Video-rate confocal endoscopy

Journal of Microscopy, Vol. 207, Pt 1 July 2002, pp. 37 42 Received 7 March 2002; accepted 18 April 2002 Video-rate confocal endoscopy Blackwell Science, Ltd T. F. WATSON*, M. A. A. NEIL, R. JUŠKAITIS, R. J. COOK* & T. WILSON *Guy s, King s & St Thomas Dental Institute, Guy s Hospital, London, SE1 9RT, U.K. Department of Engineering Science, University of Oxford, Parks Road, Oxford, OX1 3PJ, U.K. Key words. Clinical imaging, confocal, endoscopy. Summary Rigid endoscopes provide high quality optical images of reasonably accessible regions of the inner body, especially regions such as the aero-digestive and genital tracts. In order to enhance the versatility of these instruments we describe a development that permits confocal endoscopic images to be obtained along with traditional endoscopic images in real-time, from within the living patient. The system is based around a host lenslet-array tandem scanning microscope, which is capable of producing images viewed directly by eye. These types of confocal microscope are configured for fluorescence imaging together with laser illumination. Hard and soft tissues in the mouth were imaged using this combined system. Introduction Endoscopy is a vital, yet minimally invasive, operative procedure, increasingly employed for both diagnosis and management of many medical and surgical conditions. There are broadly two classes of purely optical endoscope. The first, developed in the 1960s (Cotton & Williams, 1996), uses a long flexible fibre-optic coupling between the remote lesion site and the user. This gives adequate diagnostic information, although the image quality may be compromised by both the number of (intact) elements within the fibre bundle and the light losses, which become significant when the individual fibres of a bundle decrease below 6 8 µm diameter (Cotton & Williams, 1996). An alternative class of endoscope uses a thin rigid tube enclosing a distributed lens system (see, e.g. Hopkins, 1975). These instruments are optically superior to their fibre bundle counterparts, although their use is confined to regions of the body that afford reasonably straight line access to the region of interest or keyhole surgical access, e.g. imaging ports in laparoscopic surgery. The image contrast in these instruments originates from both the surface and subsurface regions of Received 7 March 2002; accepted 18 April 2002 Correspondence to: Professor T. Wilson. Fax: +44 (0)1865 273905; e-mail: tony.wilson@eng.ox.ac.uk the translucent tissue under examination. In order to discriminate between these structural features in a controllable way, it is advantageous to be able to introduce confocal imaging. Fluorescent labelling may allow greater differentiation between cell layers capable of absorbing the dye solution. The use of digital signal detection methods, together with the absence of image degradation due to a fibre bundle, leads to higher quality image formation and manipulation. The host confocal microscope must be able to capture images in real-time and a particularly attractive configuration for real-time imaging is the tandem scanning microscope (TSM) (Petrán et al., 1968; Boyde et al., 1990; Watson & Boyde, 1991). In these instruments the illumination and detection pinhole arrays are diametrically opposed on a spinning Nipkow disc. A subsequent development permitted the same pinhole array to be used for both illumination and detection (Xiao et al., 1988). A recent development to this single sided TSM has been the introduction, by the Yokogawa company ( Japan), of a lenslet-array to permit more efficient use of the illumination light (Watson et al., 2002). The manufacturer s literature claims light transmission efficiencies of the disc of up to 60% when used for fluorescence imaging with coherent illumination sources. To enhance the capabilities of the rigid endoscope we will describe a system in which a lenslet array TSM (Yokogawa CSU10 Ultraview, Perkin Elmer Life Sciences, Cambridge, U.K.) has been incorporated. These instruments work well in fluorescence, with easy discrimination between the excitation and emitted light provided by the dichroic mirrors located between the Nipkow disc and microlens array. There is therefore little problem with unwanted reflected light from optical components within the scanning system. The aim of this study was to apply this confocal endoscope to imaging previously inaccessible regions of the aero-digestive tract. Materials and methods A Yokogawa CSU10 tandem scanning microscope was mounted on an optical bench and coupled into a rigid endoscope (Hopkins type contact telescope: Karl Storz, Tuttlingen, 2002 The Royal Microscopical Society

