Next-Generation\nPhosgene Detection: Convolutional\nNeural Network with Triphenylamine and <i>N</i>‑Salicylaldehyde\nProbes for Enhanced Sensitivity and Bioimaging

Phosgene is a highly toxic gas that is widely used in\nvarious industries,\nmaking its rapid detection essential for safety. To address this need,\nwe developed a smartphone-based technique using convolutional neural\nnetworks (CNNs) for real-time, portable phosgene detection. Unlike\ntraditional fluorescence spectroscopy, which requires specialized\nequipment and expertise, this CNN-based approach is accessible and\naffordable and offers quick analysis, making it ideal for on-the-spot\ndetection. We employed this method to identify phosgene toxicity in\nsolutions ranging from 0 to 10 ppm by analyzing images of the solutions.\nSpecifically, we used intramolecular charge transfer (ICT)-based TPAOD\nand SAHY probes to detect phosgene through turn-off and turn-on fluorescence,\nwith detection limits of 19.44 nM (0.00759 ppm) and 34.89 nM (0.00817\nppm), respectively. A lifetime study of TPAOD confirmed that the quenching\nmechanism operates through static quenching. The SAHY probe was utilized\nfor the CNN model and was also tested for cell imaging studies in\nHeLa cells.