Artificial neural network, genetic algorithm, and logistic regression applications for predicting renal colic in...

Area of Science:

• Medical informatics
• Biostatistics
• Artificial Intelligence

Background:

• Logistic regression is a standard for multivariate medical data analysis.
• Artificial intelligence (AI) models like artificial neural networks (ANN) and genetic algorithms (GA) offer potential for medical data interpretation.

Purpose of the Study:

• To apply AI models (ANN, GA) to medical data.
• To compare the performance of AI models against logistic regression.

Main Methods:

• ANN, GA, and logistic regression were applied to a dataset of 227 patients with flank pain.
• The dataset concerned patients presenting to an emergency department with symptoms suggestive of renal colic.

Main Results:

• The GA identified two decision rules for predicting urinary stones.
• ANN demonstrated the highest Area Under the Curve (AUC) at 0.867.
• ANN and GA showed competitive performance metrics compared to logistic regression.

Conclusions:

• AI techniques (ANN, GA) can predict renal colic in emergency settings.
• These AI models can establish clinical decision rules.
• AI offers a potential alternative to conventional multivariate statistical methods.