James Scheuermann, PhD

Doctoral Thesis: Development of Avalanche Amorphous Selenium for X-Ray Detectors

Abstract: Digital flat panel detectors (FPD) are replacing screen film, computed radiography and the image intensifier in diagnostic and interventional radiology. Current FPDs are limited
in low dose performance due to high noise associated with the a-Si transistor in the readout
electronics. Reducing the electronic noise is neither cost effective nor compatible with
large area fabrication techniques required for medical applications. We propose to overcome
electronic noise through photoconductive avalanche gain (gav) to amplify photo-generated
holes. The primary goal is to replace the a-Si photodiodes used in modern indirect detectors
with a-Se. Under high external bias (ESe = 70 Vµm-1), holes in a-Se can undergo impact
ionization to produce gav.

The development of solid-state avalanche a-Se has been limited by the ability to continuously
apply external bias without causing crystallization of the a-Se through Joule heating
produced from high dark current. Blocking layers must be developed to form a p-i-n junction
to reduce charge injection at the Se-metal interface. A charge transport model is created
to predict effects on temporal performance prior to detector fabrication. All blocking layers
must be deposited at room temperature to prevent crystallization of the a-Se. Novel organic
and inorganic room temperature hole blocking layers are developed on single pixel sensors.
Signal, dark current, ghosting and system lag are evaluated. In parallel, novel fabrication
techniques are proposed to permit the use of high temperature blocking layers and solution
based processing allowing for the a-Se to be deposited in the final step of fabrication. A
successful hole blocking allowed for a gav = 75 ± 5 to be achieved in the first solid-state
avalanche amorphous selenium sensor ever.

The first large area FPD to incorporate avalanche a-Se technology is then fabricated
to evaluate imaging performance. For this proof of concept, the geometry and imaging
parameters were optimized for a mammography system. The system is evaluated using
linear system modeling of the noise power spectrum and detective quantum efficiency which
are compared to experimental results with and without gav. Image quality improves as gav
is increased until the signal overcomes electronic noise thereby demonstrating the potential
improvements a-Se has to offer for digital x-ray imaging.

David A. Scaduto, PhD

Doctoral Thesis: Clinically Translating Contrast-Enhanced X-ray Breast Imaging

Abstract: Contrast-enhanced digital mammography (CEDM) and digital breast tomosynthesis (CEDBT) are radiographic breast imaging techniques currently being investigating to improve the diagnosis of breast cancer. While initial studies of these methods have shown promising results, the techniques remain nascent and are immature for clinical adoption. This work aims to address problems at several stages of the clinical translational pathway.

Methods and metrics to evaluate performance and consistency of digital breast tomosynthesis
(DBT) for clinical imaging and contrast-enhanced applications are developed. Since no
standardized procedures for measuring spatial resolution in DBT have yet been established,
various methods for characterizing spatial/axial resolution are investigated. Separately, a
geometrical calibration phantom and procedure for DBT are developed and used to study system stability necessary for dual-energy CEDBT, in which a low- and a high-energy scan must be acquired under identical geometry.

Improvements to mammographic x-ray detector performance are investigated for low-dose
imaging, necessary for DBT, and dual-energy imaging, a practical realization of contrast enhanced breast imaging. A novel technology capable of avalanche gain and equipped with high resolution field emitter array readout is investigated for low-dose imaging. Independently, a more mature detector technology, the amorphous selenium direct-conversion flat-panel imager, is evolved for dual-energy breast imaging applications. A prototype detector is fabricated and its imaging performance investigated.

Imaging techniques are developed and evaluated to reduce requisite breast compression to
facilitate contrast medium perfusion in the breast. Using a signal-difference-to-noise ratio
(SDNR) model accounting for the increase in scatter radiation due to increased effective breast
thickness, an imaging technique is developed that maintains SDNR without increasing mean
glandular dose. Several practical aspects of the implementation of dual-energy CEDM and
CEDBT protocols are considered.

Finally, the feasibility of CEDM and CEDBT for improved breast cancer diagnosis is
demonstrated in two clinical studies. The diagnostic performance these techniques is assessed by imaging 25 patients and comparing imaging findings with histopathology. Performance of
CEDM and CEDBT are compared for lesion enhancement and margin delineation. Image
subtraction strategies for CEDBT are investigated for optimal lesion enhancement and relative
contrast agent quantification.

Yue-Houng Hu, Ph.D.

Doctoral Thesis: Optimization of Digital Breast Tomosynthesis using a Cascaded Linear System Model

Abstract: Mammography has been shown to be the most and only effective means of screening cancer. Because mammography is a projection x-ray modality and is inherently planar in its imaging representation, efficacy of breast cancer detection is limited by the effect of overlapping tissue, which may obscure or confuse diagnosis of otherwise visible lesions. Digital breast tomosynthesis (DBT) is a three-dimensional (3D) imaging modality and involves the acquisition of x-ray projection images while the tube is rotated through a limited angular range (<50 degrees) over the detector. The projection images are reconstructed into an imaging volume and are viewed in 1 mm thick slices oriented parallel to the detector plane. The presentation of imaging information in 3D allows the removal of overlapping tissue to improve lesion conspicuity. The total glandular dose for a single-view DBT study should be comparable to that of a standard screening mammogram (∼1.5 mGy for a 4 cm breast).

