Propagation Phase Contrast Synchrotron Microtomography (PPC-SRμCT) is the gold standard for non-invasive and non-destructive access to internal structures of archaeological remains.
However, in this type of analysis, the virtual specimen needs to be segmented to separate different parts or materials, a process that normally requires considerable human effort. In the Automated SEgmentation of Microtomography Imaging (ASEMI) project, researchers developed a tool to automatically segment these volumetric images, using manually segmented samples to tune and train a machine learning model.
The results of the study, “Automated segmentation of microtomography imaging of Egyptian mummies,” were just published in the journal Plos One.
When used on a set of four specimens of ancient Egyptian animal mummies (from the Ptolemaic and Roman periods of 3rd century BC to 4th century AD) the team achieved an overall accuracy of 94 to 98 per cent (when compared with manually segmented slices) According to the researchers, this result is similar to the results of off-the-shelf commercial software using deep learning, which is between 97 and 99 per cent accuracy but has much lower complexity.
A qualitative analysis of the segmented output shows that these new results are close in terms of usability to those from deep learning, justifying the use of these techniques.
The authors of this study are Marc Tanti, Camille Berruyer, Paul Tafforeau, Adrian Muscat, Reuben Farrugia, Kenneth Scerri, Gianluca Valentino, V. Armando Solé, and Johann A. Briffa.
While this work demonstrates the feasibility of using machine learning for the 3D segmentation of large volumes, a number of further advances are necessary for an operational environment, the researchers state.
