• Facebook
  • Yahoo
  • Twitter
  • Google
  • Live
  • Facebook
  • Yahoo
  • Twitter
  • Google
  • Live

Research group “Statistical modelling and image analysis” – On the track to automated analysis of spectrometric, spectroscopic and image data

14 Jun | By Thomas Bocklitz
Research group “Statistical modelling and image analysis” – On the track to automated analysis of spectrometric, spectroscopic and image data
The information obtained from non-linear multimodal imaging can be translated into computational hematoxylin and eosin (HE) images by multivariate statistics. Here, a computational HE stain is generated resulting in pseudo-HE overview images that allow for identification of suspicious regions in colon tissue. A: Multimodal image; B: Pseudo-H&E image; C: H&E image
Image source: Heuke et al. Head & Neck, 2016, 38, 1545-1552, http://dx.doi.org/10.1002/hed.24477
0

The major research fields of the research group at Leibniz-IPHT Jena  and Friedrich-Schiller-University Jena are spectral data analysis, image analysis as well as correlation and combination of data derived from different measurement modalities. These computer-aided methods are powerful tools, for instance, to extract quantitative biomedical information from physical measurement data. A short overview of recent research activities in the group is given here.

With our research, we aim to develop tailored statistical and mathematical methods that enable biologists, chemists, and physicians to easily evaluate very complex Raman and infrared spectra obtained from biological samples. Employing statistical models and automated image analysis to physical measurement data provides valuable diagnostic information about the specimen, which would be otherwise hardly accessible or hidden. Biological samples such as head and neck carcinoma or colon tissue are received from our collaboration partners including the University Hospital Jena (Prof. Stallmach and Prof. Guntinas-Lichius), Charité Berlin (Prof. Lademann), University Hospital Erlangen (Prof. Waldner and Prof. Hartmann),  European Laboratory for Non-Linear Spectroscopy Firenze/LENS (Prof. Pavone), Idaho State University (Prof. Kalivas ) and the Laboratoire de Spectrochimie Infrarouge et Raman Lille/LASIR (Prof. Ruckebusch).

Are you interested in statistical modelling and image analysis of your measurement data? We are constantly looking for collaboration partners. For details, please check our offer for project partners in the Resource Finder.

 

Spectral data analysis

Within this research topic we developoptimal procedures and methods to correct and analyze different types of spectral data. Our major expertise is the analysis of Raman spectra, but we also investigate the chemometric procedures to analyze other kinds of spectral data, like NIR-spectra and MALDI-spectra. The performed studies aim to determine, which analysis and correction procedures are well suited and should be combined for an optimal and fully automatic data pipeline. Thus, we investigate the whole data pipeline and all its procedures.


Image Analysis

For a few years, we have been working intensively on the analysis of multimodal images, which are composed of coherent anti-Stokes Raman scattering (CARS), two-photon excited fluorescence (TPEF) and second harmonics generation (SHG) images, as well as other kinds of image data. In all our image related studies, we try to translate the physical measurements into biomedical information such as the concentration of lipids and collagen, or classification of tissue types in a tissue section. The data pipeline we constructed for this purpose, comprises experimental design, certain correction procedures, feature extraction and model construction. We perform investigations on all of these procedures to receive an optimal data pipeline, which delivers reliable and robust results.


Correlation of methods

The third major working field of our group is the correlation and combination of data derived from different measurement modalities. By fusing data from different measurement modalities and evaluating the data set as a whole, we are able to extract complementary information from the different measurement modalities and hence receive a more comprehensive understanding of the sample. Another aim of the investigated data fusion is the comparison of a specific technique to a reference technique and the quantification of the difference. In order to achieve these two goals, we investigate data fusion strategies and apply them to the combinations of various techniques.

Group members:


PhD students: Oleg RyabchykovNairveen AliMehul ChhallaniPranita Pradhan and Shuxia Guo

 

Acknowledgement:

Financial support of the EU, the Thüringer Kultusministerium, the Thüringer Aufbaubank, the Federal Ministry of Education and Research, Germany (BMBF), the German Science Foundation, the Fonds der Chemischen Industrie and the Carl-Zeiss Foundation are greatly acknowledged.

 

Share on Biophotonics.World:
Share on Social Media:

Comments:

No comments
You need to sign in to comment

Related Articles