3.1.

Selection of the Excitation Laser Source

In previous studies,^{24, 25} the transmission spectra of each ocular component of the human eye was measured. Table 1 shows the measured transmission characteristic at three wavelengths: 800, 1030, and $1230\mathrm{nm}$ . The average thicknesses of the human cornea, aqueous humor, lens, and vitreous humor are 0.5, 3.0, 3.2, and $15\mathrm{mm}$ ,²⁴ respectively. At $800\mathrm{nm}$ , the transmittance of each component is higher than 95% and the total transmittance through the ocular components in front of the retina runs to about 85%. The total transmittance decreases to 38% at $1030\mathrm{nm}$ and drops greatly to only 9% at $1230\mathrm{nm}$ due to the stronger water absorption. According to the previous measurement,²⁶ the water absorption coefficients at 800, 1030, and $1230\mathrm{nm}$ , are 0.0207, 0.1845, and $0.959{\mathrm{cm}}^{-1}$ , respectively. Before arriving at the retina of a human eye, most of the $1230\text{-}\mathrm{nm}$ light energy will be absorbed by the $15\text{-}\mathrm{mm}$ -thick vitreous humor, which is mainly composed of water. This indicates that to increase retina safety, an infrared (IR) excitation source with a wavelength around the order of $1230\mathrm{nm}$ or slightly longer would be a better choice, because the light penetrating the retina will be diminished, while the imaging depth will still be deep enough for imaging of the cornea.

Table 1

The transmittance of each ocular component, including cornea, aqueous, lens, and vitreous, and the total transmittance through the entire human eye at 800, 1030, and 1230nm .

3.2.

Two-Photon Fluorescence, Second Harmonic Generation, and Third Harmonic Generation Signal Intensity Measured in the Mouse Cornea

Popularly used confocal and TPF microscopies make use of the intrinsic fluorophores NAD(P)H/NAD(P) in cells to obtain the cellular morphological information of corneal tissues. Under IR excitation, most of the autofluorescence would be suppressed,²⁷ including the autofluorescence from NAD(P)H/NAD(P). Only weak TPF signals could be observed within $0\text{to}10\mu \mathrm{m}$ beneath the anterior corneal surface, but abundant SHG and THG could be found in the cornea and the lens. To investigate the relative intensities of the backward-collected SHG, THG, and TPF signals at different depths underneath the anterior corneal surface, a charge-coupled device (CCD) spectrometer was used to obtain the spectrum with range from $350\text{to}800\mathrm{nm}$ [Fig. 2a ]. Within this range, THG $\left(410\mathrm{nm}\right)$ , SHG $\left(615\mathrm{nm}\right)$ , and TPF (peaked at $650\mathrm{nm}$ ) signals could all be measured. The measured peak intensities of the SHG, THG, and 2PF signals were plotted in Fig. 2b versus the depth beneath the anterior corneal surface. In a mouse eye [Fig. 2c], the average thickness of the cornea C and the aqueous humor (AH) is about 160 and $270\mu \mathrm{m}$ , and the anterior surface of the lens is at $\sim 430\mu \mathrm{m}$ beneath the corneal surface. The thickness of each component would change in different mouse eyes. As shown in Fig. 2b, strong THG intensity was measured within $0\text{to}130\mu \mathrm{m}$ (cornea) and at $400\mu \mathrm{m}$ (the anterior part of the lens), and was hundreds of times higher than TPF signals. Within the whole depth of imaging, TPF signals were too weak to be used for imaging, and thus only SHG and THG images are shown and discussed in this work. Instead of TPF, THG, which has structural sensitivity,^{28, 29} could be useful for revealing the cellular structure of both the cornea and the lens with no energy deposition and no photodamage.

Fig. 2

(a) The $1230\text{-}\mathrm{nm}$ -excited spectrum of the cornea of an intact mouse eye was measured by a CCD spectrometer. Within a range from $350\text{to}800\mathrm{nm}$ , the epi-collected THG $\left(410\mathrm{nm}\right)$ , SHG $\left(615\mathrm{nm}\right)$ , and autofluorescence (peaked at $650\mathrm{nm}$ ) were all included. (b) The peak intensity of THG, SHG, and TPF at different depths beneath the anterior corneal surface of a mouse eye was measured and plotted versus the depth. The measured regions included the cornea (C, $0\text{to}130\mu \mathrm{m}$ ), aqueous humor (AH, $130\text{to}400\mu \mathrm{m}$ ), and lens (L, $>400\mu \mathrm{m}$ ). Within the measured regions except for the AH with no imaging contrast, only weak TPF signals were found around the anterior surface of the cornea, and the intensity of THG signals was measured to be hundreds of times higher than that of TPF signals. SHG signals were also more abundant than TPF signals in the cornea, especial in the corneal stroma. (c) The scheme of the mouse eye was shown as a reference for indicating the thickness of the cornea (C) and aqueous humor (AH), as well as the position of the lens (L).

