Systems and methods for microscopic analysis of biological material have been used for characterization and diagnosis in many applications. Fluorescence microscopy, for example, has been used for optical analysis including the histological analysis of excised tissue specimens. Optical coherence tomography has been used for three dimensional imaging of tissue structures, however, the limited resolution of existing systems has constrained its use for definitive pathological analysis. Confocal microscopy has been used for high resolution imaging and has controllable depth of field but limited imaging speed.
Multiphoton microscopy is based on the nonlinear excitation of fluorophores in which fluorescence generation is localized at the focus of excitation light. Multiphoton microscopy is used for deep tissue imaging because of its subcellular three dimensional (3D) resolution, minimal phototoxicity, and tissue penetration depth of over a few hundred micrometers. It has become useful in biomedical studies such as neuronal plasticity, angiogenesis in solid tumors, transdermal drug delivery, and non-invasive optical biopsy, for example.
A practical limitation of multiphoton microscopy is its imaging speed which typically lies in a range of less than two frames per second. While this speed is sufficient in many cases, there remain applications in which can be enhanced by improvements in imaging speed. There is a continuing need for further improvements in microscopic analysis of biological materials for numerous applications.
The present invention relates to systems and methods for the multifocal imaging of biological materials. An optical system is provided in which a plurality of optical pathways are used in combination with focusing optics to provide a plurality of focal locations within a region of interest of a material being optically measured or imaged. The detector can comprise a plurality of detector elements which are correlated with the plurality of focal locations to provide for the efficient collection of light from the material being imaged. A preferred embodiment of the invention utilizes a scanning system that provides relative movement between the material and the focal locations to provide for fast imaging of the material.
In a preferred embodiment a light source, such as a laser, is used with a multifocal optical element to provide an array of spatially separated optical pathways. The multifocal optical element can comprise a micro lens array, a diffractive optical element, or a beam splitter device, for example, such that a plurality of beams are provided that can be focused onto a plurality of focal locations within a biological material to be imaged.
An important issue in the collection of light from discrete focal spots or locations within a turbid medium such as tissue is the cross talk that can occur due to the scattering of light. This cross talk can substantially limit the usefulness of the images of the tissue that are produced. By increasing the distance between adjacent focal spots such cross talk can be reduced or eliminated, however, this reduces the resolution of the resulting image or increases the time needed to scan the tissue. Thus it is desirable to employ focal spacing of at least 10 microns and preferably more than 25 microns.
In a preferred embodiment of the invention, high speed multiphoton microscopy can measure biological systems such as, for example, kinetic processes in the cytosol of a single cell, for example, or imaging a volume of tissue. For example, high speed 3D imaging can map 3D propagation of a calcium wave and the associated physical contraction wave through a myocyte, or the rolling of luckocytes within the blood vessel of a solid tumor. High speed 3D microscopy provides for sampling a statistically significant volume of biological specimens. Since the field of view of most microscopes is limited to about 100 microns on a side with an imaging depth of 100 microns, the measurement volume is limited to only 1×10−3 mm3. While this volume is sufficient for cellular imaging, many tissues have physiologically relevant structures ranging from the cellular level up to several millimeters in size. For example, a neuron with its extensive dendritic tree can span a volume over 1 mm3 and many dermal structures such as hair follicles and sabestious glands can not be seen with images confined to an area of 100-200 micrometers. It is desirable, for example, to image a hierarchy of cardiac structures ranging from a single nucleus in a cardiac myocyte, to the distribution of muscle fibers and blood vessels, to the structure of chambers and heart valves with perfect registration across five orders of magnitude by imaging a whole mouse heart. Equally importantly, traditional 3D microscopes sample only tens to hundreds of cells and can never achieve comparable statistical accuracy and precision in many biomedical assays as techniques such as flow cytometry and image cytometry. High speed imaging can circumvent this difficulty by improving the number of cells or tissue volume to be sampled. By performing high speed multiphoton imaging, better quantitative measurements of transport pathways across the stratum corneum in transdermal drug delivery applications can be made, for example.
Systems and methods have been developed to enhance multiphoton imaging speed. A first method increases the scanning speed by using a high-speed scanner such as a polygonal mirror scanner or a resonant mirror scanner instead of a galvanometer-driven mirror scanner. This achieves an increase of scanning speed of more than 10 frames per second in the imaging of typical tissue specimens. In general, the system can operate at frequencies in a range of 1 to 500 Hz. This method can be used for turbid tissue imaging since it is not sensitive to the scattering of emission photons. A second method increases the imaging speed by parallelizing the multiphoton imaging process. It scans a sample with a multiple of excitation foci instead of forming only a single focus. These foci are raster scanned across the specimen in parallel where each focus needs to cover a smaller area. The emission photons from these foci are collected simultaneously with a spatially resolved detector. One advantage of this method is that the imaging speed is increased by the number of excitation foci generated, without increasing the power of excitation light per each focus. High speed scanning systems needs higher power to compensate for the signal reduction per pixel due to the decrease of pixel dwell time. Images can be obtained by selecting the depth of focus to be positioned in a plane within the tissue or sample at a depth in a range of 10 microns to 500 microns.
In another embodiment, fiber optics can be used to couple the light source to the microlens array or other beam splitting element. The system can be implemented as a handheld optical probe for the diagnosis of dermal, cervical or colorectal cancer, for example.
The brain is an inherently three dimensional organ composed of many subregions. Accurate segmentation of brain morphology of small mammals is currently challenged by the lack of techniques which can sample the brain at high resolution over a large volume. The current method of choice, serial section reconstruction, is laborious, time consuming, and error prone. The device and methods described herein can quickly image brains or thick tissue sections of brains in 3D at sufficient resolution and over a large enough volume to provide 3D images suitable for classification of brain morphology and biochemical composition. The brain can be further stained by dyes, such as nuclear dyes DAPI or Hoescht, either through intravital injection, transgenic expression, or ex vivo methods, to facilitate classification of regions. Automatic segmentation routines can also be used to improve the classification and automate portions of the process.
Accurate measurement of vasculature is important to characterize many biomedical for vasculature related diseases. For instance, proangiogenesis therapies are useful in such areas as tissue engineering, wound healing, bone fractures and coronary heart disease. Anti-angiongenesis treatments are important in processes as cancer, blindness, and rheumatoid arthritis. Unfortunately traditional histopathological analysis of tissue sections is wholly inadequate to characterize the vasculature of a tissue or organ as blood vessels form complex, multiscale 3D networks, with feature spanning from the submicron to centimeter scale. The device and methods described in the patent are capable of acquiring high quality 3D datasets over 3D tissue and organ samples suitable for characterization of the vasculature of the tissue. To aid visualization of the vasculature, the tissue can be stained by contrast agents which bind to the epithelial wall of the blood vessels, or fill the interior of vessels. Automatic segmentation routines can also be used to improve the classification and automate portions of the process.
