UC Santa Cruz
Cancer Cell Growth Measurement Using Computer Vision
- Author(s): Pourshafiee, Amir
- Advisor(s): Teodorescu, Mircea
- et al.
We have devised a pipeline of computer vision and machine learning algorithms and developed a cell growth rate software. The algorithm has the ability to perform single cell detection.
Our cell segmentation mechanism uses the Canny edge detection method to detect the foregrounds of an image; after processing and filtering the foregrounds of the image, the living cell images are stored. In each iteration, we run a very fast object recognition algorithms to detect and match the image of the cells that have had been stored in the previous iteration with the ones that are stored in the recent iteration. If the algorithm fails to find any matches stored from the older iteration, we run a slower algorithm with that provides a higher likelihood of recognition. After finding the matches we calculate the rate at which the cell's area and perimeters have changed.
We also made a few attempts to perform automated cell injection and have made our program forward compatible with the future automated injection software or devices.