Precise simulation of 3D fluorescence microscope image formation
| Authors | |
|---|---|
| Year of publication | 2007 |
| MU Faculty or unit | |
| Citation | |
| Description | The results of present biomedical research strongly depend on the results of image analysis algorithms applied to 2D as well as 3D biomedical image data acquired using fluorescence microscopy. Unfortunately, the latter results are often imprecise and unreliable due to gradual error accumulation throughout the long chain of operations applied to the input image (degradations caused by optics, electronics as well as data crunching). Moreover, the results can not be compared to the ground truth (GT) because GT is not known. Hence, results of different image analysis methods can not be verified or compared to each other. In some papers, this problem is partially solved by estimating GT by experts in the field (biologists or physicians). However, in many cases such GT estimate is very subjective and strongly varies among different experts. In order to overcome these difficulties we have created a toolbox that can generate 3D models of artificial biological objects along with their corresponding images degraded by specific optics and electronics. Image analysis methods can then be applied to such simulated image data and their results (such as segmentation or measurement results) can be compared with GT derived from input models of objects (or measurements on them). In this way, image analysis methods can be compared to each other and their quality (based on difference from GT) can be computed. The present version of the simulation toolbox can generate cells in 3D using deformation of simple shapes and adding texture to the cell interior. Further, it can simulate optical degradations using convolution with supplied point spread function as well as CCD camera artifacts such as impulse hot pixel noise, additive readout-noise or Poisson photon-shot noise. We have also dealt with the task of evaluating the quality of the simulated images in terms of their similarity to real image data. We have tested several similarity criteria such as visual comparison, intensity histograms, central moments or entropy. The talk will provide a short overview of 3D microscope image formation, mention specifics and problems of fluorescence mode, present examples of applications and finally describe the simulation process and evaluation of the quality of simulated images. |
| Related projects: |