Leland Stanford Junior University

University of California

Lei Xing

Christopher Barry Locke

Karl Kenneth Bush

Michael J Hadsell

Lei Zhu

Paul J Keall

Amit Ramakant Sawant

Yan Wu

Liyue Shen

Michael Sano

Norbert Joseph Pelc

Steven M Conolly

Rebecca Fahrig

Paul Keall

Greig Scott

Peter G Maxim

Kim Butts Pauly

Guillem Pratx

Jing Wang

Cesare Jenkins

Yelin Suh

Bin Han

Rodney Wiersma

Billy W Loo, Jr

Benjamin Pooya Fahimian

Ruijiang Li

Charles Huang

Ming Ma

Dominik Jan Naczynski

Yong Yang

Wei Zhao

Colin M Carpenter

Conroy Ghin Chee Sun

Yunzhi Ma

Bowen Meng

Byung Chul Cho

Wu Liu

Gary Luxton

Weihua Mao

Andrei Pugachev

Michael Hadsell

A method for diagnostic imaging reconstruction uses a prior image xfrom a scan of a subject to initialize parameters of a neural network which maps coordinates in image space to corresponding intensity values in the prior image. The parameters are initialized by minimizing an objective function representing a difference between intensity values of the prior image and predicted intensity values output from the neural network. The neural network is then trained using subsampled (sparse) measurements of the subject to learn a neural representation of a reconstructed image. The training includes minimizing an objective function representing a difference between the subsampled measurements and a forward model applied to predicted image intensity values output from the neural network. Image intensity values output from the trained neural network from coordinates in image space input to the trained neural network are computed to produce predicted image intensity values.

Implicit neural representation learning with prior embedding for sparsely sampled image reconstruction and other inverse problems

A method for tomographic imaging comprising acquiring [] a set of one or more 2D projection images [] and reconstructing [] a 3D volumetric image [] from the set of one or more 2D projection images [] using a residual deep learning network comprising an encoder network, a transform module and a decoder network, wherein the reconstructing comprises: transforming [] by the encoder network the set of one or more 2D projection images [] to 2D features []; mapping [] by the transform module the 2D features [] to 3D features []; and generating [] by the decoder network the 3D volumetric image [] from the 3D features []. Preferably, the encoder network comprises 2D convolution residual blocks and the decoder network comprises 3D blocks without residual shortcuts within each of the 3D blocks.

Single or a few views computed tomography imaging with deep neural network

A method for quantitative magnetic resonance imaging (MRI) includes [] performing an MRI scan using a conventional pulse sequence to obtain a qualitative MR image; and [] applying the qualitative MR image as input to a deep convolutional neural network (CNN) to produce a quantitative magnetic resonance (MR) relaxation parametric map. The qualitative MR image is the only image input to the deep neural network to produce the quantitative MR relaxation parametric map. The conventional pulse sequence may be a Spoiled Gradient Echo sequence, a Fast Spin Echo sequence, a Steady-State Free Precession sequence, or other sequence that is commonly used in current clinical practice.

Qualitative and quantitative MRI using deep learning

A treatment planning prediction method to predict a Dose-Volume Histogram (DVH) or Dose Distribution (DD) for patient data using a machine-learning computer framework is provided with the key inclusion of a Planning Target Volume (PTV) only treatment plan in the framework. A dosimetric parameter is used as an additional parameter to the framework and which is obtained from a prediction of the PTV-only treatment plan. The method outputs a Dose-Volume Histogram and/or a Dose Distribution for the patient including the prediction of the PTV-only treatment plan. The method alleviates the complicated process of quantifying anatomical features and harnesses directly the inherent correlation between the PTV-only plan and the clinical plan in the dose domain. The method provides a more robust and efficient solution to the important DVHs prediction problem in treatment planning and plan quality assurance.

Dosimetric features-driven machine learning model for DVHs/dose prediction

A method for radiation therapy treatment planning includes: a) defining a decision variable search space by projecting an initial seed point onto a pareto front, where the pareto front and decision variable search space are defined in the same coordinate space; b) projecting search points in the decision variable search space to the pareto front using a one-dimensional search algorithm to produce projected points on the pareto front; c) updating the search points by performing a gradient-free optimization that evaluates a treatment planning scoring function at the projected points on the pareto front; and d) repeating steps (b), (c) to search within the decision variable search space until a convergence criterion is satisfied, thereby producing a pareto optimal and clinically acceptable treatment plan.

Pareto optimal projection search (POPS) for automated radiation therapy treatment planning

Retrospective magnetic resonance imaging (MRI) uses a deep neural network framework [] to generate from MRI imaging data [] acquired by an MRI apparatus using a predetermined imaging protocol tissue relaxation parametric maps and magnetic/radiofrequency field maps [] which are then used to generate using the Bloch equations [] predicted MRI images [] corresponding to imaging protocols distinct from the predetermined imaging protocol. This allows obtaining a wide spectrum of tissue contrasts distinct from those of the acquired MRI imaging data.

Retrospective tuning of soft tissue contrast in magnetic resonance imaging

A ratio of reversible electroporation and irreversible electroporation may be controlled by selecting a symmetric waveform or asymmetric waveform to either minimize or enhance irreversible effects on cells in the target tissue. Combined reversible and irreversible electroporation includes inserting one or more therapeutic electrodes into a target tissue, introducing an electroporation compound into the target tissue, selecting a pulse waveform that is either 1) asymmetric bipolar that has positive and negative pulses with different durations, or 2) symmetric bipolar that has positive and negative pulses with the same duration, and delivering to the target tissue a series of electrical pulses having the selected pulse waveform.

