Automated quantification of drug-induced cell death via vacuolation scoring in microscopic images.

Cell death plays a critical role in maintaining cellular homeostasis. The accurate evaluation of cell death is essential to determine the efficacy of a drug. Understanding the extent of cell death helps assess the potency of the chosen drug, which is a critical aspect of drug discovery research. The traditional method of verifying drug efficacy manually by observing the cells after drug application is time-consuming, labour-intensive, and prone to human error, highlighting the need for more precise and automated approaches. We introduce here the relevance of detecting paraptosis - a novel, non-apoptotic form of programmed cell death characterised by extensive cytoplasmic vacuolation. Non-small cell lung cancer cells (A549) treated with ginger extract (GE), which is known to induce paraptosis, were used as the model system for this study. The rate of cell death induced by GE is found to increase in a dose- and time-dependent manner. The cytoplasmic vacuolation is due to dilation of endoplasmic reticulum (ER) with excessive ER stress and downregulation of Alix, a paraptosis marker. In this work, we propose an automated solution that exploits recent advances in artificial intelligence to effectively segment and isolate individual cells, and to quantify their textural, morphological, and statistical characteristics. These are then used by a machine learning classifier to predict the cell death score. Such a score can be used effectively for drug screening. The suggested framework integrates Mask R-CNN for cell segmentation and an XGBoost classifier for evaluating the segmented areas. Therefore, this automated system can be a valuable tool to assess the degree of vacuolation formed while screening for paraptosis-inducing drugs, thereby providing insights into the level of cytotoxicity. The approach has the potential to enhance the scalability and robustness of automated drug screening and possible biomedical image analysis to a great extent.