A5078330200- Medical Imaging Techniques and Applications
- Radiomics and Machine Learning in Medical Imaging
- Advanced MRI Techniques and Applications
- Radiation Detection and Scintillator Technologies
- Advanced X-ray and CT Imaging
- Advanced Radiotherapy Techniques
- Generative Adversarial Networks and Image Synthesis
- Cell Image Analysis Techniques
- Atomic and Subatomic Physics Research
- Robotic Path Planning Algorithms
- Underwater Vehicles and Communication Systems
- Aerospace Engineering and Energy Systems
Stanford University
2019-2024
Siemens (United States)
2021-2024
University of Wisconsin–Madison
2017
We present an approach that allows the Georgia Tech Miniature Autonomous Blimp (GT-MAB) to detect and follow a human. This accomplishment is first Human Robot Interaction (HRI) demonstration between uninstrumented human robotic blimp. GT-MAB ideal platform for HRI missions because it safe humans can support sufficient flight time experiments. However, due complex aerodynamic influence on blimp, following task with single on-board camera challenging problem. integrate Haar face detector KLT...
For photon attenuation correction, current positron emission tomography systems combined with magnetic resonance imaging (PET/MR) typically use methods based on MR image segmentation subsequent assignment of empirical coefficients in PET reconstruction. Delineation bone images has been challenging, especially the head and neck areas, due to difficulty separating from air. In this article, we study deep learning techniques that assist MR-based correction (MRAC) process for PET/MR systems,...
Simultaneous PET/MRI combines two powerful and complementary modalities, providing multiparametric information to assess the anatomic as well biochemical basis of disease. However, high cost current commercial integrated systems limit long-term potential, accessibility, availability PET/MRI. A portable PET insert provides a cost- effective alternative achieve simultaneous for sites that are already equipped with MR systems. Therefore, we developing radiofrequency (RF)-penetrable...
We present a context-aware generative deep learning framework to produce photon attenuation and scatter corrected (ASC) PET images directly from non-attenuation non-scatter (NASC) images. trained conditional adversarial networks (cGAN) on either single-modality or multi-modality (NASC+MRI) input data map NASC pixel-wise continuously valued ASC designed evaluated four cGAN models including Pix2Pix, attention-guided (AG-Pix2Pix), vision transformer (ViT-GAN), shifted window (Swin-GAN)....
Attenuation correction is an important for quantitative PET image reconstruction. Current PET/MR attenuation methods involve segmenting MR images acquired with zero-time echo (ZTE) or Dixon sequences and assigning known coefficients to different tissues. This work builds upon our previous where we explore a novel deep learning method of map (μ-map) generation using conditional generative adversarial network (cGAN) that allows continuous coefficients. We develop the use cGAN directly convert...
Accurate photon attenuation correction (AC) is essential for quantitative PET image reconstruction. MR-based AC of combined PET/MR systems challenging because MR images do not directly correspond to coefficients, as the case CT images. Deep learning (DL)-based methods have shown promise alternative solutions. In this work, we evaluate accuracy using a DL model purpose on reconstructed head and neck We use conditional generative adversarial network (cGAN) known pix2pix translate input into...
Attenuation correction is essential for accurate PET images, as it corrects photon loss due to scatter or absorption in tissue, which can reduce signal intensity by up 90%. Deep Learning methods have been proposed generate pseudo-CT images from MR AC PET. Despite their potential, generating MRI data remains a challenge issues such poor image quality, limited tissue heterogeneity, patient-specific anatomical variations, and pathologies, well time-consuming processing. We present novel...
In this work, we evaluate the feasibility of a simplified approach performing MR-based attenuation correction (MRAC) by generating AC PET images directly from non-attenuation corrected (NAC) and MR using machine learning model known as conditional generative adversarial network (cGAN). We for head neck region PET/MR PET/CT dataset. trained tested converting Dixon Water, Fat, NAC (multi-channel input) to PET. Our achieved mean pixel value difference 2.06% in region, NRMSE 1.29%, PSNR 37.55,...
In simultaneous positron emission tomography and magnetic resonance (PET/MR) imaging, MR radio-frequency (RF) coils are placed on the patient to receive signal. These can produce an undesirable attenuation of PET signal by as much 17% in certain cases. Currently, photon correction (AC) flexible RF is not typically performed commercial PET/MR systems. To correct for this attenuation, position must be determined. This work proposes a simple effective solution problem using three 2D optical...