissue samples excised 2 cm beyond the resection margins. The samples were collected from patients undergoing a pancreatectomy between July 2008 and August 2012, and postsurgical sample processing involved fixation in formalin, dehydration by alcohol, and embedding in a block of paraffin wax for preservation. Prior to MPM imaging, four sections of $5\text{-}\mu \mathrm{m}$ thickness were sliced from each paraffin block sample by a microtome. Each section was deparaffinized with xylene and rehydrated with a series of alcohol washes, and the processed samples were sandwiched between a microscope slide and cover glass for MPM imaging. After MPM imaging, the sections were stained with H&E stains according to standard procedures for histological examination with standard light microscopy.²⁴ MPM images and H&E-stained images were compared and analyzed by two certified pathologists.

2.3.

MPM Imaging Instrumentation

The multiphoton imaging system used in this study contained a high-throughput scanning inverted Axiovert 200 microscope (LSM 510 META; Zeiss, Oberkochen, Baden-Württemberg, Germany) and a mode-locked femtosecond Ti:sapphire laser (110 fs, 76 MHz) tunable from 700 to 980 nm (Mira 900-F; Coherent, Santa Clara, California).^23,25 For high-resolution imaging, a Plan-Neofluar objective ($40\times$ and $\mathrm{NA}=0.75$, Zeiss) was employed in MPM examination. The excitation wavelength (${\lambda }_{ex}$) used in this study was 810 nm.

Previously, we found that the histological sample processing enhanced the TPEF signal of cytoplasm and nucleoli in epithelial cells and collagen in stromal cells, but destroyed the noncentrosymmetric molecular structure of collagen resulting in loss of the intrinsic second-harmonic generation signal.²³ In addition, we found that the sample processing also led to the loss of NAD(P)H and FAD, which destroyed the TPEF signals that can be used as standard biomarkers of epithelium and whose fluorescence ratio (redox ratio) was indicative of the cellular metabolic state.

To account for these alterations in emission signals caused by histological sample processing, the image acquisition channel 3 under single-track channel mode setting was used to collect multiphoton emission signals in this study. In the optical path, a main dichroic beam splitter (KP650) was set in order to separate the emissions from the excitation light, and BG-39 glass was selected as second dichroic beam splitter in order to block any stray excitation light. The emission light was detected in the wavelength range between 350 and 710 nm by a photon counting photomultiplier tube module. Large-area images were obtained by using an optional HRZ 200 fine-focusing stage (HRZ 200; Carl Zeiss, Inc., Oberkochen, Baden-Württemberg, Germany). All images had a 12-bit pixel depth. The images were obtained at $2.56\text{-}\mu \mathrm{s}/\text{pixel}$.

The light microscope used for imaging H&E-stained sections was a standard bright field light microscope (Eclipse Ci-L, Nikon Instruments Inc., Tokyo, Japan) with a CCD camera (Nikon, DS-Fi2, Japan). An objective (Plan-fluor $40\times$, Nikon) was used in histological examination.

2.4.

Quantification of Morphological Features

To quantitatively assess the differences in morphological features between normal and three types of pancreatic neoplastic tissues, the N/C ratio was defined as the measurement of the size of the nucleus in relation to the area of the cytoplasm within cells of each tissue. Five random fields in each MPM image were selected, and cells within each field were quantified. Cells with cytoplasm and nuclei demonstrating complete and clearly discernable outlines were chosen, circled, and measured using an optional fine-focusing stage (HRZ 200 stage, Carl Zeiss). N/C ratios were expressed as the mean value followed by its standard deviation ($\text{mean}\pm \mathrm{SD}$). In addition, the N/C ratios were measured again following H&E staining to verify the reliability of the MPM technique.

