Browsing by Author "Zhao, Yongyi"

Now showing 1 - 3 of 3
  • Thumbnail Image
    Item
    GDOT: Gated Diffuse Optical Tomography
    (2021-04-29) Zhao, Yongyi; Veeraraghavan, Ashok
    Time-of-flight diffuse optical tomography (ToF-DOT) is a promising technology for non-invasive functional neuroimaging. Such an imaging device could lead to wearable brain-computer interfaces (BCIs) for prosthetics and faster communication. Unfortunately, ToF-DOT suffers from low spatial resolution and poor temporal resolution. The low spatial resolution is caused by the scattering of light in biological tissue and complex electronics, which limit the detector size. The poor temporal resolution is due to the large volume of measurements that must be transferred and processed. To address these challenges, we propose Gated Diffuse Optical Tomography (GDOT). GDOT uses time-gating to mitigate the effects of optical scattering while reducing the hardware and computational complexity associated with collecting full light transport transients. To demonstrate the capabilities of our proposed system, we conducted a simulated performance analysis and image reconstruction on experimental data collected from our prototype system. We showed over two orders of magnitude reduction in the algorithm runtime and enhanced image reconstruction quality using our GDOT technique.
  • Thumbnail Image
    Item
    Harnessing Scattered Light Transport: Computational Imaging Systems for One-Way Visibility and Polarimetric and Time-of-flight Inverse Rendering
    (2024-04-15) Zhao, Yongyi; Veeraraghavan, Ashok
    This thesis explores two challenges for imaging in the presence of optical scatterers: retrieving optical properties of the scene through novel imaging modalities and establishing optical asymmetry using scattering. In many optical imaging applications, such as non-invasive functional neuroimaging, light scattering leads to reduced performance, including reduced contrast and spatial resolution. Traditional imaging relies solely on light-intensity measurements. This may be insufficient for inverse rendering, i.e. retrieval of geometric and optical properties of the target from captured images. The first part of this thesis explores how time-of-flight measurements can improve the spatial resolution for imaging through volumetric-scattering media. This time-of-flight imaging approach, time-of-flight diffuse optical tomography (ToF-DOT), uses both physics-based and data-driven models to improve the algorithm runtime, measurement collection latency, and spatial resolution of the image reconstruction. I show that ToF-DOT can improve the state-of-the-art for imaging through densely scattering media. The second part of this thesis explores polarimetric inverse rendering. This technology, polarization-aided neural decomposition of radiance (PANDORA), combines polarimetric imaging with neural rendering algorithms to achieve state-of-the-art inverse rendering results. PANDORA can perform material estimation and 3D geometry estimation, and can handle complex surface interactions, such as subsurface scattering. Finally, the third part of this thesis explores how light scattering can be used to develop an optically asymmetric plume. Such a plume would degrade the image quality more heavily in one viewing direction compared to the opposite direction. This work explores how differentiable rendering and end-to-end optimization can be used to design this plume.
  • Thumbnail Image
    Item
    Unrolled-DOT: an interpretable deep network for diffuse optical tomography
    (SPIE, 2023) Zhao, Yongyi; Raghuram, Ankit; Wang, Fay; Kim, Stephen Hyunkeol; Hielscher, Andreas H.; Robinson, Jacob T.; Veeraraghavan, Ashok
    SignificanceImaging through scattering media is critical in many biomedical imaging applications, such as breast tumor detection and functional neuroimaging. Time-of-flight diffuse optical tomography (ToF-DOT) is one of the most promising methods for high-resolution imaging through scattering media. ToF-DOT and many traditional DOT methods require an image reconstruction algorithm. Unfortunately, this algorithm often requires long computational runtimes and may produce lower quality reconstructions in the presence of model mismatch or improper hyperparameter tuning.AimWe used a data-driven unrolled network as our ToF-DOT inverse solver. The unrolled network is faster than traditional inverse solvers and achieves higher reconstruction quality by accounting for model mismatch.ApproachOur model “Unrolled-DOT” uses the learned iterative shrinkage thresholding algorithm. In addition, we incorporate a refinement U-Net and Visual Geometry Group (VGG) perceptual loss to further increase the reconstruction quality. We trained and tested our model on simulated and real-world data and benchmarked against physics-based and learning-based inverse solvers.ResultsIn experiments on real-world data, Unrolled-DOT outperformed learning-based algorithms and achieved over 10× reduction in runtime and mean-squared error, compared to traditional physics-based solvers.ConclusionWe demonstrated a learning-based ToF-DOT inverse solver that achieves state-of-the-art performance in speed and reconstruction quality, which can aid in future applications for noninvasive biomedical imaging.