Thesis (Ph. D.)--Harvard-MIT Division of Health Sciences and Technology, 2010. / This electronic version was submitted by the student author. The certified thesis is available in the Institute Archives and Special Collections. / Cataloged from student submitted PDF version of thesis. / Includes bibliographical references (p. 127-132). / Clinical voice specialists make critical diagnostic, medical, therapeutic, and surgical decisions by coupling visual observations of vocal fold tissue motion with auditory-perceptual assessments of voice quality. The details of the relationship between vocal fold tissue motion and the voice produced are not fully understood, and there is recent evidence that the diagnostic significance of asymmetries during vocal fold vibration may be over-interpreted during clinical voice assessment. An automated system based on high-speed videoendoscopy recordings was developed to objectively quantify vocal fold vibratory asymmetry with initial validation from manual markings and visualperceptual judgments. Efficient estimation of these measures was possible due to recent technological advances in high-speed imaging of the larynx that enabled the capture and processing of high-resolution video (up to 10,000 images per second) of rapid vocal fold vibrations (100-1000 times per second). Synchronized recordings of the acoustic voice signal were made to explore physiological-acoustic relationships that were not possible using clinical stroboscopic imaging systems. In an initial study of asymmetric vibration in 14 patients treated for laryngeal cancer, perturbations in the voice signal were most associated with asymmetry that changed across vibratory cycles, while the overall level of asymmetry did not contribute to degradations in voice quality measures. / (cont.) Thus, since stroboscopic imaging is only able to capture vibratory asymmetry that occurs periodically, voice clinicians are not able to observe the time-varying nature of asymmetry that presumably affects acoustic perturbations to a higher degree. The impact of asymmetric vibration on spectral characteristics was explored in a computational voice production model and an expanded group of 47 human subjects. Surprisingly, in both model and subject data, measures of vocal fold vibratory asymmetry did not correlate with spectral tilt measures. In the subject data, left-right phase asymmetry and closing quotient exhibited a mild inverse correlation. This result conflicted with model simulations in which the glottal area waveform exhibited higher closing quotients (less abrupt glottal closure) with increasing levels of phase asymmetry. Results call for further studies into the applicability of traditional spectral tilt measures and the role of asymmetric vocal fold vibration in efficient voice production. / by Daryush Dinyar Mehta. / Ph.D.

Thesis (Ph. D.)--Harvard-MIT Division of Health Sciences and Technology, 2008. / This electronic version was submitted by the student author. The certified thesis is available in the Institute Archives and Special Collections. / Includes bibliographical references (p. 191-200). / Electronic transmission of protected health information has become pervasive in research, clinical, and public health investigations, posing substantial risk to patient privacy. From clinical genetic screenings to publication of data in research studies, these activities have the potential to disclose identity, medical conditions, and hereditary data. To enable an era of personalized medicine, many research studies are attempting to correlate individual clinical outcomes with genomic data, leading to thousands of new investigations. Critical to the success of many of these studies is research participation by individuals who are willing to share their genotypic and clinical data with investigators, necessitating methods and policies that preserve privacy with such disclosures. We explore quantitative models that allow research participants, patients and investigators to fully understand these complex privacy risks when disclosing medical data. This modeling will improve the informed consent and risk assessment process, for both demographic and medical data, each with distinct domain-specific scenarios. We first discuss the disclosure risk for genomic data, investigating both the risk of re-identification for SNPs and mutations, as well as the disclosure impact on family members. Next, the deidentification and anonymization of geospatial datasets containing information about patient home addresses will be examined, using mathematical skewing algorithms as well as a linear programming approach. Finally, we consider the re-identification potential of geospatial data, commonly shared in both textual form and in printed maps in journals and public health practice. We also explore methods to quantify the anonymity afforded when using these anonymization techniques. / by Christopher A. Cassa. / Ph.D.

