Fatal fall from a height: is it possible to apply artificial intelligence techniques for height estimation?

Area of Science:

• Forensic pathology
• Biomechanical engineering
• Data science

Background:

• Falls from height result in severe trauma, with fall height significantly influencing injury severity.
• Estimating fall height is crucial for reconstructing accident scenarios and legal investigations.
• Current methods for estimating fall height may lack precision.

Purpose of the Study:

• To investigate the utility of autopsy injury patterns in estimating the height of falls using machine learning algorithms.
• To develop and validate predictive models for fall height estimation based on post-mortem findings.

Main Methods:

• Retrospective analysis of 455 autopsy cases involving falls from height.
• Classification of cases into 7 height-based groups (6m to 24m+).
• Application of machine learning algorithms (Linear Regression, Support Vector Regressor, Kernel Ridge, Decision Trees, Random Forests) with and without dimensionality reduction techniques (PCA, SVD, ICA).

Main Results:

• Machine learning models demonstrated good performance in estimating fall height.
• The best mean absolute error achieved was 4.37 ± 1.27 meters when dimensionality reduction was not applied.
• Models incorporating autopsy injury patterns showed predictive capability for fall height.

Conclusions:

• Autopsy injury patterns, when analyzed with machine learning, can serve as a valuable tool for estimating fall height.
• The proposed methodology offers a data-driven approach to enhance forensic investigations of falls from height.
• Further refinement of algorithms and data may improve the accuracy of fall height estimation from autopsy findings.