We demonstrate the potential of a new optical imaging system to directly obtain a longitudinal slice of a biological sample. The system, based on a single-shot optical correlator, operates a time-to-space conversion and an optical time-gating by sum-frequency generation in a nonlinear crystal. Owing to the high speed acquisition of the technique, internal structures of in-vivo tissues can be imaged at video rate. With this apparatus, we recorded longitudinal images of ex vivo mouse ear and in vivo human skin with a depth resolution of approximately 15 μm.
©2002 Optical Society of America
The widespread availability of suitable laser sources and detectors has contributed to the rapid development of new optical technologies for biomedical imaging. These methods can reach a very high resolution image and/or good penetration into tissues (~ 1mm).
“Scanning” microscopy comprising linear microscopy (confocal microscopy , optical coherence microscopy (OCM) ) or non-linear microscopy (two-photon microscopy , CARS microscopy ) are widely employed for imaging biological tissues. These efficient methods use a strong focalization of the light source onto the sample, which gives a high transversal resolution, typically 1 μm, limited by the light diffraction. However, a fast scanning system is necessary to reconstruct the 2D image of the sample at a sufficient speed. For example, with a fast OCM configuration, an image acquisition of a few hertz can be obtained, corresponding to approximately 1 Mpixels/s. Nevertheless, this acquisition speed remains limited for video rate clinical diagnosis.
An alternative imaging configuration consists to directly image an entire surface of the sample, with a collimated beam, instead of using a single point as in the systems cited above. For example, optical nonlinear processes such as sum-frequency generation  or parametric amplification  make possible to instantaneously visualize the depth of a sample by time-gating imaging. Another technique using directly a 2D acquisition is the wide-field optical coherence tomography providing a high speed acquisition. However, this method is limited by the low dynamical range of the CCD camera [6,7]. Unfortunately, all these 2D techniques only provide transversal images of the sample. In general, the entire volume of the sample is reconstructed by assembling these different transversal slices. Then, longitudinal slices can be extracted to visualize a slice of the tissue from the surface to the depth, as with a histology photograph.
As a result, the necessity to directly obtain a longitudinal slice of the sample without any scanning process and without too much loss of resolution has emerged. Using ultrasonic waves, non-linear echography can solve this problem but the resolution remains too low (typically 100μm), even if a high-frequency ultrasounds (20 MHz) is used . In this paper, we present for the first time, at the best of our knowledge, an optical technique which permits to get a longitudinal slice of an in vivo biological sample at laser shot repetition rate. After a description of the principle of the method and a presentation of the imaging system, some imaging applications will be presented to demonstrate the capacities of the system for realtime imaging.
2. Principle and imaging setup
The method is based on the so-called “time-gating” imaging which makes it possible to discriminate between early arriving ballistic photons and time-delayed scattered photons. Our system employs a femtosecond sum-frequency generation in a non-linear crystal between a strong pump pulse and the backscattered probe pulse. In that case, the nonlinear interaction between the pump pulse at optical frequency ω and the probe pulse at the same frequency generates a sum-frequency signal at frequency 2ω, which corresponds to the cross-correlation between the pump and the probe pulses. Using the sum-frequency generation in the nonlinear crystal, the imaging system can provide in a single laser shot procedure a longitudinal image of the sample.
The system is based on the ability of a classical single-shot autocorrelator, widely used to measure the duration of picosecond and femtosecond laser pulses, to operate a time-to-space conversion. After being reflected by a biological sample, a probe pulse supports a temporal stretching due to the different internal reflections occurring at different depths. In the nonlinear crystal, if the backscattered probe pulse interacts with the pump pulse in a non-collinear configuration, the temporal stretching can be converted into a spatial stretching.