3.3.

Third Harmonic Generation Imaging of the Mouse Eye

According to the histological results of the cornea,^{30, 31} the scheme of the mouse eye and the structures of the cornea and the lens are shown in Fig. 3a . The cornea can be roughly divided into three main components: the corneal epithelium (EP), corneal stroma (CS), and corneal endothelium (ED). The EP and ED are cellular layers, and the CS between them is composed of connective tissues. The EP covers the anterior surface of the cornea and is only 5 to 6 cell layers thick. The EP cells of the superficial layer (close to the anterior surface of the cornea) are squamous and their nuclei are flattened with a larger diameter. The EP cells of deeper layers are columnar and their nuclei are globular with a smaller diameter. On the other hand, the ED is a monolayer of flattened and polygonal ED cells. They line the posterior surface of the cornea and play an important role for governing the fluid and solute transport across the posterior surface of the cornea. Loss of the ED cells (endothelial dystrophy) can lead to loss of corneal transparency, and an imaging tool with high sectioning power to resolve this thin layer is required for diagnosis. Between these two layers is the CS, consisting of hundreds of layers of regularly arranged collagen fibers along with sparsely populated keratocytes [arrows in Fig. 3a]. For transparency, the collagen fibers run parallel in the same layer but have a lattice arrangement within different layers. Beneath the cornea is the aqueous humor (AH), mainly composed of water and followed by the lens (L). The lens consists of three main parts: the lens capsule (LC), lens epithelium (LE), and lens fibers (LF). The membrane-like LC forms the outermost layer of the lens and completely surrounds the lens. The cellular LE is located in the anterior part of the lens between the LC and the LF, which forms the bulk of the lens, and can generate new lens fibers.

Fig. 3

(a) The scheme of the mouse eye (middle scheme), the axial structure of the cornea (top enlarged scheme), and the lens (bottom enlarged scheme). As shown in the axial structures of the cornea and lens, the cornea can be roughly divided into the corneal epithelium (EP), corneal stroma (CS), and corneal endothelium (ED), while the lens consists of the lens capsule (LC), lens epithelium (LE), and lens fiber (LF). The imaging plans of (b) through (g) were indicated in (a). (b) through (g) THG images of the ocular tissues of a mouse eye obtained at different depths of (b) 10, (c) 25, (d) 40, (e) 100, (f) 120, and (g) $430\mu \mathrm{m}$ beneath the anterior corneal surface. (b) The squamous and flattened EP cells were revealed with a larger average ND $(\sim 13\mu \mathrm{m})$ and IND $(\sim 19\mu \mathrm{m})$ . (c) The transition shape of the EP cells from squamous to columnar shape was shown with a smaller average ND $(\sim 7\mu \mathrm{m})$ and IND $(\sim 12.2\mu \mathrm{m})$ . (d) The columnar EP cells at the bottom of the EP were revealed with a small average IND $(\sim 7.7\mu \mathrm{m})$ close to the average ND $(\sim 7\mu \mathrm{m})$ . (e) shows the THG image of the CS, where both the boundaries of the collagen fibers and the keratocytes in the CS were found. (f) The uniformly sized polygonal ED cells were observed in the monolayer ED. The nuclei of the ED cells are indicated by arrowheads. (g) At a depth of $430\mu \mathrm{m}$ , the lens fibers with a width of $\sim 5\mu \mathrm{m}$ (along the arrow) can still be highly resolved through THG signals. THG is represented by a purple pseudocolor. Scale bar: $50\mu \mathrm{m}$ . (Color online only.)

Figures 3b, 3c, 3d, 3e, 3f, 3g show the THG images of the ocular tissues obtained at different depths beneath the anterior corneal surface. The imaging plan of each image is indicated in Fig. 3a. From these THG images, the morphology of the EP cells was successfully revealed at depths of 10, 25, and $40\mu \mathrm{m}$ [Figs. 3b, 3c, 3d], and the results were consistent with the histology results described in the former paragraph.³¹ Based on the THG contrast arising from the cytoplasmic organelles,³² the cytoplasm of the EP cells appeared bright in contrast to the dark nuclei. At $10\mu \mathrm{m}$ [Fig. 3b], the squamous and flattened shape of the EP cells could be revealed through both the larger average nuclear diameter (ND) $(\sim 13\mu \mathrm{m})$ and the longer average internuclear distance (IND) $(\sim 19\mu \mathrm{m})$ , which indicated the larger diameter of the cells. At $25\mu \mathrm{m}$ [Fig. 3c], the nuclei were shown to have a smaller average ND of $\sim 7\mu \mathrm{m}$ , and the average IND of $\sim 12.2\mu \mathrm{m}$ indicated the transition shape of the cells from squamous to columnar shape. At $40\mu \mathrm{m}$ [Fig. 3d], the bottom of the epithelium, the columnar shape of the EP cells was revealed through a smaller average IND of $\sim 7.7\mu \mathrm{m}$ , which was quite close to the ND of $\sim 7\mu \mathrm{m}$ .