When quantifying the morphological features of pancreatic epithelium, the localization and function of each cell were taken into account so that the comparisons would be made between similar cells. For example, acinar epithelial cells and ductal epithelial cells had striking morphological differences due to their unique functions within the pancreas.²⁶ Pancreatic colloid carcinoma was a kind of acinar-derived cancer, and pancreatic mucinous cystic neoplasm or pancreatic ductal adenocarcinoma was a kind of ductal-derived cancer. In this study, the colloid-producing epithelial cells in pancreatic colloid carcinoma (acinar-derived) were compared with acinar epithelial cells from a normal pancreas. The mucin-producing epithelial cells in pancreatic mucinous cystic neoplasm (ductal-derived) and cancerous ductal epithelial cells in pancreatic ductal adenocarcinoma (ductal-derived) were compared with ductal epithelial cells from a normal pancreas.

Datasets were analyzed using the Shapiroe–Wilk test, and results followed a Gaussian distribution when the $P$-value was greater than 0.05. The one-way ANOVA and least-significant difference analyses were used to determine statistically significant differences in the ratios of N/C between normal and neoplastic tissues and between benign and malignant tumors. Statistical analysis was performed using SPSS software (version 16.0). Exact $P$-values were computed, and differences were considered to be statistically significant when the $P$-values were less than 0.05.

3.1.

Patient Demographics and Specimen Characteristics

Normal pancreatic tissues from 10 patients were analyzed and compared with pancreatic neoplasms from 25 patients in this pilot study. The patient demographics and neoplastic characteristics were summarized in Table 1.

Table 1

Patient demographics and tissue characteristics.

3.2.

MPM Diagnostic Features of Normal Pancreas

In total, 36 sections of pancreatic mucinous cystic neoplasm, 40 sections of pancreatic ductal adenocarcinoma, 24 sections of pancreatic colloid carcinoma, and 40 sections of normal pancreatic samples were examined using MPM by detecting the peak multiphoton fluorescence intensity. A representative MPM image of the normal tissue was shown in Fig. 1(a), revealing regular tissue architecture and cell morphology. It clearly presented the typical arrangement of normal architecture of pancreatic acinus [dashed line in Fig. 1(a)] and pancreatic duct [solid line in Fig. 1(a)]. Concentric fibrous bundles were observed around the ducts with occasionally short and thick collagen fibers [pentagram in Fig. 1(a)].²⁷ The acinar cells [dashed arrow in Fig. 1(a)] were uniform and maintained their polarity with no evidence of nuclear overlapping. Also, the bright nucleolus was easily identifiable within the acinar cell. The interlobular duct [dotted line in Fig. 1(a)] was surrounded with fibrous connective tissue, and cuboidal epithelium constituted the ductal walls. These observations of cellular architecture were comparable with the corresponding H&E-stained image shown in Fig. 1(b), where the positions of the pancreatic acinus [dashed line in Fig. 1(b)], pancreatic duct [solid line in Fig. 1(b)], collagen fibers [pentagram in Fig. 1(b)], acinar cells [dashed arrow in Fig. 1(b)], and interlobular duct [dotted line in Fig. 1(b)] were identical to the corresponding positions in the MPM images.

Fig. 1

Comparison of MPM and H&E images in normal pancreatic tissue. (a) The representative MPM image ($40\times$) of an unstained section clearly reveals the normal tissue architecture and cell morphology, including typical organization of pancreatic acinus (dashed line) and pancreatic duct (solid line), the interlobular duct (dotted line), collagen fibers (pentagram), and the acinar cell (dashed arrow). (b) The corresponding H&E image ($40\times$) of fixed, stained pancreatic tissue shows a similar cell arrangement: the pancreatic acinus (dashed line), pancreatic duct (solid line), interlobular duct (dotted line), collagen fibers (pentagram), and the acinar cell (dashed arrow).

3.3.

MPM Diagnostic Features of Pancreatic Benign Tumor

The representative large-area MPM image of the pancreatic mucinous cystic neoplasm in Fig. 2(a) showed that the lesion was mainly composed of a single layer of columnar, mucin-producing epithelium [dashed arrow in Fig. 2(a)]. The epithelium was supported by ovarian-type stroma [triangle in Fig. 2(a)]. Occasionally, the dense ovarian-type stroma was replaced by a dense collagenized stroma [pentagram in Fig. 2(a)].²⁸ The structure of the lesion was rather disordered, while there were blood vessels [dotted arrow in Fig. 2(a)] distributed within the lesion tissue. It was comparable with the corresponding H&E image shown in Fig. 2(b), presenting mucin-producing epithelium [dashed arrow in Fig. 2(b)], ovarian-type stroma [triangle in Fig. 2(b)], collagenized stroma [pentagram in Fig. 2(b)], and blood vessels [dotted arrow in Fig. 2(b)] in identical positions as the MPM image.

