Chip-Scale Fluorescence Microscope

Efthymios Papageorgiou

EECS Department
University of California, Berkeley
Technical Report No. UCB/EECS-2019-31
May 10, 2019

http://www2.eecs.berkeley.edu/Pubs/TechRpts/2019/EECS-2019-31.pdf

CMOS image sensors have been widely used for decades, largely displacing CCD technology in cameras, scanners, telescopes, and a variety of medical sensors. Photodiodes are formed in the IC substrate and the technology has been repurposed to make more advanced structures with higher sensitivity, such as pinned or avalanche photodiodes. Less work has gone into generating optical elements using other features of the CMOS process, namely the metal interconnect. The precision deposition and spacing of the metal layers allows a variety of optical structures to be fabricated within the image sensor to perform both spatial and wavelength filtering. This is especially useful for applications where large-scale optical components, such as lenses or fiber optic bundles, cannot be used. For most intraoperative imaging applications and especially for modern minimally-invasive or robotic-assisted cancer surgery, these large optical elements restrict an imager's ability to thoroughly scan an entire tumor cavity.

This thesis presents an image sensor incorporating angle-selective gratings for resolution enhancement in contact imaging applications. Optical structures designed in the CMOS metal layers above each photodiode form the angle-selective gratings that limit the sensor angle of view to plus or minus 18 degrees, rejecting background light and deblurring the image. The imager is based on a high-gain capacitive transimpedance amplifier pixel using a custom 11fF MOM capacitor, achieving 8.2V/s/pW sensitivity. The pixel also includes a leakage current minimization circuit to remove signal-dependent reset switch leakage and the corresponding dark current is 14aA/um^2. The resulting 4.7mm by 2.25mm sensor (80 by 36 pixels) is designed specifically for intraoperative cancer imaging in order to solve the pervasive challenge of identifying microscopic residual cancer foci in vivo.

A custom amorphous silicon optical wavelength filter is used alongside the image sensor in order to perform fluorescence imaging. The filter rejects excitation light and has over five orders of magnitude rejection at 633nm. Fluorescence emission light above 700nm is allowed to transmit through. Several clinically-tested fluorophores, such as IR700DX, are compatible with this wavelength range. The filter is only 15um thick and is suitable for in vivo contact imaging applications as it maintains the same high rejection for all angles of incident light.

We demonstrate imaging and detection of foci containing less than 200 cancer cells labeled with fluorescent biomarkers in 50ms and signal-to-noise ratios greater than 15dB using the custom image sensor and optical filter. We also demonstrate the detection of microscopic cancer using human tissue and residual tumor in mice models. The absence of large optical elements enables extreme miniaturization, allowing manipulation within a small, morphologically complex, tumor cavity.

Advisor: Bernhard Boser


BibTeX citation:

@phdthesis{Papageorgiou:EECS-2019-31,
    Author = {Papageorgiou, Efthymios},
    Title = {Chip-Scale Fluorescence Microscope},
    School = {EECS Department, University of California, Berkeley},
    Year = {2019},
    Month = {May},
    URL = {http://www2.eecs.berkeley.edu/Pubs/TechRpts/2019/EECS-2019-31.html},
    Number = {UCB/EECS-2019-31},
    Abstract = {CMOS image sensors have been widely used for decades, largely displacing CCD technology in cameras, scanners, telescopes, and a variety of medical sensors. Photodiodes are formed in the IC substrate and the technology has been repurposed to make more advanced structures with higher sensitivity, such as pinned or avalanche photodiodes. Less work has gone into generating optical elements using other features of the CMOS process, namely the metal interconnect. The precision deposition and spacing of the metal layers allows a variety of optical structures to be fabricated within the image sensor to perform both spatial and wavelength filtering. This is especially useful for applications where large-scale optical components, such as lenses or fiber optic bundles, cannot be used. For most intraoperative imaging applications and especially for modern minimally-invasive or robotic-assisted cancer surgery, these large optical elements restrict an imager's ability to thoroughly scan an entire tumor cavity.

This thesis presents an image sensor incorporating angle-selective gratings for resolution enhancement in contact imaging applications. Optical structures designed in the CMOS metal layers above each photodiode form the angle-selective gratings that limit the sensor angle of view to plus or minus 18 degrees, rejecting background light and deblurring the image. The imager is based on a high-gain capacitive transimpedance amplifier pixel using a custom 11fF MOM capacitor, achieving 8.2V/s/pW sensitivity. The pixel also includes a leakage current minimization circuit to remove signal-dependent reset switch leakage and the corresponding dark current is 14aA/um^2. The resulting 4.7mm by 2.25mm sensor (80 by 36 pixels) is designed specifically for intraoperative cancer imaging in order to solve the pervasive challenge of identifying microscopic residual cancer foci in vivo. 

A custom amorphous silicon optical wavelength filter is used alongside the image sensor in order to perform fluorescence imaging. The filter rejects excitation light and has over five orders of magnitude rejection at 633nm. Fluorescence emission light above 700nm is allowed to transmit through. Several clinically-tested fluorophores, such as IR700DX, are compatible with this wavelength range. The filter is only 15um thick and is suitable for in vivo contact imaging applications as it maintains the same high rejection for all angles of incident light.

We demonstrate imaging and detection of foci containing less than 200 cancer cells labeled with fluorescent biomarkers in 50ms and signal-to-noise ratios greater than 15dB using the custom image sensor and optical filter.  We also demonstrate the detection of microscopic cancer using human tissue and residual tumor in mice models. The absence of large optical elements enables extreme miniaturization, allowing manipulation within a small, morphologically complex, tumor cavity.}
}

EndNote citation:

%0 Thesis
%A Papageorgiou, Efthymios
%T Chip-Scale Fluorescence Microscope
%I EECS Department, University of California, Berkeley
%D 2019
%8 May 10
%@ UCB/EECS-2019-31
%U http://www2.eecs.berkeley.edu/Pubs/TechRpts/2019/EECS-2019-31.html
%F Papageorgiou:EECS-2019-31