Combating Nicotine Addiction Using Smartwatch-Derived Biometrics and Privacy-Preserving Distributed Machine Learning
Brandon Wang, Plano West Sr. High School, TX
E-cigarette use (known as “vaping”) is on the rise in the United States, especially among teens, which has led to high nicotine addiction rates and illness. However, current addiction treatments, including support groups, are often ineffective, and most attempts to quit fail, ending in relapse. To address this problem, I plan to use health data measured from an off-the-shelf smartwatch, including heart rate and movement behavior, to predict strong urges to smoke and when they will occur for users trying to quit vaping, and preemptively warn them to help them prevent relapsing. No commercially available devices to predict these withdrawal effects exist. In order to make predictions, smartwatch data will be combined with user demographic information and then used to train a smartphone-based LSTM neural network. In order to preserve patient data privacy while training the neural network, I will implement an innovative combination of the Federated Learning and Secure Multi-Party Computation algorithms to collaboratively train a neural network over multiple devices without ever sharing raw patient data. Finally, an app interface will be created that provides warnings, access to counseling hotlines, and other resources whenever smoking urges are predicted. Multistage testing will be conducted through both digital simulations and with actual users trying to quit to ensure the device’s real-world effectiveness.
From Pixel to Paragraph: A Deep Artwork Paragraph Generator
Audrey Cui, Monta Vista High School, CA
Art influences society by shaping our perspective and sense of self. Since not all pieces of art are extensively captioned, I wonder: would it be possible for computers to understand and then describe artwork? Automated artwork analysis will make art more accessible to the visually impaired and facilitate semantic searches for artwork conveying specific themes, enabling a wider audience to enjoy art and experience what artists communicate through their artwork. The goal of my project is to develop an artificial neural network system that interprets input artwork and generates a paragraph describing objects and low level features present in the artwork, as well as ideas and emotions the artwork conveys. I will develop a visual-semantic embedding space that learns the relationship between artwork image features and art analysis paragraph sentences. The embedding module retrieves the most relevant sentences to the input artwork features. These sentences are fed into a generative adversarial network with a novel sentiment filter to generate the final art analysis paragraph, whose relevance to the input artwork is evaluated with a SPICE score. By expanding upon a human's interpretation of artwork with new and sometimes surprising insight, my project makes progress towards developing a creative AI.
Point-of-Care Biomonitor of Isoprostane Levels in Neurodegenerative Disease Patients
Youssef Kallel, Corona del Mar High School, CA
Nearly 6 million individuals within the United States suffer from a form of a neurodegenerative disease, with the most common strains by far being Alzheimer’s and Parkinson’s. While all of these diseases have particularly high mortality rates, continual tracking of their progression iscrucial to the selection of symptomatic treatments and therapies designed to slow and ease thepain caused by the affliction’s advancement on a case-by-case basis. I seek to design a point-of-care biomonitoring sensor meant for clinical application in order to allow the patient and their physician to chart the progression of their neurodegenerative disease, most notably including multiple sclerosis, Alzheimer's disease, Huntington's disease, and Creutzfeldt-Jakob disease amongst others, curating their therapies accordingly. I propose using an electrochemical-based, point-of-care biosensor sensitive to the biomarker of 8-F2-isoprostanes, compounds established as being directly related to the levels of oxidative stress within an individual. Oxidative stress is defined as the influence of reactive oxygen/nitrogen specieswhich destroy neurons and glia, resulting in the pathogenesis of neurodegenerative diseases. The biosensor would be composed of a silver/silver chloride (Ag/AgCl) reference electrode, a platinum (Pt) counter electrode, and a gold (Au) working, or sensing electrode. Integrating artificial intelligence and an electrochemical system, the biomonitoring sensor would store a patient’s historical basal levels for the concentration of 8-F2-isoprostanes in their blood’s plasma,
sensing for progressive diversions from this historical level and the degree to which their concentration has shifted relative to the lapse in time between measurements.
A Novel Device for the Inexpensive Automation of Manual Wheelchairs
Ferhan Jemal, Annandale High School, VA
The price of medical care, including medical equipment like electric wheelchairs, has been steadily increasing for several decades. Electric wheelchairs cost between $1,200 minimal to a maximum price greater than $15,000 dollars. Considering that the majority of handicapped Americans are unemployed, the cost of wheelchairs will likely place a financial burden on those people seeking to purchase an electric wheelchair. I will be approaching this problem of unaffordability by providing a mechanism to automate manual wheelchairs while substantially decreasing the overall costs of electric wheelchairs. People that are in search of mobility will have the option to purchase an automation device for a manual wheelchair for approximately $200.
SwimSafe: A Revolutionary All in One Pool Technology System
Leila Idrissi, Solomon Schechter School of Westchester, NY
Swimming is one of the most common leisure activities and has been proven to havemany different health benefits such as cardiovascular fitness and alleviating stress (Department of Health & Human Services). However, swimming pools are also dangerous without the right protection. According to the CDC (Center for Disease Control), one in 10 people die from drowning every day and drowning is the leading cause of death of children aged 1-4 years. Society needs an easy to use and functioning product that will properly address the high risk problem of accidental drowning. My goal is to solve this problem by creating, SwimSafe, a technologically advanced automatic pool cover and non-swimmer bracelet that will prevent children from accidentally falling in. The cover will be made out of clear sheets of polycarbonate material, which maintains aesthetic of the pool unlike traditional covers. When a child falls on the pool cover they will trigger a weight sensor (with a 15-pound minimum), which will send a notification to the parent through the SwimSafe app. Additional features include a pH and chlorinesensor to monitor the chemical ratio in the pool and attachable barrier to prevent the swimmer from entering the deep end. Finally, the pool app will connect to a non-swimmer bracelet and notify the parent if the child has fallen into the pool when the cover is off. The SwimSafe cover will provide pool owners with security and easy maintenance, while maintaining the beauty and functionality of the swimming pool.
A Preventative Solution to the Formation of Diabetic Foot Ulcers in Diabetic Polyneuropathy Patients
Rohan Sanda, Tamalpais High School, CA
Diabetic polyneuropathy (DPN) is one of the most common complications of diabetes mellitus (diabetes). DPN is a disorder associated with impaired peripheral nerve function and reduced blood circulation in appendages. Even minor trauma to appendages can lead to the formation of chronic ulcers (non-healing wounds) that may require expensive medications, treatments, and even amputations. The formation of diabetic foot ulcers (DFUs) is the most prevalent consequence of DPN and results from improper foot placement and friction in shoes, which often begin unknown to the patient due to the impaired sensation. To significantly reduce the chances of the forming DFUs, I propose a preventative technology that can be placed in the shoes of patients with DPN. The technology will consist of a force-detecting insole that can monitor the orientation of the user’s foot in their shoe. An algorithm will evaluate how pressure is distributed across the shoe and, in the case of a potentially damage-inducing foot position, alert the user via a smartphone and audible signal to change the orientation of their feet to reduce or eliminate foot-to-shoe friction. By developing a technology that can be integrated into everyday life, I hope to provide a patient-friendly solution to preventing DFUs and improving quality of life for patients with DPN.