Thesis (Ph. D.)--Harvard-MIT Division of Health Sciences and Technology, 2011. / Cataloged from PDF version of thesis. / Includes bibliographical references (p. 102-110). / Autocrine signaling plays essential roles in providing self-sustaining growth signals to cancer cells. Since the introduction of the autocrine hypothesis in 1980s, the contribution of autocrine signaling in cancer medicine has been limited to cancer tissues with adequately characterized mitogenic pathways. Its closed-loop nature and complex interplay with other environmental cues prevents the experimental study of unknown autocrine loops, requiring specific perturbing agents to inhibit the underlying ligand/receptor interactions. Recent studies reported the ability of drug-resistant cancer cells to acquire mitogenic signals from previously neglected autocrine loops, causing tumor recurrence. Methods that can evaluate autocrine-loop dependency in more diverse cancer tissues will help identify other means that autocrine signaling employs to maintain cancer growth. This thesis presents the use of cell-patterning methods as a tool for modulating intrinsically generated diffusive signaling cues. Such technology enables the investigation of autocrine loops without the need for specific therapeutics or prior knowledge of underlying ligand/receptor pairs. To achieve this goal, the first aim of this thesis is to determine characteristics of autocrine signaling that pertain to modulation of intercellular spacing, using existing investigation methods. In addition to demonstrating the limitation of conventional methods in examining unknown autocrine loops, we showed that changes of intercellular spacing in randomly plated culture cannot specifically modulate autocrine activity, due to the concurrent changes of other environmental cues. The second aim of this thesis is to establish engineering tools for 1) ensuring modification of only autocrine loops with the modulated cell arrangement and 2) providing prediction of autocrine activity changes with varying intercellular spacing. We illustrated cell-patteming approaches for introducing spatial regularity to standard cell culture. We then developed a stochastic model to predict changes of ligand/receptor binding with varying cell arrangement designs. We determined the spatial requirement for autocrine activity to transition from the isolated to the communicative mode. The model also helps determine cell-patterning designs that can potentially maintain uniform impacts of non-diffusive signaling cues while enabling specific modulation of autocrine signaling. In the last aim of this thesis, we evaluated the ability of regularly-shaped cell arrays to demonstrate the impact of autocrine signaling in supporting cancer growth. In comparison to randomly-plated culture, the cellpatterning platform exhibited growth change with altering intercellular spacing that better corresponds with the predicted and measured changes of autocrine ligand capture. With increasing global cell density, we also showed that regularly-shaped cell arrays acquire more uniform distribution of local cell density, while the randomly-plated cells exhibit distinct changes of local cell density. We present in this thesis the first method for the modulation of combined autocrine activity while ensuring minimal concurrent alteration of non-diffusive cues without the need of specific perturbing agents. / by Somponnat Sampattavanich. / Ph.D.

Thesis (Ph. D.)--Harvard-MIT Division of Health Sciences and Technology, 2010. / Cataloged from PDF version of thesis. / Includes bibliographical references (p. 143-144). / Hepatitis C virus (HCV) infection accounts for approximately 40% of chronic liver disease in the United States and results in an estimated 8,000-10,000 deaths annually. Simulations suggest that in the next decade morbidity and mortality associated with HCV infections will result in approximately 200,000 deaths and direct medical expenditures of over $10 billion. Furthermore, recent WHO estimates of worldwide prevalence suggest that up to 2% of the world population is infected with HCV, representing between 120 and 200 million people. For reasons that are still poorly understood, the current standard of care is effective only in a subset of patients, and depends on both patient-related and disease-related characteristics. Sustained virological response (SVR) - HCV RNA in patient plasma drops below detectable levels at week 24 following completion of treatment, which is thought to be indicative of curing the disease - is attainable in only ~50% of patients. In recent years, HCV production has been shown to be inextricably linked to lipid metabolism and to the secretion of very low density lipoproteins (VLDL) from hepatocytes. This suggests that by modulating lipid metabolism in the cell, viral production may be reduced in a clinically relevant manner. / (cont.) This work begins by characterizing the link between VLDL secretion and HCV production in the Huh7.5.1/JFH-1 system. We proceed to examine the effects of naringenin, a grapefruit flavonoid, on the production of HCV. Naringenin has been shown previously to reduce VLDL secretion from hepatocytes, and we demonstrate its ability to block HCV production, as well. We explore the mechanism of naringenin's effect on HCV, and show that the flavonoid prevents the assembly of infectious viruses in the cell. Despite previous success by several groups in describing the mechanisms involved in naringenin's effect on VLDL secretion, these mechanisms are thought to account only for ~50% of the observed inhibition, suggesting a deeper understanding of the underlying principals is still lacking. We suggest that naringenin exerts its metabolic, and consequently, antiviral effects through modulation of nuclear recepetor (NR) activity. NRs are a superfamily of ligand-regulated transcription factors known to have an important role in maintaining the homeostasis of metabolites. We show that naringenin activates peroxisome proliferator activated receptors (PPARs), NRs known to drive [beta]-oxidation; and inhibits the liver X receptor (LXR), known to drive lipogenesis and cholesterol synthesis. Despite naringenin's promise as a possible treatment for HCV, its low bioavailability, limits its clinical potential. We conclude this work by showing that naringenin's solubility and bioavailability - and thus, clinical relevance - can be greatly enhanced by complexation with [beta]-cyclodextrins. / by Jonathan Goldwasser. / Ph.D.