A large percentage of deaths are due to metastasis. Unfortunately, the migration of cancer cells from the primary tumor to secondary sites is a multi-step process which is not well understood. Standard histopathological analysis is ill-suited to study metastasis and suffers from a number of limitations. First, it is extremely difficult to find rare metastatic cancer within a 3D bulk tissue using traditional 2D histopathology. In many instances traditional 2D histopathology is unable to find evidence of the presence of metastatic cancer cells in an organ of animal. However, it is known that many subjects eventually develop tumors at a later time. It is clear that traditional histopathology cannot effectively detect rare cells. Another limitation is that the present histopathology methods provide limited information about the 3D spatial arrangement of cancer cells with the 3D vasculature of the organ. It is known that one of the critical steps in metastasis is extravasation into the surrounding stroma from the vasculature so it is essential to be able to visualize this spatial relationship between cancer cell and the endothelial blood vessel wall. Preferred embodiments of the present invention are capable of acquiring high quality 3D datasets over 3D tissue and organ samples suitable for characterization of the metastases. To aid visualization of the metastases, the cancer cell can be stained by dyes or labeled with proteins such as OFP. Automatic segmentation routines can also be used to improve the classification and automate the localization of the cancer cells and tumors.
In order to understand the effects of a drug on an organism, analysis at the tissue, whole organ, and whole organism level is vitally important. ADME, efficacy and toxicology effects are known to have strong spatial variations on the morphological, cellular and biochemical state of a tissue. Even within a specific tissue type, the response can be nonuniform due to variations in the transport and distribution of a drug throughout tissue, epigenetic expression, and cellular activity. The devices and methods described herein can be used to provide morphological, biochemical and spectroscopic information about the state of a tissue across multiple length scales, from subcellular, whole tissue, whole organ and even entire organism, in response to the treatment of a molecular agent. Efficacy, ADME, and toxicology information can be derived which provides a fuller and more accurate description to predict the actual effect of drug candidate at the organism level.
As the input power of excitation light increases, the signal is increased proportionally to the square of input power, S(t)∝[P(t)]2. However, there is a limitation in input power level due to finite lifetimes of fluorophores.
In the multiphoton excitation of fluorophores with a pulsed laser, the fluorophores, which are excited with the last pulse, stay in the excitation state for a few nano-seconds (depending on the fluorophore). Some excited fluorophores may not be excited again with the next pulse of excitation light (12 ns later in case of the laser having 80 MHz pulse repetition rate). Therefore, signal level becomes saturated with a higher input power than the limited input power level. The limitation on the input power level is related to the excitation probability of a single fluorophore with a single pulse, Ppulse. It is formulated in the following expression with the condition that excitation light is focused with an objective into a fluorophore of an absorption coefficient (δa).
The nominal conditions are that the excitation light has the wavelength (λ=800 nm), the pulse width (rτ=200 fs), the pulse repetition rate (fP=80 MHz), and the average power, Po(t). The numerical aperture (NA) of the lens objective is 1, (NA=1). The fluorophore has the absorption coefficient, (δa=10 GM), where 1 GM is 10−50 cm4×s/photon. In order to avoid the saturation, Ppulse must be less than 0.1 in general (Ppulse<0.1). With these conditions, the input power (Po(t), with which Ppulsebecomes close to the saturation limit, is approximately 6 mW, (Posat(t)≅6 mW) in this example. In case the concentration of the fluorophores is 10 μM, the number of emission photons collected per second is approximately 3×107 photons/s with the assumption that the collection efficiency of emission photons is approximately 0.01 (εcol=0.01). Assuming that each pixel needs 300 photons and each image comprises of 256×256 pixels, the frame rate that can be achieved with the input power under the saturation limit is 1.5 frames/s. Although the higher frame rate is achieved with specimens of higher fluorophore concentration, it is clear that there is a limitation in input power level due to fluorophore saturation.
MMM increases the frame rate by scanning with multiple excitation foci. Therefore, MMM can achieve the higher frame rate, while the input power for each excitation focus is kept below the saturation limit. For example, the MMM system, which scans with an 8×8 array of excitation foci, can achieve the frame rate of 96 frames/s (=1.5 frames/s×64 foci). In a preferred embodiment it is desirable to collect at least 15 frames per second and preferably 30 frames per second or more. One practical limitation in MMM is that more input power is required to generate multiple excitation foci. The power requirement to generate 64 foci is 384 mW (=64 foci×6 mW per each focus). Since available laser sources can output approximately 2 W of power, enough power is available for MMM.
The limit of optical imaging depth in tissues is limited by photon interaction with tissue constituents. Photon scattering is a dominant factor in multiphoton microscopy whereas the effect of photon absorption is relatively negligible. Scattering of excitation photons reduces the amount of fluorescence generated at its focus, because less excitation photons reach the focal volume. The emission photons from the focus are also scattered so that they may not be collected by the optics in the detection path or spatially dispersed in the imaging plane where detectors are positioned. Since the excitation light has a longer wavelength than the emission light, the excitation light typically experiences less scattering than emission light. The effect of photon scattering is expressed by the mean free path length of scattering, P which is the depth constant in exponential decay of unscattered photons, S(z)∝ exp(−z/ls).
Intralipid emulsion can be used as a tissue phantom with similar optical properties as tissue. The optical properties of 2% intralipid are mean free path length at excitation wavelength (780 nm) of 167 μm, (lexs≅167 μm) and at emission wavelength (515 nm) of 62.5 μm, (lems≅65 μm). Since it is known that only ballistic excitation photons contribute multiphoton excitation in the depth of a few times of scattering length, the amount of multiphoton excitation decays with the mean free path length of 84 μm (=167 μm/2) with the consideration that two-photon excitation is a quadratic process. Conventional multiphoton microscopy is based on the scanning of a single excitation focus and the signal is collected using a detector without spatial resolution such as a conventional photomultiplier tube (PMT). The PMT has a large detection area and can collect most of the signal generated at the excitation focus including a large fraction of the scattered photons. Therefore, conventional multiphoton microscopy is relatively immune to the scattering of emission photons by the tissue. However, for an MMM system that utilizes a CCD detector to distinguish the signals originated from each of the foci, the scattering of emission photons seriously degrades the SNR of the instrument for deep tissue imaging. The CCD camera has relatively slow readout speed and typically integrates all the emission photons during the acquisition of each frame. Because a CCD camera contains pixels in which each pixel covers a 0.1 μm2 region in the specimen plane, scattered emission photons deflected from their original paths are not collected in the correct pixel but are distributed broadly across the imaging plane. The distribution of scattered emission photons is very broad with its FWHM of 40 μm in the depth of 233 lems. These scattered photons result in a degradation of image SNR by more than one order of magnitude when imaging depth is over 2×lems, compared with conventional multiphoton microscopy.