As an example, Fig. 1 illustrates the non-collinear interaction in the nonlinear crystal between the pump pulse and two probe pulses delayed by the time Δτ, symbolizing two reflections of the incident probe pulse from the sample at two different depths. As a result, two sum-frequency signals are generated in the direction corresponding to the phase matching conditions. These two signals are translated in space by a quantity Δz given by:
where n0(λ) is for the refractive index of the nonlinear crystal at the wavelength λ, c is for the speed of the light in vacuum and Φ is for the angle between the two incident pulses on the crystal. This formula establishes the time-to-space conversion operated by the method. The direction of the spatial shift Δz corresponds to the first coordinate (Z-axis) of the longitudinal image obtained by a CCD camera collecting the sum-frequency light emerging from the crystal. This coordinate simply represents the depth of the sample. Along this coordinate, the sample surface and the internal sub-structures can be visualized.
To understand the acquisition of the second coordinate of the longitudinal image of the sample, Fig. 2 represents the whole imaging configuration of the system. The light source is an amplified Ti:sapphire laser emitting pulses as short as 40 fs with a central wavelength at 800 nm, an energy up to 400 μJ and a pulse repetition rate of 1 kHz . The diameter of the output beam is approximately 1 cm. The output of the laser is divided into two identically polarized beams (pump and probe beams) by the beam splitter (BS). The probe beam is focalized with a cylindrical lens (CL1, f=40 mm) onto the sample. The illuminated area is roughly (1 cm × 20 μm) and the incident power is about 200 mW. It means that a light line is formed at the sample surface. The direction of this line represents the second coordinate (X-axis) of the 2D longitudinal image acquired by the CCD camera. To get this image, the illuminated area of the sample is imaged by the spherical lens (L2, f=100 mm) onto a 1mm BBO crystal. The pump pulse passes through an optical delay line (DL) and is combined with the backscattered probe pulse in the crystal. The angle between the two beams in the air is Φ=30°. In the crystal, the previously described non-collinear sum-frequency generation is operated between the backscattered probe pulse and a pump pulse. The temporal overlap of the two pulses can be adjusted with the optical delay line. Finally, a second spherical lens (L3, f=100 mm) is used to collect the sum-frequency light at 400 nm and to image the crystal plane onto a Hamamatsu CCD camera (12-bit, 1280*1024 pixels, 9 frames/s, cooled at +5°C). As the configuration is non-collinear, after filtering of the fundamental light at 800 nm, the sum-frequency signal is detected without any background noise. Taking into account 12-bit acquisition, the dynamical range of the detection system is SD=20log(212)= 72dB. It’s worth to stress that it is a background free measurement, consequently all the dynamical range of the detection system is used to measure the sum-frequency signal.
The optical setup makes it possible to image at laser shot repetition rate. In our case, a 1 kHz acquisition rate could be achieved. However, the Hamamatsu camera does not allow such high speed acquisition. Nevertheless, the 10 frames/s acquisition rate and the 1024×1024 pixels image size provide a nearly 10 Mpixels/s acquisition speed, which is already 10 times faster that the fastest OCM system available, as mentioned in the introduction. The maximal depth range that we can image in a single short procedure, e.g. without changing the optical pathway of the pump beam with the delay line can be find using the formula :
Zmax equals to 450 μm, limited by the A=5 mm aperture of the BBO crystal and the angle Φ=30° between the two incident beams for an average skin refractive index nsample=1.45.
To determine the imaging resolution of our system, we placed a calibrated U.S. Air force Test Pattern (AFTP) at the sample position. A transversal resolution of approximately 35 μm has been measured by analyzing the image obtained by the CCD camera. This resolution is limited by the numerical aperture of the imaging lenses L2 and L3. Also, we evaluated the longitudinal resolution (depth resolution) in the air. For that purpose, we calibrated the Z-axis by intercepting half of the transverse profile of the pump beam by a thin glass plate (e=155 μm) introducing a time-delay Δt = (n-1)e/c (n is for the refractive index of the glass) within the transverse profile of the pump pulse. By measuring the FWHM of the reflected signal, we evaluated the depth resolution of the system to approximately 20 μm in the air. Considering a refractive index of 1.45 for a biological sample, the depth resolution into the tissues can be reduced to 15 μm. This result is close to the theoretical depth resolution of the system described in details in another paper, which will be published elsewhere . This resolution is determined by the pulse duration of the output laser pulses and the dispersion (group velocity dispersion and higher order dispersion terms) of both the pump and probe pulses traveling through the optical components of the system.
