Ionic Cell Microscopy: A new modality for visualizing cells using microfluidic impedance cytometry and generative artificial intelligence

• Published in: Biosensors and bioelectronics. X Vol. 24; p. 100619
• Main Authors: Kokabi, Mahtab, Rather, Gulam M., Javanmard, Mehdi
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.08.2025; Elsevier
• Subjects: Cancer imaging; Generative AI; Impedance cytometry; Label-free diagnostics; Microfluidic device; Generative AI; Label-free diagnostics; Impedance cytometry; Cancer imaging; Microfluidic device
• ISSN: 2590-1370; 2590-1370
• DOI: 10.1016/j.biosx.2025.100619

This study introduces a novel approach to cancer cell imaging by integrating microfluidic sensor technology with artificial intelligence (AI). We developed a custom microfluidic device with polydimethylsiloxane (PDMS) microchannels and integrated electrodes to capture electrical impedance data. The device was fabricated using photolithography, electron beam evaporation, and lift-off techniques. Instead of traditional imaging methods, electrical impedance signals were used to reconstruct cell images. A generative AI model with eight hidden layers processed 191 impedance values to accurately reconstruct the shapes of cancer cells and control beads. Our approach successfully reconstructed images of MDA-MB-231 breast cancer cells, HeLa cells, and beads, achieving 91 % accuracy on the test dataset. Validation using the Structural Similarity Index (SSI) and Mean Structural Similarity Index (MSSIM) produced scores of 0.97 for breast cancer cells and 0.93 for beads, confirming the high precision of this method. This label-free, impedance-based imaging offers a promising solution for cancer diagnostics by accurately reconstructing cell shapes and distinguishing cell types, particularly in point-of-care applications. •A novel microfluidic device was developed, incorporating polydimethylsiloxane (PDMS) microchannels and precisely positioned electrodes to capture electrical impedance data from cancer cells.•Generative artificial intelligence (AI) was employed to reconstruct detailed cell images from electrical impedance signals, achieving high accuracy in visualizing MDA-MB-231 and HeLa cancer cells, as well as control beads.•The deep learning model, featuring eight hidden layers, processed impedance data with 191 distinct values to distinguish between cancer cells and control samples, achieving 91 % accuracy on the test dataset.•Image reconstruction was validated using Structural Similarity Index (SSI) and Mean Structural Similarity Index (MSSIM), with MSSIM scores of 0.97 for breast cancer cells and 0.93 for control beads.•This study introduces a pioneering label-free imaging modality that integrates microfluidic impedance cytometry with AI, showcasing the potential for enhanced cancer diagnostics and point-of-care applications.