Neural network based hyperspectral imaging for substrate independent bloodstain age estimation
Nicola Giulietti1, Silvia Discepolo2, Paolo Castellini2
1Department of Mechanical Engineering, Politecnico di Milano, Via La Masa 1, Milan 60131, Italy.
Forensic Science International
|June 18, 2023
Summary
Determining bloodstain age is crucial for crime scene investigations. This new hyperspectral imaging technique uses artificial intelligence for accurate, substrate-independent age estimation, even on novel materials.
Area of Science:
- Forensic Science
- Biophotonics
- Artificial Intelligence
Background:
- Accurate bloodstain age determination is vital for forensic investigations.
- Current methods using reflectance spectroscopy face challenges with substrate influence and uncertainty.
- A reliable, substrate-independent method for bloodstain aging is needed.
Purpose of the Study:
- To develop a hyperspectral imaging technique for substrate-independent bloodstain age estimation.
- To utilize artificial intelligence for accurate age determination, overcoming substrate interference.
Main Methods:
- Acquisition of hyperspectral images of bloodstains.
- Employing a neural network to identify bloodstain pixels.
- Using an AI model to process reflectance spectra, remove substrate effects, and estimate age.
- Training and validation on diverse substrates over 385 hours.
Main Results:
- Achieved an absolute mean error of 6.9 hours for bloodstain age estimation over 385 hours.
- Demonstrated a mean absolute error of 1.1 hours for bloodstains within the first 48 hours.
- Successfully validated the method on an untested material (red cardboard) with consistent accuracy.
Conclusions:
- The developed hyperspectral imaging and AI technique provides accurate, substrate-independent bloodstain age estimation.
- This method significantly advances forensic capabilities by overcoming previous limitations.
- The approach shows promise for real-world crime scene applications.
