Pendar Technologies is tackling the challenge posed by the rapid evolution of synthetic opioids, particularly fentanyl and its analogs, by developing a deep learning classifier for our Pendar X10 handheld Raman spectrometer. The effort would significantly advance the safety and detection capabilities of law enforcement officers and first responders in addressing the opioid crisis.Our approach involves three key development aspects: - Data Enhancement: We will expand our extensive Raman spectral signatures dataset, currently comprising over 200 synthetic opioids, 650+ illicit drugs, and thousands of other spectra. This includes collaborating with the Global Forensic and Justice Center for the acquisition of new seized drug spectra, generating synthetic fentanyl spectra with Prof. Gomez-Bombarelli (MIT), and creating training mixture spectra from pure drug and cutting agent spectra. - Hybrid Algorithm Development: We're creating a two-step algorithm. The first step uses the Pendar X10's existing library and algorithm to identify known substances and isolate unknown signatures. The second employs a conditional convolutional neural network classifier, utilizing insights from the first step to differentiate between chemical classes like fentanyl analogs and common confusers. - Testing and Validation: We'll develop a rigorous testing framework using existing hardware and data from seized drugs.Phase I focuses on building a foundational framework and proving feasibility, including an enriched dataset and initial machine learning models. Phase II will further refine and validate these models for commercial-grade accuracy. Upon successful completion of both phases, we plan to integrate this AI-enhanced detection into the Pendar X10 as a software update.