Methods for enhancing and modulating reversible and irreversible electroporation lesions by manipulating pulse waveforms

A radiation therapy treatment method includes providing a patient model, dosimetric constraints, delivery motion constraints, and delivery coordinate space of a radiation delivery device, where the delivery coordinate space is represented as a mesh with vertices connected by edges, where the vertices correspond to directions of a beam eye view (BEV) of the radiation delivery device. BEV region connectivity manifolds are constructed from the patient model, the dosimetric constraints, the delivery coordinate space, and existing beam trajectories, wherein each of the BEV region connectivity manifolds represents connections between contiguous 2D target regions. Beam trajectories are selected based on region connectedness information in the BEV region connectivity manifolds, the dosimetric constraints, the delivery motion constraints, and the existing beam trajectories. Radiation is delivered using the radiation delivery device in accordance with the selected beam trajectories.

Trajectory optimization in radiotherapy using sectioning

A radiation therapy treatment method includes providing a patient model, dosimetric constraints, delivery motion constraints, and delivery coordinate space of a radiation delivery device, where the delivery coordinate space is represented as a mesh with vertices connected by edges, where the vertices correspond to directions of a beam eye view (BEV) of the radiation delivery device, where each BEV has corresponding area elements resulting from beam collimation. BEV region connectivity manifolds are constructed from the patient model, the dosimetric constraints, the delivery coordinate space, and existing beam trajectories, wherein each of the BEV region connectivity manifolds represents connections between contiguous 2D target regions, where each of the 2D target regions is defined at each of the vertices of the delivery coordinate space. Beam trajectories are selected based on region connectedness information in the BEV region connectivity manifolds, the dosimetric constraints, the delivery motion constraints, and the existing beam trajectories. Radiation is delivered using the radiation delivery device in accordance with the beam trajectories.

A method of radiation therapy planning and delivery is provided that includes introducing a demand metric, using a volumetric modulation arc therapy system, to select a spatially optimized and non-uniform set of station or control points to be used for intensity modulation, where the set of station or control points are uniform or non-uniform, and selecting a set of control points at different gantry angles, using the volumetric modulated arc therapy system, and using the demand metric to prioritize which station or control point receives intensity modulation, where the volumetric modulated arc therapy system is differentially boosted at selected angles by inserting additional segments to the station, where the inserted additional apertures dosimetrically boost desired regions in a planning target volume to improve planning target volume coverage in a single arc rotation while sparing sensitive structures.

Station parameter optimized radiation therapy (SPORT): a novel scheme for treatment planning and delivery in radiation therapy

A method of real-time radiotherapy beam visualization is provided that includes disposing a free-form flexible scintillating sheet on a subject of interest, irradiating the subject of interest with a source of ionizing radiation, where the free-forming flexible scintillating sheet emits light when irradiated by the therapeutic photon beam, collecting the emitted light and collecting ambient light reflected from the subject of interest and surrounding objects using a camera, where the collected light is converted to image data by the camera, where the image data is communicated to an appropriately programmed computer, and processing the image data to determine beam characteristics and the characteristics of the subject of interest, using the appropriately programmed computer, where the beam characteristics and the characteristics of the subject of interest are displayed in real-time to a machine operator enabling real-time verification of treatment delivery.

Visualizing radiation therapy beam in real-time in the context of patient's anatomy

A method of electronic portal imaging (EPID) for radiation dosimetric measurement and quality assurance (QA) in radiation therapy is provided that includes imaging a calibration photon beam and a treatment beam using an EPID imager disposed between a plastic water build-up phantom and a plastic water back-scatter phantom that is thicker than the plastic water build-up phantom, correcting the EPID images using an off-axis correction map that takes into account a horn-shape of a photon fluence profile, deconvolving the corrected EPID images using a Monte Carlo simulated EPID response kernel, the deconvolved corrected EPID images provide a primary photon fluence map, reconstructing a relative dose map using the primary photon fluence map convolving with a Monte Carlo simulated pencil-beam dose distribution kernel, and generating an absolute dose map using cross calibration to a 10&#xd7;10 cm2 square field that is compared with a planned dose distribution for QA analysis.

EPID dosimetry method and system for radiation therapy

Molecular imaging of radioluminescent nanoparticle probes injected into biological tissue is performed by irradiated the tissue with ionizing radiation to induce radioluminescence at optical wavelengths, preferably at predetermined near infrared wavelengths. The optical light is detected and processed to determine a spatial distribution of the probes. The radioluminescent nanoparticles may be inorganic or organic phosphors, scintillators, or quantum dots. Imaging systems realizing this technique include tomographic systems using an x-ray beam to sequentially irradiate selected regions, systems with a radioactive source producing the ionizing radiation from outside the tissue, such as with a beam, or inside the tissue, such as with an endoscope or injected radiopharmaceutical. The optical signals may be detected by a photodetector array external to the tissue, a photodetector integrated with an endoscope or mammographic paddle, integrated into a capsule endoscope, or an array positioned near the biological tissue.

Molecular imaging using radioluminescent nanoparticles

Growing community of inventors

Palo Alto, CA, United States of America

Lei Xing