Automated prozone effect detection in ferritin homogeneous immunoassays using neural network classifiers

Area of Science:

• Clinical Chemistry
• Immunochemistry
• Artificial Intelligence in Diagnostics

Background:

• Turbidimetric homogeneous immunoassays are widely used for plasma component determination.
• The prozone effect (high-dose hook effect) can cause false-negative results in high-concentration samples.
• Existing methods for detecting the prozone effect are not cost-effective for routine laboratory use.

Purpose of the Study:

• To develop a cost-effective algorithm for detecting the prozone effect in turbidimetric immunoassays.
• To improve the reliability of plasma ferritin determination, especially for pathologically high concentrations.
• To prevent false-negative results caused by the prozone effect.

Main Methods:

• Development of a neural network classifier system to analyze reaction kinetics.
• Training and testing the neural network using 1500 determinations and 77 patient samples.
• Utilizing reaction kinetics analysis to identify the prozone effect immediately after measurement.

Main Results:

• The neural network system successfully identified the prozone effect by analyzing reaction kinetics.
• False-negative results due to the prozone effect can be filtered immediately without re-running samples.
• The system demonstrated the ability to handle high hook effect rates (5-12%) safely.

Conclusions:

• A neural network classifier provides a safe and immediate method for detecting the prozone effect in turbidimetric immunoassays.
• This technology enhances the sensitivity and reliability of plasma ferritin determination, even with elevated concentrations.
• The developed system offers a cost-effective solution to avoid false-negative results in routine laboratory diagnostics.