JMIR Publications

ABSTRACT

Background:

In recent years, the progress and generalization surrounding portable ultrasonic probes has made ultrasound a useful tool for physicians when making a diagnosis. With the advent of machine learning and deep learning, the development of a computer-aided diagnostic system for screening renal ultrasound abnormalities can assist general practitioners in early detection of pediatric kidney diseases.

Objective:

We therefore sought to evaluate the diagnostic performance of deep learning techniques to classify kidney images as normal and abnormal.

Methods:

We chose 330 normal and 1,269 abnormal pediatric renal ultrasound images for establishing a model for artificial intelligence (AI). The abnormal images involved hydronephrosis, cysts, stones, hyper-echogenicity and space occupying lesions. We performed pre-processing of the original images for subsequent machine learning. To classify both diseased and normal kidneys, support vector machine classifiers were built on the extracted features using transfer learning from a pre-trained deep learning model. The classifier and its diagnosis performance were measured using an area under the receiver operating characteristic curve (AUC), accuracy, specificity and sensitivity.

Results:

The deep learning model 94M parameters in size based on ResNet-50 was built for classifying normal and abnormal images. The accuracy of the validated images of normal, hydronephrosis, cyst, stone, hyperechogenicity and space occupying lesions were 90.9 %, 94.17%, 91.6%, 93.17%, 89.92% and 91.3%, respectively.

Conclusions:

We have established the usefulness of a computer-aided model for automatic classification of pediatric renal ultrasound images in terms of normal and abnormal categories.