Thesis (Ph. D.)--Harvard-MIT Division of Health Sciences and Technology, 2005. / Includes bibliographical references (leaves 172-183). / In embryonic stem (ES) cell tumors, the hypoxia-inducible transcription factor, HIF- 1[alpha], has been shown to be a tumor suppressor, and HIF-1[alpha]-expressing cells have been shown to localize preferentially in vivo to regions near tumor vasculature. These differences were proposed to be due to increased hypoxia-induced apoptosis and growth arrest of HIF-1[alpha]-expressing ES cells. This thesis presents a careful investigation into the localization of ES cells in vitro and in vivo with respect to hypoxia. A sandwich culture system was utilized in which controlled gradients of oxygen and nutrients are developed in the vicinity of the tumor cells. A diffusion-consumption model was utilized to predict the oxygen and glucose concentration profiles within the system. Oxygen and glucose consumption rates were measured and used as inputs into the model, and the concentration profiles were found to depend on a single experimental parameter, the cell density within the system. The optimum cell density was found in which stable, measurable oxygen gradients develop over 2-3 mm. The model demonstrated excellent agreement between the predicted oxygen concentration profiles and experimentally determined oxygen gradients. In vitro, there was no difference in localization with respect to hypoxia between tumor cells expressing or lacking HIF-1[alpha]. / (cont.) In addition, there was no difference in apoptosis, proliferation, or migration of the tumor cells in vitro based on HIF-1[alpha] expression. Likewise, a quantitative study on localization of tumor cells within tumors in vivo demonstrated no difference between localization of HIF-1[alpha]-expressing vs. HIF-1[alpha]-lacking ES cells within tumors with respect to blood vessels or hypoxia. These results differ from previous studies, perhaps due to clonal variation of the cell phenotype or the interplay of other complex environmental factors that were not considered in this study. Interestingly, the HIF-1[alpha]-lacking cells were found to exhibit increased tumor growth relative to the HIF-1[alpha]-expressing cells, perhaps due to a normalization of the blood vessels within the HIF-1[alpha]-lacking tumors. These studies reveal the complex role of HIF-1[alpha] in tumor growth and tumor cell localization, as well as develop a useful quantitative experimental model for studying the role of the microenvironment in tumors or in embryonic stem cell biology. / by David M. Cochran. / Ph.D.

Thesis (S.M.)--Harvard-MIT Division of Health Sciences and Technology, 2010. / Vita. Cataloged from PDF version of thesis. / Includes bibliographical references (p. 49-51). / This study investigates the effect of age, gender, medical condition, and daily free text input on classification accuracy for resuscitation code status. Data was extracted from the MIMICII database. Natural language processing (NLP) was used to evaluate the social section of the nurses' progress notes. BoosTexter was used to predict the code-status using text, age, gender, and SAPS scoring. The relative impact of features was analyzed by feature ablation. Social text was the greatest single indicator of code status. The addition of text to medical condition features increased classification accuracy significantly (p<0.001.) N-gram frequency was analyzed. Gender differences were noted across all code-statuses, with women always more frequent (e.g. wife>husband.) Visitors and contact were more common in the less aggressive resuscitation codes. Logistic regression on medical, age, and gender features was used to determine gender bias or ageism. Evidence of bias was found; both females (OR=1.47) and patients over age 70 (OR=3.72) were more likely to be DNR. Feature ablation was also applied to the social section of physician discharge summaries, as well as to annotated features. The addition of annotated features increased classification accuracy, but the nursing social text remained the most individually predictive. The annotated features included: children; living situation; marital status; and working status. Having zero to one child; living alone or in an institution; being divorced or widow or widower; and working, working in white collar job, or being retired, were all associated with higher rates of DNR status, and lower rates of FC status. Contrarily, living with family; being married; and being unemployed, were all associated with lower rates of DNR status, and higher rates of FC status. Some of these findings were gender and/or age dependent. / S.M.

