Browsing by Author "Kapadia, Anuj J"

This work is an investigation into collimator designs for a deuterium-deuterium (DD) neutron generator for an inexpensive and compact neutron imaging system that can be implemented in a hospital. The envisioned application is for a spectroscopic imaging technique called neutron stimulated emission computed tomography (NSECT).

Previous NSECT studies have been performed using a Van-de-Graaff accelerator at the Triangle Universities Nuclear Laboratory (TUNL) in Duke University. This facility has provided invaluable research into the development of NSECT. To transition the current imaging method into a clinically feasible system, there is a need for a high-intensity fast neutron source that can produce collimated beams. The DD neutron generator from Adelphi Technologies Inc. is being explored as a possible candidate to provide the uncollimated neutrons. This DD generator is a compact source that produces 2.5 MeV fast neutrons with intensities of 1012 n/s (4π). The neutron energy is sufficient to excite most isotopes of interest in the body with the exception of carbon and oxygen. However, a special collimator is needed to collimate the 4π neutron emission into a narrow beam. This work describes the development and evaluation of a series of collimator designs to collimate the DD generator for narrow beams suitable for NSECT imaging.

A neutron collimator made of high-density polyethylene (HDPE) and lead was modeled and simulated using the GEANT4 toolkit. The collimator was designed as a 52 x 52 x 52 cm3 HDPE block coupled with 1 cm lead shielding. Non-tapering (cylindrical) and tapering (conical) opening designs were modeled into the collimator to permit passage of neutrons. The shape, size, and geometry of the aperture were varied to assess the effects on the collimated neutron beam. Parameters varied were: inlet diameter (1-5 cm), outlet diameter (1-5 cm), aperture diameter (0.5-1.5 cm), and aperture placement (13-39 cm). For each combination of collimator parameters, the spatial and energy distributions of neutrons and gammas were tracked and analyzed to determine three performance parameters: neutron beam-width, primary neutron flux, and the output quality. To evaluate these parameters, the simulated neutron beams are then regenerated for a NSECT breast scan. Scan involved a realistic breast lesion implanted into an anthropomorphic female phantom.

This work indicates potential for collimating and shielding a DD neutron generator for use in a clinical NSECT system. The proposed collimator designs produced a well-collimated neutron beam that can be used for NSECT breast imaging. The aperture diameter showed a strong correlation to the beam-width, where the collimated neutron beam-width was about 10% larger than the physical aperture diameter. In addition, a collimator opening consisting of a tapering inlet and cylindrical outlet allowed greater neutron throughput when compared to a simple cylindrical opening. The tapering inlet design can allow additional neutron throughput when the neck is placed farther from the source. On the other hand, the tapering designs also decrease output quality (i.e. increase in stray neutrons outside the primary collimated beam). All collimators are cataloged in measures of beam-width, neutron flux, and output quality. For a particular NSECT application, an optimal choice should be based on the collimator specifications listed in this work.

Objectives

Methods

Results

Conclusion

Key points

X-ray imaging technology has been used for a multitude of medical applications over the years. The typically measured X-ray transmission data, which records shape and density information by measuring the differences in X-ray attenuation throughout a material, have been used in the imaging modalities of radiography and computed tomography (CT), but there are cases where this information alone is not enough for diagnosis. In contrast, X-ray diffraction (XRD) is another X-ray measurement modality, one that typically does not produce spatially resolved 2D/3D images, but instead investigates small spatial spots for assessing material properties/molecular structures based on scattered X-rays. While XRD measurements of human breast tissue have previously suggested differences between signatures of cancerous and benign tissues, the typical diffraction system architectures do not support fast, large field of view imaging that is necessary for medical applications.In this work, an XRD imaging system was developed that can scan a 15x15 cm2 field of view in minutes with an XRD spatial resolution of 1.4 mm2 and momentum transfer (q) resolution of 0.02 Å-1. An X-ray fan beam was used to collect a 15 cm line of XRD measurements in a single snapshot, while a coded aperture is placed between imaged objects and detector, enabling XRD spectra for individual pixels along the fan beam extent to be recovered from the multiplexed measurement. Simulations were used to identify a suitable geometry for the system, while newly designed phantoms and test objects were used to evaluate the resolution/measurement quality. Upon finishing the design, construction, and characterization of the imaging system, studies on cancerous and benign tissue simulant phantoms were conducted to develop and identify top performing machine learning classification algorithms in a well-controlled study. With a shallow neural network (SNN) developed that achieved ≈99% accuracy on XRD image data, studies progressed to real human tissues. With these developments achieved, the final study was conducted where 22 human breast lumpectomy specimens were scanned and the SNN algorithm was modified for identification of human breast cancer. For 15 primary lumpectomy cases used for training and testing, an accuracy of 99.7% was achieved, with an ROC curve AUC of 0.953 and precision-recall curve AUC of 0.771. On the remaining 7 corner/rare cases present that were held out from initial training/testing (as an external dataset), an accuracy of 99.3% was achieved by the SNN, suggesting high performance along with a need for further representation of rare tissue cases in the training process to improve classifier generalization to new lumpectomy cases. This work demonstrates that fast, large field of view XRD imaging of thin samples on a millimeter spatial scale can be achieved using coded apertures. Further, the work shows that machine learning algorithms can complement this imaging modality by making great use of the multitude of input features available when each image pixel contains a full spectrum of XRD intensity vs angle values, allowing for algorithms to differentiate between cancerous and healthy tissue with higher accuracy (99.7%) compared to simple classification approaches (97.3%). Due to this promising potential, future work should seek to further the technology, by improving the spatial/spectral resolution, scan speed, and adding depth resolution, while applying the technology to useful medical tasks including (but not limited to) intraoperative surgical margin assessment, in-vivo imaging for biopsy vetting, and improved radiation therapy tumor localization.