The sensibility was determined by imaging a calibrated U.S. Air Force Test Pattern (USAF-TP) in a 1% Intralipid solution. The USAF-TP was imaged up to 2.5 mm in the Intralip solution, which corresponds to an attenuation of the ballistics photons by a factor of 4.10-6 and consequently to a sensibility of S=10log(4.10-6)= 54dB. The sensibility was principally limited by the reference pulse second-harmonic scattering in the nonlinear crystal.
3. Ex vivo imaging of a mouse ear
In order to test the depth resolution, we did an ex vivo imaging of a mouse ear, which had a thickness of about 330 μm. The symmetrical organization of the mouse ear makes it possible to describe the constitutive components of the two opposite sides, separated by a dividing plane consisting of the median cartilage. In Fig. 3(a), the photograph obtained by photonic microscopy of hematoxylin-eosin stained histology (HE) reveals the central cartilage (C) wrapped in a dense conjunctive capsule (cc), the epidermis (E) with the stratum corneum (sc) and the dermis (D) with internal sub-structures.
Using the longitudinal imaging system, we are able to reveal the constitutive tissues of the sample (Fig. 3(b)). The probe beam is incident from the top of the figure. The central black band obviously corresponds to the cartilage (C). A double high intensity band, corresponding to the dense conjunctive capsule (cc) surrounds this region. Symmetrically to this region, two high intensity bands are recognizable, corresponding to the superficial epidermis (E) on each side of the ear. Finally, the dermis (D) is also clearly identified, on each side of the cartilage, with some oblong areas of low intensity which may be attributed to scattering structures usually found in the dermis, such as blood vessels or extra cellular matrix. This analysis proves that the non-invasive imaging system makes it possible to observe the internal constitutive tissues of a biological sample, with a sufficient depth resolution.
4. In vivo imaging of human skin
The single shot longitudinal imaging system has also been tested for imaging in vivo biological tissues. In dermatology, for example, it is of primary importance to develop non-invasive methods to study the different layers of human skin from the stratum corneum up to the underlying dermis. Also, detection of tumors such as malignant melanoma have justified many efforts.
A photograph obtained by photonic microscopy of hematoxylin-eosin stained histology (HE) reveals the different constitutive tissues of an ex vivo human skin (Fig. 4(a)). The epidermis (E), with the stratum corneum (sc) at the interface air-sample, and the dermis (D) are clearly distinguished.
With the longitudinal imaging system, we have imaged the skin of a volunteer in the region of the forearm (Fig. 4(b)). For this experiment, the forearm of the volunteer is just put on the focalization line of the cylindrical lens CL1. This image is recorded in real-time at 9 frames/s, limited by the acquisition rate of the Hamamatsu digital camera. The probe beam is incident from the top of the figure. At the surface of the skin, the stratum corneum (sc) is clearly visible, represented by the high intensity band. Obviously, the depth resolution is not enough to measure correctly the thickness of this layer. Then, the rest of the epidermis (E) is observed, represented by the darker band. Finally appears a brighter area corresponding to the underlying dermis (D). Especially, the epidermal-dermal junction is clearly identified.
An intensity profile along the line indicated in Fig. 4(b) makes it possible to precisely measure the thickness of the epidermis at this position. The evolution of the image intensity along that line is plotted in Fig. 4(c). The thickness of the epidermis is 130 μm at this position and internal structures can be observed inside the epidermis at a depth of 40 μm below the skin surface. Then the intensity slowly decreases in the dermis up to 400 μm in depth. Moreover, the signal intensity in the epidermis is lower than in the dermis, indicating that the probe light can propagate more easily in the epidermis than in the dermis. This shows that the scattering coefficient μs of the epidermis is smaller than that of the dermis, in agreement with previous studies concerning the optical properties of the human skin .