The major limitation of CCD-based MMM system lies in its small pixel area. For conventional wide field imaging, a large number of CCD pixels are needed to maintain good resolution while covering a good size field of view. A 100 □m size image will require about 107 pixels to be imaged at full optical resolution (300 nm). The situation is very different for MMM imaging. Since a femtosecond light source can only provide at maximum 2-4 watts of optical power and typically about 50-100 mW are required at each focus to generate an efficient muliphoton excitation process for deep tissue imaging. An MMM system can realistically and effectively scan about 20-40 foci in parallel with tissue specimens. Since these foci are raster scanned across the specimen, the image resolution is determined by the excitation point spread function (PSF) of the light and is not sensitive to the detector pixelation. In particular, a preferred embodiment uses an MMM system having photon detectors containing only as many elements as the number of excitation foci. The need for fewer elements allows the use of a detector with a significantly larger pixel area while maintaining a reasonable device size. A multi-anode PMT (MAPMT) is a preferred detector for this purpose.
A preferred embodiment of the present invention uses an MAPMT instead of the CCD camera for the signal collection from multiple foci. The MAPMT is similar to conventional PMTs with a good quantum efficiency (over 20% in the blue/green spectral range), negligible read noise and minimal dark noise with cooling. MAPMT has a cathode and dynode chain with a geometry that ensures that the spatial distribution of photons on the cathode is reproduced accurately as electrons distribution at the anode. The anode of the multi-anode PMT is divided rectilinearly into its elements providing spatial resolution for the simultaneous collection of signals from multiple locations. In one example, a MAPMT, which has an array of 8×8 pixels (H7546, Hamamatsu, Bridgewater, N.J.) is used. Note that a flat panel detector having a pixel area of sufficient size can also be used. For example, a binnable CMOS or CCD imaging sensor can be operated to read out binned images at comparable frame rates with an effective pixel size corresponding to that of a MAPMT.
A preferred embodiment of the invention uses the imaging systems as described herein in conjunction with a system for sectioning a sample such as a tissue sample that is described in greater detail in U.S. patent application Ser. No. 10/642,447, by So, et al. filed Aug. 15, 2003, the entire contents of which is incorporated herein by reference.
The schematic of a preferred embodiment of the imaging system 10 in accordance with the invention is shown in
Emission photons are generated at the array of excitation foci in the specimen and are collected by the same objective lens forming an array of emission beam-lets. In case of a CCD-based MMM, the emission beam-lets are reflected on a long-pass dichroic mirror 38 (650dcxxr, Chroma Technology, Brattleboro, Vt.) and get focused in optional CCD camera 28 (PentaMax, Princeton Instruments, Trenton, N.J.) with a lens (L3). The CCD camera integrates emission photons during the scanning time of each frame to generate images. In case of a preferred embodiment using an (without the CCD) MAPMT, the emission beam-lets travel back to the scanner mirror 30 retracing the excitation paths. The emission beam-lets are reflected by the scanner mirror. The emission beam-lets are de-scanned and their propagation directions remain stationary irrespective of the movement of the scanner. The emission beam-lets are reflected by a long-pass dichroic mirror 32 (650dcxxr, Chroma Technology, Brattleboro, Vt.) and are focused after lens (L4). A short-pass filter (E700SP, Chroma Technology, Brattleboro, Vt.) blocks any strayed excitation light. The focused emission beam-lets are collected at the center of corresponding channels of a MAPMT 22 (H7546, Hamamatsu, Bridgewater, N.J.). The emission photons coming from the array of excitation foci are collected by the MAPMT. An image is formed by the temporal encoding of the integrated signal with the known raster scanning pattern using image processor or computer 24 and is electronically stored in memory and/or displayed using display 26.
The pair of L2 and L4 lenses magnifies the array of emission foci so that individual emission beamlets are focused at the center of corresponding elements of the MAPMT. Further, since the emission beam-lets are descanned, they remain stationary. Since the emission beam-lets have to go through more optical elements, loss of emission photons occurs. The transmission efficiency is approximately 0.7. The signals from the MAPMT are collected by a stack of four multi-channel photon counter card (mCPhC) which has 64 channels for simultaneous signal collection. Each mCPhC has 18 channels of photon counter circuits and can be housed 25 with a digital interface to the computer 24. The mCPhC expandable so that 64 channels are readily implemented by using 4 cards in parallel. The mCPhC has a 32-bit parallel interface with a computer for high-speed data transfer. Currently, the speed is limited by the speed of the computer PCI bus. Transfer rate can be more than one hundred frames (320×320 pixels, 16 bit images) per second.
Since the scattered emission photons have the spatial distribution of 40 μm as its FWHM at the imaging depth of 2×lems, the sensitivity of the microscope is partly determined by the effective detector area, the area in the sample plane from which a detector collects emission photons. Since microscopes are telecentric systems, the effective detector area is linearly related with the detector size in the image plane. With a magnification, M, and a linear dimension of detector, LD, the linear dimension of effective detector area (LE) is LE=LD/M. In general, the larger the effective detector area, the more effective the detector can collect scattered emission photons. In the case of using a 20 × magnification objective, a 10 mm diameter standard PMT has an effective detector area of 500 μm diameter that is significantly larger than the width of the scattered emission photon distribution. Therefore, standard PMTs have good collection efficiency of scattered emission photons and allow very effective deep tissue imaging. In the case of a spatially resolved detector, each pixel can be treated as an individual detection element. For a CCD camera with 20 μm×20 μm pixels, each pixel has an effective detector area of 1 μm×1 μm for 20 × magnification. Therefore, the CCD-based MMM system cannot utilize these scattered emission photons which are distributed uniformly throughout the image contributing to the background noise. In this example of the MAPMT-based MMM system, the effective detector area of each channel is 45 μm×45 μm. Therefore, the MAPMT can collect significantly more scattered emission photons into the correct channels than the CCD camera, because its effective detector area, or detector element collection area, is comparable with, or corresponds to, the width of the scattered photon distribution from each focal area (45 microns×45 microns). MAPMT-based MMM system can be easily converted to the conventional multiphoton microscope which is based on single-focus scanning and signal collection with PMTs. In the set up of conventional multiphoton microscope, the excitation beam is not expanded and goes directly onto the scanner without the combination of the microlens array and lens Li. The rest of excitation beam path is the same as MAPMT-based MMM. Specimens are scanned with a single excitation focus. The emission light collected by the objective lens is reflected on a dichroic mirror. The reflected emission beam shrinks with a pair of lenses and is collected by a detector (PMT). An image is formed by the temporal encoding of the integrated signal with the known raster scanning pattern.
