In the last decade, high-content screening based on multivariate single-cell imaging has been proven effective in drug discovery to evaluate drug-induced phenotypic variations. Unfortunately, this method inherently requires fluorescent labeling which has several drawbacks. Here we present a label-free method for evaluating cellular drug responses only by high-throughput bright-field imaging with the aid of machine learning algorithms. Specifically, we performed high-throughput bright-field imaging of numerous drug-treated and -untreated cells (N = ~240,000) by optofluidic time-stretch microscopy with high throughput up to 10,000 cells/s and applied machine learning to the cell images to identify their morphological variations which are too subtle for human eyes to detect. Consequently, we achieved a high accuracy of 92% in distinguishing drug-treated and -untreated cells without the need for labeling. Furthermore, we also demonstrated that dose-dependent, drug-induced morphological change from different experiments can be inferred from the classification accuracy of a single classification model. Our work lays the groundwork for label-free drug screening in pharmaceutical science and industry.

According to WHO, about 10 million new cases of thrombotic disorders are diagnosed worldwide every year. Thrombotic disorders, including atherothrombosis (the leading cause of death in the US and Europe), are induced by occlusion of blood vessels, due to the formation of blood clots in which aggregated platelets play an important role. The presence of aggregated platelets in blood may be related to atherothrombosis (especially acute myocardial infarction) and is, hence, useful as a potential biomarker for the disease. However, conventional high-throughput blood analysers fail to accurately identify aggregated platelets in blood. Here we present an in vitro on-chip assay for label-free, single-cell image-based detection of aggregated platelets in human blood. This assay builds on a combination of optofluidic time-stretch microscopy on a microfluidic chip operating at a high throughput of 10 000 blood cells per second with machine learning, enabling morphology-based identification and enumeration of aggregated platelets in a short period of time. By performing cell classification with machine learning, we differentiate aggregated platelets from single platelets and white blood cells with a high specificity and sensitivity of 96.6% for both. Our results indicate that the assay is potentially promising as predictive diagnosis and therapeutic monitoring of thrombotic disorders in clinical settings.

The drug susceptibility of leukemia cells in whole blood is evaluated by using extreme-throughput imaging flow cytometry with deep learning.

The development of reliable, sustainable, and economical sources of alternative fuels to petroleum is required to tackle the global energy crisis. One such alternative is microalgal biofuel, which is expected to play a key role in reducing the detrimental effects of global warming as microalgae absorb atmospheric CO 2 via photosynthesis. Unfortunately, conventional analytical methods only provide population-averaged lipid amounts and fail to characterize a diverse population of microalgal cells with single-cell resolution in a non-invasive and interference-free manner. Here high-throughput labelfree single-cell screening of lipid-producing microalgal cells with optofluidic timestretch quantitative phase microscopy was demonstrated. In particular, Euglena gracilis, an attractive microalgal species that produces wax esters (suitable for biodiesel and aviation fuel after refinement), within lipid droplets was investigated. The optofluidic time-stretch quantitative phase microscope is based on an integration of a hydrodynamic-focusing microfluidic chip, an optical time-stretch quantitative phase microscope, and a digital image processor equipped with machine learning. As a result, it provides both the opacity and phase maps of every single cell at a high throughput of 10,000 cells/s, enabling accurate cell classification without the need for fluorescent staining. Specifically, the dataset was used to characterize heterogeneous populations of E. gracilis cells under two different culture conditions (nitrogen-sufficient and nitrogen-deficient) and achieve the cell classification with an error rate of only 2.15%. The method holds promise as an effective analytical tool for microalgae-based biofuel production. V C 2017 International Society for Advancement of Cytometry

No abstract