published_or_final_version / Medicine / Master / Doctor of Medicine

The best way to conquer breast cancer is early detection of the disease. Research studies show that earlier detection results in the increase of life span of the affected person. Traditional two-dimensional mammography is the most prevalent method used in detecting breast cancer. Recently, a three-dimensional digital breast tomosynthesis (DBT) system has been created, which is hopeful to surpass the technology of traditional mammography systems. The DBT system can provide three-dimensional information, allowing physicians to reduce the amount of false negative screening in addition to better monitoring of breast cancer and to catch lesions that may be otherwise cancerous. In this research, the View Angle (VA) and number of projection images (N) were investigated and compared with parallel imaging configurations using two reconstruction algorithms, including Back Projection (BP) and Shift-And-Add (SAA). Modulation Transfer Function (MTF) analyses were conducted with both algorithms, in order to determine which method displayed better image qualities to ultimately improve the detection of breast cancer.

No description available.

Once diagnosed, cancer is treated by surgical resection, chemotherapy, radiation therapy, or a combination of these therapies. It is intuitive that physically and completely removing a solid tumor would be an effective treatment. A complete resection of the tumor mass, defined by surgical margins that are clear of neoplasia, is prognostic for a decreased chance of cancer recurrence and an increased survival rate. In practice, complete resection is difficult. A surgeon primarily has only their senses of touch and sight to provide "real-time" guidance in the removal of a tumor while in the operating room. Preoperative imaging can guide a surgeon to a tumor but does not give a continuous update of surgical progress. Intraoperative pathology is limited to a few slides worth of samples: a product of its time-consuming nature and the limited time a patient can remain under general anesthesia. Technologies to guide a surgeon in effecting complete resection of a tumor mass during the surgical procedure would greatly increase cancer survival rates by lowering rates of cancer recurrence; such a technology would also reduce the need for follow-up chemotherapy or radiation therapy. Here, we describe a prototype instrumentation system that can provide intraoperative guidance with exogenous optical contrast agents. The instrumentation combines interactive point excitation, local spectroscopy, and widefield fluorescence imaging to enable low-cost surgical guidance using FDA-approved fluorescent dyes, semiconductor quantum dots (QDs), or surface-enhanced Raman scattering (SERS) nanoparticles. The utility of this surgical system is demonstrated in rodent tumor models using an FDA-approved fluorescent dye, indocyanine green (ICG), and is then more extensively demonstrated with a pre-clinical study of spontaneous tumors in companion canines. The pre-clinical studies show a high sensitivity in detecting a variety of canine tumors with a low false positive rate, as verified by pathology. We also present a fundamental study on the behavior of quantum dots. QDs are a promising fluorophore for biological applications, including as a surgical contrast agent. To use QDs for in vivo human imaging, toxicity concerns must be addressed first. Although it is suspected that QDs may be toxic to an organism based on the heavy-metal elemental composition of QDs, overt organism toxicity is not seen in long-term animal model studies. We have found that some reactive oxygen species (ROS) generated by the host inflammatory response can rapidly degrade QDs; in the case of hypochlorous acid, optical changes to the QDs are suggestive of degradation occurring within seconds. It is well-known that QDs are sequestered by the immune system when used in vivo---we therefore believe that QD degradation through an inflammatory response may represent a realizable in vivo mechanism for QD degradation. We demonstrate in an in vitro cell culture model that immune cells can degrade QDs through ROS exposure. Knowledge of the degradative processes that QDs would be subject to when used in vivo informs on adaptations that can be made to the QDs to resist degradation. Such adaptations will be important in developing QD-based contrast agents for image guided surgery.

