An efficient method for building a database of diatom populations for drowning site inference using a deep learning...

Area of Science:

• Forensic Science
• Environmental Science
• Microbiology

Background:

• Diatom analysis is crucial for identifying drowning sites but traditional methods are inefficient.
• Existing diatom testing methods are laborious, time-consuming, costly, and require specialized expertise.

Purpose of the Study:

• To develop an artificial intelligence (AI)-based system for efficient diatom identification and classification.
• To create a rapid method for analyzing diatom profiles to infer drowning locations.

Main Methods:

• Developed an AI system for species-level diatom identification and classification.
• Collected diatom profile data from Shanghai's Huangpu and Suzhou Rivers within two days.
• Evaluated diatom profile similarities between lung tissues and water samples using a modified Jensen-Shannon (JS) divergence measure in animal experiments.

Main Results:

• The AI system identified and classified diatoms at the species level.
• Diatom profile analysis achieved a prediction accuracy of 92.31% in identifying drowning sites.
• The method demonstrated high efficiency and simplicity for water monitoring.

Conclusions:

• The AI-based diatom analysis system offers a more efficient and simpler alternative to traditional methods.
• This technology can aid in seasonal or monthly water monitoring of diatom populations in interconnected river systems.
• The system assists law enforcement in narrowing investigations to confirm the origin of immersed bodies.