Using dependency/association rules to find indications for computed tomography in a head trauma dataset

Area of Science:

• Medical informatics
• Data mining
• Clinical decision support

Background:

• Clinical head trauma datasets require efficient analysis for accurate patient management.
• Traditional data mining methods may not always yield the most concise or effective decision criteria.

Purpose of the Study:

• To introduce and evaluate a novel binary-based data mining technique, the Boolean analyzer (BA), for analyzing clinical datasets.
• To develop a set of criteria for identifying minor head trauma patients who require computed tomography (CT) scans.

Main Methods:

• The Boolean analyzer (BA) algorithm was employed to partition a clinical head trauma dataset.
• Dependency/association rules were generated in the form of Boolean expressions.
• A probabilistic interestingness measure (PIM) was used to order rules based on event dependency.

Main Results:

• The BA technique generated a five-variable set of criteria for CT scan decisions in head trauma patients.
• These BA-derived criteria were more sensitive but less specific than traditional seven-variable Chi-square criteria.
• The BA method proved effective in identifying relevant patient subgroups.

Conclusions:

• The Boolean analyzer (BA) offers a powerful and potentially broader applicable data mining approach in the medical domain.
• The developed criteria provide a concise and sensitive method for guiding CT scan decisions in minor head trauma cases.
• This study highlights the potential of BA for enhancing clinical data analysis and decision-making.