Microscopic saw mark analysis: an empirical approach.

Area of Science:

• Forensic Science
• Digital Microscopy
• Statistical Analysis

Background:

• Microscopic saw mark analysis is a standard qualitative forensic technique.
• Potential sources of error in this method are not well-researched.
• Variability and outcome errors can impact the reliability of conclusions.

Purpose of the Study:

• To propose and evaluate classification trees and random forest classifiers for mitigating errors in microscopic saw mark analysis.
• To provide a statistically sound method for improving the accuracy of tool mark identification.
• To address the limitations of current qualitative approaches.

Main Methods:

• Developed and applied statistical models (classification trees, random forest) to analyze saw marks.
• Created 58 experimental saw marks on human femurs using four different saw types.
• Utilized a Keyence digital microscope for high-resolution imaging and analysis.
• Weighed variables based on discriminatory value to build decision trees.

Main Results:

• The statistical models successfully analyzed experimental saw marks.
• The developed decision trees demonstrated an outcome error rate between 8.62% and 17.82%.
• The approach effectively weighed variables by their discriminatory power.

Conclusions:

• Classification trees and random forest classifiers offer a statistically robust method to reduce errors in microscopic saw mark analysis.
• This approach enhances the reliability and objectivity of tool mark comparisons in forensic investigations.
• The study provides a foundation for more accurate and reproducible microscopic saw mark interpretation.