Demographic data indicate that 60% of 6.7 million annual global cancer mortalities and 54% of 10.8 million new patients are in developing nations, unable or unwilling to avail of invasive screening tests that are the current norm. For most cancers, survival rate is strongly dependent on early detection, highlighting the need for improved screening methods. Studies have shown that cancers can be identified based on distinct sub-cellular morphological features and expression levels of specific molecular markers. Since 85% of cancers are known to originate in the epithelium, portable in vivo imaging techniques providing sub-cellular detail in tissue up to depths of 250 μm could help improve access to biopsy-free examination in low-infrastructure environments. The resultant early detection could dramatically improve patient prognosis, while reducing screening costs, treatment delay, and occurrences of unnecessary and potentially harmful medication. This dissertation investigates handheld instrumentation for laser-scanning confocal microscopy (LSCM) and its applicability to breast cancer detection and subsequent image-guided management. LSCM allows high-resolution mapping of spatial variations in refractive index or tumor marker expression within a single cell layer situated few hundred micrometers beneath the tissue surface. The main challenge facing miniaturization lies in the mechanism of beam deflection across the sample. The first part of the dissertation presents a fast, large-angle, high-reflectivity two-axis vertical comb driven silicon micromirror fabricated by a novel method compatible with complementary metal-oxide-semiconductor processing employed in the semiconductor industry. The process enables integration of rotation sensors on the chip to adaptively correct for aberrations in beam scanning while significantly reducing fabrication costs and barriers to market acceptance. The second part of the dissertation explores the integration of this micromirror with other optical and electronic components into a handheld laser-scanning confocal microscope. Applicability of the probe to epithelial breast cancer screening via reflectance and fluorescence imaging is investigated. Finally, enhanced imaging modalities based on the micromirror are presented. 3D cellular-level in vivo imaging via rapid swept-source optical coherence tomography is demonstrated. A method for “objective-less” microendoscopy, potentially resulting in substantially reduced probe dimensions, employing reflective binary-phase Fresnel zone plates monolithically integrated on the surface of the micromirror is presented. / text

Radiological imaging already has a key role in the detection and management of patients with metastatic colorectal cancer (mCRC). With the evolution of personalised medicine there is a need for non-invasive imaging biomarkers that can detect early tumour response to targeted therapies. Translation from bench to bedside requires a multicentre approach that follows an agreed development roadmap to ensure that the proposed biomarker is precise (reproducible/ repeatable) and accurate in its characterisation of a meaningful physiological, pathological or post treatment response. The following thesis (organized in the alternative format with experimental studies written as individual complete manuscripts) investigates methods to improve precision and accuracy of the Apparent Diffusion Coefficient (ADC), a proposed quantitative imaging biomarker with a potential role in characterisation of post treatment responses in mCRC. The first objective was to establish baseline multicentre reproducibility (n=20) for ADC. A change in ADC greater than 21.1% was required to determine a post treatment response. Using a statistical error model, the dominating factors that influenced reproducibility were motion artefact and tumour volume. In the second study these factors were addressed using a single centre cohort with pre and post treatment data. Correcting for errors due to motion and tumour volume improved sensitivity from 30.3% to 1.7%, so a post treatment response was detected in 6/12 tumours compared to 0/12 using the baseline approach. In the third study, motion correction was implemented and the statistical error model was applied successfully to a multicentre cohort of 15 patients (1.9% sensitivity). The results of this thesis highlights that with careful consideration and correction of factors that negatively influence sensitivity, ADC is a potential imaging biomarker for use in post treatment response for patients with mCRC.

