(228cx) Classification and Quantitative Analysis of Normal and Apoptotic Cells from Fluorescence Microscopy Images

Fluorescence microscopy is a well-developed tool to study in vitro behaviour of cells. However, microscopy experiments can generate a great amount of images of cells with varying image qualities (Waters 2009). The manual quantification and analysis of these data is time consuming. Hence, accurate and automatic analysis of cells images such as Chinese Hamster Ovary (CHO) cells can be very useful.

Mammalian cells are prone to apoptosis (programmed cell death), which is a key metabolic event that can restrict the growth of cells and decreases the productivity in a bioreactor (Rulter, Spearman & Braasch 2014). The accurate detection of apoptotic cells will help identifying the critical factors that trigger apoptosis. This knowledge may be used for delaying apoptosis and potentially increase the productivity (Taatjes, Sobel & Budd 2008).

Morphological changes in cells are highly indicative of the occurrence of apoptosis (Henry, Hollville & Martin 2013). For example, shrinkage and blebbing of the cytoplasmic membrane are found to be significant characteristics of apoptotic cells, which may cause cells to lose normal, smooth and circular shapes. Blebbing during apoptosis is generally associated to swell of the cell membrane into spherical bubbles. Hence, microscopic observation of morphological changes can be used to discern normal from apoptotic cells. However, cells may exhibit highly variable values of these morphological measures due to the dynamic nature of apoptosis.

This work presents a new image processing and quantitative analysis method that can automatically differentiate apoptotic from normal cells, while maintaining the computational time at a reasonable level. The proposed method involves three consecutive steps: (i) a coarse segmentation that can be used to identify the number of cells in a given image of cells; (ii) a fine segmentation step to detect the boundaries of cells and to identify particular morphological features related to these boundaries; and (iii) a support vector machine (SVM) based classification model that uses the morphological features identified in the fine segmentation step (step ii) to distinguish apoptotic cells from normal cells.

References

Waters, JC 2009, 'Accuracy and precision in quantitative fluorescence microscopy', The Journal of Cell Biology, vol 185, no. 7, pp. 1135-1148.

Rulter, M, Spearman, M & Braasch, K 2014, 'Monitoring cell growth, viability and apoptosis', in R Portner (ed.), Animal cell biotechnology, method and protocols, Springer, Hamburg, Germany.

Taatjes, DJ, Sobel, BE & Budd, RC 2008, 'Morphological and cytochemical determination of cell death by apoptosis', Histochemistry and Cell Biology, vol 129, no. 1, pp. 33-43.

Henry, CM, Hollville, E & Martin, SJ 2013, 'Measuring apoptosis by microscopy and flow cytometry', Methods, vol 61, no. 2, pp. 90-97.