Senior Projects and Theses

The design and construction of an all-reflective laser-scanning microscope with confocal and wide-field capabilities is presented. This instrument has the potential to image thick biological samples using a technique called optical sectioning. The all-reflective design allows for fluorescence imaging across a wide range of wavelengths by eliminating aberrations associated with lenses. The resolution of this microscope was estimated using a resolution test target and was better than 2.5 μm. The effectiveness of the instrument was investigated with fluorescent dye-soaked lens tissue and a thin plant sample.

In the context of biomedical imaging, practitioners are constantly searching for easier, more accurate, and more efficient methods for imaging thick samples. In my project I investigate one of these methods called structured illumination microscopy (SIM). Using a fluorescence microscope and structured light, three images are taken. These images have striping only appearing on parts of the sample at the focal plane of the objective lens. Using matrix functions, a Python program subtracts out of focus light.

Typically in fluorescence microscopes, LEDs work to illuminate samples, however in the case of looking at dim samples, the LED light leaks through the optical filters set in place to block out the LED light. This makes it difficult to see the light emitted by the sample itself. By using a laser beam to illuminate the sample, the narrow range of wavelengths coming from the laser itself are easy to block using an optical filter so that the camera captures only the fluorescence emitted by the sample. The downside of laser light is that its coherent illumination interferes to create a speckle pattern on the sample, meaning that the sample does not get illuminated evenly or fluoresce evenly. The speckle pattern is clearly visible on the sample and it causes dim samples to be even more difficult to see. By adding in a motorized ground glass diffuser, which essentially continually randomizes the light from the laser beam, we can more evenly illuminate the samples. The camera then averages out the various speckle patterns as the diffuser rotates, creating a clearer, smoother image. When looking at fluorescent beads with an accepted width of 1.0 μm, this microscope measured the beads to be 2.3 μm across.

Measurements of two-photon absorption spectra are presented for the long-wavelength dye IRDye 800CW in the excitation wavelength range of 1025 nm – 1700 nm. A long-wavelength dye absorbs one photon in the 650 nm – 950 nm range and emits a photon in the same range; this process is termed fluorescence. The two-photon absorption spectrum for this long-wavelength dye has not been previously characterized. The spectrum reveals that the wavelength that results in peak two-photon absorption for this dye is 1285 nm ± 10 nm. This is critical information for applications of two-photon imaging. IRDye 800CW is already being used in one-photon fluorescence guided surgery, and knowing the wavelength of peak two-photon absorption will improve the efficiency of the procedure.

In this experiment, we added an array of components to the college’s laser scanning confocal microscope with the intention of expanding the range of dyes that can be imaged using the microscope. We intended to use our components to image fluorescent beads in order to measure the point spread function of the microscope. In order to do so, we added a 782 nm laser, 730 nm LED, a camera and a detector to the microscope, which enabled us to image dyes that emit in the near-IR ranges. We were able to generate images of dyed tissue paper and mouse brain samples using our detector. Using the microscope’s 640 nm laser and our own detector enabled us to measure the point spread function of the microscope. We found the maximum resolution using the 640 nm laser and our detector to be 0.63 μm ± 0.07 μm. Ultimately, these efforts allowed us to expand the capabilities of the school’s laser scanning confocal microscope, though future work is necessary to improve the quality of image generated by our apparatus. We hope that our work will enable future research in the fields of fluorescence based imaging techniques and fluorescence guided surgery.

The shape of a laser pulse is measured both in time and in space, in order to be used in the investigation of long wavelength fluorescence dyes.. Due to the brief duration of the laser pulse, direct measurement is impossible, instead the shape in time is measured using interferometric autocorrelation in which the pulse is interfered with itself in a Michelson interferometer. The shape in space is measured in two experiments, first using the knife-edge method, then through photographic imaging of the laser pulse with an infrared camera. Measurements were taken for the laser pulse at output wavelengths from 1100 nm to 1800 nm in increments of 50 nm. The interferometric autocorrelation experiment yielded an average FWHM of 140 fs ± 20 fs. The knife-edge method determined an average width of 1.6 mm ± 0.2 mm, and the photographic imaging method an average width of 1.4 mm ± 0.3 mm. These measurements will allow future work in research on long wavelength fluorescence dyes intended for use in fluorescence guided surgery.

