pattern recognition with supervised artificial neural networks coupled with discretewavelet transforms for opioid discrimination using the colorimetric sensor array

Colorimetric sensor arrays offer a fast and accurate method for detecting and identifying various chemicals by analyzing changes in color or fluorescence using spectroscopic devices [1]. colloidal plasmonic nanoparticles, primarily composed of gold and silver, exhibit vibrant colors in the visible spectrum due to local plasmon resonance (lspr) [2]. utilizing nanoparticles, colorimetric sensor techniques rely on alterations in optical properties caused by their aggregations and morphological changes [3]. the aim of this study was to use supervised artificial neural networks and discrete wavelet transforms to discriminate between different opioids using a colorimetric sensor array. silver nanoparticles of varying sizes were employed to differentiate between methadone, morphine, and tramadol based on their reaction kinetics with the nanoparticles. the data obtained from the experiment was preprocessed using discrete wavelet transforms to reduce its complexity. neural network-based supervised pattern recognition methods, including cpann, skn, and xy-fused networks, were then utilized for discrimination after optimizing the network parameters using genetic algorithms. the best discrimination performance was achieved using the skn network and the daubechies (db9) mother wavelet. this resulted in an accuracy of 98% when distinguishing between the drugs in real serum samples that were spiked with the analytes.