Prediction of Future Risk of Moderate to Severe Kidney Function Loss Using a Deep Learning Model-Enabled Chest...

Area of Science:

• Medical Imaging
• Artificial Intelligence in Healthcare
• Nephrology

Background:

• Chronic kidney disease (CKD) is a significant global health issue.
• Early detection and intervention are crucial for managing CKD progression.
• Existing predictive models often require extensive clinical data.

Purpose of the Study:

• To evaluate a deep learning model (DLM) for predicting moderate to severe kidney function decline.
• To assess the DLM's ability to utilize raw chest X-ray (CXR) data for CKD prediction.
• To identify high-risk populations for targeted early intervention.

Main Methods:

• A DLM was developed and fine-tuned using a large dataset of 79,219 patients.
• The model was trained on chest X-ray (CXR) data and patient demographics.
• Retrospective analysis with up to 5-year follow-up was conducted for validation.

Main Results:

• The DLM achieved high concordance index (C-index) values (0.903 internal, 0.851 external).
• High-risk groups identified by the DLM showed significantly higher rates of CKD stage 3b progression (19.2% vs. 0.9%).
• The DLM accurately predicted increased risk of end-stage renal disease (ESRD)/dialysis and mortality.

Conclusions:

• A DLM utilizing CXR data can effectively predict CKD stage 3b progression.
• This AI-powered approach offers a promising tool for early identification of at-risk individuals.
• Early intervention in high-risk populations identified by the DLM can potentially improve patient outcomes.