Positron emission mammography (PEM) can offer a non-invasive method for the diagnosis of breast cancer. Metabolic images from PEM using 18F-fluoro-deoxy-glucose, contain unique information not available from conventional morphologic imaging techniques like X-ray radiography. In this work, the concept of Clear-PEM, the system presently developed in the frame of the Crystal Clear Collaboration at CERN, is described. Clear-PEM will be a dedicated scanner, offering better perspectives in terms of position resolution and detection sensitivity.


Breast cancer is the most frequent malign neoplasm in women. According to the American Cancer Society, one in each nine women will develop invasive breast cancer in their lifetime(1). About 30 y ago the introduction of the regular breast cancer screening by X-ray mammography, allowing an early detection of this illness, reduced the mortality to 29%(2). At present, the American Cancer Society and the American College of Radiology recommend an annual screening mammography starting at age 40, even for women not included in any risk group. Conventional X-ray mammography has an overall sensitivity (number of true positive/total positive) of 80%, depending on the breast type. For fatty breasts, a sensitivity of 95% can be achieved with a lower limit in the size of a detectable tumour of 5 mm, while for dense breasts the sensitivity drops to 70% with a lower limit in size of 10–20 mm. With respect to (classic) PET, it has been shown that the sensitivity does not depend on the breast density. Moreover, the localisation of tumours, otherwise not seen on X-ray mammography or ultrasonography, can be done with PET. In addition, the specificity (number of true negatives/total negatives) of conventional X-ray mammography is rather low, typically 30%, while studies indicate a specificity over 95% for PET. In fact, there is a large domain of application for a dedicated positron emission mammography (PEM). The primary task will be the diagnosing of primary malign lesion on the breast, and the control of recurrence in patients with previous morphological problems, such as those made by surgery or other cause. The assessment of the treatment response is also a possibility. Chemotherapy is being increasingly used in locally advanced breast cancer. The kind of response to this treatment may be assessed with PEM, avoiding unnecessary side effects to non-responders. The system should also be applied to inconclusive cases after an X-ray examination where it should provide information that is as detailed as possible. PEM also must cover both the breast and the lymph node area, so the detector geometry has to fit two different anatomical constraints.

Since biopsy gives the ultimate answer when diagnosing breast cancer, the PEM scanner should be coupled to a stereotactic biopsy device. Breast examinations will be performed with patients in a prone position. This position allows for the largest decoupling of the breast from the body and provides a more comfortable positioning for the patient. A scanner table with holes for the breast must be then foreseen. The system will also allow one to examine the axilla by placing the patients axilla between the PEM plates.


The choice of the PEM geometry takes several factors into consideration: sensitivity, image reconstruction, cost and so on. A two-parallel-plate preliminary configuration has been presently favoured by prototype studies. This configuration allows for a better adaptation to the breast geometry by variable plate-separation-distance. Also axilla examination will be easier with this geometry. Owing to close-proximity of the breast, PEM will allow a smaller injected dose in the patient and less examination time. The proposed PEM camera aims at a spatial resolution between 1 and 2 mm (compared to ∼5 and 10 mm for a whole body PET)(3). To achieve these resolutions, each camera plate will be constituted by a matrix of scintillating Lu-based crystals with a 2 × 2 mm2 cross section and avalanche photo-diodes (APD) will read the light signal. These photodetectors have several advantages, Such as higher quantum efficiency or better lateral uniformity, over position-sensitive photomultiplier tubes (PMT) and they also allow the construction of compact assemblies, which is an important feature for mammography. The data acquisition system under development aims at a maximum event rate of 1 MHz, which is compatible with an examination time of ∼10 min that is necessary to obtain clinical significant images.


In a PEM system, the fraction of accepted two-photon events (detection sensitivity) must be as high as possible in order to reduce the injected dose in the patient and allow for less examination time. In particular, a PEM device should optimise the two-photon event sensitivity for photons coming from the breast. Our goal is to achieve a typical value of 10% in the centre of the field of view (FOV) when the PEM plates are 10 cm apart.

Only a fraction of the total injected activity will be fixed in the breast. This fraction is responsible for the true coincidence rate (events originating in the FOV). Although not well known at present, the available literature(4) indicates a value of ∼0.5% of the injected activity. The activity fixed in other body parts will be seen as background events (random coincidences) that will affect the image quality. A good compromise would be the reduction of these background coincidences to a level below 10% of the true rate. Preliminary studies point towards a true coincidence rate lying between 40 and 250 kHz, for a total activity of 10 mCi, depending on the PEM-plate-separation and breast uptake fraction. A total single event (one photon) rate in the detector up to 3 MHz is expected, depending on the detector shielding.

The PEM detector can be integrated together with an X-ray mammograph and a stereotatic biopsy system. The integration of these two devices in the system will extend the examination capabilities. Two parallel plates will constitute the PEM detector itself, with an adjustable separation distance between the plates. Separation distances between 6 and 40 cm are foreseen.

The two plates can rotate around the PEM axis, allowing one to take data in several orientations as needed for the reconstruction of tomographic images.

To exam the axilla region (or the breast in the front–back configuration), the PEM detector has to be rotated 90° and an image is produced with one plate below the table and the other over the patient shoulder (or back).


