Fraunhofer Institute for Computer Graphics Research IGDAs part of a joint industry project with the University Hospital Frankfurt, software for automated diatom analysis was developed. The solution uses advanced machine-learning technologies to support forensic examination of samples through efficient and precise image analysis. The application enhances forensic work with fast, reliable, and user-friendly analysis processes and is ready for immediate use in scientific and forensic investigations.
Innovative Image Analysis for Forensic Medicine
The developed software automates the identification and classification of diatoms in image data. By integrating a specially trained neural network, the analysis is performed efficiently and with high accuracy. The solution supports both standard image formats and the CZI microscopy format from ZEISS, enabling broad applicability in forensic practice. Batch processing allows the parallel evaluation of multiple samples, significantly increasing efficiency in routine laboratory workflows.
User-Friendliness and Technical Excellence
The software’s graphical user interface is intuitively designed and offers features such as hit lists with filtering options, visual presentation of analysis results including zoom capability, and automatic calculation of various statistics. Users can correct classification results directly and adjust parameters individually. Export formats include Excel (xlsx) with integrated image data and JSON, ensuring structured documentation and simplifying further processing. The software is based on license-free open-source frameworks such as OpenCV and the ONNX Runtime and is available as installable software including brief documentation.
