Limor Eger (PhD ’12)
Prof. W. Clem Karl, Prof. Prakash Ishwar, Dr. Synho Do (Mass. General Hospital), Dr. Homer Pien (Mass. General Hospital)
Funding: ALERT- A Department of Homeland Security Center of Excellence
Background: X-ray computed tomography (CT) is widely used for medical diagnosis and for security purposes such as baggage inspection. CT gives a three-dimensional image of the scanned object based on its x-ray attenuation. The attenuation depends on the material being scanned and is also a function of the energy of the incident x-ray photons. In multi-energy computed tomography (MECT), multiple energy-selective measurements of the attenuation of the object are obtained. MECT measurements can be used to infer information about the chemical composition of the object, potentially allowing discrimination of materials which are indistinguishable in the conventional single-spectrum case.
Description: In this project we investigate MECT reconstruction algorithms for the purpose of material identification and explosives detection. In general, the input to such algorithms is a set of measurements collected by MECT systems and the output is a set of material-specific parameters, such as effective atomic number and density. Multi-energy equations can be easily written and solved for monochromatic energy spectra and perfect detectors but become complex when considering realistic spectra, detector sensitivity, and system non-linearity. The project aims to advance MECT detection algorithms by: (i) introducing physics-driven model-based statistical methods for extraction of information from MECT measurements, (ii) developing an understanding of the information available in MECT data that is relevant to discrimination of explosive threats, and (iii) developing and applying learning-based methods to multi-energy material discrimination.
Results: Preliminary work has been performed on extensions to current image-based multi-energy techniques, a new statistical MECT reconstruction method based on the Extended Kalman Filter and on feature-selection for discriminating between explosives and non-explosives. Initial results show promise for improved material-discrimination.
L. Eger, S. Do, P. Ishwar, W. C. Karl, and H. Pien, “A learning-based approach to explosives detection using multi-energy X-ray computed tomography,” Submitted to ICASSP 2011.