As the longitudinal imaging system makes it possible video rate imaging, the skin of the volunteer can be observed in real-time. Fig. 5 is a movie recording of the experiment. Even if the volunteer moves his arm during the recording, the images appears still clear and well contrasted owing to the single laser shot acquisition of the images. On a practical point of view, this constitutes an important characteristic of the imaging system, since the patient does not need to be completely immobile during the imaging session.
For the first time to our knowledge, we present an optical system enable to operate real-time 2D longitudinal imaging of biological tissues. The system, based on a single-shot correlator and a femtosecond laser source, has been used to image biological tissues at video rate demonstrating that the method could be perfectly adapted for video rate clinical diagnosis, especially in the domain of dermatology, odontology, etc.
The authors thank Prof. J.E. Surlève-Bazeille (Laboratoire Facteur de Défense et Régulation Cellulaire, Bordeaux I University) for the HE histology study of the human skin, and Dr. F. Salin (CELIA, Bordeaux 1 University) for the use of the femtosecond Ti:sapphire laser source.
References and links
1. B.R. Masters and P.T.C. So, “Confocal microscopy and multi-photon excitation microscopy of human skin in vivo,” Opt. Express 8, 2–10 (2001), http://www.opticsexpress.org/oearchive/source/27001.htm [CrossRef] [PubMed]
2. D. Huang, E.A. Swanson, C.P. Lin, J.S. Schuman, W.G. Stinson, W. Chang, M.R. Hee, T. Flotte, K. Gregory, C.A. Puliafito, and J.G. Fujimoto, “Optical coherence tomography,” Science 254, 1178–1181 (1991). [CrossRef] [PubMed]
3. A. Zumbusch, G.R. Holtom, and X.S. Xie, “Three-dimensional vibrational imaging by coherent anti-Stokes Raman scattering,” Phys. Rev. Lett. 82, 4142–4145 (1999). [CrossRef]
4. E. Abraham, E. Bordenave, N. Tsurumachi, G. Jonusauskas, J. Oberlé, C. Rullière, and A. Mito, “Real-time two-dimensional imaging in scattering media by use of a femtosecond Cr4+:forterite laser,” Opt. Lett. 25, 929–931 (2000). [CrossRef]
5. G. Le Tolguenec, E. Lantz, and F. Devaux, “Imaging through scattering media by parametric amplification of images: study of the resolution and the signal-to-noise ratio,” Appl. Opt. 36, 8292–8297 (1997). [CrossRef]
6. S. Bourquin, P. Seitz, and R.P. Salathé, “Optical coherence tomography based on a two-dimensional smart detector array,” Opt. Lett. 26, 512–514 (2001). [CrossRef]
7. E. Bordenave, E. Abraham, G. Jonusauskas, N. Tsurumachi, J. Oberlé, C. Rullière, P.E. Minot, M. Lassègues, and J.E. Surlève Bazeille, “Wide-field optical coherence tomography: imaging of biological tissues,” Appl. Opt. (in press).
8. S. Diridollou, M. Berson, V. Vabre, D. Black, B. Karlsson, F. Auriol, J.M. Gregoire, C. Yvon, L. Vaillant, Y. Gall, and F. Patat, “An in vivo method for measuring the mechanical properties of the skin using ultrasound,” Ultrasound in Medicine & Biology 24, 215–224 (1998). [CrossRef] [PubMed]
9. V. Bagnoud and F. Salin, “1.1 Terawatt, kilohertz femtosecond laser,” in Technical Digest CLEO ’99, (Institute of Electrical and Electronics Engineers, New York, 1999), pp. 71–72.
10. E. Bordenave, E. Abraham, G. Jonusauskas, J. Oberlé, and C. Rullière, “Single-shot correlation system for longitudinal imaging in biological tissues,” submitted to Opt. Commun.
11. C.R. Simpson, M. Kohl, M. Essenpreis, and M. Cope, “Near-infrared optical properties of ex vivo human skin and subcutaneous tissues measured using Monte Carlo inversion technique,” Phys. Med. Biol. 43, 2465–2478 (1998). [CrossRef] [PubMed]