Combining geometric and probabilistic reasoning for computer-based penetrating-trauma assessment

Area of Science:

• Medical imaging and computational analysis
• Trauma assessment and injury prediction

Background:

• Ballistic penetrating trauma to the chest and abdomen presents complex diagnostic challenges.
• Current assessment methods may benefit from advanced computational tools.

Purpose of the Study:

• To evaluate a novel computer system, TraumaSCAN, for assessing ballistic penetrating trauma.
• To determine the efficacy of integrating three-dimensional geometric and probabilistic reasoning in trauma assessment.

Main Methods:

• Development of the TraumaSCAN system integrating geometric and probabilistic reasoning with Bayesian networks.
• Retrospective analysis of 26 gunshot wound cases to evaluate system performance.
• Calculation of areas under the receiver operating characteristics (ROC) curve with and without observed findings input.

Main Results:

• TraumaSCAN demonstrated strong performance in assessing 11 conditions, with median ROC areas ranging from 0.883 (no findings) to 0.941 (with findings).
• The system's diagnostic accuracy showed significant improvement when observed findings were incorporated for certain conditions.
• The best ROC area achieved was 1.00 without findings and 0.992 with findings, indicating high diagnostic capability.

Conclusions:

• The computer-based approach combining geometric and probabilistic reasoning is a promising tool for assessing ballistic penetrating trauma.
• TraumaSCAN's ability to integrate anatomical likelihood and injury consequences offers potential for improved trauma evaluation.