In recent years, a great deal of research has been devoted to DBT on a number of different prototype and commercial units. Currently, there are no widespread standards for DBT acquisition geometry or system settings. Subtleties in these imaging parameters, however, may have profound effects on image quality. To this end, a direct-conversion x-ray detector model and a DBT model were developed based on a cascaded linear systems assumption to investigate the effects of various imaging and system parameters, including acquisition geometry, glandular dose, and detector physics. The primary objective of this thesis is to investigate the physics involved in DBT, understanding the effects of artifact propagation through the imaging process and overlying tissue as a deterministic source of noise, as well as the application the CLSM for the optimization of advanced imaging techniques such as non-uniform angular dose distributions and contrast enhanced (CE) imaging.

First, a previously designed and validated CLSM for amorphous selenium (aSe) direct panel digital mammographic detectors and DBT systems was employed in order to understand the effect of a number physical processes. Characteristic DBT artifact spread and its propagation through DBT imaging chain was investigated as was the effect of overlapping tissue acting as a deterministic source of noise (structural noise).

Finally, the CLSM was modified to include the effect of structural noise on both DM and DBT imaging. A partially isocentric acquisition geometry was modeled to more directly match results garnered from a prototype system. Advanced techniques such as non-uniform angular dose distribution and CE imaging (for both DM and DBT applications) were studied. The modeled results were validated using experimental measurements from the prototype systems, comparing physical, Fourier domain metrics such as noise power spectrum (NPS), modulation transfer function (MTF), and DQE. Additionally, these techniques were optimized to produce maximum object detectability by implementing Fourier domain imaging metrics into a formulation of the ideal observer signal-to-noise ratio (SNR), which for a simple signal known exactly (SKE) background known exactly (BKE) detection task, is known as the detectability index, d’.

Link to fulltext →

Yue-Houng Hu, Ph.D.

Doctor of Philosophy, Biomedical Engineering (2014)

Bo Zhao, Ph.D.

Doctoral Thesis: Breast tomosynthesis with amorphous selenium digital flat panel detector

Abstract: Screening mammography has become the most effective technique for early detection of breast cancer. A major limitation of conventional mammography is the overlapping breast structure that obscures the cancer lesion. Breast tomosynthesis, a three-dimensional (3-D) imaging technique, is a promising method to retrieve depth information. Recent developments in digital mammography make this advanced imaging method possible. In breast tomosynthesis, multiple projection images are acquired when the x-ray tube travels within a limited angular range, typically <50°. The projection images are reconstructed into tomographic slices that are parallel to the detector surface. To keep the total dose comparable to that used in mammography, only a fraction of the conventional mammography exposure is delivered during swift acquisition of each view. The digital detector has to be able to produce high quality images at low exposure and high frame rate. Both the acquisition parameters and reconstruction algorithm will have a direct impact on reconstructed image quality.

The primary objective of this thesis is to investigate breast tomosynthesis from the physics point of view.

Firstly, since the detector for tomosynthesis is modified from that for mammography, the two-dimensional (2-D) imaging performance of the detector was optimized for breast tomosynthesis. The limiting factors that affect breast tomosynthesis were determined using an amorphous selenium (a-Se) Full Field Digital Mammography (FFDM) detector which was characterized at tomosynthesis exposure (as low as 1 mR) by measurement of the resolution and noise characteristics parameters including modulation transfer function (MTF), noise power spectrum (NPS), and detective quantum efficiency (DQE). The high frame rate as experienced in tomosynthesis poses a tremendous challenge to the temporal performance of the detector. The temporal imaging performance was investigated by measurement of the lag and ghosting properties of the detector. The measurement was also performed on a-Se samples in order to understand the inherent physics of lag and ghosting. We concluded that decreasing electronic noise and increasing the x-ray to charge conversion gain in the a-Se layer can be effective methods to improve the detector performance at very low exposures as experienced in tomosynthesis. Careful selection of operation conditions, i.e., electric field and proper reset procedure to clear the trapped charges in the a-Se layer will help improve the temporal performance of the detector.

Secondly, a 3-D cascaded linear system model was developed for breast tomosynthesis to investigate the effects of different imaging system parameters on the reconstructed image quality. The model was validated using linear system parameters of the 3-D NPS and in-plane presampling MTF measured on a prototype breast tomosynthesis system equipped with a-Se digital mammography detector. It was found that the model agreed well with measurement. An ACR phantom was imaged to investigate the effects of limited angular range and detector operational modes on reconstructed image quality. The image quality of the objects (mass and calcifications) on the phantom show good correlation with the quantitative measurement of NPS and MTF. We conclude that the model can be used to accurately predict the imaging performance for varied acquisition and system parameters and ultimately used for the optimization of breast tomosynthesis.