Thesis (Ph. D.)--Harvard-MIT Division of Health Sciences and Technology, 2007. / Includes bibliographical references (p. 117-135). / Children with dyslexia struggle with learning to read despite adequate intelligence, motivation, and schooling. Over the years, there has been a growing consensus about the role of phonological processing in reading disability. Poor readers typically do worse than their normal reading peers on tasks that require phonological processing which has been linked, directly or indirectly, to their speech perception abilities. The work in this thesis combined behavioral, MEG, and EEG methods to examine how normal and reading-impaired children, 7-13 years of age, perceive speech under varying degrees of phonological contrast (1 vs. 3 phonetic features). In a series of auditory word perception experiments, good and poor readers were found to do worse in accuracy and/or reaction times in phonologically similar (i.e., 1-feature contrast) than phonologically dissimilar (i.e., 2 or 3-feature contrast) conditions. Despite the similar behavioral performance and EEG responses for the two groups, a region of interest (ROI) based MEG approach revealed differences in the brain activation of the two groups in superior temporal regions at 140 to 300 ms. / (cont.) In the auditory word discrimination task, differences in activation were found in good readers but not poor readers, as a function of the degree of phonological contrast, reflecting poor readers' lack of sensitivity to the phonological characteristics of the word stimuli. In the sentence plausibility judgment task, the impaired phonological processing abilities of the poor readers may have led them to rely more on top-down sentence context to perceptually disambiguate phonologically confusing terminal words, thereby deceiving them into accepting the phonologically similar incongruent sentences as being congruent. This may account for the poor reader group's reduced brain activation in the phonologically demanding condition in the sentence task. The results of the experiments are consistent with a phonological view of reading disability according to which children with reading impairments have poorly defined phonological representations. / by Daniel T. Wehner. / Ph.D.

Thesis (M. Eng.)--Harvard-MIT Division of Health Sciences and Technology, June 2011. / "June 2011." Cataloged from PDF version of thesis. / Includes bibliographical references (p. 75-79). / For this project, N-terminal and C-terminal peptide library fusions were designed,bconstructed, and screened in order to improve the repression achievable with a novel gene regulatory system. This system, based on the interaction between proteins and proteinbinding RNA aptamers, takes advantage of the reversible interaction between TetR and its RNA aptamer binding partner 5-1.2 to modulate gene expression. With no tetracyclines present, TetR preferentially binds to aptamer 5-1.2 in the mRNA of a gene of interest with low nanomolar affinity and represses translation. Tetracyclines such as aTc induce a conformation change in TetR, prevent TetR binding to aptamer 5-1.2, and induce gene expression. Therefore, TetR binds aptamer 5-1.2 in an aTc-dependent manner, allowing inducible control of gene expression through the TetR-aptamer system. Initial characterization showed a regulatory range of 78% or approximately 5 fold in S. cerevisiae. The aim of this project is to improve repression levels achievable with the TetR-aptamer system by creating libraries of N-terminal and C-terminal peptide fusions to TetR and screening for increased repression in S. cerevisiae. The N-terminal and C-terminal library fusions were constructed from synthesized oligonucleotide fragments and a baseline TetR vector containing library insertions sites at both the N-terminal and C-terminal ends. The library fragments contain 20 random amino acids and a standard SSG linker peptide flanked by both single-cutting restriction enzyme sites and 40 bases of homology to the library insertion sites on the baseline TetR vector, allowing for construction by both restriction/ligation cloning in bacteria and yeast homologous recombination. Both libraries were constructed using restriction/ligation cloning after initial experiments determined optimized conditions for PCR, digest, purification, ligation, and electrocompetent bacterial transformation to achieve a maximum efficiency, fidelity, and purity. The N-terminal and C-terminal libraries produced have a combined diversity of 2.5x 105 variants. These library variants were screened using a plate-based assay with URA3 as a reporter gene. A selection with 5-fluoroorotic acid (5-FOA) was performed to identify library variants with improved repression. Since 5-FOA is a competitive inhibitor of URA3, cells that have URA3 expression cannot live on media containing 5-FOA. Preliminary experiments determined that 0.035% 5-FOA is the threshold for growth for the baseline / (cont.) TetR-aptamer system. Library variants containing 5-1.2-URA3 were grown on media containing a gradient of 5-FOA concentrations between 0.03% and 0.06% and compared to baseline TetR growth. Thirty-one library variants grew at a 5-FOA concentration greater than the baseline threshold for growth. These library hits underwent testing to further characterize their repression, inducibility, and library sequence. The 31 colony hits were streaked on discrete concentrations of 5- FOA to determine colony-specific 5-FOA thresholds and on uracil dropout media in the presence and absence of aTc to screen for preserved inducibility. Of the original 31 colonies, 25 passed both plate assays, growing on 0.045% 5-FOA and in a Tc-dependent manner on uracil dropout media. These 25 colonies were sequenced and analyzed using MEME to detect any conserved motifs. Twenty of the 25 sequenced colonies contained correct and unique libraries, 3 with N-terminal libraries and 17 with C-terminal libraries. Both the N-terminal and C-terminal library sequences had significant motifs. For the Nterminal sequences, all three contained the same 14 nucleotide motif, and 16 of the 17 Cterminal sequences also contained a 21 nucleotide motif. However, the unbiased selection for improved TetR functionality likely returned hits that use multiple mechanisms of action to enhance repression. Therefore, out of a diverse library pool containing 2.5x 105 variants, 20 unique library variants conferred increased repression on the TetR-aptamer system while maintaining inducibility in the presence of tetracycline. / by Jessica Karen Wong. / M.Eng.

