University:
Boston University
Course:
SAR HP 151 | Introduction to Health Professions
Academic year:

2020
Views:

61

Pages:

16
Author:

Greta Andrews

Boston University Slideshow Title Goes Here AI in Chemical and Biomedical Imaging Augmented Clinical Observation in 2013 Reality for UltrasoundBoston University Slideshow Title Goes Here guided Kidney Stone Removal FDA clearance 2020 Preclinical and clinical validation Acusee R&D (~ 2 years) Computational Chemical Microscopy Advanced Instrumentation + Data Science to break the tradeoff between Speed, bandwidth, and SNR “no free lunch” SNR Bandwidth Bandwidth + Data Science Spee d Spee d SNR Fingerprint Stimulated Raman Scattering (SRS) is Informative but Noisy Mia PaCa 2 at C-H Region Boston University Slideshow Title Goes Here Mia PaCa 2 at Fingerpri nt Region Weak but more information Strong but crowded Denoising Fingerprint SRS by a Spatial-Spectral Residual Net Mapping biofuel pinene and limonene inside E.coli SS-ResNet GT (100 averages) After training the network, denoising takes ~10 s Raw image Visualization of a Limonene Synthesis Metabolon inside Living Bacteria by fingerprint SRS imaging DH1 strain 1p Cell body C-H Acetyl-CoA atoB HMG S HMG R Mevalonat e M K PMK PM D IPP 3 4 1 8 2 L S trGPPS 1 670-900 cm- Limonene 5 μm 76 5 1 0 ~760 cm-1 GPP Limonene WT id i 9 Limonene DMAPP … trGP PS LS … trGP PS × LS … trGP PS LS × … × × trGP PS LS Curating Imaging Data for Computational Models in Biomechanics and Mechanobiology Boston University Slideshow Title Goes Here Focus: methods for creating interpretable structurefunction computational models for biomechanics and mechanobiology. current interest: large datasets image analysis software (some ML) curated data combined mechanistic/ML models computational models Boston University Slideshow Title Goes Here large datasets curated data computational models automated image analysis: Boston University Slideshow Title Goes Here validation with synthetic data: large datasets curated data reproducible pipeline computational models integrating data-driven and mechanistic models is ongoing… Boston University Slideshow Title Goes Here lab website (includes more info on non-bio work) large datasets curated data computational models Boston University Slideshow Title Goes Here Deep Learning Augmented Computational Imaging for Biomedical Microscopy & Neural Imaging BU Computational Imaging Systems Lab Computational phase imaging Computational Neural Imaging Boston University Slideshow Title Goes Here Imaging in scattering media Deep learning for microscopy Deep learning for neural imaging 56 Computational Miniature Mesoscope (CM2) for large-scale neural imaging Boston University Slideshow Title Goes Here CM2 achieves single-shot cortex-wide, cellular-resolution 3D imaging by integrating novel miniature optics and advanced computational algorithms Collaboration with David Boas, Ian Davison Self-supervised denoising for in vivo spatiotemporal neural imaging data Mean frame raw Single Boston University Slideshow Title Goes Here frame raw Single frame denoised input DeepVID output In vivo two-photon voltage imaging DeepVID achives 15X improvement in SNR Collaboration with Jerry Chen Deep learning for information extraction from multi-modal, multiscale imaging data Boston University Slideshow Title Goes Here 𝑿𝑿: OCT scattering coefficients Relating label-free OCT measurements to Gallyas stain measurements for quantifying myelin content and neuron density in human brains 𝒀𝒀: Gallyas of adjacent slice