Hyperspectral depth reconstruction by an iso-point approach for understanding of the light matter interaction on a macroscopic scale in turbid media.

• Applicant: Dr.-Ing. Martin Hohmann
• Subject Area: Joining and Separation Technology
• Term: since 2022
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 503621669

Hyperspectral imaging is becoming more and more a tool moving from pre-clinical research towards clinical application. While the high spectral resolution of hyperspectral imaging enables the collection of many physiological parameters, at the same time, the high spectral resolution hampers direct analysis of the hyperspectral images by visual inspection. One way of overcoming this is machine learning. It represents one of the main approaches of the advancement of hyperspectral imaging towards clinical applications. In general, many aspects of macroscopic interaction of light with turbid media are still incompletely understood or unknown. Even though it is well known that the depth of an inclusion changes the back reflection spectra at the surface, until recently, it was not known that there is a wavelength at which the intensity of the reflection is not affected by the depth of the inclusion, which can be called an iso-point. This could be shown in the DFG-funded project. This iso-point can be used to determine the depth of an inclusion. So far, however, this could only be shown qualitatively.The planned research project therefore focuses on the quantitative transfer of hyperspectral images into depth maps of inclusions using the iso-point approach. For example, three-dimensional angiographic information can be obtained remotely and completely non-invasively with a single image. The aim is for the method to work for a wide range of scattering and absorption coefficients of the tissue under investigation. Ideally, the principle of the iso-point can be understood so well in the research project that a reconstruction from an RGB image is possible. In this case, the presented method could easily be used in clinical facilities all over the world for the detection of diseases, as only a standard RGB-camera is needed. But even without the ability to use RGB images, recent advances in commercially available hyperspectral cameras allow for the reconstruction of depth information with a much cheaper and simpler setup compared to photoacoustic tomography or optical coherence tomography. Since a tumor, for example, alters the three-dimensional angiographic structure, the presented method has the potential for broad medical applications in tumor detection and evaluation.

• DFG Programme: Research Grants
• Co-Investigator: Professor Dr.-Ing. Michael Schmidt