Thesis (Ph. D.)--Harvard University--MIT Division of Health Sciences and Technology, 2005. / Includes bibliographical references (p. 229-233). / This thesis addresses the design and evaluation of algorithms to improve speech reception for cochlear-implant (CI) users in adverse listening environments. We develop and assess performance metrics for use in the algorithm design process; such metrics make algorithm evaluation efficient, consistent, and subject independent. One promising performance metric is the Speech Transmission Index (STI), which is well correlated with speech reception by normal-hearing listeners for additive noise and reverberation. We expect the STI will effectively predict speech reception by CI users since typical CI sound-processing strategies, like the STI, rely on the envelope signals in frequency bands spanning the speech spectrum. However, STI-based metrics have proven unsatisfactory for assessing the effects of nonlinear operations on the intelligibility of processed speech. In this work we consider modifications to the STI that account for nonlinear operations commonly found in CI sound-processing and noise reduction algorithms. We consider a number of existing speech-based STI metrics and propose novel metrics applicable to nonlinear operations. A preliminary evaluation results in the selection of three candidate metrics for extensive evaluation. In four central experiments, we consider the effects of acoustic degradation, N-of-M processing, spectral subtraction, and binaural noise reduction on the intelligibility of CI-processed speech. We assess the ability of the candidate metrics to predict speech reception scores. / (cont.) Subjects include CI users as well as normal-hearing subjects listening to a noise-vocoder simulation of CI sound-processing. Our results show that: 1) both spectral subtraction and binaural noise reduction improve the intelligibility of CI-processed speech and 2) of the candidate metrics, one method (the normalized correlation metric) consistently predicts the major trends in speech reception scores for all four experiments. / by Raymond Lee Goldsworthy. / Ph.D.

Thesis (S.M.)--Harvard-MIT Division of Health Sciences and Technology, 2004. / Includes bibliographical references (leaves 66-67). / In order to automate data extraction from electronic medical documents, it is important to identify the correct context of the extracted information. Context in medical documents is provided by the layout of documents, which are partitioned into sections by virtue of a medical culture instilled through common practice and the training of physicians. Unfortunately, formatting and labeling is inconsistently adhered to in practice and human experts are usually required to identify sections in medical documents. A series of experiments tested the hypothesis that section identification independent of the label on sections could be achieved by using a neural network to elucidate relationships between features of sections (like size, position from start of the document) and the content characteristic of certain sections (subject-specific strings). Results showed that certain sections can be reliably identified using two different methods, and described the costs involved. The stratification of documents by document type (such as History and Physical Examination Documents or Discharge Summaries), patient diagnoses and department influenced the accuracy of identification. Future improvements suggested by the results in order to fully outline the approach were described. / by Reejis Stephen. / S.M.