Oropharyngeal Squamous Cell Carcinoma (OPSCC) is by far the most predominant form of head and neck cancer in the United States. The survival rate for OPSCC is very high, which, while fortunate, yields many patients who are left with the late term toxicities consequent of their treatment. This project aimed to use patient-reported outcome (PRO) data from two sources – the PRO-CTCAE and the QLQ-C30 – along with the dosimetric data of patients that have already been treated, in order to characterize retrospectively a relationship between patient dosimetric data and the severity of response of PRO data. In particular, PRO data was used as a way to characterize the severity of patient-experienced xerostomia and dysphagia. Additionally, this data was used to fit the radiobiological parameters for two normal tissue complication probability (NTCP) models: the Lyman-Kutcher-Burman (LKB) model, and the Relative Seriality (RS) model. Overall, it was found that the PRO-CTCAE data was more robust than the QLQ-C30 data in its characterization. Based on the PRO-CTCAE data, the V52 (volume which receives at least 52 Gy) of the combined constrictors and the V59 of the superior pharyngeal constrictor show the strongest relationship with patient-reported dysphagia. Additionally, the V27 of the contralaterals and the V12 of the contralateral parotid show the strongest relationship with patient-reported xerostomia. Furthermore, it was found that the dose response curves for both NTCP models fit the data with similar accuracy.

Proposed is an investigation into the effect of lower-energy source on the tumor representation of an image acquired by a neutron-based spectroscopic imaging modality, Neutron Stimulated Emission Computed Tomography (NSECT).

The NSECT experiments were performed at a shielded neutron source of the Triangle Universities Nuclear Laboratory (TUNL), which was proficient at creating neutron beams of energy up to 20MeV. However, this neutron generator is not feasible for clinical use due to its large size. Smaller compact sources such as deuterium-deuterium (DD) neutron generators are attractive alternatives that can produce neutrons of sufficient energy to stimulate isotopes of interest in the human body. However, DD generator is not competent at producing neutrons of high energy. Thus, the focus of this work is to evaluate the effect of lower-energy neutrons, such as 2.5MeV and 3.2MeV, on the NSECT images.

The experiments were modeled and simulated in this work using a Monte Carlo toolkit, Geant4. In Geant4 space, an anthropomorphic phantom of cancerous tissue was scanned by a simulated neutron source. During scanning, the phantom was translated to cover the whole field of view (FOV) and rotated over 180 degrees for the purpose of tomographic imaging. Neutrons and gammas emitted were captured by a virtual detector, which could identify the energy and position of each particle. Information of position and energy of gammas detected resulted in a sinogram for an array of energies, created by selecting the energy characteristic to a specific element. Using the sinograms, two-dimensional maps of the spatial concentration of the element could be reconstructed through a reconstruction algorithm and the elemental concentration revealed the internal geometry of the phantom.

Images were generated when the phantom was scanned by 5MeV, 3.2MeV and 2.5MeV neutron sources. Comparison of tumor parameters in these images indicates that a neutron source of lower energy could degrade the tumor representation in a NSECT image on the aspects of concentration, brightness and underestimation of the tumor size. Then further investigations with 50,000, 100,000, 150,000 and 200,000 neutron events were performed respectively in the 3.2MeV-source case and 2.5MeV-source case in order to test whether the number of neutrons is correlated to the quality of the reconstructed images. Improvement of tumor representation, for example, a clearer tumor region and more accurate tumor size information, shows that increase in the number of incident neutrons has a positive effect on the reconstructed image. This work demonstrates the effect that low energy neutrons have on the image and verifies the feasibility of using low-energy neutrons as the source in NSECT breast imaging.

