Automatic Segmentation Ultrasound-based Radiomics Technology in Diabetic Kidney Disease

Noninvasive Detection of Diabetic Kidney Disease Based on Automatic Segmentation Ultrasound-based Radiomics Technology

Diabetic kidney disease is a common complication of diabetes and the main cause of end-stage renal disease. In this study, the investigator plan to enroll nearly 500 participant with/without DKD and to develop an automatic segmentation ultrasound based radiomics technology to differentiating participant with a non-invasive and an available way.

Study Overview

• Status: Completed
• Conditions: Diabetic Kidney Disease
• Intervention / Treatment: Diagnostic test: ultrasonic imaging

Detailed Description

Ultrasound examination is a convenient, cheap and non-invasive method for kidney examination. However, the ability of conventional ultrasound to distinguish diabetic kidney disease from normal kidney is limited, and it is difficult to accurately distinguish between diabetic kidney disease and normal kidney only with the naked eye. In recent years, computer science has developed rapidly and artificial intelligence has been developing continuously. Much progress has been made in applying artificial intelligence in data analysis. Machine learning is a direction of generalized artificial intelligence, its main characteristic is to make the machine autonomous prediction and create algorithm, so as to achieve autonomous learning. kidney disease and deep learning are two different approaches in the field of machine learning. In this study, image omics and deep learning were used to analyze the images. Image omics extracts traditional image features, including shape, gray scale, texture, etc., and uses machine learning (pattern recognition) models to classify and predict, such as support vector machine, random forest, XGBoost, etc. Deep learning directly uses the convolutional network CNN to extract features, and completes classification and prediction in combination with the full connection layer, etc.

This study aims to explore the detection of diabetic kidney disease and its pathological degree based on automatic segmentation ultraound-based radiomics technology, mining of internal information of ultrasound images, and form a set of non-invasive monitoring of diabetic kidney disease complications development system, especially in primary medical institutions, has a broad clinical application prospect.

• Study Type: Observational
• Enrollment (Actual): 499

Contacts and Locations

Study Locations

• China
  • Anhui
    • Fuyang, Anhui, China, 236200
      • The People's Hospital of Yingshang
  • Tianjin
    • Tianjin, Tianjin, China, 300000
      • Tianjin Third Central Hospital
  • Zhejiang
    • Hangzhou, Zhejiang, China, 310000
      • Department of Ultrasound, Second Affiliated Hospital, School of Medicine, Zhejiang University

Participation Criteria

Eligibility Criteria

• Ages Eligible for Study: 18 years to 80 years (Adult, Older Adult)
• Accepts Healthy Volunteers: No
• Genders Eligible for Study: All

• Sampling Method: Non-Probability Sample

Study Population

1. patients with diabetes
2. patients with or without DKD

Description

Inclusion Criteria:

• patients with clinical diagnosis of T2DM and DKD were enrolled.
• patients with clear B mode ultrasound imaging in both side of kidney (left and right).
• No missing value in the vital clinical data such as eGFR and UACR.

Exclusion Criteria:

• Patients with large kidney space occupying disease such as kidney renal cyst and tumor were excluded.
• Ultrasound images with severe shadow or incomplete kidney border were excluded.

Study Plan

How is the study designed?

Number of groups / cohorts

2

Cohorts and Interventions

Group / Cohort	Intervention / Treatment
Experimental group; Experimental group1：DKD patients with Type 2 diabetes patients with DKD Experimental group2：High level DKD patients with diabetic kidney disease Stage III and IV.	Diagnostic test: ultrasonic imaging; Two-dimensional ultrasound images of the patient's kidneys were obtained by ultrasound imaging.
Control group; Control1：T2DM patients with Type 2 diabetes Control2：Low level DKD patients with diabetic kidney disease Stage I and II.	Diagnostic test: ultrasonic imaging; Two-dimensional ultrasound images of the patient's kidneys were obtained by ultrasound imaging.

What is the study measuring?

Primary Outcome Measures

Outcome Measure	Measure Description	Time Frame
AUC	The area under curve (AUC) of radiomics model for differentiating DKD and T2DM or high level and low level DKD patients	6 months

Secondary Outcome Measures

Outcome Measure	Measure Description	Time Frame
Miou	The mean intersection over union (Miou) of DL-based auto-segmentation in different medical centers	6 months
mPA	The mean pixel accuracy (mPA) of DL-based auto-segmentation in different medical centers	6 months

Collaborators and Investigators

• Sponsor: Second Affiliated Hospital, School of Medicine, Zhejiang University
• Investigators: Study Chair: Pintong Huang, Department of Ultrasound, The Second Affiliated Hospital of Zhejiang University

Study record dates

Study Major Dates

• Study Start (Actual): June 1, 2021
• Primary Completion (Actual): December 1, 2021
• Study Completion (Actual): December 1, 2021

Study Registration Dates

• First Submitted: August 24, 2021
• First Submitted That Met QC Criteria: August 24, 2021
• First Posted (Actual): August 27, 2021

Study Record Updates

• Last Update Posted (Actual): February 16, 2022
• Last Update Submitted That Met QC Criteria: February 15, 2022
• Last Verified: August 1, 2021

More Information

Terms related to this study

• Keywords: ultrasound; multi-center; diabetic kidney disease; deep learning; radiomics
• Additional Relevant MeSH Terms: Urologic Diseases; Endocrine System Diseases; Diabetes Complications; Diabetes Mellitus; Kidney Diseases; Diabetic Nephropathies
• Other Study ID Numbers: 2021-0465

Drug and device information, study documents

• Studies a U.S. FDA-regulated drug product: No
• Studies a U.S. FDA-regulated device product: No