\documentclass{article} %% Better math support: \usepackage{amsmath} %% Bibliography style: \usepackage{mathptmx} % Use the Times font. \usepackage{graphicx} % Needed for including graphics. \usepackage{url} % Facility for activating URLs. \usepackage{enumitem} % to be able to enumerate over letters, etc \usepackage{fancyhdr} % For customizing headers %% Set the paper size to be A4, with a 2cm margin %% all around the page. \usepackage[a4paper,margin=2cm]{geometry} %% textcomp provides extra control sequences for accessing text symbols: \usepackage{textcomp} \newcommand*{\micro}{\textmu} %% Here, we define the \micro command to print a text "mu". %% "\newcommand" returns an error if "\micro" is already defined. \newcommand{\Lim}[1]{\raisebox{0.5ex}{\scalebox{0.8}{$\displaystyle \lim_{#1}\;$}}} %this one puts the text of a limit under the limit writing %% This is an example of a new macro that I've created to save me %% having to type \LaTeX each time. The xspace command provides space %% after the word LaTeX where appropriate. \usepackage{xspace} \providecommand*{\latex}{\LaTeX\xspace} %% "\providecommand" does nothing if "\latex" is already defined. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% %% Start of the document. %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%% \pagestyle{fancy} \rhead{Lewis Guignard EE230 Final Lab Report Page \thepage} \cfoot{} \date{\today} \title{EE230 Final Lab Report: Structured Illumination and Two Photon Microscopy} \author{Lewis Guignard} \begin{document} {\fontsize{0.8 cm}{1 em}\selectfont EE230 Final Lab Report: \newline Structured Illumination and Two Photon Microscopy} \vspace{1 cm} \newline In this report, we review two papers, each implementing a modern technique of microscopy in reference to a biological application. We will first investigate \textbf{Structured Illumination} \cite{Kner2009}, and then investigate \textbf{Two Photon Microscopy}\cite{Liang2009}. \section{Structured Illumination} \begin{enumerate}[label = \alph*.] \item \textbf{The Work} In this paper, a new technique of Structured Illumination Microscopy (SIM) is introduced, allowing higher speeds of 3D imaging and the introduction of multicolor imaging, greatly increasing the utility of the technique. The authors are concerned that, although Structured Illumination is one of the most attractive sub-diffraction limited microscopy methods, due to its low light power requirements, and not having to use specialized fluorophores (and thus too adversely manipulate the sample), it is slow and monochromatic, making in vivo measurements difficult at best. This is of special importance to researchers in biology, for to be able to resolve biological processes in vivo and in real time is of great advantage to the science. In vivo processes are not easily replicated in vitro; a lot of variables come into play can manipulate an experiment's outcome, and extrapolating results back to the in vivo case is tenuous at best. Further, to have the ability to use differential coloring, or two fluorophores, greatly increases the biological information. When limited to monochromatic fluorescent imaging, say of mitochondria, one may see the mitochondria but will have difficulty resolving cell structure, or the mitochondria's relative location therein. Previous work has brought high speed temporal 3 dimensional imaging using the Stochastic Optical Reconstruction Microscopy technique (STORM) \cite{Manuscript2013}. While the STORM technique showed image acquisition in as little as 0.5 second periods, the light intensity to achieve such short exposures required was several orders of magnitude higher than that required by the SIM technique described by our authors. Other techniques such as STED provide much higher spatial resolution, and can give reasonable temporal resolution, however it requires incident light of watt-density many orders of magnitude higher than SIM (on the order of $MW/cm^{2}$ as opposed to $W/cm^{2}$), and so can be very destructive to the sample, therefore bending the definition of 'in vivo.' \cite{Rust2006} The work of Reto Fiolka et al \cite{Rycroft2013} brings higher speed to structured illumination microscopy, high enough to measure some biological processes (time periods of 8.5 s). They also introduce new polarization and grating techniques to allow two color imaging in SIM, without the need for specialized fluorophores. This potentially brings the structured illumination toolset to the in vivo application arena, allowing a high-speed, minimally invasive microscopy tool to be added to the in-vivo toolset. \item \textbf{The Method} The method of SIM works with the concept of increasing the volume of the Optical Transfer Function (OTF), which in turn is the Fourier Transform of the Point Spread Function (PSF), which is itself the object a particular optical system will generate when interacting with a point source, or point object. One can imagine in the frequency domain, the objective is to gain higher frequency components (and thus increase resolution). To increase the frequencies that are recorded, optical gratings are used. One can imagine the