In Drug testing and analysis
On-scene drug detection is an increasingly significant challenge due to the rapidly changing drug market as well as the risk of exposure to potent drug substances. Conventional colorimetric cocaine tests require touching and handling of the unknown material and are prone for false positive reactions on common pharmaceuticals used as cutting-agents. In this work, the novel application of 740 - 1070 nm small wavelength range NIR spectroscopy to confidently detect cocaine in case samples is demonstrated. Multi-stage machine learning algorithms are applied to exploit the limited spectral features and predict not only the presence of cocaine but also predict a concentration and sample composition. A model based on >10,000 spectra from case samples yielded 97% true positive and 98% true negative results. The practical applicability is shown on over 100 case samples not included in model design. One of the most exciting aspects of this on-scene approach is that the model can almost instantly adapt to changes in the illicit-drug market by updating meta-data with results from subsequent confirmatory laboratory analysis. These results demonstrate that advanced machine learning strategies applied on limited range NIR spectra from economic handheld sensors can be a valuable procedure for rapid on-site detection of illicit substances by investigating officers. Besides forensics, this interesting approach could also be beneficial for screening and classification applications in the pharmaceutical, food-safety and environmental domains.
Kranenburg Ruben F, Verduin Joshka, Weesepoel Yannick, Alewijn Martin, Heerschop Marcel, Koomen Ger, Keizers Peter, Bakker Frank, Wallace Fionn, van Esch Annette, Hulsbergen Annemieke, van Asten Arian C
cocaine, indicative testing, k-Nearest Neighbors, machine learning, near-infrared