A CCD BASED DIGITAL DETECTOR FOR WHOLE-BREAST DIGITAL MAMMOGRAPHY

1 Introduction

We are developing a 18x27cm digital x-ray detector for whole-breast mammography based on CCD imaging technology [1]. In this paper we report the performance of a 10x20 cm, 2-CCD prototype we have built and tested. Similar to a film system, this detector uses a phosphor screen to convert an x-ray image to visible light. In the digital system the light photons are transferred onto a charge-coupled device (CCD) by a demagnifying fiber optic taper. Dedicated electronics then transfer the image to a PC based viewing station. The dynamic range of the detector is roughly 375 times greater than that of Kodak Min-R film [2]. This difference increases our ability to discriminate low contrast differences in tissue, and allows image exposure to be determined solely by the desired signal-to-noise ratio and acceptable patient dose, whereas in a non-digital system, exposure is determined by film’s narrow linear response range.

The detector is based on a modular design using a small number of CCDs and optical tapers to build up a full field image. Each module consists of a fiber optic taper that transfers the image from the phosphor converter to a CCD chip. The 10x10 cm front surface of the tapers are placed next to each other with a space of approximately 25mm at the interface. This design allows us to achieve good resolution and signal-to-noise ratio without loss of information between the tapers. By using a minimum number of CCDs and employing a modular design, we have attempted to limit the cost and mechanical complexity of the imager.

In this paper, we describe the detector design and report the spatial resolution, dynamic range, and signal-to-noise ratio of the prototype. We also report results of testing the detector on an ACR approved mammographic phantom and present patient images obtained with the detector.

2 Detector Design

The detector design, as shown in Figure 1, involves the conversion of x-ray photons into visible photons by a phosphor screen held in contact with the large end of a demagnifying fiber optic taper. The phosphor converter is a film of terbium doped

Gd₂O₂S deposited on an aluminized Mylar backing to a density of 18mg/cm². The phosphor screen is held in contact to the fiber optic taper by a 4mm thick foam compressed to half its original thickness with a rigid piece of carbon fiber plastic.

In the prototype version described here we align next to each other two fiber optic tapers, each with an imaging area of 10x10cm, to form a total imaging area of 10x20cm. The tapers have a demagnification ratio of 3.5:1 and transmit approximately 6% of the incident photons. Light transmitted through the taper is coupled to the CCD through a 1:1 fiber optic faceplate epoxy-bonded to the front surface of the CCD.

(a) The two module prototype in operation at the University of Virginia.

(b) Overview of detector components.

Each CCD is placed in thermal contact with a copper heat transfer block, which is cooled by a thermoelectric (TE) cooling module. The heat from the TE cooling module is dissipated by water flowing through a heat exchanger block and the water is kept at constant temperature by a closed cycle chiller. There is a 25mm gap at the interface between the tapers. As this space is less than the 50mm effective pixel size of the system, information at the interface region is collected by each neighboring taper. Therefore, while the system suffers a decreased signal-to-noise ratio at an interface, there is no spatial discontinuity in the image.

We use front illuminated CCDs (Thomson THX 7899) with 2048 by 2048 14mm pixels. The effective pixel size, determined by multiplying the CCD pixel size by the taper demagnification ratio, is 49mm. The full well capacity of the CCD is roughly 250,000 electrons and the signal for each pixel is digitized with a 16-bit analog to digital converter. Utilizing 4 amplifiers per chip allows full image readout in approximately 6.3 seconds.

Once the image is collected, three steps are taken to remove artifacts of the imaging system: 1) a dark image is subtracted; 2) the subtracted image is then corrected for non-uniform response of the phosphor, fiberoptics and CCD; and 3) this image is corrected for geometric distortions introduced by the fiberoptics. These corrections take approximately two seconds to perform using a PC with a Pentium 200 processor.

3 Detector Characterization and Imaging Results

The read noise is approximately 15 electrons at the readout rate of 166 kHz. With the detector cooled to -10? C, the dark noise is approximately 6.7 electrons/pixel/sec. Under standard mammographic conditions (28kVp Mo source, 25mm Mo filter) the system exhibits a total noise of roughly 16 electrons/pixel. Using antiblooming electronics, the CCD exhibits a full well depth of 250,000 electrons, resulting in an effective linear dynamic range of roughly 15,000. The system gain is 3 electrons generated in the CCD per 17 keV x-ray photon incident on the phosphor. The system MTF is shown in Figure 2. The MTF drops to near zero at roughly the Nyquist frequency of 10 lp/mm. The noise power spectrum (NPS) describes the spatial-frequency dependence of the system noise and is shown in Figure 2. The DQE, shown in Figure 3, is highest at low frequencies and diminishes to near 0 at 10 lp/mm. This pattern is repeated over the dynamic range of the detector.

The performance of the system was first evaluated using the ACR mammographic phantom. Under standard mammographic conditions all but the smallest spec group is visible. We have also taken initial patient images at the University of Virginia, an example of which is shown in Figure 4, alongside a film image of the same subject.

The results presented in this paper demonstrate that this detector has low noise, a high DQE, a large dynamic range, and good resolution. Our simple design allows for large area mammographic imaging without the loss of information in regions between fiber optic tapers.

4 References

[1] D. Rosen, M.J. Stanton, W. Phillips, M.B. Williams, D. O'Mara and H. Qian (1997) Full Field Digital Mammography Using CCD Technology, RSNA97, infoRAD Presentation

[2] M.B. Williams, P.A. Mangiafico, P.U. Simoni, M.J. Stanton, W. Phillips, D. Rosen, (1997) Workstation Display of Images From a Prototype Detector for Digital Mammography, RSNA97, infoRAD Presentation

This work was supported by NIH grants CA66202 and CA69452

 Figure 2. System MTF and NPS                                          Figure 3. DQE as a Function of Frequency and Intensity

Figure 4. Digital (left) and Film (right) Images of the Same Subject