Real-Time Intraoperative Fluorescence Imager for Microscopic Residual Tumor in Breast Cancer Efthymios Papageorgiou, Dr. Mekhail Anwar (UCSF), Prof. Bernhard Boser (UC Berkeley)
Outline Microscopic cancer: background and motivation Angle-selective fluorescence contact imaging ASIC Amorphous silicon absorption filter Integration & Conclusion
Microscopic Cancer: Background & Motivation
Microscopic residual cancer Large Tumor: Typically 1-3cm diameter 5cm Mammogram Human Tissue
Microscopic residual cancer Tumor Left Behind: 10s-1000s of cells >200 cells known to be harmful (AJCC Handbook) 1cm Resection cavity: 3-30cm 2 surface area Human Tissue
Residual cancer is a widespread problem Chance of Cancer Returning (%) No Residual Tumor Cancer Left Behind
Impact of residual tumor in breast cancer 230,000 women diagnosed each year in U.S.
Impact of residual tumor in breast cancer 230,000 women diagnosed each year in U.S. 65%, 150,000 undergo lumpectomies
Impact of residual tumor in breast cancer 230,000 women diagnosed each year in U.S. 65%, 150,000 undergo lumpectomies 16.5%, 40,000 have residual tumor Further treatment: Additional surgery ($40,000)
Detection: sensor requirements Human Tissue
Detection: sensor requirements Fit in 1-3cm diameter resection cavity Human Tissue
Detection: sensor requirements Fit in 1-3cm diameter resection cavity Image entire 3-30cm 2 rapidly Human Tissue
Detection: sensor requirements Fit in 1-3cm diameter resection cavity Image entire 3-30cm 2 rapidly Clusters of >200 cells (AJCC) If >200 cells, should be removed Human Tissue
Detection: sensor requirements Fit in 1-3cm diameter resection cavity Image entire 3-30cm 2 rapidly Clusters of >200 cells (AJCC) Should not disrupt surgeon s typical workflow If >200 cells, should be removed Human Tissue
Medical Imaging Techniques MRI: CT: philips.com usa.healthcare.siemens.com No cancer-specific markers
Medical Imaging Techniques MRI: CT: philips.com usa.healthcare.siemens.com No cancer-specific markers Rely on non-specific factors (i.e., asymmetry)
Medical Imaging Techniques MRI: CT: philips.com usa.healthcare.siemens.com No cancer-specific markers Rely on non-specific factors (i.e., asymmetry) Can only detect ~1cm 3 tumor = 1 billion cells
Fluorescence imaging Targeted molecular labels: Antibody + Fluorophore
Fluorescence imaging Targeted molecular labels: Antibody + Fluorophore Labeled cancer cells
Fluorescence imaging Targeted molecular labels: Antibody + Fluorophore Excitation source Excitation Light Labeled cancer cells Emission Light
Fluorescence imaging Targeted molecular labels: Antibody + Fluorophore Excitation source Wavelength filter Excitation Light Labeled cancer cells Emission Light
Fluorescence imaging Targeted molecular labels: Excitation source Antibody + Fluorophore Sensor Wavelength filter Excitation Light Labeled cancer cells Emission Light
Normalized Emission Targeted fluorescence labels Clinically tested fluorophores in vivo IR700DX IR800CW ICG Normalized Emission, 633nm Excitation For testing purposes Quantum dots (Qdot 705) Bind with targeted antibody Breast: Trastuzumab (Herceptin) Wavelength (nm)
Fluorescence microscope High resolution Bulky Cannot enter tumor bed Restricted to LOS Fluorescence microscope 1 Microscope >2m Miss sidewalls! Rigid optics Breast tumor bed Operating Table 1 Van Dam et al., Nature Medicine, 2011.
Fiber optic-based imagers Inside tumor bed Fiber bending radius limited to >2cm for fiber bundle diameter >3mm 1 Fiber optic bundle Cannot image all points of complex tumor cavity! Breast tumor bed 1 D. Shin et al., PLoS One, June, 2010.
Contact imaging using CMOS IC Can mount on a flexible probe, place directly against tissue No physically constraining optics Scalable CMOS Image Sensor Breast tumor bed Cancer Cell Fluorescent tag
Blur and background light Sensor Two objects
Blur and background light Two images Sensor Two objects
Blur and background light Two images Sensor Two objects Gap
Blur and background light Two images Sensor Two objects Images blur Gap
Blur and background light Two images Sensor Object can be lost Two objects Images blur Gap Cancer cell cluster
Angle-Selective Contact Imaging ASIC
Blur Angle-selective gratings (ASG) No optics
Blur Angle-selective gratings (ASG) No optics Blinders
Blur Angle-selective gratings (ASG) No optics Top View Cross-Section Metal Oxide Single Photodiode Side View Blinders
Block off-axis light Blocked Passes through
Depth (μm) Depth (μm) Depth (μm) ASG 2D Simulations 0 4 0 incidence 8 0 20 40 60 80 100 0 4 30 incidence 8 0 20 40 60 80 100 0 4 8 60 incidence 0 20 40 60 80 100 Distance (μm)
ASG perspective side view M5 M1-4 6.8μm
ASG orthographic top view 45μm 2.4μm
10μm 500μm System overview Background Signal Cancer Signal Tissue Biomarker Cancer Fluorophore Quartz Wavelength Filter 6.8μm Excitation Light CMOS Chip ASG over PD 55μm
