Machine Learning-Enabled Multimodal Electrochemical Sensor Integrated with Ionic Liquids to Detect Volatile Organic Compounds

Non-invasive or minimally invasive disease diagnostics have revolutionized how various medical conditions are detected and monitored. Diseases like cancer cause changes in metabolic pathways and release unique metabolites. Some volatile metabolites are excreted through airways, making them measurable in exhaled breath. Identifying these volatile organic compounds (VOCs) is a promising approach for cancer diagnostics. The prevailing approach for identifying VOCs with high sensitivity and selectivity in exhaled breath involves gathering the sample using collection cartridges or containers, followed by laboratory-based analysis utilizing Gas Chromatography–Mass spectrometry (GC-MS). Nevertheless, this approach necessitates sample preparation, transportation, and the desorption of gases, all contributing to substantial turnaround times and augmenting the overall expenses of the procedure. Due to these, researchers have tried developing portable, low-cost sensors that detect multiple VOCs. The current research primarily relies on electronic noses based on metal oxide semiconductor (MOS) sensor arrays, which detect changes in conductivity in the presence of volatile organic compounds (VOCs). However, a major limitation of MOS sensors is their poor selectivity. In contrast, our electrochemical sensor utilizes ionic liquids (ILs) as the sensing medium, offering significantly improved selectivity. This enhancement is achieved by tailoring the ionic composition or selecting ILs with different gas solubilities. Moreover, our sensor integrates multiple electrochemical techniques—including electrochemical impedance spectroscopy (EIS), cyclic voltammetry, and amperometry—to measure parameters such as conductivity, charge transfer resistance, capacitance, and diffusion. These diverse measurements generate richer datasets, which can be leveraged by AI/ML algorithms to further boost both sensitivity and selectivity. To incorporate the ILs and perform in-situ detection, we developed a sensor architecture integrating microfluidics and microelectrode technology. This multi-layered design, essentially a membrane contactor on a microscale, consists of 5 layers. At the bottom is a glass substrate with planar interdigitated gold microelectrodes fabricated using photolithography and a physical vapor deposition process. The next layer is a double-sided tape with a microchannel (for confining the liquid) with a length of 65 mm, a width of 500 μm, and a thickness of 130 μm cut out using a Cricut machine. Next is the gas-permeable hydrophobic Polytetrafluoroethylene (PTFE) membrane with a pore size of 0.45 μm. The membrane confines the liquid within the microchannel while allowing the gas to pass through. The next layer is a double-sided tape with a microchannel for gas flow with the exact dimensions (65mm x 500 μm x 130 μm) of the liquid microchannel. Finally, the top glass slide has gas inlet and outlet ports. This multi-layered design enables interaction between the gas and the ionic liquid (IL) through the membrane and allows the detection of perturbations in the IL using microelectrodes connected to a potentiostat. Commercially available ionic liquids—1-ethyl-3-methylimidazolium bis(trifluoromethylsulfonyl)imide ([EMIM][TFSI]) and 1-butyl-3-methylimidazolium hexafluorophosphate ([BMIM][PF 6 ])—were incorporated into this microfluidic platform for the selective detection of model VOCs, specifically acetone and toluene. Various electrochemical techniques, including EIS, cyclic voltammetry, differential pulse voltammetry, and amperometry, were employed to capture the perturbations induced by these VOCs in both ILs. The resulting multi-modal dataset was used to train a support vector machine (SVM) model, bypassing the need for a traditional electrochemical or equivalent circuit model. This approach enabled clear differentiation between acetone and toluene signals. Additionally, support vector regression (SVR) was applied to develop calibration curves for quantitative sensing.