US-12622620-B2 - Pediatric hydronephrosis severity from ultrasound images

Abstract

A method for predicting a severity and appearance of renal obstruction, including receiving, via processing circuitry, a plurality of medical images corresponding to a kidney of a patient, identifying, the kidney in the plurality of medical images, selecting, at least one relevant image from the plurality of medical images based on a relationship to renal function, standardizing the at least one relevant image, determining, via a deep learning model at least one risk score based on the at least one relevant image, and determining, a final ultrasound-based risk score based on the at least one risk score, wherein the final ultrasound-based risk score is a determination of renal obstruction and/or a probability of renal obstruction. Clinical and/or demographic patient information can be used along with the final ultrasound-based risk score to determine a final risk score to predict the severity and appearance of renal obstruction.

Inventors

• Marius George LINGURARU
• Pooneh ROSHANITABRIZI
• Hans G. Pohl

Assignees

• CHILDREN'S NATIONAL MEDICAL CENTER

Dates

Publication Date

20260512

Application Date

20220411

Claims (20)

1. 1 . A method for predicting a severity and appearance of renal obstruction, the method comprising: receiving, via processing circuitry, a plurality of medical images corresponding to a kidney of a patient; identifying, via the processing circuitry, the kidney in the plurality of medical images; selecting, via the processing circuitry, at least one relevant image from the plurality of medical images, the at least one relevant image being selected based on a relationship to renal function; standardizing, via the processing circuitry, the at least one relevant image; determining, via a deep learning model executed by the processing circuitry, at least one risk score based on the at least one relevant image; and determining, via the processing circuitry, a final ultrasound-based risk score based on the at least one risk score, wherein the final ultrasound-based risk score is an indication of a presence of renal obstruction and/or a probability of renal obstruction.
2. 2 . The method of claim 1 , wherein identifying the kidney in the plurality of medical images includes segmenting the kidney or localizing a boundary of the kidney.
3. 3 . The method of claim 1 , wherein standardizing the at least one relevant image includes standardizing a field of view, an intensity, and/or a size of the at least one relevant image.
4. 4 . The method of claim 3 , wherein standardizing the field of view includes standardizing an orientation of the kidney in the at least one relevant image along a longest axis of the kidney in a coronal view.
5. 5 . The method of claim 1 , wherein the at least one relevant image includes a middle slice image.
6. 6 . The method of claim 5 , wherein the at least one relevant image further includes at least one adjacent slice image, and wherein the at least one adjacent slice image is selected based on a correlation between the at least one adjacent slice image and the middle slice image.
7. 7 . The method of claim 1 , wherein the final ultrasound-based risk score is a maximum, a product, or a median of the at least one risk score, or wherein the final ultrasound-based risk score is determined via at least one neural network.
8. 8 . The method of claim 1 , wherein the final ultrasound-based risk score is a weighted fusion of the at least one risk score and wherein a weight of the at least one risk score is based on a difference between the at least one relevant image and a middle slice image, wherein the difference is based on correlation analysis or a deep learning-based approach.
9. 9 . The method of claim 1 , further comprising determining, via the processing circuitry, a final risk score based on the final ultrasound-based risk score and clinical patient information, wherein the final risk score is a label of a presence of renal obstruction and/or a probability of renal obstruction.
10. 10 . The method of claim 9 , wherein the final risk score is determined using a neural network.
11. 11 . An apparatus for predicting a severity and appearance of renal obstruction, comprising: processing circuitry configured to: receive a plurality of medical images corresponding to a kidney of a patient; identify the kidney in the plurality of medical images; select at least one relevant image from the plurality of medical images, the at least one relevant image being selected based on a relationship to renal function; standardize the at least one relevant image; determine, via a deep learning model, at least one risk score based on the at least one relevant image; and determine a final ultrasound-based risk score based on the at least one risk score, wherein the final ultrasound-based risk score is an indication of a presence of renal obstruction and/or a probability of renal obstruction.
12. 12 . The apparatus of claim 11 , wherein the processing circuitry is further configured to segment the kidney or localize a boundary of the kidney to identify the kidney in the plurality of medical images.
13. 13 . The apparatus of claim 11 , wherein the processing circuitry is configured to standardize a field of view, an intensity, and/or a size of the at least one relevant image.
14. 14 . The apparatus of claim 13 , wherein the processing circuitry is configured to standardize an orientation of the kidney in the at least one relevant image along a longest axis of the kidney in a coronal view to standardize the field of view.
15. 15 . The apparatus of claim 11 , wherein the at least one relevant image includes a middle slice image.
16. 16 . The apparatus of claim 15 , wherein the at least one relevant image includes at least one adjacent slice image and wherein the processing circuitry is configured to select the at least one adjacent slice image based on a correlation between the at least one adjacent slice image and the middle slice image.
17. 17 . The apparatus of claim 11 , wherein the final ultrasound-based risk score is a weighted fusion of the at least one risk score and wherein a weight of the at least one risk score is based on a difference between the at least one relevant image and a middle slice image, wherein the difference is based on correlation analysis or a deep learning-based approach.
18. 18 . The apparatus of claim 11 , wherein the processing circuitry is further configured to determine a final risk score based on the final ultrasound-based risk score and clinical patient information, wherein the final risk score is a label of a presence of renal obstruction and/or a probability of renal obstruction.
19. 19 . The apparatus of claim 18 , wherein the final risk score is determined using a neural network.
20. 20 . A non-transitory computer-readable storage medium for storing computer-readable instructions that, when executed by a computer, cause the computer to perform a method for predicting a severity and appearance of renal obstruction, the method comprising: receiving, via processing circuitry, a plurality of medical images corresponding to a kidney of a patient; identifying, via the processing circuitry, the kidney in the plurality of medical images; selecting, via the processing circuitry, at least one relevant image from the plurality of medical images, the at least one relevant image being selected based on a relationship to renal function; standardizing, via the processing circuitry, the at least one relevant image; determining, via a deep learning model executed by the processing circuitry, at least one risk score based on the at least one relevant image; and determining, via the processing circuitry, a final ultrasound-based risk score based on the at least one risk score, wherein the final ultrasound-based risk score is an indication of a presence of renal obstruction and/or a probability of renal obstruction.