Fig. 2

Comparison of MPM and H&E images in pancreatic mucinous cystic neoplasm. (a) The representative MPM image ($10\times$) of an unstained section reveals a lesion composed of a single layer of columnar, mucin-producing epithelium (dashed arrow), ovarian-type stroma (triangle), collagenized stroma (pentagram), and a blood vessel (dotted arrow). (b) The corresponding H&E image ($10\times$) shows the similar features that correlate with the MPM image: the mucin-producing epithelium (dashed arrow), ovarian-type stroma (triangle), collagenized stroma (pentagram), and a blood vessel (dotted arrow).

3.4.

MPM Diagnostic Features of Pancreatic Malignant Tumors

The representative large-area MPM image of the pancreatic ductal adenocarcinoma in Fig. 3(a) revealed that the cancer tissue was composed of nonuniform and irregular glands [solid arrows in Fig. 3(a)]. Necrosis [arrow head in Fig. 3(a)] was observed in the glandular cavity. Blood vessels [dotted arrow in Fig. 3(a)] composed of randomly arranged, vascular endothelial cells were observed between the glands. Bands of fibrous stroma surrounding malignant glands and vessels were observed with a marked increase in collagen compared with normal pancreatic tissue [pentagram in Fig. 3(a)]. Deposition of a collagen-rich stroma around a malignant gland may represent a physiologic effort to confine the cancer.²⁹ More importantly, high-magnification MPM images showed cancer cells concentrated in lesion tissue [dashed arrows in Fig. 3(c)]. Tumorigenic cells displayed marked cellular and nuclear pleomorphisms characterized by irregular size and shape, enlarged nuclei and bright nucleoli. Unlike normal pancreatic tissue, the intercellular space between individual cells in malignant tissue was not readily discernable. These features were readily correlated with the corresponding H&E images shown in Figs. 3(b) and 3(d) including the irregular glands [solid arrows in Fig. 3(b)], necrosis [arrow head in Fig. 3(b)], blood vessels [dotted arrow in Fig. 3(b)], collagen fibers [pentagram in Fig. 3(b)], and cancer cells [dashed arrows in Fig. 3(d)].

Fig. 3

Comparison of MPM and H&E images in pancreatic ductal adenocarcinoma. (a) The representative MPM image ($10\times$) of an unstained section reveals the irregular glands (solid arrows), necrosis (arrow head), blood vessels (dotted arrow), and collagen fibers (pentagram). (b) The corresponding H&E image ($10\times$) shows the similar features that correspond to the MPM image: irregular glands (solid arrows), necrosis (arrow head), blood vessels (dotted arrow), and collagen fibers (pentagram). (c) Magnified MPM image ($40\times$) presents the cancer cells (dashed arrows) that exhibit distinct cellular features with illustrious nucleoli. (d) The standard bright field light microscopic image ($40\times$) of the corresponding H&E-stained histological section shows cancer cells (dashed arrows) with a similar appearance to the corresponding MPM image.

In the pancreatic colloid carcinoma, the representative large-area MPM image shown in Fig. 4(a) revealed nests of cancer cells [open arrows in Fig. 4(a)] that were sporadically distributed. Glandular expansion was easily discernable [asterisks in Fig. 4(a)], and clusters of cancer cells floated in mucous lakes that were formed by the expanded stroma. Dense bands of collagen [pentagram in Fig. 4(a)] were clearly observed, separating the mucous lakes.³⁰ At high magnification, cell clusters or “nests” [open arrow in Fig. 4(c)] were composed of a large number of cancer cells [dashed arrow in Fig. 4(c)]. These cancer cells were characterized by irregular size and shape, enlarged nuclei and bright nucleoli. The features were consistent with the H&E-stained results shown in Figs. 4(b) and 4(d), which visualized the nest of cancer [open arrows in Figs. 4(b) and 4(d)], mucous lakes [asterisks in Fig. 4(b)], collagen fibers [pentagram in Fig. 4(b)], and cancer cells [dashed arrow in Fig. 4(d)].

