Embedded System for the Real-Time Fall Detection of Elderly Individuals using Thermal Imaging and Deep Learning

Overview

Falls are the leading cause of fatal injury among older adults with over 12 million falls in the United States annually. For elderly people living alone, quick assistance after a fall is critical and can reduce hospitalization rate by 26% and death rate by 80%. While there are wearable medical alert systems, many elderly individuals prefer not to wear them or may be unable to call for help if injured after a fall. The research goal of this study was to develop an unobtrusive embedded system for real-time fall detection. A Raspberry Pi and a Forward-Looking Infrared (FLIR) thermal camera were utilized. Thermal imaging preserves privacy while detecting heat signatures of humans. Deep learning was used to analyze and classify images due to its effectiveness for complex pattern recognition. The system developed uses a set of thermal images to train a convolutional neural network to identify images of fallen individuals. The effects of factors including image resolution and number of training images on the accuracy of the created neural networks were investigated. An average accuracy of 99.2% (SD 0.62%) was achieved even in the presence of non-human heat sources such as pets. A functioning wall-mounted system has been developed to detect falls and immediately call for help, potentially increasing the safety and independence of elderly people living alone at home or in assisted living facilities. The embedded system can also be used for other applications such as monitoring the night-time activity and sleep patterns of autistic children, enhancing their safety. Effective repeated filtration of antibiotics from wastewater using activated charcoal filters Lea Wang Council Rock High School South Holland, PA Mentor - Professor Jeffrey Field, Ph.D. University of Pennsylvania Perelman School of Medicine Department of Pharmacology The CDC reports that 2.8 million Americans are affected by antibiotic resistance and that 35,000 die each year as a result. Recent studies indicate that bacterial infections secondary to the novel coronavirus (a viral infection itself) have been fatal to those with antibiotic resistance. Antibiotics from human and animal consumption have polluted drinking water, enabling bacteria and fungi to develop resistance and hindering the treatment of many bacterial and fungal diseases. Many wastewater treatment facilities cannot employ expensive existing filtration technologies, so an effective, economical filtration option is needed. This study investigated the repeated use of inexpensive, scalable activated charcoal in filtering amoxicillin, a penicillin-class antibiotic, from water. The first phase of the study established an amoxicillin-in-water solution concentration vs. absorbance standard curve using a UV-vis spectrophotometer. Amoxicillin filtration was then studied over five rounds with filters containing varying ratios of charcoal to solution. The percent removal rates were determined by the standard curve. A bacteria test was conducted to visually display the effects of filtration by applying various amoxicillin solutions to E. coli col onies and conducting survival counts. A limited attempt at regenerating used charcoal thermally was conducted. A 1:10 ratio of charcoal to solution consistently removed >99.9% of amoxicillin and a 1:20 ratio filter displayed removal rates between 94.0% and 99.9%. The quantitative results of the reusability study, further validated by the bacterial study, indicated the potential of activated charcoal to filter amoxicillin and possibly other antibiotics effectively and economically. The charcoal regeneration study had limited success. Potential future work includes investigating ways to restore used charcoal and researching ways to apply these findings in treatment plants to help reduce the population affected by antibiotic resistance. Development of a Peptoid-peptide Macrocycle Inhibitor of CDK2-cyclin A as a Cancer Therapeutic Ethan Weisberg Packer Collegiate Institute New York, New York Supervising Scientist: Kent Kirshenbaum NYU Biomedical Chemistry Institute Cyclin-dependent kinases (CDKs) are critically important targets for cancer therapy, as their overactivity is often associated with tumorigenesis. The traditional approach to targeting CDKs has been the development of small molecules, which are ATP competitive as they competitively inhibit the enzyme at the ATP-binding site. This approach, however, is not ideal because of the conservation of the ATP-binding site structure among CDKs and kinases more broadly. Thus, the development of allosteric inhibitors specific to tumorigenic CDKs and CDK function is ideal. I have sought to develop a non-ATP competitive inhibitor of CDK2-cyclin A by targeting the cyclin-binding groove. The CDK2-cyclin A complex is involved in the phosphorylation of transcription factor E2F1 necessary to terminate E2F1 activity and allow the cell to enter the G2 phase and continue through the cell cycle. This function is mediated by the cyclin-binding groove (CBG), as the E2F1 substrate must bind there in order to be recognized and phosphorylated. Therefore, peptides that bind selectively to this site cause inappropriate persistence of E2F1 activity in cells with deregulated E2F1, resulting in tumor-specific E2F1-induced apoptosis. Unfortunately, peptides have weak pharmacokinetic properties because of their susceptibility to proteolytic cleavage. Therefore, I have used N-substituted glycines to build a macrocycle to target the CBG, as these macrocycles, termed peptoids, have been previously shown to have folding capabilities complementary to the protein surface, as well as pharmacokinetic properties superior to those of peptides. Here, I report the successful synthesis of the peptoid macrocycle, the synthesis of a positive control, and Rosetta docking simulations to validate the design of the macrocycle.