abstract: In medical imaging, a wide variety of methods are used to interrogate structural and physiological differences between soft tissues. One of the most ubiquitous methods in clinical practice is Magnetic Resonance Imaging (MRI), which has the advantage of limited invasiveness, soft tissue discrimination, and adequate volumetric resolution. A myriad of advanced MRI methods exists to investigate the microstructural, physiologic and metabolic characteristics of tissue. For example, Dynamic Contrast Enhanced (DCE) and Dynamic Susceptibility Contrast (DSC) MRI non-invasively interrogates the dynamic passage of an exogenously administered MRI contrast agent through tissue to quantify local tracer kinetic properties like blood flow, vascular permeability and tissue compartmental volume fractions. Recently, an improved understanding of the biophysical basis of DSC-MRI signals in brain tumors revealed a new approach to derive multiple quantitative biomarkers that identify intrinsic sub-voxel cellular and vascular microstructure that can be used differentiate tumor sub-types. One of these characteristic biomarkers called Transverse Relaxivity at Tracer Equilibrium (TRATE), utilizes a combination of DCE and DSC techniques to compute a steady-state metric which is particularly sensitive to cell size, density, and packing properties. This work seeks to investigate the sensitivity and potential utility of TRATE in a range of disease states including Glioblastomas, Amyotrophic Lateral Sclerosis (ALS), and Duchenne’s Muscular Dystrophy (DMD). The MRC measures of TRATE showed the most promise in mouse models of ALS where TRATE values decreased with disease progression, a finding that correlated with reductions in myofiber size and area, as quantified by immunohistochemistry. In the animal models of cancer and DMD, TRATE results were more inconclusive, due to marked heterogeneity across animals and treatment state. Overall, TRATE seems to be a promising new biomarker but still needs further methodological refinement due to its sensitivity to contrast to noise and further characterization owing to its non-specificity with respect to multiple cellular features (e.g. size, density, heterogeneity) that complicate interpretation. / Dissertation/Thesis / Masters Thesis Biomedical Engineering 2018

Cancer is the second deadliest disease in the United States with an estimated 1.69 million new cases in 2017. Medical imaging modalities, such as computed tomography (CT) and magnetic resonance imaging (MRI), are widely used in clinical medicine to detect, diagnose, plan treatment, and monitor tumors within the body. Advances in imaging research related to cancer assessment have largely relied on consented human patients, often including varied populations and treatments. Tumor bearing mouse models have been highly valued for basic science research, but imaging focused applications are limited by the translational ability of micro imaging systems. Pig models are well suited to bridge the gap between human cohorts and mouse models due to similar anatomy, physiology, life-span, and size between pigs and humans. These models provide the opportunity to advance medical imaging while simultaneously characterizing progressive changes resulting from an intervention, exposure, or genetic modification. We present a foundation for effectively characterizing disease models in pigs, susceptible to tumor development, using longitudinal medical image acquisition and post-processing techniques for quantification of disease. Longitudinal, whole-body protocols were developed with CT and MRI. Focus was placed on systematic process, including transportation, anesthesia and positioning, imaging, and environmental controls. Demonstration of the methodology was achieved with six pigs (30-85 kg) with four to seven imaging time points acquired per animal. Consistent positioning across time points (CT to CT) and within time points (CT to MRI) was assessed with distance measures obtained from the skeleton following rigid registration between images. Alignment across time points was achieved with an average value of 16.51 (± 12.46) mm observed all acquired measurements. For consistent, retrievable, and complete qualitative assessment of acquired images, structured reports were developed, including assessment of imaging quality and emphasis on tumor development throughout the body. Reports were used to perform a systematic, semi-qualitative comparison of CT and MRI lung assessment with an overall agreement of 72% in detection of disease indicators. A multi-level registration algorithm was developed to align anatomic structures of interest in the acquired longitudinal datasets. The algorithm consisted of initialization followed by repeated application of a core registration framework as the input data reduced in image field of view. It was applied to align regions of interest in the brain, upper right lung, and right kidney. Validation was performed with overlap (range = [0.0,1.0], complete overlap = 1) and distance measures (range = [0.0, ∞], perfect match = 0.0) of corresponding segmentations with overall results of 0.85 (± 0.11) and 0.41 (± 0.83) mm, respectively. An extension of the algorithm was created, demonstrating the ability to incorporate directional growth and feature extraction measurements into longitudinal tumor progression monitoring. Techniques were applied to a phantom dataset showing solid tumor growth and transition from a non-solid to part-solid lesion in the lungs. Finally, the developed methods – imaging, structured reporting, registration, and longitudinal feature extraction – were applied to four different porcine models pre-disposed to tumor development. 1) A genetically modified Li-Fraumeni (TP53R167H/+/TP53R167H/R167H) background model showing the development of osteosarcoma and lymphoma. 2) A TP53R167H/+ animal with exposure to crystalline silica showing progression of silicosis in the lungs. 3) TP53R167H/+/TP53R167H/R167H animals with exposure to radiation for targeted sarcoma development and 4) TP53R167H/+ pigs with conditional KRASG12D/+ mutation activated in the lung and pancreas. Whole-body and targeted imaging protocols were developed for each model and qualitatively interpreted by a radiologist using structured reports. Multi-level registration was used to align identified tumors and longitudinal features were extracted to quantitatively track change over time. Overall, the developed methods aided in the effective, non-invasive characterization of these animals.