Figure 3e shows the THG image of the CS at a depth of $100\mu \mathrm{m}$ . In the CS, much information can be obtained through THG signals. As mentioned before, there are collagen fibers, and the keratocytes exist in the CS. Based on the THG contrast arising from interfaces^{33, 34} and cytoplasmic organelles,³² not only the boundaries of the collagen fibers but also the keratocytes could be revealed in Fig. 3e. Beneath the CS, Fig. 3f shows the THG image of the ED at $120\mu \mathrm{m}$ , and the uniformly sized polygonal cells in this monolayer could be revealed. The nuclei of the endothelial cells appear dark [arrowheads in Fig. 3f], while the cytoplasm appear bright. Passing through the AH with no THG contrast, at a depth of $430\mu \mathrm{m}$ beneath the anterior corneal surface, the lens fibers [Fig. 3g] with a width of $\sim 5\mu \mathrm{m}$ [along the arrow in Fig. 3g] could be highly resolved through THG signals. Moreover, it is important to notice that we did not observe significant SHG signals from the lens fibers. To show the details more clearly, some contrast adjustments have been applied to the original images, which were obtained with the same PMT voltage, and the actual intensity of the THG contrast at different depths could be found in Fig. 2b. Although the attenuation coefficient of the $1230\text{-}\mathrm{nm}$ excitation light in the cornea, lens, and the AH is about $2.158{\mathrm{cm}}^{-1}$ ,²⁰ $1.001{\mathrm{cm}}^{-1}$ ,²⁰ and $0.959{\mathrm{cm}}^{-1}$ (water absorption coefficient),²² the THG intensity kept around the same order within the cornea and the THG intensity was ten times stronger in the lens [Fig. 2b]. Even at a depth slightly greater than $700\mu \mathrm{m}$ , the structure of the lens fibers could still be revealed through THG, but the higher PMT voltage and the contrast adjustment were needed. Since the total thickness of the human cornea is about $535\mu \mathrm{m}$ ,³⁵ a penetration depth greater than $700\mu \mathrm{m}$ indicates the ability to investigate even the deepest part of the human cornea.

3.4.

Multimodality (Second and Third Harmonic Generation) Imaging of the Mouse Eye

The corneal stroma consists of hundreds of layers of regularly organized collagen fibers, including mainly type-1 collagen, but also types 3, 5, and 6.³⁶ Accounting for the transparency of the cornea, the collagen fibers run parallel to each other but at large angles to the fibers in the next layer.³⁶ Since SHG is highly sensitive to collagen fibers, especially type-1 collagen,^{13, 14, 15} the arrangement of collagen fibers in the stroma were expected to be revealed by SHG microscopy. As shown in the SHG and THG images of the CS [Figs. 4a and 4b ], in addition to the collagenous structure of the CS shown by SHG as expected, some more information can be given by THG. Combining Figs. 4a and 4b together to form Fig. 4c, it can be found that THG and SHG would overlap in some regions [shown white in Fig. 4c] but not overlap in other regions. As mentioned in the former paragraph and has been confirmed by transmission THG microscopy (not shown), THG contrast in the CS is dominated by the boundary of the collagen fibers and the keratocytes. Since the collagen fibers have SHG contrast, the colocalized THG signals can be easily distinguished while the combined color will appear white, as shown in Fig. 4c. Taking advantage of the combined SHG and THG modality, the keratocytes can thus be easily identified with their pure THG color, together with size and shape characteristics. Figure 4d shows the scheme of the mouse eye as a reference, and an optical axial section, including SHG and THG signals [Fig. 4e], was obtained in the region indicated by square $e$ in Fig. 4d. The total depth of Fig. 4e is $750\mu \mathrm{m}$ . No adjustments of the PMT voltage and contrast have been applied within the whole imaging depth. The axial structure of both the cornea (C), including EP, CS, and ED, and the lens (L), including LC, LE, and LF, could both be resolved in this axial section. It indicates a high penetrability of $\sim 700\mu \mathrm{m}$ and a high axial sectioning power of our IR-based harmonic generation microscopic imaging of the ocular tissues.

