Multi-instance Deep Learning of Ultrasound Imaging Data for Pattern Classification of Congenital Abnormalities of the Kidney and Urinary Tract in Children
Related Experiment Video
Updated: Dec 20, 2025
Author Spotlight: Developing a Bedside Protocol for Kidney and Genitourinary Ultrasonography
Published on: June 21, 2024
Shi Yin1, Qinmu Peng2, Hongming Li3
1School of Electronic Information and Communications, Huazhong University of Science and Technology, Wuhan, China; Department of Radiology, Perelman School of Medicine, University of Pennsylvania, Philadelphia, PA.
Insights
A new deep learning method accurately diagnoses posterior urethral valves (PUV) in children using ultrasound images. This automated approach improves diagnostic speed and reliability for PUV detection.
Area of Science:
- Pediatric Urology
- Medical Imaging Analysis
- Artificial Intelligence in Medicine
Background:
- Posterior urethral valves (PUV) are a common cause of bladder outlet obstruction in male infants.
- Accurate and timely diagnosis of PUV is crucial to prevent long-term renal damage.
- Current diagnostic methods rely on expert interpretation of ultrasound images, which can be time-consuming and subjective.
Purpose of the Study:
- To develop and validate a multi-instance deep learning (MIDL) method for automated diagnosis of PUV.
- To compare the performance of the MIDL method with traditional single-image deep learning classifiers.
- To assess the ability of the MIDL method to identify relevant anatomical features in ultrasound images.
Main Methods:
- A dataset of 6062 ultrasound images (3504 sagittal, 2558 transverse) from 86 children with PUV and 71 controls with mild unilateral hydronephrosis was used.
- A multi-instance deep learning classifier was trained and evaluated using both sagittal and transverse views.
- Classification activation mapping was employed to visualize discriminative regions identified by the deep learning models.
Main Results:
- The MIDL classifier achieved superior performance compared to single-image classifiers, with an area under the receiver operating characteristic curve (AUC) of 0.961 ± 0.026 when using both sagittal and transverse views.
- The MIDL method demonstrated high classification accuracy (0.925 ± 0.060), specificity (0.986 ± 0.032), and sensitivity (0.873 ± 0.120).
- Classification activation mapping confirmed that the deep learning model identified clinically relevant anatomical features for PUV diagnosis.
Conclusions:
- The developed MIDL method offers an automatic, accurate, and reliable tool for diagnosing PUV in children.
- This AI-driven approach can significantly aid clinicians in distinguishing infants with PUV from those with unilateral hydronephrosis.
- The MIDL technique enhances the efficiency and precision of ultrasound image analysis for pediatric urological conditions.
Objective:
To reliably and quickly diagnose children with posterior urethral valves (PUV), we developed a multi-instance deep learning method to automate image analysis.
Methods:
We built a robust pattern classifier to distinguish 86 children with PUV from 71 children with mild unilateral hydronephrosis based on ultrasound images (3504 in sagittal view and 2558 in transverse view) obtained during routine clinical care.
Results:
The multi-instance deep learning classifier performed better than classifiers built on either single sagittal images or single transverse images. Particularly, the deep learning classifiers built on single images in the sagittal view and single images in the transverse view obtained area under the receiver operating characteristic curve (AUC) values of 0.796 ± 0.064 and 0.815 ± 0.071, respectively. AUC values of the multi-instance deep learning classifiers built on images in the sagittal and transverse views with mean pooling operation were 0.949 ± 0.035 and 0.954 ± 0.033, respectively. The multi-instance deep learning classifiers built on images in both the sagittal and transverse views with a mean pooling operation obtained an AUC of 0.961 ± 0.026 with a classification rate of 0.925 ± 0.060, specificity of 0.986 ± 0.032, and sensitivity of 0.873 ± 0.120, respectively. Discriminative regions of the kidney located using classification activation mapping demonstrated that the deep learning techniques could identify meaningful anatomical features from ultrasound images.
Conclusion:
The multi-instance deep learning method provides an automatic and accurate means to extract informative features from ultrasound images and discriminate infants with PUV from male children with unilateral hydronephrosis.
Related Concept Videos
01:28Imaging Studies I: Kidney, Ureter, and Bladder Studies
01:24Imaging Studies II: Ultrasonography
01:20Ultrasound I: Abdominal Ultrasonography
Abdominal ultrasonography, commonly known as abdominal ultrasound, is a vital, non-invasive medical imaging technique widely used in healthcare.
Procedure:
This diagnostic tool allows the clinician to visually inspect internal structures within the abdomen, including vital organs such as the liver, gallbladder, pancreas, kidneys, and spleen.
The abdominal ultrasound process begins with applying a special gel to the patient's skin over the abdomen. This gel enhances the...
01:17Ultrasonography
During an ultrasonography procedure, a handheld device called...
01:27Imaging Studies IV: Magnetic Resonance Imaging
01:22Imaging Studies V: Intravenous Urography and Retrograde Pyelography