Introduction: RECA (Radiotherapy Enhanced with Cherenkov photo-Activation) is a novel treatment that induces a synergistic therapeutic effect by combining conventional radiation therapy with phototherapy using the anti-cancer and potentially immunogenic drug, psoralen. Psoralen is photo-activated in-situ by UVA (UltravioletA, 320-400nm) Cherenkov Light (CL), produced in tissue directly by the treatment beam. In this study, we develop methods to image and quantify relative CL production (primarily in the UVA range) from a range of tissue and phantom materials upon photon irradiation. These methods are further applied to identify a tissue-equivalent optical phantom, mimicking CL production in the UVA range, in order to facilitate further RECA experiments.

Methods: The imaging system included a deep-cooled, high-sensitivity CCD camera, equipped with either a visible range lens (sensitive to 400-700nm photons) or a UVA-compatible lens assembly and a band-pass filter (sensitive to 320-400nm photons). CL emission was quantified in bulk tissue samples, solid waters (SW brown and white), and agarose gels in a series of experiments. The samples and imaging equipment were placed in a dark, light-blocking chamber to avoid contamination from other light sources. In addition, the camera was carefully positioned with respect to the LINAC head and was also shielded using lead bricks to minimize radiation noise.The samples were then irradiated with clinical photon beams, while simultaneously being imaged by the camera.

Results: In the visible range, solid water had similar CL emission to that from bulk tissue samples (34% less than the maximum and 44% higher than the minimum UVA emitting tissue). A 25% reduction in radiation noise in the UVA spectrum was achieved using lead block shielding of the camera. In the UVA range at 15MV, white SW emitted 66±5%, 64±5% and 76±3% less UVA than chicken, pork loin and pork belly respectively. Similar under-response was observed at 6MV. Brown SW had 21±8% less UVA emission than white SW at 15MV, and no significant emission at 6MV. Agarose samples (1% by weight) doped with 250ppm India Ink exhibited equivalent UVA CL emission to chicken breast (within 8%).

Conclusion: The results confirm that for the same absorbed dose, SW emits lessUVA light than the tissue samples, indicating that prior in-vitro studies utilizing SW as the CL-generating source may have underestimated the RECA therapeutic effect. Agarose gel doped with 250ppm India Ink is a convenient tissue equivalent phantom for further work.

TBD

Neutron stimulated emission computed tomography (NSECT) is a non-invasive, tomographic imaging technique with the ability to locate and quantify elemental concentration in a tissue sample. Previous studies have shown that NSECT has the ability to differentiate between benign and malignant tissue and diagnose liver iron overload while using a neutron beam tomographic acquisition protocol followed by iterative image reconstruction. These studies have shown that moderate concentrations of iron can be detected in the liver with moderate dose levels and long scan times. However, a low-dose, reduced scan time technique to differentiate various liver diseases has not been tested. As with other imaging modalities, the performance of NSECT in detecting different diseases while reducing dose and scan time will depend on the acquisition techniques and parameters that are used to scan the patients. In order to optimize a clinical liver imaging system based on NSECT, it is important to implement low-dose techniques and evaluate their feasibility, sensitivity, specificity and accuracy by analyzing the generated liver images from a patient population. This research work proposes to use Monte-Carlo simulations to optimize a clinical NSECT system for detection, localization, quantification and classification of liver diseases. This project has been divided into three parts; (a) implement two novel acquisition techniques for dose reduction, (b) modify MLEM iterative image reconstruction algorithm to incorporate the new acquisition techniques and (c) evaluate the performance of this combined technique on a simulated patient population.

The two dose-reduction, acquisition techniques that have been implemented are; (i) use of a single angle scanning, multi-detector acquisition system and (ii) the neutron-time resolved imaging (n-TRI) technique. In n-TRI, the NSECT signal has been resolved in time by a function of the speed of the incident neutron beam and this information has been used to locate the liver lesions in the tissue. These changes in the acquisition system have been incorporated and used to modify MLEM iterative image reconstruction algorithm to generate liver images. The liver images are generated from sinograms acquired by the simulated n-TRI based NSECT scanner from a simulated patient population.

The simulated patient population has patients of different sizes, with different liver diseases, multiple lesions with different sizes and locations in the liver. The NSECT images generated from this population have been used to validate the liver imaging system developed in this project. Statistical tests such as ROC and student t-tests have been used to evaluate this system. The overall improvement in dose and scan time as compared to the NSECT tomographic system have been calculated to verify the improvement in the imaging system. The patient dose was calculated by measuring the energy deposited by the neutron beam in the liver and surrounding body tissue. The scan time was calculated by measuring the time required by a neutron source to produce the neutron fluence required to generate a clinically viable NSECT image.

