Automatic Cell Tracking in Time Series of Holographic Microscopy Images
Thesis: Finished.
Supervision by:
Markus Rempfler,
Prof. Dr. Bjoern Menze
In cooperation with: Valentin Stierle and Dr. Philipp Paulitschke (
LMU)
Background
|
Cell growth and migration play key roles in cancer: abnormal cell growth can lead to formation of tumors and cancer cells can spread to other parts of the body, a process known as metastasis. In vitro (i.e. “in the petri dish”) experiments are essential for studying cancer as well as developing and improving anti-cancer drugs.
Typically, in vitro experiments are performed with light microscopes. However, recent advances in CMOS sensor technology have created a promising alternative: Holographic Microscopy (HM). In HM partially coherent light is scattered by the sample, in this case the cell. The scattered light interferes with the unscattered wavefront and the resulting interference pattern -- the hologram -- is recorded with a CMOS sensor. The components required for HM are extremely small and cheap. HM therefore provides the means for a wide range of applications where a light microscope would be either to big or simply too expensive.
Image analysis algorithms are crucial to get from the large amount of holographic images to the clinically relevant information on cell growth and cell migration. Counting the number of cells and detecting cell division events in a time series of images gives access to the dynamics of cell growth. Locating and tracking individual cells in each images provides information about cell motility. Machine learning based methods have recently proven to be very effective for such tasks (see References below), however, little research has been conducted on using them on HM image series.
|
Objective
In this project, you will devise an algorithm that performs
cell recognition and
tracking, and
event classification, as illustrated in Fig. 2. To achieve this, you will:
- Review literature on related cell detection/tracking methods and assess their feasibility for the analysis of HM image series.
- Adapt and implement one or more algorithms, and ...
- Validate them on real annotated data sets.
You will work closely together with the
physicists at LMU, who are developing the HM sensor and acquired a collection of image series that will be used for validation of the algorithm.
Requirements
- Strong interest in machine learning/deep learning and computer vision.
- Good programming skills (Python, C++).
- Ability to work independently.
- Experience with OpenCV? is an advantage.
- Experience with medical image processing can be an advantage but is not a necessity.
Contact
If you are interested in the project or if you have any questions please contact
Markus Rempfler.
References
Cell detection/tracking with machine learning: