COMPUTER-BASED DETECTION OF APOPTOTIC CELL MORPHOLOGY

Apoptosis is an erratic physiological process whereby cells undergo a self-destruction process. A quantitative analysis of apoptosis, namely counting the number of cells undergoing apoptosis, is required not only to understand the mechanism of this process but also to quantitatively evaluate the effect of anticancer drugs as many anticancer drugs attempt to achieve their effects by inducing the cancer cell to apoptosis. Currently, counting cells undergoing apoptosis is generally carried out manually by the naked eyes. Although these methods are accurate and appropriate in some cases, they can be extremely time consuming and labor intensive. Also, manual counting is subjective and difficult to achieve high throughput. There is compelling necessity for developing an automatic apoptosis-counting method. In this paper, we described an image-based method, which is applied in HCT116 human colon adenocarcinoma cell culture to count the number of apoptosis induced by anticancer drug oxaliplatin. Our methods are mostly based on well-known and mature algorithms in image processing and pattern recognition such as the contrast-limited adaptive histogram equalization, pixel classification, size distribution detection by granulometries, watershed algorithm, and artificial neural networks. We evaluated our segmentation results using the area overlap measurement, and the best achieved performance that can reach over 98%. We also evaluated our apoptosis detection results against the results of four biomedical experts and compared the results with other existing efforts. The final performance shows that our method is more reliable and accurate for detecting apoptosis than the previous methods.