The ASEMI (Automated segmentation of microtomography imaging) project, led by Professor Johann Briffa, Head of the Department of Communications & Computer Engineering at the University of Malta, developed a tool using artificial intelligence techniques to automatically segment volumetric microtomography images, with particular application to Egyptian mummies. The project was in partnership with the European Synchrotron Radiation Facility (ESRF) in Grenoble, France, which has developed unique expertise in the application of synchrotron imaging to palaeontology and archaeology.

Microtomography is an X-ray imaging technique based on the same principle as the medical scanner, using synchrotron radiation instead of a conventional X-ray source to capture volumetric scans at much higher resolution and quality. This technology is able to provide 3D images to visualise the structure of materials in a non-invasive and non-destructive way, with applications in cultural heritage, materials research, life sciences, biomedical research and palaeontology. 

Rendering of the mummy of a puppy. A: external aspect of the remaining bandages, B: rendering of the preserved soft parts of the puppy, C: rendering of the bones, D: detail on the soft parts showing its weak preservation state through holes certainly caused by pests or putrefaction, E: rendering of the decidual and permanent dentition of puppy showing its young age.

One of the most recent applications is in Egyptology, through the investigation of animal mummies. In this application, trained specialists manually segment the volume into the various component materials, such as textiles, organic tissues, balm resin, ceramics, and bones. Depending on the complexity and size of the dataset, this process is very time consuming, typically taking several weeks for a small animal mummy. With the construction of the new beamline BM18, the same process is expected to be applied to human mummies, which would take considerably longer.

The main result of the ASEMI project is the development of a tool to automatically perform this laborious process. Using the developed software, the human specialist only needs to manually segment a small sample of the volumetric image. This is used to train and automatically optimise a machine learning system, which can then segment the whole volume in a fraction of the time previously required. The accuracy obtained by the ASEMI segment approaches the results of off-the-shelf commercial software using deep learning, at a much lower complexity. Following the principles of “Open Innovation, Open Science, Open to the World”, the developed algorithms, data sets, and results have been made freely available to the general public.

 3D rendering of the mummy of an ibis in a ceramic jar. A: general rendering of the external aspect of the mummy, B: internal content of the mummy made visible thanks to the transparency of the sealed jar, C: focus on the textile surrounding the animal, focus on the ibis inside the mummy D: showing its soft part, E: showing the ibis bones.

This project has received funding from the ATTRACT project funded by the EC under Grant Agreement 777222.

The full academic paper is published on PLOS ONE and can be accessed here. Links to the developed software and the scanned volumes can be found in the paper.