38 T. F. WATSON ET AL. original design of the instrument. This is introduced directly into the light guides of the endoscope via a flexible fibre optic cable (Fig. 1) and therefore bypasses the laser illumination channel of the confocal system. In this conventional mode of operation the tissue is illuminated via separate light guides within the endoscope. The image contrast, which is essentially dark-field in nature, is obtained from light backscattered from the sample. Although this incoherent light returning from reflective features passes through the confocal unit to the camera, it is important to realize that the resultant image is non-confocal. An argon-ion laser was used as the light source for the confocal images, with appropriate dichroic filters. Results Fig. 1. The Yokogawa CSU10 tandem scanning microscope head, mounted on an optical bench and coupled to a rigid (colpo-hysteroscopy pattern) Hopkins type endoscope via a 60 mm focal length relay lens. The endoscope itself contained an internal focus mechanism and a 30 forward angle lens for imaging structures lateral to the axis of the instrument. A video-rate CCD camera was attached to the confocal unit imaging port for image capture. Direct operator viewing of the images was also available via the viewing eyepiece. Fluorescence illumination was derived from a 488 nm emission wavelength argon-ion laser. Incoherent white light illumination was also used for imaging; the white light conducted to the site via an appropriate (Karl Storz) flexible light guide and the endoscope s epi-illumination pathways. Germany) via a 6 cm focal length relay lens (Fig. 1). The endoscope contained an internal focusing mechanism and a 30 forward angle lens for imaging structures lateral to the axis of the instrument. A video-rate CCD camera (Cohu, California, U.S.A.) was attached to the confocal unit for final image capture and display. Furthermore, direct viewing of the images was also available. Fluorescein in an aqueous solution was applied to areas of interest on two volunteers, using cotton buds. The keratinized skin on the dorsal and palmar surfaces of the hands were stained and examined first, allowing the researchers to learn to drive the new configuration using simple, well-understood anatomical features. Furthermore, this allowed us to assess the (minimal) heat emissions at the working end of the instrument, before imaging our more difficult and sensitive internal regions. The intra-oral tissues under examination consisted of the dorsal and ventral surfaces of the posterior third of the tongue, the oral vestibular lining mucosa especially the inner lip and retromolar fossa regions. Images were also made of extracted teeth restored with a rhodamine-labelled adhesive material. Contact between the endoscope end lens and the subject was achieved with either saliva or glycerol as coupling media, thereby eliminating strong surface reflections. Conventional endoscopic images were obtained using the incoherent illumination source, incorporated within the Before using the instrument internally, the thermal output from the working endoscope tip was assessed as safe by imaging the skin of the dorsal and palmar aspects of fingers, using both illumination modes. Non-contact imaging using incoherent white light illumination provided clear conventional endoscopic images of the fingertip. By abutting the soft tissues with the working tip of the endoscope, contact microscopic imaging was then undertaken. Serial confocal fluorescence images of the keratinized epithelium of the palmar aspect of a finger tip clearly identified stained desquamating keratinocytes at the surface, and associated adnexal structures such as sweat duct orifices (Fig. 2). Using glycerol as the coupling medium, serial confocal optical sections allowed the operators to follow features of interest such as these ducts, penetrating the keratinized epithelium, at least to the level of the basement membrane (Fig. 2). At this level, alternating light and dark banding was seen across the field, reflecting the gross ridge and trough convolutions of the basement membrane level, associated with the ridge and trough patterns of the overlying fingerprint (Fig. 2). The deeper, stained keratinized mucosal troughs appear brighter when compared to the presumably unstained crests of mesodermal tissues extending between them, and therefore are included in the same confocal optical plane. Having established validity and thermal safety of the configuration, intra-oral imaging was then undertaken from both dorsal and ventral aspects of the posterior third of the tongue, inner cheeks, lips and retromolar regions using both confocal laser and incoherent epi-illumination sources (Fig. 3). As a proof of principle instrument, rigid bench mounting was necessary to maintain optical alignment, so the subject was required to orientate his head and oropharynx around the fixed microendoscope tip. Direct subject viewing of the live CCD images aided both the location of interesting features as well as posture maintenance during real-time imaging sequences of these hitherto impossible anatomical sites for in vivo imaging in an intact subject. Confocal fluorescence images from the dorsum of the posterior third of the tongue (Fig. 4) show the keratinized epithelium of

