Mangrove forests are rich ecosystems that support our planet and the mankind in many unique ways. Unfortunately, these mangroves are declining at a rapid rate due to deforestation and other activities of the mankind. Monitoring and tracking of these mangrove trees is essential for their conservation. Machine Learning can be used for this purpose but to take advantage of the power of machine learning, image data needs to be captured for these mangrove ecosystems. This data collection is done using Unmanned Aerial Vehicles like drones in this research. Manually labeling the acquired image data for machine learning applications is a tedious and time-consuming task. This called for the development of architectures which could learn from limited labeled data and take advantage of the large amounts of unlabelled data. Such architectures are the semi-supervised semantic segmentation architectures and are studied in this research. We have shown how different semi-supervised models like the self-learning based Pseudo-Labelling architecture, the graph-based label propagation architecture and the deep leaning UNet-Autoencoder architecture perform on the task of mangrove segmentation in the aerial imagery. In order to evaluate different models we mainly look at the Intersection over Union because of it’s popularity in segmentation tasks. Overall, we see that the deep learning UNet-Autoencoder architecture performs the best with an average IoU of 0.78. Conceivably, the performance of each of the models improves as more labelled data is provided for training. The highest IoU obtained in this research is 0.9 with the UNet-Autoencoder when as much as 75% of the data provided is labelled.

Forest Tales: Experiments in Narratech, elaborates the process of producing a femi-queer eco-cinema retelling of the South Asian epic Ramayana as a Sitayana. Both cinema and the Ramayana are world-building technologies that have been integral to different iterations of empire and accumulation through time. In their major strain, both traditions tell stories of human exceptionalism; in their minor strains, they both have the potential to spark our revolutionary imaginations. Speaking with Sita - daughter of the earth - has long been a mode of speaking truth to power from the position of marginalized communities. Following this method, Sita emerges as forest in my project, and it is through their voice that we make sense of this tradition. In another prong of the project, the vulture Jatayu - an important figure in the Ramayana epic, who gives his life for Sita - is resurrected with a prosthetic wing, only to face the extinction of his kind in the present.

The dissertation includes three section: the first (chapters one and two) offers an in-depth introduction to the two foundational themes of Forest Tales – the Ramayana and cinema; the second engages my interest in the larval (which emerges from my distributed intelligence with artist/scholar Praba Pilar) as an experimental method for making performance, and as a theoretical and ethical position that grounds these experiments; the final section brings my practice into the present through my ethnographic fieldwork with forest communities in the city of Mumbai. Between the fivechapters that comprise these sections are three photo essays that engage my practice more directly: 1. Forest Tales: A Larval Cinema, offers a survey of the various iterations and incarnations of the embodied eco-cinema experiments at the heart of this dissertation; 2. Mother Nurture documents a bio-social sculpture project created in collaboration with the Hypha Collective, and 3. Bombay Futures captures the latent and contested visions for Bombay’s estuarine ecologies through a podcast series.

In the end, this dissertation approaches stories not merely as narratives but as narratech: both shaped by the technologies of storytelling (which have their own latent narratives), and simultaneously technologies that shape the world. This method of inquiry demands that we not only re-story our pasts/presents/futures, but also remediate the technologies that we use to narrate and make sense of them, for contemporary media technologies (digital and otherwise) are deeply embedded in capitalism's 'hydrocarbon imagination' (Bozak 2012).

