Exploring new dimensions in impedance cytometry: Integrating machine learning to enhance cellular visualization and particle characterization

Cancer is a complex disease that affects millions of people worldwide, presenting on going challenges in diagnosis and treatment. Cancer cells are generally classified into two types: suspension cells, which grow in liquid (e.g., leukemia and lymphoma), and adherent cells, which attach to surfaces and form clusters (e.g., breast and lung cancer). Differentiating between these cell types is crucial for identifying malignant and benign cells. This distinction plays a key role in shaping how the disease progresses and responds to treatment. This research introduces a label-free method to classify cancer cells by analyzing their electrical properties. In this study, I designed and fabricated custom microfluidic devices with precision-engineered electrodes to measure electrical properties of cells. Three suspension and three adherent cancer cell lines were examined. The results revealed clear differences: suspension cells, which are structurally simpler, showed lower impedance values, while adherent cells, being larger and more complex, exhibited higher impedance values. This method provides a fast, cost-effective, and reliable way to identify cancer cell types without relying on chemical labeling. By streamlining cancer cell classification, this approach could enhance diagnostic accuracy and improve treatment planning.Building on our progress in classifying cancer cells by their electrical properties, we aimed to enhance accuracy by integrating optical features. In addition to electrical measurements, we explored the combination of optical features and artificial intelligence (AI) to improve particle classification. Particle classification plays a vital role in various scientific and technological fields, from distinguishing bacteria and viruses to identifying cancer cells. In this study, we developed a multimodal approach that integrates electrical and optical data for more accurate classification. We classified three different sizes of control beads and a breast cancer cell line, leveraging both modalities to distinguish them. Machine learning algorithms were applied to evaluate the effectiveness of this approach. Our results demonstrate the advantages of multimodal analysis over single-modality methods. By combining electrical and optical features, we achieved an average test accuracy of 94.9%, compared to 66.4% with electrical data alone and 90.7% with optical data alone. This highlights the complementary nature of both measurements and their potential to improve classification performance. By integrating electrical sensing and optical imaging, our strategy offers deeper insights into particle properties. This approach advances cancer cell analysis and contributes to the development of more accurate and efficient diagnostic tools.Expanding the frontier of cancer cell analysis, this research introduces a novel approach that bridges imaging and impedance analysis. To overcome the limitations of traditional imaging techniques— including high costs, time-consuming processes, and reliance on complex instruments— we developed a method to reconstruct cell images directly from electrical signals. Our study harnesses the synergy of microfluidic sensor technology and advanced artificial intelligence (AI) to reconstruct cell images from electrical impedance signals. A generative AI model transforms impedance peak data, collected through a custom microfluidic device, into detailed visual representations of cells. This marks a departure from conventional image-based analysis, introducing electrical impedance as a foundation for image reconstruction. In this methodology, I designed a deep learning model that reconstructs cell visualizations from the full impedance spectrum, capturing distinct impedance signatures. The approach demonstrated 91% accuracy in reconstructing images of breast and HeLa cancer cells, as well as control beads. This research not only advances cancer cell imaging but also highlights the potential of integrating MEMS technology with AI for innovative biomedical applications. By transforming electrical impedance signals into high-fidelity cell images, this work opens new possibilities for enhanced diagnostics and cancer research.