King's College London

One of the most common tasks in criminal investigation is to determine from which tissue source a biological fluid stain originates. As a result, there are many tests that are frequently used to determine if a stain is blood, semen or saliva by exploiting the properties of certain molecules present within the fluids themselves. These include chemical reagents such as the Kastle-Meyer or Acid Phosphatase tests, as well as other techniques like the use of alternative light sources. However, most of the tests currently available have some major drawbacks. In this study, a handheld near-infrared spectrometer is investigated for the specific identification of deposited bloodstains. First, a calibration was carried out by scanning over 500 positive (blood present) and negative (blood absent) samples to train several predictive models based on machine learning principles. These models were then tested on over 100 new positive and negative samples to evaluate their performance. All models tested were able to correctly classify deposited stains as blood in at least 81% of tested samples, with some models allowing for even higher classification accuracy at over 94%. This suggests that handheld near infrared devices could offer great opportunity for the rapid, low cost and non-destructive screening of body fluids at scenes of crime.