Mobile ultrawideband radar monitoring thoracic fluid levels and cardio-respiratory function
A mobile bodily imaging system for determination of a lung fluid content is provided, comprising: a stepped-frequency radar generator that generates one or more tone and transmits the one or more tone into a chest of the patient; a sparse deconvolution inversion algorithm to detect reflection points indicating tissue interfaces within the patient and create a sparsity-based image; an initial model generated using Markov Chain Monte Carlo or Deep Neural Network estimates; and a final model generated using a neural renderer, wherein the neural renderer is provided with confirmed data that permits the neural renderer to predict the form of the reflection points, wherein the neural renderer creates an image that is compared to the sparsity based image, and wherein the neural renderer creates refined images in a number of iterations and/or based upon a predetermined threshold of a difference between the refined neural renderer image and the sparsity-based image.
1 . A mobile bodily imaging system for determination of a lung fluid content in a lung of a patient, comprising:
an ultrawideband pulse generator that generates one or more ultrawideband pulse and transmits the one or more ultrawideband pulse into a chest of the patient;
an ultrawideband RF sensor positioned on the patient's chest adjacent to the patient's skin to receive reflected backscatter waves from the patient's chest;
a sparse deconvolution inversion algorithm used to detect reflection points indicating tissue interfaces within the patient's body and create a sparsity-based image;
an initial model generated using one or more of Markov Chain Monte Carlo or Deep Neural Network estimates; and
a final model generated using a neural renderer,
wherein the neural renderer is previously provided with confirmed data that permits the neural renderer to predict a form of the sparse deconvolution reflection points given previously established model parameters,
wherein the neural renderer creates an image that is compared to the sparsity-based image, and
wherein the neural renderer creates refined images in one or more of a predetermined number of iterations and based upon a predetermined threshold of a difference between the refined neural renderer image and the sparsity-based image.
2 . The mobile bodily imaging system of claim 1 , wherein the final model is a graph with an indication of the tissue interfaces within the patient's body.
3 . The mobile bodily imaging system of claim 2 , wherein the tissue interfaces include one or more of an interface between a patient's skin, fat, muscle, bone, and lung tissues.
4 . The mobile bodily imaging system of claim 1 , wherein the initial model includes multiple iterations of adjusting a forward model of multilayer propagation based upon the one or more of the Markov Chain Monte Carlo or Deep Neural Network estimates.
5 . The mobile bodily imaging system of claim 4 , wherein the forward model of multilayer propagation is compared to the sparsity-based image.
6 . The mobile bodily imaging system of claim 1 , wherein the neural renderer accounts for effects in the final model such as a non-ideal antenna response, a spherical wave propagation, and a non-homogenous tissue media.
7 . The mobile bodily imaging system of claim 1 , wherein the sparsity-based image and neural renderer image are subjected to encoding feature extraction.
8 . The mobile bodily imaging system of claim 7 , wherein the encoding feature extraction is achieved using one or more lung fluid detection networks.
9 . The mobile bodily imaging system of claim 8 , wherein the one or more lung fluid detection networks produce lung fluid indicators.
10 . The mobile bodily imaging system of claim 1 , wherein the final model indicates thicknesses of tissues of the patient.
11 . The mobile bodily imaging system of claim 1 , wherein the final model indicates a percentage of lung fluid content in the lung of the patient.
12 . A method for determining a lung fluid content in a lung of a patient, comprising:
providing an ultrawideband pulse generator that generates one or more ultrawideband pulse and transmits the one or more ultrawideband pulse into a chest of the patient;
positioning an ultrawideband RF sensor on the patient's chest adjacent to the patient's skin to receive reflected backscatter waves from the patient's chest;
using a sparse deconvolution inversion algorithm to detect reflection points indicating tissue interfaces within the patient's body and creating a sparsity-based image;
generating an initial model using one or more of Markov Chain Monte Carlo or Deep Neural Network estimates; and
generating a final model using a neural renderer,
previously providing the neural renderer with confirmed data that permits the neural renderer to predict a form of the sparse deconvolution reflection points given previously established model parameters,
creating an image using the neural renderer, wherein the image is compared to the sparsity-based image, and
creating refined images using the neural renderer in one or more of a predetermined number of iterations and based upon a predetermined threshold of a difference between the refined neural renderer image and the sparsity-based image.
13 . A mobile bodily imaging system for determination of a lung fluid content in a lung of a patient, comprising:
a stepped-frequency radar generator that generates one or more tone and transmits the one or more tone into a chest of the patient;
an ultrawideband RF sensor positioned on the patient's chest adjacent to the patient's skin to receive reflected backscatter waves from the patient's chest;
a sparse deconvolution inversion algorithm used to detect reflection points indicating tissue interfaces within the patient's body and create a sparsity-based image;
an initial model generated using one or more of Markov Chain Monte Carlo or Deep Neural Network estimates; and
a final model generated using a neural renderer,
wherein the neural renderer is previously provided with confirmed data that permits the neural renderer to predict a form of the sparse deconvolution reflection points given previously established model parameters,
wherein the neural renderer creates an image that is compared to the sparsity-based image, and
wherein the neural renderer creates refined images in one or more of a predetermined number of iterations and based upon a predetermined threshold of a difference between the refined neural renderer image and the sparsity-based image.
14 . A method for determining a lung fluid content in a lung of a patient, comprising:
providing a stepped-frequency radar generator that generates one or more tone and transmits the one or more tone into a chest of the patient;
positioning an ultrawideband RF sensor on the patient's chest adjacent to the patient's skin to receive reflected backscatter waves from the patient's chest;
using a sparse deconvolution inversion algorithm to detect reflection points indicating tissue interfaces within the patient's body and creating a sparsity-based image;
generating an initial model using one or more of Markov Chain Monte Carlo or Deep Neural Network estimates; and
generating a final model using a neural renderer,
previously providing the neural renderer with confirmed data that permits the neural renderer to predict a form of the sparse deconvolution reflection points given previously established model parameters,
creating an image using the neural renderer, wherein the image is compared to the sparsity-based image, and
creating refined images using the neural renderer in one or more of a predetermined number of iterations and based upon a predetermined threshold of a difference between the refined neural renderer image and the sparsity-based image.