Artificial Neural Network Approach in Laboratory Test Reporting:  Learning Algorithms.

Area of Science:

• Laboratory medicine
• Biochemical testing
• Artificial intelligence in healthcare

Background:

• Standardization and error reduction in laboratory medicine require predefined processes.
• Efficient evaluation of critical biochemical test values is crucial for patient safety.

Purpose of the Study:

• To develop an experimental, improvable decision algorithm model for rapid, multi-factor evaluation of critical biochemical test results.
• To leverage artificial neural networks for enhanced laboratory diagnostics.

Main Methods:

• Utilized Weka software and artificial neural network methods to build the experimental model.
• Trained the model using data from Dokuz Eylül University Central Laboratory.
• Validated the model's performance using separate test sets and statistical assessment.

Main Results:

• The decision algorithm, after three training iterations, showed no discrepancies with laboratory specialist verification.
• Achieved a sensitivity of 91% and a specificity of 100%.
• The estimated κ score was 0.950, indicating high agreement.

Conclusions:

• This study presents the first artificial neural network-based experimental assessment and decision algorithm for laboratory diagnostics.
• Integration of the trained algorithm into laboratory information systems offers potential for reduced workload and maintained patient safety.