moire pattern for a macro comparison. Two simple gratings of similar wavelengths, when laid over each other, will yield a longer-wavelength pattern. We can see then, that if the wavelengths of each component are below the resolution limit of a given system, the resultant miore pattern may not be. SIM takes advantage of this by illuminating the sample with an optical pattern that is one spatial Fourier component, at a freqency on the edge of the OTF function mentioned above. This is usually achieved by interfering multiple laser sources, which are all s-polarized with respect to the sample surface (which they illuminate normally). This can be seen in Figure \ref{fig:SI_setup}. For speed, a new method of generating the initial pattern is to use a ferroelectric liquid crystal Spatial Light Modulator (SLM), which can change patterns faster than traditional methods (~.5 ms). This SLM diffracts incoherent laser light, and the 0th and 1st orders are filtered and used for illumination. The speed advantages of the ferroelectric SLM also allow it to switch between pattern spacing optimized for different wavelengths of laser quite rapidly, opening the door for multicolor imaging. This highly coherent s polarized light (in three beams) is then passed through a liquid crystal cell. The different laser sources interfere, giving a one dimensional spatial Fourier component of light at the sample. \begin{figure} \centering \includegraphics[width=0.9\linewidth]{./SI_setup} \caption{SI setup} \label{fig:SI_setup} \end{figure} By choosing the frequency of the incident light interference pattern, which is ideally on the boundary of the OTF, one teases out high frequency information in the resultant interference of the sample and the incident light. Taking several angles of this interference will allow one to build a pseudo OTF that is bounded by twice the frequencies as the original, essentially doubling the resolution of the system. Computation is used to back out the image from each interference pattern obtained (usually 5 for a given depth in the sample). For samples thin enough to illuminate through, the sample is then shifted a small amount, so that the focal plane lies at a new point within it. Using this method, imaging can be performed to build a 3D model of the sample. The authors introduce two novel improvements to the traditional SIM setup. The first is the introduction of the ferroelectric SLM that allows rapid adjustment of diffraction gratings, allowing two color imaging. This technique was not fully utilized by the authors, for the emission filter blocks had to be switched between one light source and another (taking around 40 ms), and so the authors chose to take all images with one wavelength of light before switching sources. The second 'improvement' to the SIM system by the authors is the use of a liquid crystal and quarter wave-plate for polarization, allowing different polarization angles to be generated in 1 ms timescales. Together, these two improvements greatly increase the speed at which SIM can be taken, and thus allow ideas like sequential color imaging to be possible. SIM was directly discussed in class, and the author's augmentation of it make the methods concrete. Theory of point spread functions, Fourier Transforms, miore patterns and fluorescence microscopy all play a pivotal role in this technique, and so its largely in-line with the course information. \item \textbf{The Results} I found the Movies provided in supplementary information, especially movies S2 and S5 to be the most compelling, for they show most directly the benefits of high speed in vivo imaging, to record biological processes. Care must be taken to keep in mind the colors are about 4 s time-shifted from each other, a relatively large shortcoming of the imaging system. Figure \ref{fig:SI_setup} above shows the general setup of the SIM system. As a novice in the field, I found the acronyms myriad, and the location of the sample not so obvious. If I could change only one thing about this paper, it would be to work on this figure, to clear up the sample location, and also add in here what the authors did to change the SIM system for the better. \item \textbf{The Future} The authors propose future augmentations to the measurement system, where they would not take the different color images sequentially, but take each color at each spatial filter and depth focus. This would be achieved by either using two cameras, each filtered for the respective color, or using a notch excitation filter, to reduce recording of scattered excitation light. I found it very limiting that the images have to be completed in one color before moving to the next, on account of the ~40ms of changing the excitation filter for each wavelength. This means that for a given focal plane, images in one color will be projected onto images in another color taken 4s later, yielding movies that aren't so coherent. This could be worked around by limiting the choices of fluorophores to those where the emission wavelength of the fluorophore corresponding to the first light source is too low to excite the second fluorophore. In