# US 2021/0264574 A1 · Correcting image blur in medical image

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 2021/0264574 A1
- Document ID: US20210264574A1
- Status: Published application
- Kind: A1
- Application number: 17/318,363
- Filed: 2021-05-12
- Published: 2021-08-26
- Earliest claimed: 2018-11-14
- Assignees: Rutgers, The State University of New Jersey
- Applicants: Rutgers, The State University of New Jersey
- Inventors: Christine I. Podilchuk; Richard Mammone
- Web page: https://patentgrep.com/patent/US20210264574A1

## CPC

- G06T 5/73: Deblurring; Sharpening
- G06T 5/50: using two or more images, e.g. averaging or subtraction
- G06T 5/60: using machine learning, e.g. neural networks
- G06T 5/70: Denoising; Smoothing
- G06T 7/0012: {Biomedical image inspection}
- G06T 7/13: Edge detection
- G06T 2207/10004: Still image; Photographic image
- G06T 2207/10132: Ultrasound image
- G06T 2207/20081: Training; Learning
- G06T 2207/20084: Artificial neural networks [ANN]
- G06T 2207/20104: Interactive definition of region of interest [ROI]
- G06T 2207/20192: Edge enhancement; Edge preservation
- G06T 2207/20216: Image averaging
- G06T 2207/30004: Biomedical image processing
- G06T 2207/30168: Image quality inspection
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL

## Abstract

A device to correct an image blur within a medical image is described. An image analysis application executed by the device receives the medical image from a medical image provider. Next, the image blur is detected within the medical image by analyzing the medical image. The medical image is subsequently processed with a deep learning model to correct the image blur. In response to the processing, a de-blurred medical image is generated. The de-blurred medical image is provided for a presentation or a continued analysis.

## Claims

### Claim 1

A device to train a deep learning model to correct an image blur in a medical image, wherein the device is configured to:
receive a training input set of a plurality of training input medical images and a plurality of expected output medical images associated with the plurality of training input medical images;
process the plurality of training input medical images with a deep learning model to correct image blur;
generate a plurality of de-blurred training input medical images; and
train the deep learning model based at least in part on an analysis of the plurality of de-blurred training input medical images and the plurality of expected output medical images.

### Claim 2 (depends on 1)

The device of claim 1, wherein each training input medical image includes a medical ultrasound image.

### Claim 3 (depends on 1)

The device of claim 1, wherein each training input medical image includes a three dimensional image.

### Claim 4 (depends on 1)

The device of claim 1, wherein processing the training input medical images further includes a process to:
evaluate a metadata of the medical image;
identify an annotation associated with the medical image within the metadata, wherein the annotation designates an averaging process used to generate the medical image from a plurality of scanned images of a scanning session of a biological structure of a patient.

### Claim 5 (depends on 1)

The device of claim 1, wherein processing the training input medical images further includes a process to:
receive a selection of a region of interest (ROI) of the medical image from a user; and
analyze the ROI to identify the image blur within the ROI.

### Claim 6 (depends on 1)

The device of claim 1, wherein each training input medical image is processed with the deep learning model in a real-time or offline.

### Claim 7 (depends on 1)

The device of claim 1, wherein each training input medical image and a subsequent image of a time sequence based scanning session of a biological structure of a patient are processed with the deep learning model in a real-time or offline.

### Claim 8 (depends on 1)

The device of claim 1, wherein the training input set of the deep learning model includes averaged images.

### Claim 9 (depends on 8)

The device of claim 8, wherein each of the training input medical images includes a noise reduced average of medical scan images captured during an imaging session.

### Claim 10 (depends on 9)

The device of claim 9, wherein one or more edges of an object of interest (OI) within each of the training input medical images is blurred as a result of the noise reduced average of the medical scan images.

### Claim 11 (depends on 8)

The device of claim 8, wherein the expected output set of the deep learning model includes de-blurred images corresponding to the averaged images.

### Claim 12 (depends on 11)

The device of claim 11, wherein one or more edges of an object of interest (OI) within each of the de-blurred images are sharpened.

### Claim 13 (depends on 1)

The device of claim 1, wherein the image provider includes a medical imaging device.

### Claim 14 (depends on 13)

The device of claim 13, wherein the medical imaging device is configured to:
capture the medical image during a capture session to scan a biological structure of a patient.

### Claim 15 (depends on 1)

The device of claim 1, wherein the image provider includes a camera component.

### Claim 16 (depends on 15)

The device of claim 15, wherein the camera component is configured to:
capture the medical image from a display device associated with a medical imaging device, wherein the display device is configured to display a scanned image of a biological structure of a patient.

### Claim 17

A mobile device for training a deep learning model to correcting an image blur in a medical ultrasound image, the mobile device comprising:
a memory configured to store instructions associated with an image analysis application, a processor coupled to the display component, the camera component, and the memory, the processor executing the instructions associated with the image analysis application, wherein the analysis application includes:
a neural network module configured to:
receive a training input set of a plurality of training input medical images and a plurality of expected output medical images associated with the plurality of training input medical images;
process the plurality of training input medical images with a deep learning model to correct image blur;
generate a plurality of de-blurred training input medical images; and
train the deep learning model based at least in part on an analysis of the plurality of de-blurred training input medical images and the plurality of expected output medical images.

### Claim 18 (depends on 17)

The mobile device of claim 17, wherein processing the training input medical images includes one or more operations to:
identify one or more edges of an object of interest ((ill) within the medical ultrasound image, wherein the one or more edges are blurred by the noise reduced average of the ultrasound session images; and
sharpen the one or more edges of the OI based on the deep learning model.

### Claim 19

A method of correcting an image blur in a medical ultrasound image, the method comprising:
receiving, by a processor, a training input set of a plurality of training input medical images and a plurality of expected output medical images associated with the plurality of training input medical images;
processing, by a processor, the plurality of training input medical images with a deep learning model to correct image blur;
generating, by a processor, a plurality of de-blurred training input medical images; and
training, by a processor, the deep learning model based at least in part on an analysis of the plurality of de-blurred training input medical images and the plurality of expected output medical images.