VIDEO-RATE CONFOCAL ENDOSCOPY 39 Fig. 3. Intra-oral imaging was achieved by the subject aligning the 30 forward angled endoscope tip with the structures of interest within his mouth. In this example, showing real-time imaging of the right retromolar fossa mucosa, evidence of the fluorescein solution used for confocal fluorescence imaging of the posterior buccal sulcus mucosa, lateral to the subject s last standing molar, can also be seen. filiform papillae and demonstrates cellular detail from the confocal optical sections through the crowns of fungiform papillae (Figs 4 and 5). The fluorescein dye pooled in the gutters around the bases of lingual structures such as the papillae, outlining them and providing scattered illumination within the subject, permitting cellular detail to be distinguished from within these structures. This contrasts sharply with the dark central shadow and edge effect seen in occasional air bubbles within the fluorochrome dye itself (Fig. 5). Comparison of matched pair images of the oral soft tissue structures served well to highlight the benefits of both illumination schemes within the single instrument design. The inner aspect of the human lower lip bears a myriad of sub-mucosal accessory salivary glands, whose tubular ducts release mucous saliva, lubricating labial movements against the teeth. Figure 6 shows that combined, concurrent confocal fluorescence and white light epi-illumination imaging is counterproductive because the incoherent illumination image is far brighter Fig. 2. Serial confocal fluorescence images of the keratinized epithelium of the palmar aspect of a finger tip, showing (a) labelled desquamating keratinized squames, (b) a cluster of four sweat ducts just below the epithelial surface, and (c) four ducts deeper within the structure of the finger. The duct lumens show as dark voids, whereas the top right-hand duct has some fluorochrome within its lumen, which has adopted the well known corkscrew or helical configuration (c). The alternating light and dark banding across the field in (c) demonstrates the convolutions of the basement membrane interface between epithelium and deeper structures, reflecting the ridge and trough patterns of the fingerprint. The deeper, stained keratinized mucosal troughs are brighter when compared to the unstained crests of mesodermal tissues included in the same confocal optical section. Fieldwidth 1.5 mm.

40 T. F. WATSON ET AL. Fig. 4. Confocal fluorescence image from the dorsum of the posterior third of the tongue, showing the keratinized epithelium of compressed filiform papillae to the left and centre of the field, with an optical section through the crown of a fungiform papilla clearly demonstrated on the right. Fieldwidth 1 mm. Fig. 5. Confocal fluorescence image of the dorsum of the tongue at the junction of the anterior two-thirds and posterior one-third. Fluorescein dye has pooled in the gutters around fungiform papillae (p), within which cellular detail can be discerned. The central round structure a fungiform papilla is also demarcated and cellular detail can be distinguished from within its structure, unlike the air bubble seen to the left. The keratinized tips of filiform papillae can be seen within the pooled fluorochrome, which has further outlined folds in the lingual mucosa in the lower part of the field of view. Fieldwidth 1 mm. Fig. 6. Matching paired images of the submucosal structures of the inner aspect of a live human lower lip. (a) A subsurface image captured with both confocal fluorescence and non-confocal, incoherent white light epiillumination. The image contains minimal evidence of vascular markings (v) on the left side of the field and limited information concerning the minor salivary gland duct structures, so prolific in the vestibule of the lower lip. The same field illuminated purely in confocal fluorescence mode (b), however, shows the darker lumenae of the salivary ducts but all evidence of deeper vascular structures is lost within this shallow optical section. Fieldwidth 1.5 mm. than the fluorescence image. The resulting image contains minimal evidence of either vascular, cellular or salivary duct structures, so prolific in the vestibule of the lower lip. However, the same site seen in pure confocal fluorescence mode shows the dark lumenae of the salivary ducts deep within the lip s structure. However, minimal evidence of deeper vascular structures is seen with this optical sectioning mode (Fig. 6).