Simulation studies indicate that this NSECT system can detect, locate, quantify and classify liver lesions in different sized patients. The n-TRI imaging technique can detect lesions with wet iron concentration of 0.5 mg/g or higher in liver tissue in patients with 30 cm torso and can quantify lesions at 0.3 ns timing resolution with errors ≤ 17.8%. The NSECT system can localize and classify liver lesions of hemochromatosis, hepatocellular carcinoma, fatty liver tissue and cirrhotic liver tissue based on bulk and trace element concentrations. In a small patient with a torso major axis of 30 cm, the n-TRI based liver imaging technique can localize 91.67% of all lesions and classify lesions with an accuracy of 88.23%. The dose to the small patient is 0.37 mSv a reduction of 39.9% as compared to the NSECT tomographic system and scan times are comparable to that of an abdominal MRI scan. In a bigger patient with a torso major axis of 50cm, the n-TRI based technique can detect 75% of the lesions, while localizing 66.67% of the lesions, the accuracy of classification is 76.47%. The effective dose equivalent delivered to the larger patient is 1.57 mSv for a 68.8% decrease in dose as compared to a tomographic NSECT system.

The research performed for this dissertation has two important outcomes. First, it demonstrates that NSECT has the clinical potential for detection, localization and classification of liver diseases in patients. Second, it provides a validation of the simulation of a novel low-dose liver imaging technique which can be used to guide future development and experimental implementation of the technique.

The overarching goal of this project was to investigate organ dose assessment and variability using Monte Carlo methods to study two areas of medical imaging – fluoroscopy and computed tomography. Namely, these studies were intended to (1) provide estimates of the dose incurred by fluoroscopy-guided spinal injection procedures, and (2) investigate dose heterogeneity in chest and abdominopelvic computed tomography (CT) scans for a range of patient sizes. Fluoroscopy dose estimates were calculated using GEANT4, by recreating the patient procedures of six lumbar-sacral epidural injections. Computed tomography dose was estimated with a GPU-accelerated Monte Carlo package, MCGPU. Both simulations used a library of digital human (XCAT) phantoms, which were previously derived from real-patient CT scans. The fluoroscopy simulations suggest that smaller patients have a higher effective dose per dose area product, and the overall results agreed with previous experimental measurements. Variation of absorbed dose within a given organ was calculated for chest and abdominopelvic CT protocols. It was found that the 95th percentile dose can be over 11 times the mean organ dose in pediatric and adult phantoms. Furthermore, if the organ dose is calculated using only voxels within the beam or all the voxels within an organ, the result can change the result by a factor of 8. The change in dose was found to be higher for organs that have smaller fractions within the beam. Several models of tissue-weighted dose were also investigated, following similar methods to those used for effective dose. It was found that these tissue-weighted dose calculations can vary by up to 13% depending on whether the out of field dose is included. We also found that the results were not significantly affected by the pitch or projections per rotation. The results have shown that dose-volume details may be hidden by average dose estimates and suggested the need to consider intra-organ dose heterogeneity in CT dose calculations, particularly in the case of sensitive tissues (e.g., bone marrow) and populations (e.g., pediatric).

Proposed is a method for investigating optimal acquisition parameters in NSECT, neutron stimulated emission computed tomography, for good image quality and low dose for diagnosing liver and breast cancers. These parameters include the number of angles, number of translations per angle, beam width, and beam width spacing. These parameters will affect dose, which will increase with increasing total neutron flux. Therefore, a balance must be achieved for the parameters mentioned above, to yield a desired dose limit and tolerable spatial resolution necessary for liver and breast cancer diagnosis.

Using Monte Carlo simulation toolkit GEANT4, the effects of beam spread due to neutron elastic scatter was explored. Then, a geometrical water torso phantom with slanted edge solid iron phantom was run for different acquisition parameters, and an MTF was taken to determine resolution for each set. For dose considerations, two anthropomorphic voxelized phantoms, one with liver cancer lesions, and one with breast cancer lesions, were scanned with the same parameter sets, and organ doses and DVHs, dose volume histograms, was computed for each set. In addition, images of the phantom in the lesion plane were reconstructed for those parameter sets showing best resolution and lowest dose.

It is found that beam spread due to elastic scatter off of Hydrogen atoms is negligible for all beam widths. For optimal resolution in the primary breast phantom, it was found that acquisition parameters of a 5 mm beam, with no gaps, with any of the five angles provided the superior resolution. For the optimal resolution in the liver, it was found that down sampling angles and introducing gaps between projections greatly affected image accuracy and resolution. Also, the 5 mm beam width provided better geometrical accuracy, but the 1 cm bream width provided slightly better resolution.

Organ doses are computed for the primary organ and organs at risk for each parameter set at 500 K neutrons per projection. For a scan of the full volume of the liver, liver organ doses ranged from 25.83 to 0.19 mSv. For the same scan, the organ doses for the heart ranged from 0.18 to 0.05 mSv. For a scan with the same pool of acquisition parameters of the full volume of the breast, breast organ doses ranged from 49.87 to 0.38 mSv. Furthermore, the DVHs for both scans showed a very steep drop-off at low dose bins for secondary organs at risk and a reasonable drop-off for the primary organ.