Description

CROSS-REFERENCE TO RELATED APPLICATIONS The present application claims priority to U.S. Provisional Application No. 63/173,755, filed Apr. 12, 2021, which is incorporated herein by reference in its entirety for all purposes. BACKGROUND Field of the Disclosure The present disclosure is generally directed to systems, methods, and apparatuses for predicting kidney function. More specifically, embodiments of the present disclosure involve systems, methods, and apparatuses for assessing renal obstruction, including identification of a presence of renal obstruction and determination of a severity or a likely severity of renal obstruction. Description of the Related Art Hydronephrosis is a dilatation of the renal collecting system resulting from a buildup of fluid that cannot be drained from the kidney. Congenital hydronephrosis is common in children and is typically evaluated using visual assessment of ultrasound images. For example, the Society for Fetal Urology (SFU) has developed a grading system to visually assess hydronephrosis severity using ultrasound images. However, the grading system has been shown to lack apparent correlation with other functional imaging modalities that provide information about renal function. Renal obstruction, such as at ureteropelvic junction, can cause hydronephrosis. Untreated obstruction can result in permanent loss of kidney function. Variability in ultrasound images, especially for pathological kidneys, can result in difficulty determining renal obstruction and dysfunction using ultrasound. In addition, visual assessment of ultrasound images is a subjective process that usually requires expert interaction. As a result, the presence and severity of renal obstruction is typically evaluated using more invasive techniques such as diuresis renography. Diuresis renography is costly and can require sedation in children, while ultrasound imaging is comparatively quicker and used as a routine procedure. The foregoing Background description is for the purpose of generally presenting the context of the disclosure. Work of the inventors, to the extent it is described in this background section, as well as aspects of the description which may not otherwise qualify as prior art at the time of the filing, are neither expressly or impliedly admitted as prior art against the present disclosure. SUMMARY According to one embodiment, the present disclosure relates to a method for predicting a severity and appearance of renal obstruction, including receiving, via processing circuitry, a plurality of medical images corresponding to a kidney of a patient, identifying, via the processing circuitry, the kidney in the plurality of medical images, selecting, via the processing circuitry, at least one relevant image from the plurality of medical images, the at least one relevant image being selected based on a relationship to renal function, standardizing, via the processing circuitry, the at least one relevant image, determining, via a deep learning model executed by the processing circuitry, at least one risk score based on the at least one relevant image, and determining, via the processing circuitry, a final ultrasound-based risk score based on the at least one risk score, wherein the final ultrasound-based risk score is a determination of renal obstruction and/or a probability of renal obstruction. According to another embodiment, the present disclosure relates to an apparatus for predicting a severity and appearance of renal obstruction, including processing circuitry configured to receive a plurality of medical images corresponding to a kidney of a patient, identify the kidney in the plurality of medical images, select at least one relevant image from the plurality of medical images, the at least one relevant image being selected based on a relationship to renal function, standardize the at least one relevant image, determine, via a deep learning model, at least one risk score based on the at least one relevant image, and determine a final ultrasound-based risk score based on the at least one risk score, wherein the final ultrasound-based risk score is a determination of renal obstruction and/or a probability of renal obstruction. According to another embodiment, the present disclosure relates to a non-transitory computer-readable storage medium for storing computer-readable instructions that, when executed by a computer, cause the computer to perform a method for predicting a severity and appearance of renal obstruction, the method including receiving, via processing circuitry, a plurality of medical images corresponding to a kidney of a patient, identifying, via the processing circuitry, the kidney in the plurality of medical images, selecting, via the processing circuitry, at least one relevant image from the plurality of medical images, the at least one relevant image being selected based on a relationship to renal function, standardizing, via the processing circuitry, the at least one releva