This research focuses towards the development of a clinical decision support system (CDSS) based on cellular and tissue image analysis and classification system that improves consistency and facilitates the clinical decision making process. In a typical cancer examination, pathologists make diagnosis by manually reading morphological features in patient biopsy images, in which cancer biomarkers are highlighted by using different staining techniques. This process is subjected to pathologist's training and experience, especially when the same cancer has several subtypes (i.e. benign tumor subtype vs. malignant subtype) and the same cancer tissue biopsy contains heterogeneous morphologies in different locations. The variability in pathologist's manual reading may result in varying cancer diagnosis and treatment. This Ph.D. research aims to reduce the subjectivity and variation existing in traditional histo-pathological reading of patient tissue biopsy slides through Computer-Aided Diagnosis (CAD). Using the CAD, quantitative molecular profiling of cancer biomarkers of stained biopsy images are obtained by extracting and analyzing texture and cellular structure features. In addition, cancer sub-type classification and a semi-automatic grade scoring (i.e. clinical decision making) for improved consistency over a large number of cancer subtype images can be performed. The CAD tools do have their own limitations and in certain cases the clinicians, however, prefer systems which are flexible and take into account their individuality when necessary by providing some control rather than fully automated system. Therefore, to be able to introduce CDSS in health care, we need to understand users' perspectives and preferences on the new information technology. This forms as the basis for this research where we target to present the quantitative information acquired through the image analysis, annotate the images and provide suitable visualization which can facilitate the process of decision making in a clinical setting.

The statistical evaluation of molecular imaging approaches for detecting, diagnosing, and monitoring molecular response to treatment are required prior to their adoption. The assessment of fluorescence-enhanced optical imaging is particularly challenging since neither instrument nor agent has been established. Small animal imaging does not address the depth of penetration issues adequately and the risk of administering molecular optical imaging agents into patients remains unknown. Herein, we focus upon the development of a framework for OAIQ which includes a lumpy-object model to simulate natural anatomical tissue structure as well as the non-specific distribution of fluorescent contrast agents. This work is required for adoption of fluorescence-enhanced optical imaging in the clinic. Herein, the imaging system is simulated by the diffusion approximation of the time-dependent radiative transfer equation, which describes near infra-red light propagation through clinically relevant volumes. We predict the time-dependent light propagation within a 200 cc breast interrogated with 25 points of excitation illumination and 128 points of fluorescent light collection. We simulate the fluorescence generation from Cardio-Green at tissue target concentrations of 1, 0.5, and 0.25 µM with backgrounds containing 0.01 µM. The fluorescence boundary measurements for 1 cc spherical targets simulated within lumpy backgrounds of (i) endogenous optical properties (absorption and scattering), as well as (ii) exogenous fluorophore crosssection are generated with lump strength varying up to 100% of the average background. The imaging data are then used to validate a PMBF/CONTN tomographic reconstruction algorithm. Our results show that the image recovery is sensitive to the heterogeneous background structures. Further analysis on the imaging data by a Hotelling observer affirms that the detection capability of the imaging system is adversely affected by the presence of heterogeneous background structures. The above issue is also addressed using the human-observer studies wherein multiple cases of randomly located targets superimposed on random heterogeneous backgrounds are used in a “double-blind” situation. The results of this study show consistency with the outcome of above mentioned analyses. Finally, the Hotelling observer’s analysis is used to demonstrate (i) the inverse correlation between detectability and target depth, and (ii) the plateauing of detectability with improved excitation light rejection.