Spectral Data Fusion for the Detection of Existing and Emerging Synthetic Opioid Analogs

Physical Sciences Inc. (PSI) proposes to develop a multimodal data fusion algorithm that is based on deep learning algorithms to detect previously unseen fentanyl analogs. The algorithm will be deployed on a hardware kit which contains a handheld Raman spectrometer and mass spectrometer (MS). The Raman spectrometer will be a handheld adaptation of PSI’s existing through-container Raman chemical sensor based on a spatial heterodyne spectrometer, which has improved limits of detection compared to conventional dispersive Raman spectrometers. A commercial off the shelf MS will be utilized to minimize additional hardware development cost. The proposed data fusion algorithm will leverage prior work performed by PSI on one-dimensional convolutional neural networks for spectral identification of chemicals with Raman spectrometers. The Phase I effort will focus on developing the detection algorithm separately for both the Raman and MS data channels, and subsequently demonstrating improved probability of detection and false alarms of the data fusion architecture compared to the independent channels. In parallel, a preliminary design for the hardware kit will be developed, which will be built and assembled in a subsequent Phase II program.