Two-photon fluorescence is a process in which two photons interact with a molecule, causing it to undergo an energy transition and emit light. Two-photon fluorescence requires both photons to interact with the same molecule at the same time, which is extremely rare. In order to increase the likelihood of two-photon fluorescence we use short laser pulses. To accurately measure the two-photon fluorescence of long wavelength dyes used in fluorescence guided surgery, there are certain properties of the laser pulse that need to be known. The goal of this experiment is to measure the laser spectra and pulse widths of a pulsed laser and to eventually determine the two-photon cross section of these long wavelength dyes. Using a spectrometer and the first order autocorrelation of the pulsed laser we are able to report that there is an offset of on average 12.8 ± 1.7 nm between the reported and actual wavelength values of our pulsed laser. Second order autocorrelation results yield an average pulse width of 120 ± 9 fs.

In a temporal focusing microscope, dispersion tuning via a 4f pulse shaper leads to axial shifts of the plane where pulses of an ultrashort pulsed laser are at their shortest, called the temporal focus. Synchronizing this with an electrically tunable lens (ETL) in the detection path of the camera enables remote acquisition of image stacks equivalent to those obtained using a traditional stage scan. In this paper, the limitations of both the two-photon signal generation and the penetration depth for such a system are discussed.

In a temporal focusing microscope, the spectrum of wavelengths comprising ultrashort laser pulses is separated spatially by a diffraction grating, collimated by a lens, and re-focused by an objective lens. The ultrashort pulse is only completely recollected at a single plane beyond the objective, which is called the temporal focus. By confining fluorescence to the depth of the temporal focus, it is possible to image a single depth of a fluorescent sample at a time and combine images of different depths to create a 3-dimensional rendering of the sample. In this paper, images acquired using this method with long-wavelength (1100+ nm) laser pulses are shown and the merits of such a system are discussed.

Confocal microscopy is a useful optical imaging technique for rendering three dimensional images of thick tissue. This technique utilizes fluorescence to capture several two-dimensional images at varying depths. Motorized mirrors scan the laser beam across the sample, and areas on the sample with fluorescent dye will glow. The emitted fluorescent light is directed back along the same path as the incoming scanning laser light, and the in-focus-light will be directed through a pinhole and hit a detector. The pinhole rejects all out-of-focus light, which is essential for clear 2-D images. After repeating this at different depths, a 3-D image can be reconstructed from the 2-D slices. In our project, we have yet to test samples labeled with dye, but instead utilized reflected laser light to run our tests. The ultimate goal of this project is to image infrared dyes and visible color dyes simultaneously, which cannot be done with a conventional confocal microscope, because lenses are subject to chromatic aberration. This causes lenses to focus different colors to different positions, so we cannot use the same lenses for all colors. To eliminate this issue, we use curved mirrors instead of lenses, since mirrors have a fixed focal length completely independent of color. This alteration allows different colors to be used with the same apparatus, making our design more convenient and versatile than a lens-based microscope.

The shape of a laser pulse in time can be controlled by changing the characteristics of the different frequencies that make up the laser. In this project, laser pulses were shaped by changing one such characteristic, phase, as a polynomial function of frequency using a liquid crystal spatial light modulator (SLM). To determine if the pulse was shaped as we expected, we compared theoretical and experimental autocorrelation traces. Major trends in experimental autocorrelation traces qualitatively agreed with theoretical higher-order traces, so we concluded that phase shifts were likely being applied as expected. However, noise in the traces suggest imprecision in pulse shaping due to the dimensions of the SLM and the geometry of the pulse shaper. One future application of the shaped laser pulses is in temporally focusing microscopy, where the shape of the laser pulse determines the location of imaged formed.

Most imaging technology is limited to 2-dimensions, lacking the depth dimension of the object. To address this issue, this project acquires the depth dimension through the use of beamsplitters, which split the light emitted from our sample into different paths to reach the camera. Because the light path lengths through each beamsplitter are different, different depths of the object are in focus on different parts of the camera sensor simultaneously. These regions on the camera represent different “slices,” and a single camera image can then be re-compiled into a 3D rendering of the object. The average focal distance between adjacent images, or images that appear in-focus consecutively, is 23.2 μm with a standard deviation of 9 μm. We calculate the resolution of our microscope to be 1.4 microns, which is limited by the size of the camera pixels. We successfully resolved the smallest features of a USAF target, which were 2.2 microns in size.