Fig. 4

Comparison of MPM and H&E images in pancreatic colloid carcinoma. (a) The representative MPM image ($10\times$) of an unstained section reveals the nest of cancerous tissue (open arrows), mucous lakes (asterisks), and collagen fibers (pentagrams). (b) The corresponding H&E image ($10\times$) shows the similar features that correlate with the MPM image: the clusters of malignant cells (open arrows), mucous lakes (asterisks), and collagen fibers (pentagrams). (c) Magnified MPM image ($40\times$) demonstrates that the nest of cancerous tissue (open arrow) is formed by a large number of cancer cells (dashed arrow). (d) The corresponding H&E light microscopic image ($40\times$) reveals the similar cellular architecture that correlates with the MPM image: the cluster of malignant tissue (open arrow) containing cancer cells (dashed arrow).

3.5.

Quantitative Analyses of Normal and Neoplastic Pancreatic Tissues

To quantitatively assess the morphological differences between normal and neoplastic pancreatic cells, a total of 20 cells in five random fields were examined within each MPM image. The N/C ratios from normal pancreatic cells or three types of neoplastic tissues were calculated and analyzed (Fig. 5). The N/C ratio was $0.25\pm 0.05$ ($N=40$) in normal pancreatic acinar epithelium cells, but the ratio increased to $0.85\pm 0.04$ ($N=24$) in colloid-producing epithelium cells found in pancreatic colloid carcinomas. The N/C ratio was $0.48\pm 0.04$ ($N=40$) in normal pancreatic ductal epithelium cells, while it was $0.59\pm 0.05$ ($N=36$) in mucin-producing epithelium cells of pancreatic mucinous cystic neoplasms. However, the N/C ratio increased to $0.72\pm 0.10$ ($N=40$) in cancerous ductal epithelium cells isolated from pancreatic ductal adenocarcinomas. We observed a statistically significant difference between the N/C ratio of pancreatic malignant tumor cells (colloid carcinoma) and normal pancreatic cells (acinus) ($P<0.001$). In addition, there was a statistically significant increase in the N/C ratio of pancreatic malignant tumors (ductal adenocarcinoma) compared with normal pancreatic cells (duct) ($P<0.001$) and benign tumors (mucinous cystic neoplasm) ($P<0.001$). We also observed a statistically significant increase in the N/C ratio of benign tumors (pancreatic mucinous cystic neoplasm) compared with normal pancreatic tissue ($P=0.014$). These results demonstrate that the N/C ratios obtained from MPM images successfully delineated malignant from normal tissues, with the ratios from malignant tumors being significantly greater than that of the normal pancreas or the pancreatic benign tumors. Quantitative analysis of benign tumors indicated that the N/C ratio was also significantly greater than the ratio of normal pancreas. Together, these data indicated the power of MPM imaging in distinguishing between tumors with differing malignant potential.

Fig. 5

The N/C ratios of normal tissue and three types of pancreatic neoplasms ($\text{mean}\pm \mathrm{SD}$) measured from MPM or standard H&E stained images. Error bars indicate standard deviations.

The quantitative results of the N/C ratios measured from H&amp;E images were also presented in Fig. 5 for comparison. Similar to the results calculated from MPM images, there was significant difference in the N/C ratios between colloid-producing epithelial cells of pancreatic colloid carcinomas (malignant tumor) and normal pancreatic acinar epithelium cells (compare $0.88\pm 0.08$ versus $0.38\pm 0.08$, $P<0.001$). There was also a significant difference in the N/C ratios between cancerous ductal epithelium cells of pancreatic ductal adenocarcinomas (malignant tumor) and normal pancreatic ductal epithelial cells (compare $0.80\pm 0.06$ versus $0.51\pm 0.09$, $P<0.001$) or mucin-producing epithelial cells of pancreatic mucinous cystic neoplasms (benign tumor) (compare $0.80\pm 0.06$ versus $0.65\pm 0.10$, $P<0.001$). Moreover, a statistically significant difference was also observed between pancreatic mucinous cystic neoplasms (benign tumor) and normal pancreas (compare $0.65\pm 0.10$ versus $0.51\pm 0.09$, $P<0.001$). In addition, differences in the N/C ratios between MPM and H&E images were not found to be statistically significant (data not shown).