CCD-based MMM have limitations for turbid tissue imaging by measuring the effect of emission photon scattering on PSF (scattering function). Scattered emission photons form additional intensity distribution around the PSF constructed with ballistic unscattered photons. Their intensity distribution is quite wide with its FWHM of 40 μm at the imaging depth of 2×lems. FWHM of the total PSF (including the intensity distribution due to scattered emission photons) is not changed due to scattering up to such depth because the wide distribution of scattered emission photons does not contribute to FWHM. In terms of contrast, signal decay in CCD-based MMM with the increase of imaging depth is higher than that of SMM by an order of magnitude at 2×lems. Also the wide distribution of scattered emission photons contributes as noise and causes loss of contrast by another order of magnitude at the depth.
Imaging dermal structure based on autofluorescence has been performed using the system of the present invention. Endogenous fluorophores have low quantum yield and low extinction coefficients compared with typical exogenous fluorescent labels. The dermal structure imaged using a preferred embodiment of the present invention has a layered structure with significantly different indices of refraction resulting in significant spherical aberration. Multiphoton imaging of dermal structures without photodamage has a pixel rate of 15 KHz with 15 mW input power. In this example an input power of 7 mW per focus at the specimen with the excitation wavelength set at 800 nm. The objective used is 20×water immersion with 0.95 NA (XLUMPLFL20XW, Olympus, Melville, N.Y.). With a frame rate of 2.5 fps for a 320×320 pixel image (4 KHz pixel rate) this is 10 times faster than the previous systems. The epidermis is imaged down to the basal cell layer using this MAPMT-based MMM. Representive layers from the stratum corneum, stratum granular, and the basal layer are shown in
Using both CCD and MAPMT detectors in MMM geometry, the signal decay can be measured as a function of scattering length. As the imaging depth increases, the signal is decreased due to scattering of both excitation photons and emission photons. The signal decay is measured by imaging 4 μm diameter fluorescent latex microspheres (F8858, Molecular Probes, Eugene, Oreg.) immobilized in 3D by 2% argarose gel (UltraPure Low Melting Point Argarose, Invitrogen, Carlsbad, Calif.). Intralipid emulsion (Liposyn III, Abbott Laboratories, North Chicago, Ill.) is added to the sample as a scatterer in various concentrations of 0.5 to 2%. Intralipid emulsion of 2% volume concentration is known to have similar scattering properties to those of tissues: mean free path length (ls) of scattering is 80 μm, 168 μm at the wavelength of emission (605 nm), excitation (800 nm) respectively. The scattering properties of these intralipid solutions are verified by diffusive wave measurements. Peak intensity of the sphere image is a signal in the measurement and the decay of peak intensity as a function of the imaging depth is measured at each concentration. The signal decay can also be measured with a conventional multiphoton microscope as a reference. Signal decays in the three systems are measured down to a depth of 180 μm which is equivalent to 2.25×lems (
Although a significant portion of the scattered emission photons are still distributed outside the correct pixels in MAPMT-based MMM, these photons can be effectively restored to the correct pixels based on post-acquisition image processing. Note that the photons acquired at each pixel are temporally encoded and are organized to form an image based on the known scanner position as a function of time. This is exactly how images are formed in a conventional multiphoton or confocal microscope. A primary image is formed by photons acquired at the correct pixels corresponding to the fluorophore distribution in that portion of specimen. Note that the scattered photons in the neighboring pixels are also similarly temporally encoded. Therefore, secondary “ghost” images are formed in the areas of the image covered by the neighboring pixels. As an example,
{Shd s}esl=[C]esl×{Sacq} (2)
The restored image is presented in
The performance comparison of the two MMM systems can also be evaluated for the imaging of biological tissues. The specimen is an ex-vivo brain tissue section with neurons expressing green fluorescent protein (GFP). Thy1-GFP transgenic mice are deeply anesthetized with 2.5% Avertin (0.025 ml/g i.p.) and transcardially perfused with PBS, followed by 4% paraformaldehyde. Brains are dissected and placed overnight in cold 4% paraformaldehyde. 1-mm thick coronal sections are taken by vibrotome, mounted and coverslipped on microscope slides using adhesive silicone isolators (JTR20-A2-1.0, Grace Bio-Labs, Bend, Oreg.). The specimen is imaged in 3D with both CCD-based MMM and a MAPMT-based MMM. The objective used is 20×water immersion with NA 0.95 (XLUMPLFL20XW, Olympus, Melville, N.Y.). The input laser power is 300 mW at 890 nm wavelength. The frame rate is 0.3 frames per second with 320×320 pixels. The slow frame rate is set in order to collect enough emission photons up to 120 μm deep. The total imaging depth is 120 μm with 1.5 μm depth increment. Representative images are shown in
However, additional improvements of this system can be made. First, since the MAPMT is positioned in the image plane, the location of each excitation focus corresponds to the center position of the matching pixel of the MAPMT. The effective detector area scales quadratically with the separation of the foci. Therefore, with wider foci separation, the MAPMT has higher collection efficiency for scattered emission photons. In the current configuration, the excitation foci are separated from each other by 45 μm so that the effective detector area for each channel of the MAPMT is 45 μm×45 μm. The size of imaging field with 8×8 foci becomes 360 μm×360 μm. As the excitation foci are separated more, the system becomes less sensitive to the scattering of emission photons. The maximum separation of excitation foci is limited by either the field of view of the objective or apertures of other collection optics. The 20×water immersion objective used has the field of view of 1000 μm in diameter. This allows positioning the foci as far apart as almost 100 microns in this example.
A limitation of the MAPMT-based MMM system compared with a CCD-based MMM design is that the signals are de-scanned. In the de-scanned configuration, emission photons are processed by more optical elements including the scanner mirror before they are collected at the MAPMT suffering more optical loss at each reflection. Further, the de-scanned geometry also has a longer optical path that contributed to the loss of some scattered photons due to the finite aperture of the optics. The signal collection efficiency is approximately 70% in this example due to additional optical elements. An MAPMT-based MMM system in a non-de-scanned geometry for example can recover this loss.
The MAPMT is manufactured with a current quantum efficiency of about 20% compared to 80% quantum efficiency of the CCD camera. However, MAPMT has very low noise. It has 20 dark counts per second without cooling and can be several orders of magnitude lower with cooling. Since the MAPMT has a readout rate of approximately 20 KHz, the typical dark count per pixel. is less than 1×10−3. In comparison, the CCD noise is dominated by both read noise and dark noise which are a few counts per pixel. Therefore, for very low photon count situation, i.e. dim sample or high frame rate, the MAPMT system can have superior performance. MAPMTs with higher sensitivity cathode materials such as GaAsP can provide a system with a quantum efficiency up to about 40-50%.