In choosing the optimal acquisition parameters using both resolution and dose, a metric equal to resolution times dose is used, in which low values are optimal. An upper threshold for the metric was chosen based on dose values in currently used medical imaging modalities. A pool of optimal parameter sets was then identified using the metric. To further identify the optimum, a metric estimating geometrical accuracy of the reconstructed square was used. For the breast scan, the optimal parameter set was a 1 cm beam width, with 0 mm a gap, with 12 angles. For the liver scan, the optimal parameter set was a 1 cm beam width, with a 0 mm gap, with 36 angles.

Finally, reconstructed images of the anthropomorphic scans using the super sampled geometry in the liver scan showed one lesion, using images of iron and phosphorous. With more degraded image quality, reconstructed images of the breasts using the super sampled geometry showed only the three cm lesion accurately. The images reconstructed from the optimal set identified for liver scans also showed the larger lesion, except with some noise from the presence of iron and phosphorous in other organs. The images reconstructed from the optimal set identified for the breast scans had a similar result to that of the super-sampled case, albeit with lower contrast. The least sampled case for both scans were found to be diagnostically useless. From these anthropomorphic images, this work demonstrates that in-vivo imaging of breast and liver cancers may be potentially possible with NSECT at a low dose.

Coherent scatter spectral imaging has been demonstrated as an effective way to classify healthy and malignant breast tissues. Previously in our group, data acquired using sectioned, lumpectomy specimens obtained from surgical pathology have been used to demonstrate the efficacy of this imaging method. Although effective in its current state, the system has not been optimized for use with these types of specimens (i.e. tissue types and thicknesses). Specimens obtained from lumpectomies often vary in thickness (up to 3 mm). The current X-ray tube operating parameters have been considered excessive for these tissues based on heating of the tube’s anode and the unnecessary, high quality of resulting spectra. The purpose of this work was to optimize our spectral imaging system to maintain accurate and consistent results of sectioned lumpectomy specimens while simultaneously maximizing system throughput by reducing the power requirements of the imaging system.

Teflon, adipose breast tissue, and malignant breast tissue were scanned using different combinations of X-ray source parameters (70-125 kVp, 25-500 mAs) to obtain a coherent-scatter diffraction spectrum for each measurement. Cross correlation was performed on the measured spectra to compare their quality against known, ground truth spectra from literature. In addition, a classification algorithm was developed to classify our measured spectra as one of four tissue types (adipose, normal, fibroglandular, and cancer). The locations of the spectral peaks were used to distinguish cancer from adipose and normal (50/50 fibroglandular/adipose) tissue, followed by a weighted cross correlation method used to distinguish cancer from fibroglandular tissue. Classification performance was assessed across all acquisition protocols to evaluate accuracy.

The optimal setting was identified as the minimal power supplied to the X-ray tube that resulted in the highest correlation to the ground truth spectra. The optimal setting was identified at 115 kVp, 100 mAs when using the raw spectra and 95 kVp, 50 mAs after processing the spectra. These settings result in an increase in system efficiency of at least 400% (at 115kVp, 100 mAs) compared to our current system operating protocol. Finally, the optimized system was tested using a new, unknown tumor specimen obtained from a preserved lumpectomy section.

This study successfully demonstrates the optimization of a new coherent scatter spectral imaging system based on classification using cross correlation and a weighted cross correlation method. The efficiency improvement obtained through the work allows for higher system throughput, thereby allowing enhanced data collection with shorter scans or scanning the specimens at higher resolutions.

In many parts of the developing world, easy access to volumetric imaging is not available. A Lower-Cost CT setup was proposed and found feasible by Dobbins et al., but was not yet optimized to maximize image quality while minimizing radiation dose to the patient. A combination of spectrum modeling and Monte Carlo simulations were used to compare x-ray beam parameters to determine which combination was optimal. The beam parameters considered were filter type, filter thickness, and tube peak kilovoltage (kVp). The optimization was based on the differential signal-to-noise ratio (dSNR) and the dose, using a factor referred to as dSNR Efficiency. After the three different filter materials at three different thicknesses were compared across five different kVp values, it was determined that one half-value-layer (HVL) of copper was the best filter type and thickness to achieve maximum image quality for minimal patient dose.