Temporal focusing microscopy enables imaging through the surface of biological tissue. Modeling the temporal focusing microscope’s optical system requires numerical simulations in order to determine how different diffraction gratings and lens focal lengths affect the electric field in space and frequency. The numerical simulations of a large number of points in space and frequency are computationally intensive, requiring the code be optimized for both computational and time efficiency. This research describes a computationally efficient Python code base, which enables analysis of the temporal focusing microscope system in an open-source language, allowing for widespread accessibility.

Light-field microscopy is a computational imaging method that enables 3D volumetric imaging from a single 2D image, however at a considerable computation cost and a compromised spatial resolution. A newer iteration on this technique is the Fourier light-field microscope which utilizes a completely different optical geometry that is easily scalable with a potential for much better spatial resolution, as well as lower computational cost. In this paper we outline the imaging mechanism, design parameters and computation algorithm. We will also analyze the data captured by a Fourier light-field microscope built in lab and suggest design improvements.

The goal of this project is to generate new colors of light for two-photon imaging using supercontinuum generation. Supercontinuum generation generates a wide range of frequencies from an ultrashort laser pulse traveling through glass due to the interactions between dispersion and nonlinearities. Dispersion is when a glass fiber has different refractive indices for different wavelengths, so the laser pulse broadens in time as it travels through the fiber because the different frequencies of the pulse travel at different speeds. The pulse experiences nonlinear effects as a result of high-intensity incident light causing the glass fiber to radiate new frequencies, which broadens the width of the pulse in frequency. Supercontinuum generation is the creation of almost white (broad bandwidth) light from the interaction between dispersion and nonlinearities. We have observed supercontinuum generation by sending an 800 nm laser pulse through a FemtoWHITE 800 photonic crystal fiber and measuring the output wavelength spectrum. We then compared the experimental observations to numerical simulations of supercontinuum generation performed in Mathematica. I have explained the equations and Mathematica code that I used to numerically simulate the results of sending a laser pulse through a thin optical fiber, so that this paper can be used as a manual for undergraduate researchers looking to simulate nonlinearities, dispersion, and supercontinuum generation in a photonic crystal fiber.

Two-photon microscopy is an imaging method in which two photons of a near-infrared wavelength are absorbed at a single point in a thick biological sample stained with fluorescent dye. When the photons are absorbed by the sample, a photon of a different wavelength is emitted where it is detected by a photomultiplier tube. The two-photon method allows individual depths of a thick sample to be imaged separately without physically slicing the sample into thin parts. I built a second channel on the existing microscope apparatus so that two different types of fluorescent dye can be detected simultaneously.

We present the implementation of a photothermal microscope to image sub-resolution nanoparticles. We utilize the light from a pump laser beam to heat the sample while using a second probe beam to detect the hot spots. The signal detected by the probe beam is weak, and thus we implement a chopper wheel to modulate the pump laser at a known frequency. This in turn modulates the refractive index of the absorbing structure, allowing for the separation of the high frequency signal and low frequency noise. Using a lock-in amplier, we detect this modulation from the probe beam signal, thus extracting information about the location of sub-resolution nanoparticles.

Laser experimentation requires proper alignment of the laser beam, which is typically done through an iterative manual process. Over the course of an experiment, it is common that a laser will gradually deviate from its original alignment. This deviation can result from ambient temperature fluctuations in the laboratory and changes within the laser itself, necessitating frequent realignment by the experimenter. Making use of quadrant photodiodes, an Arduino microcontroller, routine optics equipment, and standard electronics, this paper demonstrates an automatic laser alignment system that costs a fraction of those available commercially. The system is able to detect deviations in the laser path of (40 ± 20) μrad or greater.

