Automation of the kidney function prediction and classification through ultrasound-based kidney imaging using deep learning

Abstract

Prediction of kidney function and chronic kidney disease (CKD) through kidney ultrasound imaging has long been considered desirable in clinical practice because of its safety, convenience, and affordability. However, this highly desirable approach is beyond the capability of human vision. We developed a deep learning approach for automatically determining the estimated glomerular filtration rate (eGFR) and CKD status. We exploited the transfer learning technique, integrating the powerful ResNet model pretrained on an ImageNet dataset in our neural network architecture, to predict kidney function based on 4,505 kidney ultrasound images labeled using eGFRs derived from serum creatinine concentrations. To further extract the information from ultrasound images, we leveraged kidney length annotations to remove the peripheral region of the kidneys and applied various data augmentation schemes to produce additional data with variations. Bootstrap aggregation was also applied to avoid overfitting and improve the model’s generalization. Moreover, the kidney function features obtained by our deep neural network were used to identify the CKD status defined by an eGFR of <60 ml/min/1.73 m 2 . A Pearson correlation coefficient of 0.741 indicated the strong relationship between artificial intelligence (AI)- and creatinine-based GFR estimations. Overall CKD status classification accuracy of our model was 85.6% —higher than that of experienced nephrologists (60.3%–80.1%). Our model is the first fundamental step toward realizing the potential of transforming kidney ultrasound imaging into an effective, real-time, distant screening tool. AI-GFR estimation offers the possibility of noninvasive assessment of kidney function, a key goal of AI-powered functional automation in clinical practice.

Role of Artificial Intelligence in Kidney Disease

“…What is more, Kuo et al identified the CKD status defined by an eGFR of <60 ml/min/1.73 m2 based on 4,505 kidney ultrasound images by deep neural network (The AUC of the model is 0.904). A Pearson correlation coefficient of 0.741 and accuracy of 85.6% indicated the strong relationship between AI and creatinine-based GFR estimations [48].…”

Restructured society and environment: A review on potential technological strategies to control the COVID-19 pandemic

“…Machine learning and rapidly developing deep learning-based technologies have exhibited their ability to convert these big data into a usable form in biomedical applications. In general, the implementation of AI and ML in the healthcare field have improved the welfare of the patients (Lundberg et al, 2018;Saria et al, 2010;Marella et al, 2017), improved quality of healthcare and effective diagnosis (Kuo et al, 2019;Rumsfeld et al, 2016;Liang et al, 2019), and also have lowered the healthcare costs (Bates et al, 2014;Özdemir and Barshan, 2014;Lo-Ciganic et al, 2015). Though a huge amount of data is available, most of them are never used for building mathematical models that can be integrated with health care system (Weintraub et al, 2018) and it is seen that only about 15% of the hospitals are currently using these technologies even then the usage accounts only for limited purposes (Zeng and Luo, 2017).…”

Kidney Diseases: Challenges and Opportunities of the Third Millenium. How can digital health help the National Health System?

“…Kuo et al (14) developed a convolutional deep learning (AI) model for the prediction of glomerular filtrate and the stage of renal failure from ultrasound images of the kidneys. Even this effort, while providing encouraging results, needs further development in order to be implemented in the clinic.…”