The photon sensitivity of each channel is not equal and can vary up to 50%. This effect is further compounded by the Gaussian geometry of the excitation beam which results in higher excitation efficiency at the center pixels verses the edge region. This problem has been solved previously using multiple reflecting beam splitter to generate equal intensity beam-lets. The MAPMT-based MMM system can be further improved by utilizing this type of beam splitter with an additional flat field correction algorithm to remove inherent sensitivity non-uniformity of the MAPMT.
There is also cross talk between neighboring pixels of MAPMT. The typical crosstalk is minimal at about 2% when the photons are collected at the center of each pixel. However, this cross talk can be removed by post-processing of the image similar to ghost image removal discussed previously.
In MMM imaging, more power of excitation light is required. Assuming that input power of 10 mW is needed for each excitation focus, generation of 64 excitation foci requires 640 mW input power. In the imaging of turbid tissue specimens, more input power is required to compensate the signal loss due to excitation photon scattering. In case of a tissue specimen whose mean free path length is 160 μm at excitation wavelength, the input power of 2200 mW is required to image at 100 μm deep, assuming that signal level is decreased only due to excitation photon scattering and there is no change in collection efficiency of emission photons. Therefore, the current power of Ti-Sapphire laser is limited for MMM imaging and can further increase in imaging speed by the use of even more foci.
Referring to
In the multi-photon case, light from each path generates a 3D intensity distribution in its associated focus, according to the multi photon excitation process. The detectors 111,112,113 detect all the light in the ‘detection cone’ associated with their active area. This light includes light generated by the light path associated with each detector (for example, light from the first focus 151 is detected by first detector 121), as well as light that is generated in the first focus 151 but scatters around the first focus 151 on its way to the first detector 151, and light that is generated in the second and/or third foci 152, 153 and is then scattered into the detection cone of the first detection light detection path 111.
In the confocal case, a confocal pinhole is placed in front of the detectors, for instance, in
In order to reduce cross talk between the light beams due to scattering, the illumination light and the associated detection can be time multiplexed.
Still referring to
Imaging in the z plane occurs by moving the imaging device with respect to the sample, or vice-versa. The intensity information is recorded along with the z-position of the sample or device and reconstructed by an image processor. Another embodiment according to the invention provides for a multifocal, multiphoton microscope based on MAPMT, as illustrated in
As shown in
A z-piezo actuator 109B (such as MIPOS 250 SG, micro-objective positioning system, integrated strain gauge motion: 200 μm (closed loop), Piezo System Jena controllable by controller 170B is attached to the objective lens 110B in order to move it in the z direction for 3D image generation. The sample 105B is attached to a sample stage 115B, which can be moved in x, y and z directions, also controllable by controller 170B and/or computer 176B. Light reflector 134B (such as, for example, a dichroic mirror) is positioned in the detection pathways to enable multi channel imaging by a first MAPMT detector 120B and a second MAPMT detector 124B, for multi channel imaging.
An IR block filter 116B (such as e700sp Special, Multi-Photon Blocking, Block 750-1000 nm>OD 6, Chroma Technology Corp is positioned in the detection pathway to separate the long wavelength excitation light from the short wavelength detection light. The filter 116B is exchangeable with a variety of filters or can be removed completely for reflected light confocal imaging. The filter 116B can be mounted on a motorized mount, which allows it to exchanged via a controller 170B and/or computer interface 176B. A band-pass filter 117B (such as 560DCXR for detecting DAPI and FITC, Chroma Technology Corp) (560DCXR for the transmission of light generated by the excitation of GFP and Rhodamin, HQ460/40 for the transmission of light generated by the excitation of DAPI, HQ630/60 for the transmission of light generated by the excitation of Alexa 594 bandpass; Chroma Technology Corp) is positioned in front of each of the multi anode PMTs 120B, 124B, in order to detect certain spectra. The band-pass filters 117B, 117B are exchangeable with other different filters and can be mounted on a motorized mount enabling changing of filters via a controller 170B and/or computer interface 176B. The same sample region can then be imaged with a different set of band-pass filters for more than two-color imaging.
A detection-part light-shield enclosure 118B is used to shield the detection part of the apparatus from ambient light. A variable iris 119B (such as for the case of a manual version D20S—Standard Iris, 20.0 mm max. Aperature; Thorlabs. Motorized versions of equivalent devices are available as well) is positioned in the focal plane of the micro lens array 140B in order to enable single spot illumination. For 8×8 foci imaging, the iris 119B is relatively open and for fewer or single spot imaging the iris is relatively closed, so that only a few or one micro lenses illuminate the sample. The variable iris 119B does not have to be in a round shape; and it will be square in shape when only a certain array of micro lenses should be blocked to enable illumination with a view of selected foci only. The variable iris 119B can be motorized and controlled via controller 170B (such as, for example by connection 191B), and/or via the computer 176B.
A micro lens foci mask 125B (such as a thin (for example 0.3 mm) aluminum sheet in which small wholes (for example 0.5 mm wholes) are drilled, at the points of where the micro lens focuses) positioned proximate the micro lens array 140B is a pinhole mask with a large pinhole size that enables the transmission of most of the light focused by the micro lenses, but otherwise it blocks ambient and stray light from the laser.
A first reflector 131B generates a first laser reference beam 165B from the incident laser beam (for monitoring the laser illumination power, wavelength and angular pointing stability). The reference beam 165B projects upon the diode or detector 160B which generates a signal that measures the laser illumination power, wavelength and angular pointing stability.
A further embodiment of the invention provides for a scan reference beam 166B from a scan reference beam illumination source 168B to be projected via reflector 172B and reflector 132B onto the scan region, whereupon the returning scan reference beam returns via reflector 132B to pass through dichroic 172B and lens 174B to be received by detector 164B. The scan beam is provided for monitoring the scanning accuracy. Detector 164B can be a diode or CCD detector or another type detection device. As shown in
A high voltage power supply 188B supplies power to the multi anode PMTs. Multi channel photon counting cards 184B, 186B are connected to each element of the MAPMTs, with one photon counting device for every multi anode PMT element, such as, for example, MAPMT elements 120B and 124B.
A computer 176B (including input devices, such as, for example, a keyboard and mouse) can be provided in one embodiment, connected to computer display 178B. The computer 176B can be connected to controller 170B.
The computer 176B controls numerous elements of the invention either directly and/or indirectly through controller 170B, and one skilled in the art will appreciate that numerous alternative configurations can be implemented within the scope of the invention.
One embodiment provides for the computer 176B to be programmed with a processing software and for the computer 176B to control a number of optical elements through a variety of electronic interfaces. For example, without limitation, the computer 176B and/or the controller 170B can be electronically interfaced with the scanner 180B and the multi channel photon counting cards 184B, 186B to perform the steps of scanning and data acquisition. Further the computer 176B can perform imaging post-processing steps. The display 178B can be used to display the acquired images in real-time after further processing.