The design and optimisation of the PEM detector parameters are being obtained using Monte Carlo simulation techniques. Moreover, the Monte Carlo simulation also provides realistic data for reconstruction purposes. A dedicated and versatile Monte Carlo simulation framework is under construction based on GEANT4(5). The ROOT(6) toolkit has been adopted for event data storage and analysis. In the present, the developed framework consists of three autonomous modules:

PhantomFactory, PEMsim and DIGITsim

The PhantomFactory module simulates radioactive decay in different phantoms: homogeneous, heterogeneous, mathematical-type and voxel-based phantoms. Photons reaching a scoring region are stored for later tracking. The PEMsim module then performs the detector simulation. As input PEMsim uses data from the PhantomFactory module. Each photon or photon pair that interacts with the detector defines an event. The DIGITsim module simulates the signal formation process in the crystals and the response of the associated electronics. This module converts the information from PEMsim (energy, time) per event into a signal shape, adds electronic noise and performs the signal A–D conversion.

The Detection sensitivity, system count rate (prompt+accidental events), spatial resolution and zdepth-of-interaction (DoI) capability of the Clear-PEM device were investigated using Monte Carlo methods. In the PEM working region between 7 and 13 cm detector separation plate distances a sensitivity of 7–17% was found for a 18F point source placed at the centre of a 270 cm3 water phantom. Sensitivity was found to decrease by about a factor of two when the source was placed 4 cm off-axis. The system count rates were assessed using a mathematical phantom implemented in the simulation framework. Uptake in the considered organs represents the upper limits of tracer measured 1 h after an injection of 10 mCi of 18F-fluoro-deoxy-glucose (FDG). Most of the accidental coincidences were found to be produced by FDG uptake in torso (92%), liver (8%) and heart (1%), essentially because of their proximity to the detector's FOV. Results for a 350–700 keV energy window, 4 ns time window with 1 ns r.m.s. single-photon time measurement and 13 cm detectors separation distance are displayed in Table 1.

Table 1.

Count rates estimated by Monte Carlo simulation.

Single-photon rate 1.5 MHz per plate 
Accidental coincidence rate 17 kHz 
Prompt coincidence rate 36 kHz (up to 250 kHz) 
Total coincidence rate 54 kHz 
Single-photon rate 1.5 MHz per plate 
Accidental coincidence rate 17 kHz 
Prompt coincidence rate 36 kHz (up to 250 kHz) 
Total coincidence rate 54 kHz 

An increase by <5%, in single-photon and accidental coincidence rates was obtained when events from 176Lu radioactive decay (300 Bq in LuAP crystals) were added to β+18F decays. The Clear-PEM design significantly increases detection sensitivity, which should be stated in comparison with conventional PET cameras (<0.1%). Count-rate simulation results are within operation limits for the data acquisition system, and are able to read 1 MHz event rates, allowing one to fully profit from the large detector acceptance. The intrinsic spatial resolution of PEM for a point source in air placed at the centre of the FOV was estimated to be 1.2 mm FWHM. This result takes into account 18F positron range, non-collinear photon emission, crystal size and crystal identification algorithm for multi-hit events based on Compton kinematics. Depth-of-interactors simulation studies were performed for a single LuAP:Ce crystal (n=1.97). Best configuration yields a 2.4 mm FWHM DoI resolution with a light collection efficiency between 28 and 30%.


In order to assess the key aspect of the PEM detector, two experimental set-ups have been assembled. The first system is used to study DoI measurement performance. A small NaI(Tl) scintillator is used to electronically collimate the 511 keV gamma photons from a 22Na source. This set-up allows investigating the light collection as function of the depth of interaction in the crystal. One of the PEM scanner's basic components, consisting of 32 LYSO:Ce crystals coupled with to a 4 × 8 pixels avalanche photo-diode (APD) S8550 array manufactured by Hamamatsu is also under the process of being tested. A pixel array is connected to 32-channel readout arrangement, which includes a charge sensitive pre-amplifier and amplifier system developed by the Crystal Clear Collaboration. The VME readout includes a peak-sensing ADC and, as an alternative, a custom 100 MHz sampling ADC can be used.


Both simulation and laboratory tests have proved so far the concept of the Clear-PEM detector. A small-scale prototype (only one plate) is under construction which allows the testing of several physical, electronic and mechanical parameters.

P. R., A. T., N. M. and M. M. are supported by FCT grants SFRH/BD/10187/2002, SFRH/BD/10198/2002, SFRH/BD/6187/2001 and SFRH/BD/3002/2000, respectively.


Cancer Facts and Figures 1995. American Cancer Society, Atlanta, GA (
Basset, L. W., Liu, T. H., Giuliano, A. E. and Gold, R. H. The prevalence of carcinoma in palpable versus impalpable, mammographically detected lesions.
Am. J. Roentgenol.
–24 (
Lecoq, P. and Varela, J. Clear-PEM, a dedicated PET camera for mammography.
Nucl. Instrum. Meth.
  Phys. Res. A
–6 (
D. Vrasnsjent et al. Relationship between 18F-FDG uptake and breast density in women with normal breast tissue
J. Nucl. Med.
–1242 (
Available on www http://root.cern.ch/