Sprayable, Biodegradable, Thermoresponsive, Antimicrobial Hydrogel Wound Dressing

Overview

Wound dressings are critically important to wound treatment and healing. Conventional wound dressings are prefabricated, non-degradable, non-conformal, and contain silver particles. Not only are these difficult to use for irregularly shaped wounds, wounds with varying depths, and wounds with large surface areas, but they 17 require frequent changes, which disrupts healing, causes pain, and increases risk of infections. Furthermore, at high concentrations, silver ions are cytotoxic to fibroblasts and stem cells, which may hinder tissue growth and cause excessive scarring. This project is a sprayable, biodegradable, intrinsically adhesive hydrogel dressing that does not need to be changed while time-releasing antimicrobial agents. The formulation is a sprayable, reverse thermal hydrogel that is directly sprayed on the wound and releases borate, a safe anti-infective agent with broad-spectrum antimicrobial activity and wound-healing properties. It contains Poloxamer-407, a biodegradable reverse-thermal block copolymer of poly(ethylene oxide)-poly(propylene oxide)-poly(ethylene oxide), and a biodegradable poly(vinyl alcohol)–borate complex for extended-release of borate. Chitosan and polyols like mannitol and trehalose are added for homogenous incorporation of the PVA-borate complex into the Poloxamer-407 solution. The formulation has a pH 5.0-6.0, close to the optimal pH for wound healing. With optimized pH, composition, and molecular weights of chitosan and poly(vinyl alcohol), the dressing formulation has a gelling temperature of 24-26°C, enabling its rapid transition from a solution to a gel form at body temperature when sprayed on the wound. For extended applications, other topical antimicrobial agents, pain relievers, and tissue engineering and hemostatic biomaterials can be delivered by this formulation. MADLIBS: A Novel Multilingual Data Augmentation Approach for Low-Resource Neural Machine Translation Zeyneb N. Kaya Saratoga High School, Saratoga, CA Neural Machine Translation (NMT) has been established as the dominant approach for developing state-of- the-art translation systems, but its effectiveness depends substantially on the availability of large parallel corpora, resulting in a significant performance gap in low-resource settings. With the declining diversity of the world’s languages, accessibility to low-resource NMT systems is valuable for promoting inclusivity. T o address this, I propose a novel Data Augmentation (DA) method suitable for underrepresented languages, MADLIBS, Multilingual Augmentation of Data with Alignment-Based Substitution, which generates diversified sentence pairs without auxiliary data. The approach consists of three components optimized for the low-resource setting: an attention-based encoder-decoder aligner, a semi-supervised POS-tagger, and a template generator. From the templates, constructed from existing sentence pairs, aligned source-target word pairs are replaced with plausible substitutions. Experimental results on a range of linguistically diverse languages in the extremely low-resource setting show improvements in translation quality by up to +3.4 BLEU points over the baseline, and +0.5 BLEU over the current established DA method, Back-translation, even without the use of monolingual data. I surpass the results of the top state-of-the-art OPUS-MT model on the leaderboard for one task. The method demonstrates great contributions towards advancing in one the most prevalent and difficult challenges of deep learning and in one of the most complex modalities and tasks. It contributes an effective and fundamentally unique approach. It is a step forward in the current frontier of DA research. The work further presents potential for emerging methods towards preserving endangered languages with NLP .

Competition history

• JSHS 2023 Category not listed

Resources

• National JSHS abstract book ↗

Related projects

Closest projects by meaning, across every fair and year in the corpus.