Photon-counting mammography

Photon-counting mammography was introduced commercially in 2003 and was the first widely available application of photon-counting detector technology in medical x-ray imaging. Photon-counting detectors are fast enough to register single photon events and can reduce patient dose and enable quantitative imaging.

Basic principle

Photon-counting image receptors are made up of a sensor and electronics. The sensor material is a solid-state semi-conductor (e.g. silicon [31,74,77–80]), which is depleted by a bias voltage. When a photon interacts in the material, charge is released and collected by electrodes on the sensor. The electrodes are in general connected to parallel channels in an application-specific integrated circuit (ASIC) [31,86,89,97–99][92]. Each channel comprises an amplifier and a shaper, which convert the charge to a pulse with a height proportional to the energy of the impinging photon. The pulse height is measured by comparators, generally referred to as energy thresholds, which are followed by corresponding counters. The counters register the sum of all events within a specific energy window and are generally referred to as energy bins. The lower-most threshold is put below the expected incident spectrum to prevent electronic noise from being counted, while still counting all photons.

Fig. 3 shows the basic principle of a photon-counting spectral detector according to the description above.

Stryk ev detta stycke eller förkorta och lägg in informationen i nästa stycke. Fokusera på applications.

Clinical applications

Dose reduction

Photon-counting mammography allows for a reduction of patient dose while keeping image quality on par with conventional technologies, or, equivalently, improving image quality at equal dose. A study that compared photon-counting mammography to the state-wide average of the North Rhine-Westphalian mammography screening program in Germany reported a slightly improved diagnostic performance at a dose that was 40% of conventional technologies. Dose reduction is enabled by:

• Rejection of electronic noise: A low-energy threshold is put below the expected incident spectrum and prevents electronic noise from being counted. It should be noted, however, that electronic noise will still be added onto the pulse height and to some extent influence the energy resolution.
• Equal weighting of photons: Energy-integrating detectors intrinsically assign a higher weight to high-energy photons because more charge is released in the detector. This weighting is opposite to optimal because low-energy photons carry more contrast information. Photon-counting detectors, on the other hand, intrinsically weigh all photons equally, which is closer to optimal.
• Scatter rejection: The photon-counting mammography system uses a slit-scanning technique; the detector is made up of a number of thin lines and scanned across the object to acquire an image. This configuration is not intrinsic to photon-counting detectors, but it is often practical to make room for the large amount of electronics.

Energy weighting

Energy weighting was pioneered by Tapiovaara and Wagner [37],

and has subsequently been refined for projection imaging, including mammography [38,88,108–111]. In CT, energy weighting can either be applied on the projections [112], or on the reconstructed images [113]. Energy weighting improves the pixel-to-pixel contrast-to-noise ratio (CNR) by assigning a higher weight to low-energy photons. Thus, the dose efficiency is effectively improved, which enables a higher CNR at a constant patient dose, or a lower dose at a constant CNR. The effect follows as a logical consequence of the fact that the X-ray contrast drops with increasing photon energy, which holds true in energy regions away from absorption edges and where the photoelectric effect is non- negligible, including most X-ray imaging applications without contrast agents.

Several studies have investigated energy weighting with photon-

counting systems at the limit of infinite energy resolution, which results in a CNR improvement of approximately 9% in mammogra- phy [38,110], 6% for light elements (wax) in CT [114], and 24%–80% for heavier elements in CT [113,114], compared to photon counting without energy resolution (the improvement compared to integrating systems is larger). However, studies with realistic energy resolution report CNR improvements in the range 1%–5% [108,109,115] for mammography, 4% for light elements (wax) in CT [114], and 7%–30% for heavier elements in CT [113,114]. In particular, Berglund et al. modified a clinical spectral photon-counting mammography system and was able to improve the CNR of clinical images by 2.2–5.2%, which translates to a potential dose reduction at a constant CNR in the range of 4.5%–11% [111]. Even though these CNR improvements are relatively moderate, they come for free if imaging protocols are not altered for spectral acquisitions, such as in incidence-based methods (sandwich and photon-counting detectors).

