Spectral imaging (radiography)
[2][3] The first medical application of spectral imaging appeared in 1953 when B. Jacobson at the Karolinska University Hospital, inspired by X-ray absorption spectroscopy, presented a method called "dichromography" to measure the concentration of iodine in X-ray images.
[4] In the 70's, spectral computed tomography (CT) with exposures at two different voltage levels was proposed by G.N.
[5] The technology evolved rapidly during the 70's and 80's,[6][7] but technical limitations, such as motion artifacts,[8] for long held back widespread clinical use.
In recent years, however, two fields of technological breakthrough have spurred a renewed interest in energy-resolved imaging.
In a generic imaging system, the projected signal in a detector element at energy level
Linear attenuation coefficients and integrated thicknesses for materials that make up the object are denoted
Most elements appearing naturally in human bodies are of low atomic number and lack absorption edges in the diagnostic X-ray energy range.
The linear attenuation coefficients can hence be expanded as[6] In contrast-enhanced imaging, high-atomic-number contrast agents with K absorption edges in the diagnostic energy range may be present in the body.
) yields a conventional non-energy-resolved image, but because X-ray contrast varies with energy, a weighted sum (
[17] An example with a realistic detector was presented by Berglund et al. who modified a photon-counting mammography system and raised the CNR of clinical images by 2.2–5.2%.
System properties and linear attenuation coefficients need to be known, either explicitly (by modelling) or implicitly (by calibration).
In CT, implementing material decomposition post reconstruction (image-based decomposition) does not require coinciding projection data, but the decomposed images may suffer from beam-hardening artefacts because the reconstruction algorithm is generally non-reversible.
In the absence of K-edge contrast agents and any other information about the object (e.g. thickness), the limited number of independent energy dependences according to Eq.
[25] The technique can also be used to characterize healthy tissue, such as the composition of breast tissue (an independent risk factor for breast cancer)[26][27][28] and bone-mineral density (an independent risk factor for fractures and all-cause mortality).
[29] Finally, virtual autopsies with spectral imaging can facilitate detection and characterization of bullets, knife tips, glass or shell fragments etc.
[30] The basis-material representation can be readily converted to images showing the amounts of photoelectric and Compton interactions by invoking Eq.
Further, differentiation between iodine and calcium is often challenging in conventional CT, but energy-resolved imaging can facilitate many procedures by, for instance, suppressing bone contrast[33] and improving characterization of atherosclerotic plaque.
VNC images are free from iodine staining (contrast-agent residuals),[35] can save dose to the patient by reducing the need for an additional non-contrast acquisition,[36] can improve radiotherapy dose calculations from CT images,[37] and can help in distinguishing between contrast agent and foreign objects.
[42] Some contrast agents can be targeted,[43] which opens up possibilities for molecular imaging, and using several contrast agents with different K-edge energies in combination with photon-counting detectors with a corresponding number of energy thresholds enable multi-agent imaging.
Temporal differences between the exposures (e.g. patient motion, variation in contrast-agent concentration) for long limited practical implementations,[6] but dual-source CT[9] and subsequently rapid kV switching[45] have now virtually eliminated the time between exposures.
Splitting the incident radiation of a scanning system into two beams with different filtration is another way to quasi-simultaneously acquire data at two energy levels.
[46] Detection-based methods instead obtain spectral information by splitting the spectrum after interaction in the object.
[47][48] Detection-based methods enable projection-based material decomposition because the two energy levels measured by the detector represent identical ray paths.
[50] Hence, the number of energy bins and the spectral separation are not determined by physical properties of the system (detector layers, source / filtration etc.
The first commercial photon-counting application was the MicroDose mammography system, introduced by Sectra Mamea in 2003 (later acquired by Philips),[10] and spectral imaging was launched on this platform in 2013.
[51] The MicroDose system was based on silicon strip detectors,[10][51] a technology that has subsequently been refined for CT with up to eight energy bins.
[54] The relatively low photo-electric cross section can be compensated for by arranging the silicon wafers edge on,[55] which also enables depth segments.
[63] Other solid-state materials, such as gallium arsenide[64] and mercuric iodide,[65] as well as gas detectors,[66] are currently quite far from clinical implementation.
