LCM

Live Cell Monitoring – Trainable Image Analysis for Optical Monitoring of Cell Cultures

© Photo Fraunhofer FIT

(left) Cell counting in suspension. | (right) Classification of cells in adhered (red) und detached (blue).

Fraunhofer Institutes FIT, IPM, IPA, and IBMT collaborate in the project 'Live Cell Monitoring'. The objective is the development of a cell cultivation platform that automates all steps from plating to harvesting of cells. The individual process steps of cultivation are steered by optical control of the cell culture and an automatic analysis of the image data. Fraunhofer FIT develops flexible image analysis techniques for this platform.

Trainable Image Analysis

Cell cultivation places high demands on image analysis, since different cell types can have very different appearances. Even cells of a single type strongly vary depending upon cultivation stage. Therefore, image analysis algorithms must be adapted to the respective task. Current image analysis software permits customization of algorithms only on lowest level, i.e., parameters must be set manually. Biologists, however, usually do not have the technical background to select a correct setting.

With the trainable image analysis developed by Fraunhofer FIT, the biologist simply marks interesting biological structures in image data with the mouse. The software computes the common features of the selected examples and automatically determines parameters and algorithms for pertinent image analysis.

The trainable image analysis allows, for example, an automatic determination of the degree of confluence of a cell culture, counting of cells in a suspension, and an automatic classification of cell conditions.

By this approach, the configuration of image analysis for a high throughput application of the system can be achieved on the basis of few example data.

Applications

Degree of Confluence 
The growth of cells must be supervised during cultivation. If cells grow too densely, stress can render them useless for subsequent experiments. For this task, a biologist has to mark examples of two regions that can be distinguished in the image data: covered and cell-free regions. The system is trained by these examples and is then able to discriminate further image data automatically.

Cell Counting 
For harvesting, cells are brought into suspension and transferred to microplates. In subsequent experiments, the concentration of cells in the solution must be identified. This is achieved via optical counting in a small sample of the suspension. The system is first trained by the biologist marking examples of individual cells via mouse clicks. After this training the system can recognize this structure and find respective cells automatically.

Classification of Cell Conditions 
If problems occur during cultivation (bad culture medium, bad cell line), a visible change of the cells will occur, e.g., an increased detachment. Detached cells clearly differ from adhered cells. If the image analysis is trained to discriminate these two cell conditions, a quality assurance can be based on optical monitoring during cultivation.