A laser power attenuator 163B can be provided to control the laser incident power. The attenuator 163B can be controlled by the controller 170B and/or by the computer 176B in order to enable power adjustments for different samples and different locations in samples. During imaging at different depths in the sample, for example, the laser power can be automatically adjusted, so that the laser power can be increased at higher penetration depth. The attenuator 163B is integrated in order to make laser power adjustments, such as, for example, low power at the sample surface and increased power at increased penetration depth.
A third reflector 133B generates a second laser reference beam 167B from the incident laser beam (also for monitoring the laser illumination power, wavelength and angular pointing stability). This second laser reference beam 167B projects upon a second diode or detector 161B to generate a signal that measures the laser illumination power, wavelength and angular pointing stability. A laser power attenuator 163B controls the laser incident power and is integrated in order to make laser power adjustments, such as, for example, low power at the sample surface and increased power at increased penetration depth. Laser 101B is an illumination light source, such as a titanium sapphire laser (Mai Tai HP, Spectra Physics).
Multi-photon microscopy works most efficiently with short laser pulses. Owing to dispersion, the optical elements in the illumination pathway broaden the initially short laser pulse. The pulse compressor 102B is built from a pair of standard high reflectance mirrors and a pair of prisms (IB-21.7-59.2-LAFN28, Material: LaFN28; CVI Laser Corp., Albuquerque, N.M. 87123) mounted on translational and rotational stages pre-chirps the laser pulse in order to attain a short laser pulse in the focus of the objective lens.
A confocal pinhole array optionally can be placed between either of the multi anode PMT arrays 120B, 124B and the band-pass filters 117B, 117B, respectively. This option enables the system to be used for confocal microscopy or for multi-photon microscopy with confocal detection.
A telescope 103B expands the laser beam. With different expansion ratios, a different number of micro lenses can be illuminated. With a small beam expansion for example, a relatively smaller array of 2×2 micro lenses can illuminated and, thus, an array of only 2×2 foci is generated. As the beam expansion is made larger, an array of 8×8 or more micro lenses can be illuminated and, thus, an array of 8×8 or more foci is created. A further preferred embodiment employs a set of at least two mirrors 135B, 136B after the telescope 3B for precise beam alignment.
A mechanical micro lens holder 145B enables the precise positioning of the micro lens array 140B with respect to the multi anode PMTs 120B, 124B in the x, y and z directions. The holder 145B can be a motorized holder and can be controlled through a computer interface 176B, or, alternatively, can be controlled via a controller 170B, which controller in turn can be directed by computer 176B.
Mechanical multianode PMT holder 125B, 126B enables the precise positioning of the multi anode PMTs 124B, 120B, respectively, with respect to the micro lens array 140B in the x, y and z directions. The holders 125B, 126B can be motorized holders and can be controlled through a computer interface 176B, or, alternatively, can be controlled via a controller 170B, which controller in turn can be directed by computer 176B.
The computer 176B, or the controller 170B, or the computer and controller together can be configured to control automatically or to control in supervised fashion, one or more of the following elements, without limitation: the scan reference beam illumination source 168B, the sample piezo stage 115B, the objective z-piezo stage 109B, the scan reference beam detector 164B, the scanner 180B (by connection 193B), the IR block filter 116B (by connection 194B), the band-pass filters 117B (for example, by connection 195B), the laser source 101B, the laser attenuator 163B, the first laser reference beam detector 161B, the second laser reference beam detector 160B, the pulse compressor 102B, the multi-photon channel counting cards 184B, 186B, the mechanical multi-anode PMT holders 125B, 126B (for example, by connection 190B), the variable iris 119B (such as, for example, by connection 191B), and the mechanical micro lens array holder 145B (for example, by connection 192B).
The focal region has a focal pattern variation in xy plane. The foci can be distributed unevenly, e.g., the rows and columns do not have to be spaced uniformly.
Also, a system can be built, in which there are additional rows and/or columns of PMT's at the outer region of the array of detection tubes. For example, there can be more than 8×8 rows and columns in both the micro lens array and the detector. This is particularly important for detecting scattered photons of the outer foci and for using the information of the scattered photons from the outer foci for deconvolution purposes. Further, an embodiment of the invention provides for a system in which there are more detector elements than there are foci, so that a plurality of detector elements (or detection pixels) collect the photons of one, optically conjugated foci. For example, a 16×16 detector array can used as a detector device, while an array of 8×8 foci can be illuminated by an 8×8 multi lens array. Smaller and larger PMT-to-foci ratios can be utilized. In particular, a detector array in which one foci is optically conjugated to an uneven number of detector elements can be employed. This is important for detecting scattered photons in the neighboring channels of the to the focus optically conjugated detector and for using the information of the scattered photons for deconvolution purposes.
The image of the sample is formed by scanning in the optical plane (xy) when the intensity signal from the detectors is correlated with the foci positions. The foci scan the specimen in the x direction, then move an increment in the y direction, and then raster in the x direction again until the sample is fully covered at some desired resolution. During the rastering, intensity light signals are recorded by the multi anode PMT. These signals are then saved along with the foci positions in the computer and can be concurrently or afterwards displayed by the computer display or other graphics outputs. The foci positions are known by the scanner position (beam scan) or the sample position (stage scan). The smaller the step increments, the higher the resolution the final image will be. The scanning can be performed in a raster fashion, or in many other ways, such as with time multiplexed methods, or scanning simultaneously at different depths.
Referring to
3D AM-PMT MMM can be used in multi photon endoscope device in accordance with another preferred embodiment of the invention (to be added).
Referring to
Referring to
In the setup the two multi-anode PMTs will be stacked on top of each other, each serving as a spectral detection device for one line of 1×8 foci. The grating 192E properties (pitch/inch) and the focal length of the focusing lens (L4), which determines the distance between the grating and the multi-anode PMT, have to be chosen in accordance to the anticipated fluorescent probes used for staining the tissue sample. For this embodiment, a transmission grating 192E is used. Nevertheless, comparable and/or better efficiency can be achieved in embodiments that use a reflection grating or a prism.
Referring to
In addition to the primary use of this instrument for two photon microscopy, other multi-photon sensing and imaging methods can also be used with the system described herein including:
Referring to
Time multiplexed illumination and detection enable MMM microscopy with one detector only, which is gated to the excitation light pulse. In one variation of a multi-photon MMM, the illumination light source is a pulsed laser. In this case, a Ti:Sa Laser with a repetition rate of approx. 80MHz and a pulse width of approx. 100-200 fs as an example. In the standard illumination version of this MMM, all beams carry the same pulse distribution along time. As a consequence, the array of excitation foci in the focal region is formed simultaneously. For image formation, during or after at least one pulse has illuminated the sample, the beam or sample is scanned on both axis perpendicular to the optical plane; here indicated by x≥0.
