Deep Learning-Enabled Screening of Chronic Kidney Disease from Echocardiography

Area of Science:

• Cardiology
• Nephrology
• Artificial Intelligence in Medicine

Background:

• Chronic kidney disease (CKD) affects 850 million people globally, with 60% undiagnosed.
• CKD is closely linked to cardiovascular disease, highlighting the need for effective screening.

Purpose of the Study:

• To develop and validate a deep learning (DL) model for detecting CKD using parasternal long-axis (PLAX) echocardiogram videos.
• To assess the model's performance in detecting any stage of CKD across diverse patient cohorts.

Main Methods:

• A DL model was trained on 325,377 PLAX videos from 62,818 patients at Cedars-Sinai Medical Center (CSMC).
• External validation was performed on cohorts from Stanford Healthcare (SHC) and Kaiser-Permanente Northern California (KPNC).

Main Results:

• The DL model achieved an AUC of 0.756 in the CSMC test cohort.
• Strong external validation performance was observed: AUC 0.718 in KPNC and 0.719 in SHC.
• The model demonstrated robust CKD detection capabilities across different clinical sites.

Conclusions:

• The developed DL echo model offers a noninvasive method for CKD screening.
• This approach has the potential to significantly improve CKD detection rates globally.