System and method for complex input data configurations for imaging applications

Patent metadata

Field	Value
Publication number	US-2023342996-A1
Application number	US-202318305697-A
Country	US
Kind code	A1
Filing date	Apr 24, 2023
Priority date	Apr 22, 2022
Publication date	Oct 26, 2023
Grant date	—

Abstract

Systems, methods, and media for complex input data configurations for imaging applications. Complex data optimization can be provided to improve accuracy of models (e.g., neural networks) used to reconstruct medical images from raw sensor data, for example. Complex data optimization can include applying raw sensor data to an input layer of a neural network to generate an input vector ordered such that real components and imaginary components of samples in the raw sensor data are adjacent. The input vector can then be applied to convolutional layer of the neural network.

First claim

1 . A method for medical imaging, comprising:
receiving raw sensor data acquired from a patient using a medical imaging modality; applying the raw sensor data to an input layer of a neural network to generate an input vector, wherein the input layer of the neural network orders the input vector such that a real component and an imaginary component of each sample in the raw sensor data are adjacent to each other; applying the input vector to a first convolutional layer of the neural network to generate a filtered input vector; applying the filtered input vector to at least one fully connected layer of the neural network to generate a matrix; applying the matrix to at least one additional convolutional layer of the neural network different from the first convolutional layer to generate a medical image of the patient; and displaying the medical image of the patient.

2 . The method of claim 1 , wherein the first convolutional layer has a kernel of two and a stride of two.

3 . The method of claim 2 , wherein the input vector comprises a one-dimensional vector and the first convolutional layer comprises a one-dimensional convolutional layer.

4 . The method of claim 1 , wherein applying the filtered input vector to the at least one fully connected layer of the neural network to generate the matrix comprises:
applying the filtered input vector to a first hidden layer activated by a first activation function; and applying an output of the first hidden layer to a second hidden layer activated by a second activation function to generate the matrix.

5 . The method of claim 1 , wherein the raw sensor data comprises raw magnetic resonance imaging (MM) k-space data.

6 . The method of claim 1 , wherein the neural network comprises a data-driven, manifold learning neural network.

7 . The method of claim 1 , wherein applying the matrix to the at least one additional convolutional layer of the neural network different from the first convolutional layer to generate the medical image of the patient comprises:
applying the matrix to a second convolutional layer to filter the matrix in accordance with a first filter; applying an output of the second convolutional layer to a third convolutional layer to filter the output of the second convolutional layer in accordance with a second filter; and applying an output of the third convolutional layer to an output layer to generate the medical image of the patient.

8 . A non-transitory computer-readable storage medium having instructions stored thereon that, when executed by at least one processor, cause the at least one processor to implement operations comprising:
receiving raw sensor data acquired from a patient using a medical imaging system; applying the raw sensor data to an input layer of a neural network to generate an input vector, wherein the input layer of the neural network orders the input vector such that a real component and an imaginary component of each sample in the raw sensor data are adjacent; applying the input vector to a first convolutional layer of the neural network to generate a filtered input vector; applying the filtered input vector to at least one fully connected layer of the neural network to generate a matrix; applying the matrix to at least one additional convolutional layer of the neural network different from the first convolutional layer to generate a medical image of the patient; and displaying the medical image of the patient for clinical analysis.

9 . The computer-readable medium of claim 8 , wherein the first convolutional layer has a kernel of two and a stride of two.

10 . The computer-readable medium of claim 8 , wherein the input vector comprises a one-dimensional vector and the first convolutional layer comprises a one-dimensional convolutional layer.

11 . The computer-readable medium of claim 8 , wherein applying the filtered input vector to the at least one fully connected layer of the neural network to generate the matrix comprises:
applying the filtered input vector to a first hidden layer activated by a first activation function; and applying an output of the first hidden layer to a second hidden layer activated by a second activation function to generate the matrix.

12 . The computer-readable medium of claim 8 , wherein applying the matrix to the at least one additional convolutional layer of the neural network different from the first convolutional layer to generate the medical image of the patient comprises:
applying the matrix to a second convolutional layer to filter the matrix in accordance with a first filter; applying an output of the second convolutional layer to a third convolutional layer to filter the output of the second convolutional layer in accordance with a second filter; and applying an output of the third convolutional layer to an output layer to generate the medical image of the patient.

13 . The computer-readable medium of claim 8 , wherein the raw sensor data comprises raw magnetic resonance imaging (MM) k-space data.

14 . The computer-readable medium of claim 8 , wherein the neural network comprises a data-driven, manifold learning neural network.

15 . A system comprising:
a display; one or more sensors; one or more processors; and one or more non-transitory computer readable storage media having instructions stored thereon that, when executed by the one or more processors, cause the one or more processors to implement operations comprising:
receiving raw sensor data acquired from a patient from the one or more sensors;
applying the raw sensor data to an input layer of a neural network to generate an input vector, wherein the input layer of the neural network orders the input vector such that real components and imaginary components of samples in the raw sensor data are adjacent;
applying the input vector to a first convolutional layer of the neural network to generate a filtered input vector;
applying the filtered input vector to at least one fully connected layer of the neural network to generate a matrix;
applying the matrix to at least one additional convolutional layer of the neural network different from the first convolutional layer to generate a medical image of the patient; and
causing the display to display the medial image of the patient for clinical analysis.

16 . The system of claim 15 , wherein:
the first convolutional layer has a kernel of two and a stride of two; the input vector comprises a one-dimensional vector; and the first convolutional layer comprises a one-dimensional convolutional layer.

17 . The system of claim 15 , wherein applying the filtered input vector to the at least one fully connected layer of the neural network to generate the matrix comprises:
applying the filtered input vector to a first hidden layer activated by a first activation function; and applying an output of the first hidden layer to a second hidden layer activated by a second activation function to generate the matrix.

18 . The system of claim 17 , wherein applying the matrix to the at least one additional convolutional layer of the neural network different from the first convolutional layer to generate the medical image of the patient comprises:
applying the matrix to a second convolutional layer to filter the matrix in accordance with a first filter; applying an output of the second convolutional layer to a third convolutional layer to filter the output of the second convolutional layer in accordance with a second filter; and applying an output of the third convolutional layer to an output

Assignees

• Massachusetts Gen Hospital
• Univ Boston

Inventors

• Bhutto Danyal Fareed
• Rosen Matthew S
• Koonjoo Neha
• Zhu Bo

Classifications

• G06T12/20 Primary

  Inverse problem, i.e. transformations from projection space into object space · CPC title

• G06T11/006 Primary

  Physics · mapped topic

• G16H30/40

  for processing medical images, e.g. editing · CPC title

• A61B5/055

  involving electronic [EMR] or nuclear [NMR] magnetic resonance, e.g. magnetic resonance imaging · CPC title

• G06T2211/421

  Filtered back projection [FBP] · CPC title