In the example of multiplexed operation 500 shown in
The temporal delays introduced into the illumination foci mean that this supersampling condition exists even when the number of detectors is the same as the number of foci.
In anther example, alternating excitation foci pattern can be detected by a multichannel detector with smaller number of elements than number of foci.
In cases in which the repetition rate of the laser is lower that in the case of the Ti:Sa laser, a time multiplexed MMM illumination and a detection in a single channel can be used. In this particular case, the repetition rate of the laser is a hundred times lower than in previous examples. In a time multiplexed version, each beam carries a pulse which is temporally separated in regards to pulses of the other beams. In one particular case, they are separated evenly over the time period of one laser repetition, so that at evenly distributed time points, one single foci is illuminated at a time. If a fast detector is correlated with the pulse distribution and capable of detecting each pulse separately during this short time period, an MMM with only one detection element can be used. This detection element has a corresponding detection area to collect light which is generated by each individual foci during its scan. For image formation, during or after at least one pulse cycle has illuminated the sample, the beam or sample is scanned on both axis perpendicular to the optical plane. As a result, optical cross-talk is completely eliminated, as the light from the different foci is excited and detected at different time points. Applications for this case are excitation processes which appear instantly, like scattering effects (such as Second Harmonic Generation (SHG) or Coherent Ramen Anti Stokes scattering (CARS)). This configuration can be used suited for a non-de-scanning configuration. Other repetition and detection rates are possible.
In the case of using a pixellated detector such as a CCD or a CMOS imager, an array of 3×3 beams,
The above described systems and methods can be used for imaging of all semi-transparent and highly scattering materials; 2D and 3D, and in particular for imaging of human and animal tissue, cells in suspension, plants and other biological material.
The illumination can be achieved with visible light and alternated with the MMM scanning measurement or out of band illumination light can be used and the camera measurement can be taken simultaneously with the MMM measurement. This configuration can be used for large field imaging, sample guided MMM measurements, conventional staining measurements, and online MMM measurement process control, for example, bubble formation monitoring, and laser spot diagnostics.
There is a large variety of fluorescent that can be used with various embodiments of the invention dyes. In general they fall into two families: Dyes that have to be applied to stain the tissue “from the outside” and dyes, that are expressed from animals as proteins. Most commonly used dyes by external staining MitoTracker Red, DAPI, Hoechst 33342, Cy2-IgG, Alexa Fluor 546, Rhodamine 123, Alexa Fluor 488, FITC-IgG, Acridine Orange, Cy3-IgG, Issamine Rhodamine, TexasRed-Phalloidin, TexasRed-IgG, Alexa Fluor 594, Propodium Idonide. Dyes genetically expressed by genetically modified animals: green fluorescent protein (GFP) and other dyes in this family: Enhanced GFP EGFP, Yellow fluorescent protein (YFP), Enhanced YFP. Auto fluorescent imaging does not use a particular dye, but can be used as part of an imaging technique.
Besides confocal microscopy (fluorescent, as well as reflected light confocal), these include all other multi-photon microscopy techniques, such as, 2, 3, or more photon excitation microscopy, Second (SHG), Third (THG) ore more Harmonic Generation microscopy, Coherent Anti Stokes Raman Scattering (CARS) microscopy, multi photon quantum dot imaging, surface plasmon imaging and Stimulated Emission Depletion (STED) Microscopy. These techniques can be used with or without staining methods. The scattering techniques, such as SHG, THG, CARS are developed to be able to image without any staining involved.
Referring to
Referring to
Scattered light detection from a spot 251G created by multi photon excitation, detected by two large area detectors 120G, 124G, positioned next to each other, are separated by a distance: In this case, the unscattered photon 261G and the first scattered photon 271G are still collected by the active detection area 222G of the first detector 120G. The second scattered photon 281G is not lost, but is collected by the second detector 124G. This effect of light being scattered into detectors other than the optically conjugated detectors is termed “optical cross talk”.
Referring to
Referring to
In a further embodiment, changing the optical configuration of the apertures and focal length of the lenses in the optical system can create the same effect. Changing the aperture and the focal length of the micro lens array, increasing the area of the scan mirror, changing the aperture and the focal length of the lenses L1, L2, L3 and L4 has a similar effect of reducing cross talk without loss of signal. An example case is presented in tabular format in FIG. W-22(a)-(e).
The optical configurations for two alternative embodiments of the invention, Example A and Example B, are presented in
In Example A, the side aperture pitch of each micro lens is 1.06 mm and the diagonal is about 1.5 mm. For the entire array in Example A the 8 lenses per side create a side aperture of 8.48 mm and a diagonal aperture across the array of 11.99 mm. The focal distance of each micro lens is 17 mm in Example A. In the embodiment of Example B, the aperture or pitch of each micro lens in the array is 1.9 mm and its focal distance is 25 mm.
The focal lengths of the lenses L1 and L4 are 50 mm and 103.77 mm, respectively, in Example A, while in Example B they are 40 mm and 46.32 mm, respectively. When standard optical components are used, they can approximate components with a focal distance of 100 mm and 45 mm for L4 in the configurations A and B, respectively. In both the embodiments of Example A and B, the focal lengths of lenses L3 and L4 are 30 mm and 125 mm, respectively. The diameter of illumination of the back aperture of the objective lens for both Examples A and B remains approximately constant at 13.0 mm and 12.7 mm, respectively. This results in an ‘under-illumination’ of the back-aperture of the objective lens which has a back aperture of 17.1 mm in diameter. This is desirable, so an optimal (maximal) employment of the illumination light power is warranted.
As listed in
The optimal distance between the foci is influenced by three factors: (1) the optimal number of foci that are needed to generate as much light as possible (this number can be distinguished in accordance with the graph in
At a fixed number of foci, a large field objective provides an advantage for certain embodiments of the invention, because the foci can be further separated. An objective lens with large field of view enables large separation of foci and thus reduces optical crosstalk. In
In the embodiments shown in
One embodiment of a method for data post-processing according to the invention is illustrated in
Referring to
The post-processing steps can include a number of substeps, including, among other steps those illustrated in
i) after data access 332H normalizing, filtering (de-noising), blending (integrating) of multifoci subimages 334H;
ii) Registering subimages into a single image;
iii) Filtering and normalizing images produced from corrected subimages 336H;
iv) Registering, building mosaics, and blending sets of corrected images into a larger whole 338H;
v) Optionally, at this stage, lower resolution images can be created of the larger image to facilitate access, as well as images from different perspectives (xy, xz, yz).
vi) Data-compressed versions (e.g., JPEG, wavelet compression, inter alia without limitation) can be produced;
vii) Segmenting images into objects 340H This segmentation can either be manual, automated or a combination of both.
viii) Parameterizing the objects 342H (for instance, size, shape, spectral signature).
ix) Classifying objects into higher order structures/features 344H (e.g. material stress or cracks, vasculature, nuclei, cell boundaries, extra-cellular matrix, and location, inter alia, without limitation)
x) Statistically analyzing parameterized objects 346H (e.g., by correlation or other methods).
xi) Presenting results to user on display device 348H (examples include: 3D/2D images, annotated images, histograms, cluster plots, overlay images, and color coded images).