Increased understanding of the ecology of the cystic fibrosis (CF) lung microbiome has im- proved treatment options for CF lung infections by allowing clinicians to target microbes known to be associated with acute disease. The microbial community responsible for CF lung infections is a complex collection of bacteria, viruses, and fungi with varying nutritional sources and metabolisms. Oxygen is a key resources in this environment, however it has different effects on the various microbial species with some species requiring it for respiration and others inhibited by it. Episodes of acute disease known as cystic fibrosis pulmonary exacerbations (CFPEs) are now known to be associated with higher abundances of fermentative anaerobes which can propagate in low oxygen conditions. The goal of this dissertation is to characterize the CF lung microbiome quantitatively, specifically the inter- action between anaerobic and aerobic microbes, the role of oxygen, and the potential for oxygen-based therapies and optimized antibiotic usage. We begin by developing a model for the in vitro growth of CF pathogens and estimating basic growth parameters for individual pathogens. We use these parameterizations to model the competition between aerobic and anaerobic communities inside of a CF airway and to predict antibiotic treatments and oxy- gen conditions that maintain a low anaerobic population. Next, we extend our modeling of aerobic-anaerobic competition to include spatial effects and stochasticity to investigate how oxygen gradients affect community dynamics within aggregated mucus. We use our spatial models to predict physical properties such as the minimum mucus plug diameter necessary to support anaerobic growth and show that an antibiotic with a high permeability into mucus is the most effective for controlling the anaerobic community. The results from this dissertation will aid the clinical treatment of CF lung infections by predicting antibiotic treatments and oxygen therapies which can be tested in laboratory conditions and eventually be applied clinically.

Hidradenitis suppurativa (HS) is a chronic, inflammatory, debilitating disease of unknown etiology. HS can occur in people of all ethnicities and ages, and affects approximately 3-4% of the United States. To date, few studies have specifically examined the race prevalence of HS; further epidemiological research is needed to identify specific trends among HS and its racial predilections. At our center, 1.3% of African-American patients were seen for HS, compared to 18% of Caucasian patients (p<0.05), and the percent ratio of African-American versus Caucasian patients with HS was 7.22:1. Our number ratio of African-American patients versus Caucasian patients with HS was 1.19:1. Studies performed at Henry Ford Medical Center and University of Pittsburgh report ratios of 1.64:1 and 1.98:1 respectively. These data support study trends suggesting HS is more common among patients of African-American descent. A large, population-based study across the United States is needed to better assess the associations between ethnicity and HS. Examining this patient population has the potential to improve our understanding of HS pathophysiology, and will enable clinicians to better manage patients with this disease.

Traumatic dislocations of the shoulder commonly present to emergency departments (EDs). Immediate closed reduction of both anterior and posterior glenohumeral dislocations is recommended and is frequently performed in the ED. Recurrence of dislocation is common, as anteroinferior labral tears (Bankart lesions) are present in many anterior shoulder dislocations.14,15,18,23 Immobilization of the shoulder following closed reduction is therefore recommended; previous studies support the use of immobilization with the shoulder in a position of external rotation, for both anterior and posterior shoulder dislocations.7-11,19 In this study, we present a technique for assembling a low-cost external rotation shoulder brace using materials found in most hospitals: cotton roll, stockinette, and shoulder immobilizers. This brace is particularly suited for the uninsured patient, who lacks the financial resources to pay for a pre-fabricated brace out of pocket. We also performed a cost analysis for our low-cost external rotation shoulder brace, and a cost comparison with pre-fabricated brand name braces. At our institution, the total materials cost for our brace was $19.15. The cost of a pre-fabricated shoulder brace at our institution is $150 with markup, which is reimbursed on average at $50.40 according to our hospital billing data. The low-cost external rotation shoulder brace is therefore a more affordable option for the uninsured patient presenting with acute shoulder dislocation. [West J Emerg Med. 2015;16(1):–0.]

Prediction, both on the syntactic and the semantic level, is a central process in language comprehension. For instance, people predict aspects of event structure based on morphosyntactic markers on verbs: hearing has peeled directs one's attention towards a culminated event, as opposed to an ongoing event. Here, we ask how general this prediction process is, and specifically, whether it extends to cues outside the predicate, using the Hindi split-ergative system as case study. Ergativity allows properties of an event to be predicted on the subject, notably a constituent outside the Verb Phrase. In four studies, we map out the role subject marking plays for prediction of event properties in comprehension. Our results show that in some offline judgments, ergativity is a strong predictor of culminated events; but the cue provided by ergative marking is not taken into account during incremental comprehension, questioning accounts of automatically triggered culmination inferences in ergative constructions as well as providing evidence for a limit of predictive processing.