Artificial intelligence-assisted estimation of postmortem intervals in bacterially infected cadavers using...

Area of Science:

• Forensic Science
• Computational Pathology
• Artificial Intelligence

Background:

• Estimating postmortem interval (PMI) is vital but challenging due to numerous variables.
• Traditional PMI methods are subjective and less effective for extended intervals.
• Whole-slide imaging (WSI) and AI offer precise, reproducible, data-driven PMI estimation.

Purpose of the Study:

• To extend AI-based PMI estimation to bacterially infected cadavers.
• To evaluate the AI model's performance under various temperature conditions.
• To establish a practical, objective, and scalable PMI estimation method for forensic science.

Main Methods:

• Utilized whole-slide imaging (WSI) to capture digital pathology images of infected mouse cadavers.
• Applied artificial intelligence (AI) algorithms for data-driven PMI estimation.
• Tested the model on Staphylococcus aureus, Escherichia coli, and Pseudomonas aeruginosa infections at 25°C, 37°C, and 4°C.

Main Results:

• The AI model demonstrated robustness across diverse scenarios, including bacterial infections and varying temperatures.
• Achieved high performance metrics: micro- and macro-area under the curve (AUC) of at least 0.873 (patch-level) and 0.717 (WSI-level) in training/testing.
• External validation showed no less than 0.948 (patch-level) AUC, confirming model reliability.

Conclusions:

• AI-powered computational pathology provides an objective and scalable solution for PMI estimation.
• This approach enhances forensic workflows by leveraging pathological sections and AI algorithms.
• Establishes a new technical benchmark for PMI estimation in both infected and uninfected cases.