VIDEO-RATE CONFOCAL ENDOSCOPY 41 Fig. 7. The clinical and diagnostic utility of the pairing of confocal fluorescence and incoherent epi-illumination systems is further reinforced by this image taken using epi-illumination white light only. The normal branching arteriolar and capillary networks supplying both the submucosal and epithelial structures of the ventral surface of the root of the subject s tongue are shown. The image is a still frame from a real-time video sequence, in which the movements and flow patterns of individual 7 8 µm diameter red blood corpuscles in capillary loops and the larger vessels could be discerned. Fig. 8. Combined pseudo-coloured epi-illumination (green) and confocal rhodamine (red) fluorescence image showing an air bubble in a vertically sectioned dental restoration, bonded to the tooth dentine (d) by an adhesive labelled with rhodamine B fluorochrome. Whereas the nonconfocal illumination displays the three-dimensional structure of the rather large void, the confocal fluorescence signal shows some evidence of the labelled adhesive having formed an interfacial layer on the dentine surface. However, some of the material has become incorporated into the air bubble in the bulk of the semi-translucent filling material to the left of the field. The smaller round structures were small mobile air bubbles in the glycerol lens-coupling medium. Images acquired using epi-illumination alone are also of great diagnostic and research significance in this context. Figure 7 shows the normal branching network patterns of arteries and arterioles supplying the tissues of the ventral surface of the root of the subject s tongue. The flow of individual red blood corpuscles in capillary loops and larger vessels could be seen in real-time video sequences taken from the same anatomical site. The development of this tandem illumination system was also applied to semi-translucent dental hard tissues, to assess further the instrument s reflection and fluorescence performance and capabilities. An extracted human tooth bearing a cavity restored (ex vivo) using a dental composite white filling material, incorporating a rhodamine-labelled adhesive, was sectioned and the interfaces were examined. The combination of a pseudo-colour epi-illumination (green) and confocal rhodamine (red) fluorescence images (Fig. 8) produced useful information, showing the adhesive dentine interface layer and its contribution to the walls of a void within the restoration (Fig. 8). Serial confocal fluorescence images were achieved through an intentionally introduced void, whose domed surface had fractured. The specimen was a glass ionomer cement restoration, labelled with rhodamine B fluorochrome and placed in a human tooth. (Fig. 9). The images were taken using a glycerol coupling medium to reduce interfacial surface reflections, with the endoscope abutting the outer surface of the otherwise intact molar. This demonstration of the ability of the system to generate serial optically sectioned images from semi-opaque dental restorative materials, further demonstrates the flexibility of this instrument configuration. Conclusions In this study, a novel combination of a rod lens type rigid endoscope in current clinical use and a video-rate laser confocal microscope has demonstrated that remote, endoscopic confocal and high contrast incoherent illumination microscopy is achievable for both soft and hard tissues, via a single instrument. Using an appropriate coupling medium, clinically useful confocal optical sectioned fluorescence images are demonstrably achievable, both within and beneath keratinized and non-keratinizing epithelia from soft tissues deep within the body. The epithelia of the vast majority of the genital and aerodigestive tracts are non-keratinizing in healthy individuals, so the ventral surface of the tongue presents a good representative model of imaging within these tracts and is itself a previously impossible area to examine microscopically within a live and intact subject. The combination of fluorescence and incoherent imaging severely limits the available information, the latter swamping