Fig. 4

(a) SHG, (b) THG, and (c) $\mathrm{SHG}+\mathrm{THG}$ image of the corneal stroma. (a) SHG revealed the collagenous characteristic of the corneal stroma, which is mainly composed of type-1 collagens fibers. (b) THG arose from both the boundaries of the collagen fibers and the keratocytes, which were difficult to distinguish. (c) THG and SHG would overlap in some regions (shown white) but not in other regions. Since the THG signals from collagen fibers would mostly overlap the SHG signals (shown in white), the THG signals from keratocytes (arrows) could be distinguished. (c) shows the scheme of a mouse eye as a reference and the square $e$ indicates the imaging region of (d). (d) shows the $\mathrm{SHG}+\mathrm{THG}$ axial section of the ocular tissues with a depth of $750\mu \mathrm{m}$ . Without any adjustments of the PMT voltage and contrast, a penetration depth of $\sim 700\mu \mathrm{m}$ could be easily achieved. Each layer of the cornea (EP, CS, and ED) and the anterior part of the lens (LC, LE, and LF) were revealed with a high axial resolution. SHG and THG are represented by green and purple pseudocolors, respectively. Scale bar: (a), (b), and (c) $50\mu \mathrm{m}$ ; (e) $100\mu \mathrm{m}$ .

3.5.

Axial Structures Revealed with a Lateral Resolution

Based on $700\text{-}\mu \mathrm{m}$ penetrability, we moved the imaging plane to $550\mu \mathrm{m}$ beneath the anterior corneal surface [dashed arrow in Fig. 5a ]. At this depth, an intact mouse eye is optically sectioned with a total lateral diameter of $\sim 2\mathrm{mm}$ , as shown in Fig. 5b. As indicated in Fig. 5a, the intact mouse eye is sectioned with the lens in the center, and surrounding the lens is the aqueous humor (AH) and then the cornea (C). Enlarged from the squared area $c$ in Fig. 5b, due to the spherical shape of the eye, the axial structure instead of the lateral structure of the cornea is revealed by THG with a submicron lateral resolution [Fig. 5c]. Comparing Fig. 5c with Fig. 4e, both the EP cells and the fibers of the CS in Fig. 5c were shown more distinctly with higher resolution than those in Fig. 4e. In addition, enlarged from the squared area $d$ in Fig. 5b, the axial structure of the lens, including the lens capsule (arrow), lens fibers (arrowhead), and the lens epithelium (star), can all be resolved by the THG signals [Fig. 5d].

Fig. 5

(a) shows the scheme of a mouse eye to indicate the imaging plan of (b) (dashed arrow), the optical axis of the eye (dashed line), and the different incident angles [ $0^{}\mathrm{deg}$ , gray arrow; ${60}^{}\mathrm{deg}$ , black arrow in (a)]. (b) A lateral section of an intact mouse eye obtained at a depth of $550\mu \mathrm{m}$ beneath the anterior corneal surface. The intact mouse eye was optically sectioned with a lateral diameter of $\sim 2\mathrm{mm}$ . As indicated in (a), the mouse eye would be sectioned with the lens in the center, and surrounding the lens was the aqueous humor (AH) and then the cornea (C). (c) The THG image enlarged from the squared area $c$ in (b). Due to the spherical shape of the eye, the axial structure of the cornea would be revealed with a submicron lateral resolution in this lateral section. (d) The THG image enlarged from the squared area $d$ in (b). The axial structure of the lens, including the lens capsule (arrow), lens fibers (arrowhead), and the lens epithelium (star), could all be resolved by THG. Scale bar: (b) $500\mu \mathrm{m}$ ; (c) and (d) $100\mu \mathrm{m}$ .

Generally, including our previously discussed studies, the microscopies used for corneal diagnosis have a normal incident excitation [gray arrow in Fig. 5a], which means that the incident angle relative to the optical axis of the eye is $0\mathrm{deg}$ , and the excitation light will go straight to the retina. If the incident angle of the excitation light can be increased to around $45\text{to}60\mathrm{deg}$ [black arrow in Fig. 5a, an incident angle of ${60}^{}\mathrm{deg}$ ], the percent of the excitation light going to retina will be greatly reduced and retina safety will be improved as well. Due to the dome shape of the cornea, with a larger incident angle, the axial structure instead of the lateral structure of the cornea will be revealed in the lateral sections with a submicron lateral resolution, as shown in Fig. 5c. Using a $1230\text{-}\mathrm{nm}$ excitation light would be the first step to increase both eye safety and penetrability in ophthalmological study. To further improve retina safety for future IR diagnosis of the human cornea, not only a larger laser incident angle but a slightly longer illumination wavelength (not longer than $1300\mathrm{nm}$ ) and a lower excitation power, will be desired. Although THG intensity will decrease with a lower excitation power, the decreased THG contrast may be recovered by shortening the pulse width to increase the peak intensity of a single laser pulse.