In order to verify that a good dSNR efficiency using the spectrum modeling and Monte Carlo meant that the system would provide useable images, the extended cardiac-torso (XCAT) phantom was used to simulate CT images, using a ray tracer, and estimate dose, using a full LCCT Monte Carlo simulation. The ray-tracer produced x-ray projections of the XCAT phantom which were then run through a Feldkamp reconstruction algorithm to produce CT images. The full LCCT Monte Carlo simulation modeled the LCCT setup using the XCAT phantom to determine the dose to the patient. From the reconstructed CT images, it was determined that for image studies that favor air contrast higher kVp values, such as 140 kVp, are optimal. For studies that favor bone contrast, however, the lower kVp values, such as 60 or 80 kVp, are optimal. For 140 kVp images, the average effective dose, calculated using the ICRP 103 protocol, was mSv per mAs. The average effective dose for 60 kVp was mSv per mAs, and the average effective dose for 80 kVp was mSv per mAs. Further work is needed to determine optimal mAs values for different imaging studies. The LCCT setup can provide volumetric imaging to developing parts of the world that currently have no volumetric imaging, which would greatly improve the quality of readily available medical care.

Our research group has previously described the development and testing of a coherent-scatter spectral imaging system for identification of cancer using surrogate phantoms, formalin-fixed pathology tissues and, more recently, surgically resected breast tumor. Here we present the implementation of a Monte-Carlo simulation tool for optimization of the imaging system.

MC-GPU, a GPU-enabled Monte Carlo software was modified and used to simulate X-ray diffraction experiments for combinations of X-ray spectra (tungsten and molybdenum anode), kV (15-150), filtration (material and thickness) and phantom geometry and material (normal, adipose, fibroglandular, and cancerous breast tissue). For each combination, a simulated measurement of contrast-to-noise (CNR), signal strength and object detectability were assessed.

Examination of Monte Carlo simulations showed optimal spectrum characterization strategies that exploit spectral and filter characteristics to increase material identification probabilities via momentum transfer measurement. Increased detectability was shown with molybdenum energy spectra, and a higher CNR metric was observed to show better pathological assessments and findings of cancer.

This work demonstrates the utility of Monte Carlo methods and MCGPU in optimizing coherent scatter imaging systems and can be used to provide insightful information regarding the design of coherent scatter imaging systems for material classification breast tissue types.

A need exists for an intraoperative margin assessment tool that can improve the efficiency of pathological assessment by efficient classification of excised tissue boundaries. To address this need, we have developed a system that combines x-ray diffraction imaging with a neural network classifier to achieve high-resolution, high-accuracy cancer imaging. The system’s x-ray diffraction imaging component is constructed using a Coded Aperture Coherent Scatter Spectral Imaging (CACSSI) arrangement, which provides tissue-specific molecular-scale contrast, and processes this data through a multi-layer perceptron neural network. Our current system shows upwards of 84% classification accuracy compared to the accepted standard of pathological assessment. Compared to our previous system, this is a 10% improvement in classification accuracy and is achieved in less than a third of the time needed by our previous system.

Scatter in medical imaging is typically cast off as image-related noise that detracts from meaningful diagnosis. It is therefore typically rejected or removed from medical images. However, it has been found that every material, including cancerous tissue, has a unique X-ray coherent scatter signature that can be used to identify the material or tissue. Such scatter-based tissue-identification provides the advantage of locating and identifying particular materials over conventional anatomical imaging through X-ray radiography. A coded aperture X-ray coherent scatter spectral imaging system has been developed in our group to classify different tissue types based on their unique scatter signatures. Previous experiments using our prototype have demonstrated that the depth-resolved coherent scatter spectral imaging system (CACSSI) can discriminate healthy and cancerous tissue present in the path of a non-destructive x-ray beam. A key to the successful optimization of CACSSI as a clinical imaging method is to obtain anatomically accurate phantoms of the human body. This thesis describes the development and fabrication of 3D printed anatomical scatter phantoms of the breast and lung.

The purpose of this work is to accurately model different breast geometries using a tissue equivalent phantom, and to classify these tissues in a coherent x-ray scatter imaging system. Tissue-equivalent anatomical phantoms were designed to assess the capability of the CACSSI system to classify different types of breast tissue (adipose, fibroglandular, malignant). These phantoms were 3D printed based on DICOM data obtained from CT scans of prone breasts. The phantoms were tested through comparison of measured scatter signatures with those of adipose and fibroglandular tissue from literature. Tumors in the phantom were modeled using a variety of biological tissue including actual surgically excised benign and malignant tissue specimens. Lung based phantoms have also been printed for future testing. Our imaging system has been able to define the location and composition of the various materials in the phantom. These phantoms were used to characterize the CACSSI system in terms of beam width and imaging technique. The result of this work showed accurate modeling and characterization of the phantoms through comparison of the tissue-equivalent form factors to those from literature. The physical construction of the phantoms, based on actual patient anatomy, was validated using mammography and computed tomography to visually compare the clinical images to those of actual patient anatomy.

