|
Acquired drug resistance is a fundamental predicament in cancer therapy. Early detection of drug-resistant cancer cells during or after treatment is expected to benefit patients from unnecessary drug administration and thus play a significant role in the development of a therapeutic strategy. However, the development of an effective method of detecting drug-resistant cancer cells is still in its infancy due to their complex mechanism in drug resistance. To address this problem, we propose and experimentally demonstrate label-free image-based drug resistance detection with optofluidic time-stretch microscopy using leukemia cells (K562 and K562/ADM). By adding adriamycin (ADM) to both K562 and K562/ADM (ADM-resistant K562 cells) cells, both types of cells express unique morphological changes, which are subsequently captured by an optofluidic time-stretch microscope. These unique morphological changes are extracted as image features and are subjected to supervised machine learning for cell classification. We hereby have successfully differentiated K562 and K562/ADM solely with label-free images, which suggests that our technique is capable of detecting drug-resistant cancer cells. Our optofluidic time-stretch microscope consists of a time-stretch microscope with a high spatial resolution of 780 nm at a 1D frame rate of 75 MHz and a microfluidic device that focuses and orders cells. We compare various machine learning algorithms as well as various concentrations of ADM for cell classification. Owing to its unprecedented versatility of using label-free image and its independency from specific molecules, our technique holds great promise for detecting drug resistance of cancer cells for which its underlying mechanism is still unknown or chemical probes are still unavailable.
|