this way, all images at each focal plane could be taken sequentially and a nearly time independent image for a given focal plane could be captured. \end{enumerate} \section{Two Photon Microscopy} \begin{enumerate}[label = \alph*.] \item \textbf{The Work} In this paper, an advance in Two Photon Microscopy (TPM) is introduced, which increases the (at the time of printing) barrier of 1 mm for depth using the technique. This depth is more probabilistically defined as the attenuation length, or that after which the probability for a given photon to not have been absorbed is 1/e. Notice should be taken here that for the effect to be beneficial, the excitation photon has to make its way to a certain depth before being absorbed (by a fluorophore), and then the emission photon has to be able to make its way back out of the system (the same length). Increasing the depth of microscopy resolution is important for many biological applications, especially in 'in vivo' or clinical applications. The further into a tissue or sample one can see with microscopy, the less one has to interact with the sample and perturb it from a natural state. Previous work \cite{Kobat2011} \cite{Rust2006} has shown that indeed longer wavelength excitation photons increase tissue penetration depth. This paper does not specifically bring out a novel technique, but does increase the effective depth by using a longer wavelength excitation photon than previously reported (1280 nm). We find this new depth of interest as for the specific case of the mice under research, much of the mouse cortex is less than 1.6 mm deep, and so TPM can now image the complete cortex over much of the surface area. In general, more imaging depth is of utility to get more non-invasive information for a given study. \item \textbf{The Method} The method of TPM takes advantage of the probability of a given fluorophore to absorb two photons (or more in rare setups), each of half the energy required for the fluorophore's ground state electron to jump the band gap to an excited state. While the probability for this to occur is small, with enough intensity, it will indeed occur. Laser light of wavelength twice that which would normally excite the fluorophore is focused on an extremely small point in the sample. Excitation occurs only at this point, because of the low probability of excitation (and thus high intensity required for an appreciable number of excitations). In this way, high spatial sensitivity is achieved. Another benefit of TPM is the fact that mean path length through a material is directly proportional to the photon's wavelength, i.e. longer wavelengths penetrate further into a material. This is obvious when considering a case of something like radio, where wireless signals penetrate walls and other thick substances relatively unperturbed, and their wavelengths are much larger than visible light. As the wavelength of the emission photon is always longer than that of the excitation photon (maintaining the Conservation of Energy), when using TPM, the limiting wavelength is now shifted to the emission one for photon detection. As the probability of as fluorophore ground electron absorbing two photons at once is extremely low, extremely high intensities are required for this method, normally too high for in vivo work, without completely destroying the sample (kW range). For this reason, the incident light is pulsed on the fs scale, so that the overall incident power is readily handled by the tissue. Even though this paper was published in 2011, the concept of increasing the depth used is quite simple. As stated above, the penetration depth is directly proportional to the wavelength of incident light. The authors choose to use incident light of 1280 nm, which has a theoretical depth limit of just more than 1.6 mm (which was reached) This group is very active in the TPM technique \cite{Kobat2011}, and with good reason. Deeper penetration depth and lower phototoxicity than more conventional confocal microscopy give TPS a definite advantage, especially in the biological domain. The technique is relatively young, the room for improvement large, and the potential impact of increases in depth / resolution are myriad, making this a good system to optimize. The group's approach brings several advantages. First, a longer wavelength excitation photon is less dangerous for tissues. Even at the same intensities, if a red-shifted photon is absorbed by tissue other than the intended fluorophores, that tissue will be given lower energy (By $E = Hf = Hc/\lambda$), so this technique is safer for a given incident light intensity. Second, the longer wavelength brings deeper tissue penetration and so thicker sample images. Noted for this approach, but intrinsic to the TPM technique, is the long integration times required for deeper tissue imaging. This comes about from the fact that few photons make it to the deepest imaging plane, and few return through the surface of the tissue and to the detector, and is shown by the authors in Figure \ref{fig:TPS}, where one can see a somewhat linear relationship between the Log of the signal intensity