This study intends to validate the sensitivity and specificity of coded aperture coherent scatter spectral imaging (CACSSI) by comparison to clinical histological preparation and pathologic analysis methods currently used for the differentiation of normal and neoplastic breast tissues. A composite overlay of the CACSSI rendered image and pathologist interpreted, stained sections validate the ability of coherent scatter imaging to differentiate cancerous tissues from normal, healthy breast structures ex-vivo. Via comparison to the pathologist annotated slides, the CACSSI system may be further optimized to maximized sensitivity and specificity for differentiation of breast carcinomas.

One of the challenges facing clinical practice today is intra-operative margin detection in breast conserving surgeries (BCS) or lumpectomy procedures. When a surgeon removes a breast tumor from a patient during a BCS procedure, the surgically excised tissue specimen is examined to see whether it contains a margin of healthy tissue around the tumor. A healthy margin of tissue around the tumor would indicate that the tumor in its entirety has been removed. On the other hand, if cancerous tissue is at the surface of the specimen, that would indicate that the tumor may have been transected during the procedure, leaving some residual cancerous tissue inside the patient. The most effective intra-operative real-time margin detection techniques currently used in clinical practice are frozen section analysis (FSA) and touch-prep cytology. These methods have been shown to possess inconsistent accuracy, which result in 20% to 30% of BCS patients being called back for a repeat BCS procedure to remove the residual tumor tissue. In addition these techniques have been shown to be time-consuming--requiring the operating room team to have to wait at least 20 minutes for the results. Therefore, there is a need for accurate and faster technology for intra-operative margin detection.

In this dissertation, we describe an x-ray coherent scatter imaging technique for intra-operative margin detection with greater accuracy and speed than currently available techniques. The method is based on cross-sectional imaging of the differential coherent scatter cross section in the sample. We first develop and validate a Monte Carlo simulation of coherent scattering. Then we use that simulation to design and test coherent scatter computed tomography (CSCT) and coded aperture coherent scatter spectral imaging (CACSSI) for cancerous voxel detection and for intra-operative margin detection using (virtual) clinical trials. Finally, we experimentally implement a CACSSI system and determine its accuracy in cancer detection using tissue histology.

We find that CSCT and CACSSI are able to accurately detect cancerous voxels inside of breast tissue specimens and accurately perform intra-operative margin detection. Specifically, for the task of individual cancerous voxel detection, we show that CSCT and CACSSI have AUC values of 0.97 and 0.94, respectively. Whereas for the task of intra-operative margin detection, the results of our virtual clinical trials show that CSCT and CACSSI have AUC values of 0.975 and 0.741, respectively. The gap in spatial resolution between CSCT and CACSSI affects the results of intra-operative margin detection much more than it does the task of individual cancerous voxel detection. Finally, we also show that CSCT would require on the order of 30 minutes to create a 3D image of a breast cancer specimen, whereas CACSSI would require on the order of 3 minutes.

These results of this work show that coherent scatter imaging has the potential to provide more accurate intra-operative margin detection than currently used clinical techniques. In addition, the speed (and therefore low scan duration: 3 min) of CACSSI, along with its ability to automatically classify cancerous tissue for margin detection means that coherent scatter imaging would be much more cost-effective than the clinical techniques that require up to 20 minutes and a trained pathologist. With the cancerous voxel detection accuracy of a 0.94 AUC and scan time of on the order of 3 minutes demonstrated for coherent scatter imaging in this work, coherent scatter imaging has the potential to reduce healthcare costs for BCS procedures and rates of repeat BCS surgeries. The accuracy for CACSSI can be considerably improved to match CSCT accuracy by improving its spatial resolution through a number of techniques: incorporating into the CACSSI reconstruction algorithm the ability to differentiate noise from high frequency signal so that we can image with higher frequency coded aperture masks; implementing a 2D coded aperture mask with a 2D detector; or acquiring additional angles of projection data.

Although mammography is the gold standard for early screening for breast cancer, there is a need in improving its specificity. X-ray diffraction (XRD) has shown the ability to detect cancer based on its molecular properties; however, most commercial XRD systems focus on very thin targets using very low x-ray energies. To improve the imaging capabilities for thicker targets, we have developed XRD imaging systems capable of running at diagnostic X-ray energies. In this work, we evaluate the performance of our XRD system at two source configurations: 20 keV (mammography) and 60 keV (radiography), to understand and quantify the trade-off of between higher Rayleigh scatter cross section versus reduced penetration depth.

An XRD system was built using a Bremsstrahlung X-ray source, an energy discriminating X-ray detector, and customizable geometry. XRD scans were performed using adipose, fibroglandular, and carcinoma surrogate targets at two mean energies – 20 keV and 60 keV. The 20 keV configuration used a molybdenum filter and 2-mm collimated beam, whereas the 60 keV configuration used a tungsten filter with 1-mm diameter X-ray beam. Each target was scanned 5-10 times to evaluate measurement uncertainty. XRD spectra, normalized to mAs, were extracted from the detected signal and compared against known diffraction data for each material. System performance was evaluated using signal-to-noise ratio (SNR), average-percent-difference (APD) and uncertainty in each measurement.