42 T. F. WATSON ET AL. Fig. 9. Serial confocal fluorescence images from deep (a) through to most superficial cracked surface layers (f ) of a large void in a poorly mixed glass ionomer cement restoration placed in a human tooth, having been labelled with rhodamine B fluorochrome. The images were taken using a glycerol coupling medium to reduce interfacial surface reflections, the endoscope being applied the curved outer (cheek) aspect of the otherwise intact tooth. This clear demonstration of the ability of the system to generate serial optically sectioned images from semi-opaque dental restorative materials further demonstrates the flexibility of this instrument configuration. the former because of the significantly greater non-confocal light levels. However, if recorded independently, near simultaneous confocal fluorescence and incoherent epi-illumination images of the same soft tissue site via the same instrument can provide a wealth of vital diagnostic information far more rapidly than conventional flexible instrument endoscopy and histological analysis of biopsied specimens. Both the cellular structure, layers and basement membrane level of a suspicious epithelial lesion and its associated vascular architecture and flow patterns may now be examined together. In the diagnosis of malignant epithelial tumours, evidence of breaches of basement membrane integrity are of paramount importance and currently require accurate biopsy and histological processing of the sampled material, with the inherent delays of a joint surgical and laboratory diagnostic process. This new technique provides clear evidence that fluorescence optical sectioning can achieve clinically useful images at depths within and beyond the epithelium and that by changing illumination strategy, the same instrument can produce data concerning the blood supply to the same lesion. It is a common feature of malignant tumours that they require an increased blood supply to sustain their excessive growth rates. Many tumours produce angiogenic stimuli to drive the host to satisfy their needs, so the ability to examine vascular architecture and flow at the site of a lesion of doubtful basement membrane integrity further reinforces both the near immediate diagnostic, clinical screening and research potentials of this novel conventional and confocal tandem microendoscopic imaging system. The clear demonstration of the ability of the system to further generate serial optically sectioned images from uneven surfaces of semi-opaque dental restorative materials, employing confocal reflection, fluorescence and incoherent epiillumination imaging modes, indicates that further useful hard tissue research and diagnostic imaging roles may well be achievable with this instrument configuration. Typical hard and soft tissue pathological systems in which concurrent soft and mineralized or crystalized hard tissue endoscopic imaging may offer significant diagnostic and therapeutic advantages would include, for example, intraductal stones and lesions of the salivary and urinary tracts, where rigid endoscopy is currently in practical use. As the Hopkins rod endoscopes are gas-tight, the role for such diagnostic imaging could be simply extended to laparoscopic surgery too, again offering potential on the spot screening and diagnoses. A significant theoretical advance in the diagnostic and therapeutic power of these endoscopic instruments is thus demonstrably possible and within the reach of current leading edge technology. Acknowledgements The authors thank Karl Storz UK for their support and Mr Peter Pilecki at Guy s Hospital for his technical assistance. This work was conducted under MRC JREI grant G9817920. References Boyde, A., Jones, S.J., Taylor, M.L., Wolfe, L.A. & Watson, T.F. (1990) Fluorescence in the tandem scanning microscope. J. Microsc. 157, 39 49. Cotton, P. & Williams, C. (1996) Practical Gastrointestinal Endoscopy, 4th edn. Blackwell Science, Oxford, pp. 1 12. Hopkins, H.H. (1975) Improvements in or Relating to Optical Systems. UK Patent 1 534 541. Filed 30 April 1975. Petran, M., Hadravsky, M., Egger, M.D. & Galambos, R. (1968) Tandem scanning reflected light microscope. J. Opt Soc. Am. 58, 661 664. Watson, T.F. & Boyde, A. (1991) A status report on a new light microscopic technique for examining dental operative procedures and dental materials. Am. J. Dent. 4, 193 200. Watson, T.F., Juškaitis, R. & Wilson, T. (2002) New imaging modes for lenslet-array tandem scanning microscopes. J. Microsc. 205, 209 212. Xiao, G.Q., Corle, T.R. & Kino, G.S. (1998) Real time confocal scanning optical microscope. Appl. Phys. Lett. 53, 716 718.