A two-photon laser scanning microscope allows us to look inside a sample layer by layer without making an incision. However, when imaging low-signal samples, such as biological tissue, it is difficult to produce a clear image with our student-built microscope due to noise fluctuations and bleed-through from adjacent points on the sample. We present a modification that oversamples each point on the sample and forms clusters of the multiple values, and then averages the clusters before constructing the final image. This ensures that even if some of the brightness values are corrupted by noise fluctuations or neighboring pixels, their overall contribution to the final brightness value will be negligible when averaged with multiple brightness values. To measure the effectiveness of this modification, we calculate the standard deviation and signal-to-noise ratio of an image produced with the original sampling procedure and compare it to an image produced with the modified procedure. We find that when oversampling 10 times for each pixel, the average standard deviation for each pixel is 1.3 times smaller. We find that the signal-to-noise ratio is 24:61:3 for images captured with the original sampling procedure, and 29:0 1:5 for images captured with clustering sampling procedure.

Laser scanning microscopes take three dimensional images by focusing a laser beam to a single point in space, and scanning across the sample over a period of time, recording an image pixel by pixel. In each short time window, or pixel, emitted fluorescent light at a certain location is collected within a photomultiplier tube (PMT) which releases an output voltage pulse. This pulse must then be converted into a numerical value so that the pixels can combine to form an image. For this project, I am building an electronic analog circuit that integrates the signal from the PMT, giving a precise quantification of each pixel in the imaged sample in order to reconstruct an image.

Pulse shaping is a method in which the individual wavelengths that make up a laser pulse are variably delayed, which alters the properties of the laser. This has applications for two-photon microscopy, a form of 3D tissue imaging. We use an electrically tunable lens to shape ultrashort pulses. We measure broadening that matches theory to within 300 fs from 180 ± 4% fs to 2400 ± 4% fs, though due to issues with our measurement apparatus, it is possible that the actual range is greater than this.

In this project, we built an ultrashort pulse shaper. Ultrashort pulses (less than 100 femtoseconds) are formed when a wide bandwidth of frequencies interfere in phase. The primary goal of this project was to manipulate these phase relationships in order to change the shape and width of our laser pulses, which can be done using a deformable mirror. After programming the mirror, we then measured the pulses of the spectral phase shifted laser using autocorrelation. By matching these measurements with simulations and theory, we found that we could reasonably accurately shape pulses over a moderate range by adding second order dispersion, with this range being limited by our equipment.

Depending on laser pulse width, laser power, fiber shape, and fiber core size, the propagation of a pulse through an optical fiber can broaden and even distort the pulse. Chromatic dispersion, in which the refractive index is wavelength-dependent and therefore different wavelengths travel through the fiber at different speeds, dominates in large-mode-area fibers. Nonlinearities primarily occur in step-index single mode fibers. Because the light is confined to a smaller core, the intensity of the light is much greater in step-index single mode fibers, and nonlinear effects change the pulse shape proportional to the laser power. Two different fiber lengths will be used to analyze the relationship between chromatic dispersion effects and the distance through which the pulses travel through the fiber, in addition to developing a computer simulation to predict this broadening. Two types of single-mode fiber cores will be tested, large-mode-area and step-index single mode, to study nonlinearities and dispersion experimentally and through a numerical model.

Two-photon fluorescence microscopy allows for imaging deep within biological tissue without making an incision. Two-photon fluorescence occurs when two photons are simultaneously absorbed by a dye molecule and a single higher energy photon is released. The low probability of two-photon absorption ensures that fluorescence will only occur where many photons are focused at the same place at the same time. By moving the focus of the laser, a three-dimensional image is produced with micron level resolution. We have built an upright microscope that can image biological samples and created a user-friendly computer program to control it. By measuring the point spread function of the microscope, we have found the lateral resolution to be (1.6 ± 0.1) μm and the axial resolution to be (4.4 ± 0.3) μm.

Two-photon fluorescence microscopy allows for imaging deep within biological tissue without making an incision. Two-photon fluorescence occurs when two photons are simultaneously absorbed by a dye molecule and a single higher energy photon is released. The low probability of two-photon absorption ensures that fluorescence will only occur where many photons are in the same place at the same time, i.e., at the focus of a laser beam. By moving the focus of the laser, a 3D image is produced with micron level resolution. We have built an upright microscope to allow for water immersion samples, and by experimentally measuring its point spread function have found its lateral resolution to be (2.2 ± 0.2) μm.

