Current research / employment

Conventional light microscopy is commonly used and one of the most powerful techniques for the inspection and diagnosis of biological samples. However it is constrained by several factors: Only a small part of a sample is visualized at a time, magnifications are limited, samples are irreproducible, slides have to be physically present for examination, stain colours fade, and simultaneous viewing is limited. One goal of virtual microscopy is to solve these problems by replacing the direct work with the microscope and the slide with the use of digitized probes. For this purpose the slide is scanned once by a fully automated microscope and saved digitally. Such a digitized probe has none of the problems mentioned above.

Nevertheless, there are still impediments that have not been entirely solved. Considering the case of 20-fold magnification and a typical camera, in order to scan an area of 2x3 cm, roughly 4500 fields of view have to be acquired. To generate the virtual slide these fields of view have to be composed into a large mosaic. The scanning process of the slide is done by computer-controlled stages, which have to be accurately calibrated. However, motorized stages have positioning errors that are bound to accumulate during the scanning process, leading to offsets in the final virtual slide. Additionally, each field of view suffers from typical image acquisition problems: inhomogeneous illumination, non-uniform camera sensitivity, camera noise and dirt or dust contaminations on the lens and the detector. These impairments disturb the viewing experience and the medical diagnosis.

This research deals with these impediments. The emphasis is on the correct alignment of all fields of view and the blending between adjacent fields of view to compensate for image acquisition impairments. Additional focus is laid on the integration of fields of view acquired with different objective magnifications into the alignment process. Contrary to state of the art algorithms the alignment is not approached sequentially but as an optimization problem that incorporates the whole slide in a single model and is solved in one step. For this purpose equations are derived that describe the solutions between pair wise adjacent fields of view. These pair wise equations are then cast into a constrained mathematical model that describes the alignment problem of the complete slide. A solution to the mathematical model and thus the positioning of each field of view is found in one single step by global error minimization schemes.

Figure: The figure below depicts exemplary a section of a slide from Hematology acquired with different magnifications and stitched using the developed algorithm. The images were acquired using 10-fold magnification as well as 100-fold magnification. In the foreground the 100-fold images are visible. Those images provide high resolution views of leucocytes. It can be seen that all images are aligned correctly and that there is no offset between the single images. When zooming into the slide the resolution of the 100-fold images will gradually increase up to the original acquired resolution. Thus, the viewer can use these parts of the slide for medical diagnosis such as differential blood cell counts.

Publications A. Ihlow, C. Held, C. Dach, D. Steckhan, T. Wittenberg, Evaluation of Expectation Maximization for the Segmentation of Cervical Cell Nuclei., Bildverarbeitung für die Medizin 2011, , 139-144., (2011)

S. Rupp, D. Steckhan, A uniform, raytracing-based imaging model for rigid and fiber-optic endoscop, Engineering in Medicine and Biology Society (EMBC), 2010 Annual International Conference of the IEEE, , 61-67, (2010)

D. Steckhan, T. Wittenberg, A study on cervical nuclei detection algorithm in slide based cytometry, Journal of Microscopy, , in review, (2010)

D. Steckhan, D. Paulus, A quadratic programming approach for the mosaicing of virtual slides that incorporates the positioning accuracy of the microscope stage, Engineering in Medicine and Biology Society (EMBC), 2010 Annual International Conference of the IEEE, , 72-77, (2010)

D. Steckhan, T. Wittenberg, Optimized Graph-Based Mosaicking for Virtual Microscopy, Proceedings of the SPIE Medical Imaging 2009 Orlando, 7259, , (2009)

T. Bergen, D. Steckhan, T. Zerfass, T. Wittenberg, Segmentation of leukocytes and erythrocytes in blood smear images, Engineering in Medicine and Biology Society, 2008. EMBS 2008. 30th Annual International Conference of the IEEE, , 3075--3078, (2008)

D. Steckhan, T. Bergen, T. Wittenberg, S. Rupp, Efficient large scale image stitching for virtual microscopy, Engineering in Medicine and Biology Society, 2008. EMBS 2008. 30th Annual International Conference of the IEEE, Proceedings of the 30th Annual International Conference of the IEEE EMBS, , 4019-4023, (2008)

D. Steckhan, T. Zerfaß, T. Wittenberg, A comparative study of cervix nuclei detection algorithm, Cytometry A, 71A(7), 521-522, (2007)