Detection of renal allograft rejection by computer

Area of Science:

• Nephrology
• Medical Informatics
• Biostatistics

Background:

• Monitoring renal allograft recipients is crucial for early detection of kidney dysfunction.
• Traditional methods rely on clinical assessment, which can be subjective and delayed.
• Objective, data-driven approaches are needed to enhance the timeliness of dysfunction detection.

Purpose of the Study:

• To evaluate a computer program employing a multiprocess Kalman filter for detecting changes in plasma creatinine and urea concentrations.
• To compare the computer's ability to identify renal allograft dysfunction with that of an experienced renal physician.
• To assess the timeliness of dysfunction detection by the computer program versus clinical assessment and treatment initiation.

Main Methods:

• A computer program adapted with a multiprocess Kalman filter was utilized.
• The program analyzed plasma creatinine and urea concentrations in 28 renal allograft recipients.
• Detected episodes of renal dysfunction were compared with retrospective clinical assessments by a renal physician.

Main Results:

• The computer program identified 31 out of 32 clinically recognized episodes of renal dysfunction using creatinine and urea.
• Using creatinine alone, the program identified 29 episodes.
• Computer-identified dysfunction occurred significantly earlier (p<0.05) than clinical identification and a median of one day earlier (p<0.02) than treatment initiation.

Conclusions:

• The computer-assisted multiprocess Kalman filter method is effective in detecting renal allograft dysfunction.
• This computational approach offers earlier detection of kidney dysfunction compared to traditional clinical evaluation.
• The method provides a valuable tool for real-time analysis of results and precise event timing in transplant monitoring.