The normalized signal distribution resembles the normalized power map squared and shows an intensity drop of 45% toward the corner PMTs in respect to the center PMTs. The laser power was attenuated to 75.3 mW in the sample and can reach a maximal value of approx. 645 mW, resembling a power of approx. 15 mW for the center foci in the sample. Measured intensity profile can be generated by imaging a homogeneously distributed fluorescent dye under a cover slip. The intensity measurement is not only mapped by the power/intensity distribution of the foci, but also by the sensitivity of the detector array. As a result it resembles the “true” measured intensity distribution. The image consists of 192×192 pixels and was generated by an array of 6×6 foci which were scanned across a uniform fluorescent dye sample. 2D xy image normalization is carried out in different ways:
A method for multifocal multiphoton imaging of a specimen in accordance with a preferred embodiment of the present invention:
In more detail these are
(1) Sample pre processing
(2) place the sample in the region of focus of the focusing device (objective lens):
(3) determine imaging parameters
Determine imaging parameters by which include variable or multiple values for each of the parameters
(4) set imaging parameters
A computer program is fed with the imaging potentially dynamically adjustable (including feedback from the measurement) parameters and controls the imaging procedure.
(5) image
Point measurement.
1D scan: Collect data from a region of interest
2D scan:
Image a 2D region of interest by scanning the foci in parallel across the imaging plane (XY)
the 2D scanning starts for example at a corner of an area and is then scanned in a raster until the area is covered according to the imaging parameters (current implementation)
or it can be scanned in any other way (even random scan is allowed), as long as the area is covered according to the imaging parameters and the position of the foci is known by the signal sent or received by the scanner.
A key consideration for the improvement in the measurement is that the detector measures the sample for each scanning position, without overlap.
3D scan:
move the focusing device (objective lens) in reference to the sample along the optical axis (Z) and repeat the 2D imaging process. Either, the focusing device (objective lens) or the sample can be moved. Right now, the focusing device (objective lens) is moved stepwise in regards to the fixed sample.
2D imaging along the optical axis (Z) can begin at any point in the sample and end at any point of the sample within the region of interest.
For some applications it is preferable, if the movement can be reversed (starting inside of the sample and then move out), or performed in a ‘random’ fashion, covering the whole area, as long as the z-position is known.
The z position of the foci is known as the piezo position is known
The 2D scanning is done while the z-scan from one position to the next takes place or after the z-scan has completed its move to the next position.
Images of 2D sections can be done alone, without any 3D movement involved.
(6) Process images for feedback purposes (optional)
(7) display the images (optional)
(8) save the data
Process data before saving (Optionally)
(9) Post process the data (optional)
(10) save the processed data (optional)
(11) display the processed data (optional)
In addition to mechanistic applications, time-resolved measurements, either alone or in conjunction with spectral measurements, can greatly aid in distinguishing signals from different reporter probes and processes, such as simple scattering and non-linear scattering. For cytometry applications, the additional information from time-resolved measurements can potentially increase the number of probes which can be used simultaneously, provide images cell morphology by detection of second harmonic generation, and aid in deconvolution of images from highly scattering samples.
Assuming a homogeneously scattering material (which can be assumed for samples in the first approximation), the relative and absolute height of the peaks of the delta function is fixed for every channel at its neighbors at a certain imaging depth into the sample. As a result, xy images can be linearly de-convoluted.
The linear deconvolution of cross-talk is primarily a 2D process. The values of the weighting matrix depend on several factors. The optical contribution to the cross talk increases with increasing penetration depth. Furthermore, the channels have different sensitivities, there is electronic cross talk between channels that varies from channel to channel and other factors influence the amount of total cross talk between the channels.
The cross talk for each individual channel can be determined experimentally. An example is where, one focus illuminates the sample or a test object and the whole array of detectors detects the signal. At different penetration depths a cross talk matrix is measured for each channel. This matrix is then used to carry out the deconvolution. Data of such a measurement at the sample surface and at a penetration depth of 200 mm, can be used. The measurement is repeated for every channel for example by moving the iris from transmitting light from one single micro lens to the next (in this case for channels c11 to c33). Similar alternative methods are also possible, for example by illuminating with all of the foci but using a sample with large object spacing. Furthermore, models can also replace experimental determination.
An entire 2D image consists of collections of ensembles for a non linear deconvolution, of pixels, from each detector. The key point is that relationships between entire ensembles, and certain regions of pixels between ensembles, can be established to constrain the variation of the weighting matrix to aid convergence without assumptions, or with minimal assumptions, of the sample or the processes which cause the variation in the weighting matrix.
For example, continuity of the values across the boundaries of the ensemble can be generally required. In the case with the minimal assumption that the objects under observation are smaller than the region covered by an individual ensemble, the ensembles can be considered largely independent, except due to the cross-talk introduced by the weighting matrix. The ideal image can be recovered by simultaneously solving for a weighting matrix which minimizes the covariance between ensembles. In the other case where the objects under observation are of similar size or larger than the regions covered by the ensembles, minimal models of the object (such as from image morphology or segmentation of the collected image, etc . . . ) can be used to form constraints.
Additional model dependent and independent constraints can also be applied by consideration of the planes above and below the plane under evaluation. Further constraints can also be applied to the weighting matrix from either general (such as continuation, smoothness, sharpness, etc . . . ) or model based considerations.
While this invention has been particularly shown and described with reference to preferred embodiments thereof, it will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the scope of the invention encompassed by the appended claims.
This application is a continuation of U.S. patent application Ser. No. 11/442,702, filed May 25, 2006, which claims the priority of U.S. Provisional Application No. 60/684,608 filed May 25, 2005 entitled, MULTI FOCAL MULTIPHOTON IMAGING SYSTEMS AND METHODS, the whole of each of these applications being hereby incorporated by reference herein.
This invention was made with government support under grant numbers R33 CA84740 and R33 CA91354 awarded by the National Institutes of Health. The government has certain rights in this invention.
Number | Date | Country | |
---|---|---|---|
60684608 | May 2005 | US |
Number | Date | Country | |
---|---|---|---|
Parent | 11442702 | May 2006 | US |
Child | 15815536 | US |