Towards substrate-independent age estimation of blood stains based on dimensionality reduction and k-nearest neighbor classification of absorbance spectroscopic data

- Blood stain age estimation method based on spectroscopic data from dissolved blood.
- PCA provides age-correlated transformations from which age estimates can be deduced.
- Results suggest that the method can be applied independent from stain substrate.
- Estimation accuracy is comparable to established techniques.
- Future work aims at implementing environmental information to the prediction scheme.

The age determination of blood traces provides important hints for the chronological assessment of criminal events and their reconstruction. Current methods are often expensive, involve significant experimental complexity and often fail to perform when being applied to aged blood samples taken from different substrates. In this work an absorption spectroscopy-based blood stain age estimation method is presented, which utilizes 400-640 nm absorption spectra in computation. Spectral data from 72 differently aged pig blood stains (2 h to three weeks) dried on three different substrate surfaces (cotton, polyester and glass) were acquired and the turnover-time correlations were utilized to develop a straightforward age estimation scheme. More precisely, data processing includes data dimensionality reduction, upon which classic k-nearest neighbor classifiers are employed. This strategy shows good agreement between observed and predicted blood stain age (r > 0.9) in cross-validation. The presented estimation strategy utilizes spectral data from dissolved blood samples to bypass spectral artifacts which are well known to interfere with other spectral methods such as reflection spectroscopy. Results indicate that age estimations can be drawn from such absorbance spectroscopic data independent from substrate the blood dried on. Since data in this study was acquired under laboratory conditions, future work has to consider perturbing environmental conditions in order to assess real-life applicability.

94