2.25mm Chip block diagram Chip Boundary 4.7mm 80x36 Pixel Array Shared Column Current Sources 80 Reset Lines 80 Signal Lines 10:1 MUX and Output Buffers Digital Drive Digital Control 8 Reset Lines 8 Signal Lines 8 12-bit ADCs FPGA
Reset Col Signal Col Pixel architecture: CTIA Reset Switch Φ R SF C int C LR PD w/asg Φ S SF C LS
Complete pixel In Pixel 1.8V 1.8V M 11 V b1 V b1 M 4 Φ R M 13 M 12 SF Reset M 10 V b2 V b2 M 3 Φ R C LR M 9 M 8 V rep V b3 Φ RST M 7 Φ RST V M x 6 Φ RST M 5 PD w/asg C int V b3 M 2 M 1 Φ S Φ S M 15 M 14 C LS SF Signal
Lateral flux MOM integration capacitor Metal 5 Metal 3-4 and Vias C int = 11fF, Responsivity = 8.2V/s per pw Metal 2
2.25mm Chip micrograph 4.7mm 80x36 Angle-Selective Pixel Array Column current sources and output buffers Digital decoders and row select circuits
2.25mm Chip micrograph Pixel Pitch = 55μm PD Area = (44μm) 2 Fill Factor = 28% 4.7mm 80x36 Angle-Selective Pixel Array Pixel Close-up 55μm Readout Column current sources and output buffers Digital decoders and row select circuits Angle-selective grating over photodiode
Normalized Pixel Response Angle-selectivity Input light 14dB boost in SBR 0 Sensor θ 50% @ ±18 Angle of Incident Light ( )
Cell-imaging experimental setup Angle-Selective ASIC & optical wavelength filter Sensor Quartz, 0.5mm Coverslip, 0.15mm Labeled cancer cells in Matrigel, ~40μm thick Glass slide, 1.5mm
3D breast cancer cell cultures ASIC Image, T int = 50ms 500µm Microscope Image, T int = 1s 500µm
3D breast cancer cell cultures ASIC Image, T int = 50ms A: 180 cells, 15.1dB SNR A 500µm B B: 270 cells, 15.7dB SNR Microscope Image, T int = 1s A B 500µm
SNR (db) SNR vs. Cell Cluster Size SNR 15dB 160 Cells Number of Cells in Cluster
Human HER2+ breast cancer tissue #1 Microscope Image Custom ASIC Image 75ms 75ms, filtered 5s Cancer Healthy 4.4mm
Human HER2+ breast cancer tissue #2 Microscope Image Custom ASIC Image 75ms 75ms, filtered 4.4mm 5s
Amorphous Silicon Absorption Filter
Oblique illumination Necessary for in vivo contact imaging Tissue Cancer Steep! (70 ) Quartz Wavelength Filter Excitation Light CMOS Chip
Oblique illumination Necessary for in vivo contact imaging Tissue Cancer Scattering off tissue Quartz Wavelength Filter Excitation Light CMOS Chip
Thin-film interference filters Alternating layers with different indices of refraction Incident light Reflected light Transmitted light chroma.com
Interference filter angle dependence
Absorption filters
Absorption filters Colored glass 2-3mm thickness!
Transmittance Amorphous silicon absorption a-si bandgap ~1.65eV, 750nm wavelength λ edge = hc E g No Transmission Transmission Wavelength
Transmittance Transmittance Amorphous silicon absorption a-si bandgap ~1.65eV, 750nm wavelength λ edge = hc E g Urbach tail No Transmission Transmission Wavelength Cody et al., Physical Review Letters, 1981. Wavelength
Fabricated amorphous silicon filters 5μm, 10μm, 15μm thickness layers a-si:h on a 500μm thickness fused silica wafer 5μm thickness
Fabricated amorphous silicon filters 5μm, 10μm, 15μm thickness layers a-si:h on a 500μm thickness fused silica wafer 5μm thickness 700nm thickness
Measured transmittance spectra 1.4μm -1 @ 633nm 54% in passband
Transmittance spectra and IR700DX
Relative angular response, λ ex = 633nm
15 μm a-si 10 μm a-si Hillocks form during fabrication 200 μm 100 μm 200 μm 200 μm
ASIC with a-si optical filter 4.7mm
Human HER2+ Tissue and a-si filter T int = 75ms T int = 5s
Alternative semiconductor materials Material Bandgap (ev) Wavelength (nm) Gallium Phosphide (GaP) Cadmium Selenide (CdSe) Amorphous Silicon (a-si) Gallium Arsenide (GaAs) Indium Phosphide (InP) Crystalline Silicon (c-si) 2.25 550 1.74 710 1.65 750 1.43 870 1.27 980 1.11 1100
Integration
Integrate laser diodes surrounding ASIC Necessary for in vivo contact imaging Tissue Cancer Excitation Light Excitation Light Quartz Wavelength Filter Laser Diode Chip Laser Diode Chip CMOS Chip
Integrated system on Flex PCB
Conclusion Large optics can t be used in a tumor cavity A contact image sensor with ASG and amorphous silicon optical filter images the entire cavity Breast cancer detection <200 cells, 3D cultures Human HER2+ tissue Combines in vivo use, resolution, and speed Currently performing animal testing to validate the sensor in vivo
Acknowledgements The Department of Defense (DoD), the Mary Kay Foundation, the American Society of Clinical Oncology (ASCO) Hui Zhang and Dr. Catherine Park, UCSF TSMC University Shuttle Program for chip fabrication Mark Brunson and Darick Baker, University of Washington Nanofabrication Facility My colleagues: Simeon Giverts, Hossein Najafi, Burak Eminoglu, Behnam Behroozpour, Pramod Murali, Hao-Yen Tang, Andy Michaels
Commercialization through industry/university partnerships since 1986 Booth 1339 Contact Dr. Mike Cable mdcable@berkeley.edu +1 510 643 5663 bsac.berkeley.edu 2016 University of California Berkeley Sensor & Actuator Center