An analytical calculation was done to test the effects of attenuation on the 20 keV and 60 keV configurations using elastic scatter coefficients and total attenuation coefficients excluding elastic scatter for each energy for breast tissue. Using the mean-free-path and the calculated exponential attenuation, an estimate of the surviving fraction of x-rays that undergo Rayleigh scatter was calculated.

The 20 keV configuration showed 8.25 SNR, 95.96% accuracy, and 1.46% uncertainty. The 60 keV configuration showed 7.05 SNR, 94.72% accuracy, and 0.96% uncertainty. Overall, the 20 keV configuration showed 13.21% improvement in SNR compared to the 60 keV configuration. The analytical estimation of the surviving fraction of x-rays having undergone Rayleigh scatter showed that at the average compressed breast thickness of 4.4 cm – 4.8 cm, the 20 keV system performed on par with the 60 keV system while the surviving fraction of Compton scatter, which is detrimental for XRD analysis, has decreased with energy. Calculating the ratio of the surviving fraction of Rayleigh scattered x-rays to Compton scattered x-rays to estimate the SNR with depth, 20 keV x-rays showed a consistent advantage. Through this, we demonstrated the viability of low-energy XRD imaging for characterizing breast tissues. The 20 keV configuration presents a viable method to characterize breast tissues at energies relevant to mammography, representing a potential method to improve specificity in mammography.

Breast cancer surgical treatment options prove effective at treating breast cancer and reducing breast cancer death rates, prompting women to elect to surgically excise the tumor via a lumpectomy procedure. Despite women choosing lumpectomy over a mastectomy in 60% of cases, and despite the general effectiveness of the lumpectomy procedures, patient recall rates due to missed cancerous tissue are unfavorably high and variable at approximately 25% nationally. In addition, drawn-out processing times due to pathology assessment contribute to sub-optimal patient care and overly onerous costs and workload for hospitals. Therefore, it is the focus of this work to develop, evaluate, and refine a novel imaging modality to aid pathologists and pathologists’ assistants in assessing breast cancer via a more quantified means that would eventually lower the recall rates in breast cancer surgery.

Through previous work, we established a Coded Aperture Coherent Scatter Spectral Imaging (CACSSI) system, characterized several facets of the imaging setup, and evaluated its utility in breast cancer applications. Using Monte Carlo simulations, anthropomorphic breast phantoms, and human breast tissue specimens, we previously validated CACSSI’s utility in differentiating breast tissue types in a clinically relevant manner, which makes the system a promising candidate to act as a supplementary tool to implement in the pathology workflow. This work continues the previous research by applying and implementing the tissue classification ability within a short, clinically feasible timeframe (5-30 minutes) and demonstrating utility in a broader population of 12 patient-derived lumpectomy specimens. The work presented herein is broken into three subprojects: (1) Assessing various characterizations of the system (i.e. the background signal effects, the detector temperature-dependent response, the precision in consecutive scans, and the effect of formalin-fixation) to demonstrate its feasibility for the cancer detection/classification tasks; (2) Evaluating the accuracy of the system in a population of 12 excised breast tissue specimens while establishing and implementing the scan room procedures across multiples specimens; and (3) Utilizing a concurrently-developed classification scheme to more thoroughly compare the system’s fidelity and robustness against pathology-assessed outcomes, which currently serve as the clinical gold standard for breast cancer judgments.

The typical workflow included Surgical Pathology preparing the surgically excised specimens and indicating via palpation the location of the tumor. The specimen, with the preliminary tumor location marked, was then scanned in our imaging system, and spectral scatter signatures were obtained at multiple locations within the tissue. The resulting form factor spectra were then compared with reference spectra to classify the tissue as cancerous or non-cancerous (healthy). The tissue classification mapping was compared against the indicated tumor area or against pathology-stained microslides for verification of tumor diagnosis.

Formalin-fixation was found inconsequential for tissue classification, with fresh-to-formalin-fixed spectra correlations of 0.9782 and 0.9881 over 10 spot scans each for healthy and cancer tissue, respectively. The spatial resolution of the system was found to be 1.5 mm in the lateral direction and 5 mm along the beam path. Our CACSSI system was able to distinguish between cancerous and healthy areas in the tissue slices in a consistent manner, and the system was, on average, 82.93% accurate for the initial classification scheme and 83.70% accurate using a more quantitative classification scheme. Furthermore, we were able to achieve these results in a clinically relevant timeframe on the order of 30 minutes, integrating into the pathology workflow with minimal interruption. Aggregating these results CACSSI will continue to be developed for use as a clinical imaging tool in breast cancer assessment and other diagnostic purposes.