Systems and methods for automated clinical image quality assessment

Inventors: Roxana Daneshjou (Stanford, CA); Kailas Vodrahalli (Stanford, CA); James Zou (Stanford, CA); Justin Ko (Stanford, CA); Roberto A. Novoa (San Francisco, CA); Albert Chiou (Stanford, CA)

Assignee: The Board of Trustees of the Leland Stanford Junior University

G06T7/0002 G06V10/42 G06V10/764 G06T2207/20081 G06T2207/30168

View Patent ↗

Loading inventors, assignments & file history…

Monitor This Case

Get email alerts when status or documents change.

Order Certified Copies

Most orders are placed with the USPTO same day — all within 24 business hours.

Order via The Patent Place →

Pre-filled with this patent's details

Quick Facts

Patent No.: US 12682438
App. No.: 17/937,714
Granted: Jul 14, 2026
Kind: B2

External Links

USPTO Patent Center↗ Full Patent PDF↗

Abstract

Embodiments herein describe systems and methods for automated clinical image quality assessment and uses thereof are provided. Many embodiments provide instructions and feedback for clinically relevant image quality. Several embodiments provide methods to generate a model to assess image quality.

Claims (37)

1 . A machine learning model for assessing image quality, comprising:

a deep learning model comprising a linear classifier configured to identify image quality of an image;

a classical vision model configured to identify a feature quality in the image; and

a logistic classifier configured to provide an image quality as an output.

2 . The machine learning model of claim 1 , wherein the deep learning model is a plurality of deep learning models, wherein one deep learning model provides an overall quality of the image and another deep learning model provides a binary classifier for a feature of the image.

3 . The machine learning model of claim 2 , wherein the feature is selected from blur, lighting, and zoom/crop.

4 . The machine learning model of claim 1 , wherein the deep learning model is four deep learning models, wherein the first deep learning model provides an overall quality of the image and the second deep learning model, the third deep learning model, and the fourth deep learning models provide a binary classifier for a feature of the image.

5 . The machine learning model of claim 4 , wherein the feature is selected from blur, lighting, and zoom/crop.

6 . The machine learning model of claim 4 , wherein the first deep learning model is a gateway classifier.

7 . The machine learning model of claim 1 , wherein the classical vision model is a plurality of classical vision models.

8 . The machine learning model of claim 7 , wherein the plurality of classical vision models are input selected features.

9 . The machine learning model of claim 8 , wherein the features are selected by one or more of: local binary pattern, Fourier blur, Laplacian blur, Lighting, skin distribution, image cropping, and color space transform.

10 . The machine learning model of claim 1 , wherein the logistic classifier is further configured to provide an explanation for a poor quality image.

11 . The machine learning model of claim 1 :

wherein the deep learning model is four deep learning models,

wherein the first deep learning model provides an overall quality of the image,

wherein the second deep learning model, the third deep learning model, and the fourth deep learning models provide a binary classifier for a feature of the image, wherein the feature is selected from blur, lighting, and zoom/crop, and

wherein the first deep learning model is a gateway classifier;

wherein the classical vision model is a plurality of classical vision models input selected features, wherein the features are selected by one or more of: local binary pattern, Fourier blur, Laplacian blur, Lighting, skin distribution, image cropping, and color space transform; and

wherein the logistic classifier is configured to provide an explanation for a poor quality image.

12 . A method for improving image quality for clinical use, comprising:

obtaining an image of an area of concern on an individual;

providing the image to a machine learning model trained to identify image quality, where the machine learning model comprises:

a deep learning model comprising a linear classifier configured to identify image quality of the image;

a classical vision model configured to identify a feature quality in the image; and

a logistic classifier configured to provide an image quality as an output; and

receiving an image quality rating for the image from the machine learning model.

13 . The method of claim 12 , wherein the area of concern is a skin lesion.

14 . The method of claim 12 , further comprising:

obtaining a second image of the area of concern based on the image quality rating from the machine learning model;

providing the second image to the machine learning model; and

receiving an image quality rating for the second image from the machine learning model.

15 . The machine learning model of claim 12 , wherein the deep learning model is a plurality of deep learning models, wherein one deep learning model provides an overall quality of the image and another deep learning model provides a binary classifier for a feature of the image.

16 . The machine learning model of claim 15 , wherein the feature is selected from blur, lighting, and zoom/crop.

17 . The machine learning model of claim 12 , wherein the deep learning model is four deep learning models, wherein the first deep learning model provides an overall quality of the image and the second deep learning model, the third deep learning model, and the fourth deep learning models provide a binary classifier for a feature of the image.

18 . The machine learning model of claim 17 , wherein the feature is selected from blur, lighting, and zoom/crop.

19 . The machine learning model of claim 12 , wherein the logistic classifier is further configured to provide an explanation for a poor quality image.