Spectral imaging

tomosynthesis

Challenges

The main challenge of photon-counting detectors is pulse pileup [92, 101–103], which occurs when more than one photon interacts in a single detector channel within the time window for pulse detection, the so-called dead time. Pileup results in lost counts and reduced energy resolution because several pulses are counted as one, with a height that is a superposition of the true pulse heights and depends on the time difference between the events. Pileup will always be present in photon-counting detectors because of the Poisson distribution of X-ray photons (two events can occur arbitrarily close in time), but the speed of detector electronics has increased substantially during the past ten years, and acceptable pileup levels at CT count rates are coming within reach [104].

Another challenge ofphoton-counting detectors is cross-talk between adjacent channels, induced, for instance, by fluorescent X-rays or so- called charge sharing, which occurs when charge from a single photon interaction is collected by neighbouring electrodes and therefore de- tected as several events with distributed photon energy [82,90]. Such cross talk can be mitigated by anti-coincidence logic in the ASIC [92], which sorts out pulses in adjacent detector channels within a certain time window. The size of this time window is a compromise between efficiency of the technique and the probability of discriminating against true quasi-coincident events, so called chance coincidence. Chance co- incidence together with increased complexity of the electronics reduces the maximum count rate and increases power consumption, which are the main drawbacks of anti-coincidence logic. Coincident counts may be excluded all together, added to a high-energy bin (because charge sharing occurs with higher probability for high-energy photons) [31], or treated with more advanced schemes that include summation of the pulse height from adjacent detector channels and a probability- based localization of the impulse, such as implemented in the Medipix3 chip [105].

Stryk ev detta stycke eller förkorta eller lägg in informationen i före stycke. Fokusera på applications.

Detector technologies

Photon-counting image receptors are roughly made up of a sensor and electronics. The sensor material is usually a solid-state semi- conductor, such as silicon [31,74,77–80]. The advantages of silicon include high charge-collection efficiency, ready availability of high- quality high-purity silicon crystals, and established methods for test and assembly driven by the semiconductor industry [81]. The main challenge is the relatively low photo-electric cross section, which limits the the detection efficiency and leads to a large fraction of Compton scatter in the detector. The low detection efficiency is often addressed by arranging the silicon wafers edge on [82,83]. Compton scatter degrades the energy response because the full photon energy is not deposited in a single detector channel and compensation may be necessary, for in- stance, by a separate energy threshold to sort out Compton events [81]. A key characteristic of a photon-counting spectral detector is its energy resolution. The energy resolution of the MicroDose detector ranges from 2.0–2.3 keV standard deviation in the range 20–40 keV, and is shown in Fig. 2(d) for monochromatic impulses at 20 keV, 30 keV, and 40 keV, derived from a fitted detector model [31]. Newer silicon detectors have improved the energy resolution to less than 2 keV in the 40–120 keV interval [78].

Several research groups and commercial companies are investigating cadmium telluride (CdTe) and cadmium–zinc telluride (CZT) as sensor materials for photon-counting CT [27–29,84–87], and for projection imaging [88,89]. The higher atomic number of these materials result in higher absorption and less detector scattering, but, on the other hand, the higher ??-fluorescent yield leads to degraded spectral response and cross-talk between detector elements [90,91]. Also, manufacturing of macro-sized crystals of these materials poses practical challenges, and the crystals generally suffer from lattice defects and impurities that lead to charge trapping. Charge trapping limits the charge-collection efficiency for individual events [90,92], and may also cause long-term polarization effects (build-up of space charge) [93], which reduces energy resolution and detector speed.

Other solid-state materials, such as gallium arsenide [94], and mer- curic iodide [95], are currently quite far from clinical implementation, but may be expanding in the future. Gas detectors have also been investigated [96], but gas is more difficult to handle than solid materials and have limited absorption efficiency.

Stryk ev detta stycke eller förkorta eller lägg in informationen i föregående stycke. Fokusera på applications.