We have designed and built detection electronics for a laser scanning microscope. The microscope operates by scanning a laser across a sample pixel by pixel and recording the fluorescence with a photomultiplier tube. The photons that make up the fluorescence are recorded by a computer in order to digitally re-construct an image. Currently the detection system samples only once per pixel. By using an analog integration technique we hope to improve the resolution. To accomplish this we have constructed an op amp based analog circuit that integrates the voltage over a set window. To maximize the speed and signal-to-noise ratio of the system it was created to fit on a Printed Circuit Board.

Using thin films of fluorescent dye, I characterize the imaging quality of a temporal focusing microscope. Because two-photon absorption generates very few fluorescent photons, I employ photon counting using a photomultiplier tube (PMT), a comparator to discriminate between signal and noise, and a pulse-counter DAQ board. Using the PMT, I determine that the axial profile of the illumination region is Lorentzian in shape, with a width of 14 μm. After positioning the sample at the focus, I acquire images of the illumination using an sCMOS camera, yielding a lateral profile that is an asymmetric Gaussian with widths of 131 μm and 30.2 μm in each dimension. I then measure the point spread function using 1 μm-diameter fluorescent beads. Both the lateral and axial profiles of the beads are Gaussians, with dimensions of 1.2 μm, 1.2 μm, and 4.2 μm.

Using fluorescent dye and fluorescent beads, I characterize the focus of a custom-built temporal focusing microscope. Because these films generate very few fluorescent photons through two photon excitation, counting photons rather than integrating the signal from the photomultiplier tube (PMT) yields improved sensitivity to low signals. A comparator converts PMT pulses above a certain discrimination level to uniform square pulses, and a digital data acquisition board counts these pulses via a LabVIEW program. With temporal focusing, the number of photon counts will peak at the position of the focus. The axial profile is Lorentzian in shape, with a full-width-at-half-maximum (FWHM) of 17 μm. After positioning the sample at the focus, I acquire 2D-images of the illumination profile. The lateral profile is an asymmetric Gaussian with FWHMs of 131 μm and 30.2 μm. Small fluorescent beads are also visible when viewed with a camera, yielding a lateral detection resolution of at least 1 μm.

Standard microscopes produce two-dimensional images, but to obtain depth information one must physically move the sample. This method is both cumbersome and cannot be done in real time. The Microsoft Kinect is a commercial gaming device that obtains live three-dimensional (3D) data of a scene. This project investigates the feasibility of applying the different Kinect sensors to 3D microscopy. Through reverse engineering we determined that the Kinect v2 uses a time-of-flight method for depth sensing which we could not apply to microscopy due to speed limitations of the electronics. However, we determined that the Kinect v1 uses a structured light method of depth sensing, and we successfully applied this to a single magnifying lens. This setup not only magnified the image 1.8x but also magnified the depth sensing abilities of the Kinect v1 3.6x. We believe that this type of depth sensor could be economically applied to microscopy in the future.

A multi-photon microscope uses a focused laser beam to create fluorescence images pixel by pixel. This project creates a laser-scanning microscope capable of acquiring high quality, three-dimensional images. To begin, we program rotating galvanometer mirrors to deflect a beam, and rely on a series of lenses to convert the deflection angle to a lateral position on the sample. To calibrate and align this microscope setup, we test our automation programs with an alignment laser. Later on, we use a pulsed, mode locked Titanium-Sapphire laser to excite two photon fluorescence in a basic sample. Construction of an epi-detection system allows us to raster scan a laser focus around a volume several hundred microns in each direction and generate a three-dimensional image. Ultimately we generated a 3D multiphoton image using lens tissue stained with fluorescein.

A multi-photon microscope relies on a sharp focus to produce images of a fluorescent sample. This project will develop the infrastructure necessary to produce a high quality image using a simple continuous wave (cw) alignment laser. We use rotating galvanometer mirrors to deflect a beam, and a series of lenses converts the deflection angle to a lateral position on the sample. In addition, this alignment laser allows us to determine the system’s resolution and to calibrate the beam positions with respect to our LabVIEW program. Once the setup and calibration are complete, the apparatus can be readily adapted into a multi-photon system that will raster scan a pulsed laser through a fluorescent sample to generate depth-resolved images.