4.1.

Feasibility of Using MPM to Diagnose Pancreatic Neoplasms

Currently, morphological examination of stained histological sections remains the standard diagnostic approach for pancreatic cancer. We performed this study to compare the morphological features of normal pancreas and pancreatic neoplasms based on MPM images of unstained tissues. Our studies showed that the MPM imaging could clearly delineate the parenchyma and the stroma in pancreatic specimens. Specially, it revealed regular tissue architecture and cell morphology in normal pancreas, and distinctive tissular and cellular pleomorphisms in pancreatic neoplasms. More importantly, typical cancer cells, characterized by irregular size and shape and with enlarged nuclei, were clearly observed in MPM images. In short, the high-resolution MPM images clearly demonstrated apparent differences between normal pancreatic tissue and pancreatic neoplasms, as well as distinguished between benign versus malignant tumors in terms of their tissue architecture and cell morphology, and these results were comparable with H&E images.

Moreover, when we quantitatively assessed the N/C ratios obtained from MPM images, we found that the ratios for pancreatic malignant and benign tumors were clearly distinguishable from normal pancreatic tissue. In addition, we were able to distinguish between pancreatic malignant versus benign tumors using the N/C ratios from MPM images. These data were consistent with the H&E results, as well as previous reports that suggest the N/C ratio in neoplastic cells is greater than in normal cells due to the presence of enlarged nuclei that undergo an enhanced rate of mitosis.³¹ It was demonstrated that single abnormal cells, irregular nuclear membranes, and enlarged nuclei were important cytologic criteria for differentiating malignant from benign pancreatobiliary specimens.³² In addition, the extensive activity of the nucleus in neoplastic cells increases with the stage and grade of malignancy, resulting in increased N/C ratios in malignant tumors compared with benign tumors. An increased N/C ratio, marked nuclear envelope abnormalities, prominent nucleoli and hyperchromasia have been demonstrated as overt malignancy.³³

Furthermore, comparison of the N/C ratios in H&E versus MPM samples showed similar results, which verified the reliability of the MPM technique. Together, these data suggest that the application of MPM for histological investigations of biopsy specimens can be used to reliably diagnose the pancreatic neoplasms in unstained tissue slices.

4.2.

Significance of This Study

The application of optical methods with single-cell resolution provides detailed characterization of pancreatic morphology and improves the accurate detection and diagnosis of pancreatic neoplasia, which is a significant advancement in the field of histology.¹¹ Because H&E staining may cause chromatic aberration and produce environmental waste, alternative methods such as MPM imaging of unstained sections may be used to diagnose and classify pancreatic diseases including cancer without the disadvantages associated with traditional histological methods. Another advantage of MPM is its versatility, since it does not require fresh tissue samples and is an excellent tool for retrospective analyses of unstained histological sections acquired over several years. Less is understood regarding the association between early morphological changes within the pancreas and later development of pancreatic cancer, and MPM may help to establish early diagnostic markers through analysis of sample libraries obtained from early and late-stage cancer patients around the world.

In conclusion, we demonstrated that the MPM acquires valuable and complementary information from unstained tissues with the same degree of accuracy as H&E staining. MPM can be successfully used to differentiate between pancreatic neoplasms and normal pancreas, as well as distinguish pancreatic benign tumors from malignant tumors, further establishing its promise as a diagnostic tool. Future studies will utilize MPM imaging to perform a retrospective analysis of unstained pancreatic tumors obtained from a tissue library, in order to visualize changes in tumor morphology and malignancy over time in single patients.