compared to depth of tissue. These long integration times slow down any time sensitive measurements, or movies a researcher might want to take. \begin{figure} \centering \includegraphics[width=0.7\linewidth]{./TPS} \caption{Log of intensity vs tissue depth} \label{fig:TPS} \end{figure} This augmented technique brings together several concepts taught in class: TPS itself, photon scattering in a given medium, and incident intensity of light. \item \textbf{The Results} Of most use to me was Figure 1 in the paper, which shows a 3 dimensional stack of the imaging technique. One can see the lower completion of the cortex in the image, and the expansion of the slides at lower depth show the precision of artifacts at that level. This image really sums up the abilities of having such penetration depth; which at first blush might not seem that great. 1-1.6 mm of tissue depth on a person might not make it through the skin, but when looking at systems such as brain tissue, or smaller organisms, there is quite a lot of information to be had (as seen in Figure 1 of the paper) One thing I found lacking was a proper explanation of Figures 2 and 4 (see \cite{Kobat2011}): particularly the behavior of the curve at either end. At just below 0 for the depth profile, there is a spike in signal, is the light scattering back before it even touches the sample, is this interference with an optic on the mouse skull? and at the deepest distances, we see a break from the linear (in log - lin space) of the profile, where intensity is higher for 1600 (Figure \ref{fig:TPS} below) than a linear model would suggest. Where is this relative increase in intensity coming from? Is it a non-linear error in measurement? some explanation as to what os going on here would be of great use. \item \textbf{The Future} The authors propose further improving the TPS system at depth by working with incident light intensity. The integration time at depth being a shortcoming (mentioned above), can be diminished by increasing intensity used for different focal lengths (depths), without increasing the average power of laser pulses. (by using shorter pulses of higher intensity). Why this was not optimized for the current experiment I am not sure, though it may have to do with the duty cycle of the laser pump being on the fs order already, so it could be near its maximum calibration range. Bringing about this improvement would decrease the time needed to take a given image, and so would allow higher speed dynamic imaging to be taken, revealing faster or more subtle biological processes. As more and more fluorophores are being developed, I have to wonder about using even longer wavelength light for TPS. The absorption intensities may get lower and lower, but what about a 2$\mu$m source? If lasers would be difficult to arrange that could produce this source, some other means might be taken. At longer wavelengths, higher tissue penetration has certainly been correlated. I would explore this limit, even at the risk of lower intensities and thus much longer integration times. These longer wavelengths would be even less dangerous for a sample. I wonder, however, where the limit lies in which a long enough wavelength is used to completely penetrate a sample, say a human torso. Then the signal and recorder could be on opposite sides of the sample. But this regime is very much like an x-ray machine, which uses light of extremely short wavelength. Where does this switch in optimality occur, and what can be gleaned at that meeting point of each method? \end{enumerate} \newpage \section{Comparison of Methods} Structured Illumination Imaging and Two Photo Microscopy bring two slightly different microscopy benefits to the arsenal of modern tools. Structured Illumination is sub-diffraction limited, allowing to see processes on the 1-150nm scale, and does not use high intensity light. It however does not bring great depths to the possible range: (on the order of $\mu$m). Two Photon Microscopy can resolve details three orders of magnitude greater in depth ( ~1 mm), but yields lateral resolution on the order of $\mu$m, and is slow in time. So we see that each tool has its use in a specific setting: high spatial resolution of surface effects - SIM. High depth resolution of effects - TPM. My personal opinion is that spatial resolution is currently more important, no matter the thickness, as the most biological research is occurring at the molecular level, and the closer imaging can get to that point, the more utility it will have. As the TPS seemed to only report a new depth achieved, using principles previously shown \cite{Kobat2009} by their own work, I find the SIM paper to be more novel and useful. They also took a more rigorous approach, noting both the basic science of the current status quo in more detail, and the improvements made upon that. This makes the improvements made to SIM more suited for the spatial and temporal resolving power reported than the increased depth of TPM for mouse cortex imaging (which still in areas is greater than the depth obtained). \bibliographystyle{plain} \bibliography{EE230_Final_Lab} \end{document}