Abstract
Technical Report
The technical capstone focuses on transforming neuronal morphological analysis from a slow, subjective process into a more objective, scalable, and high-throughput workflow. Neurodevelopmental disorders (NDDs) arise from underlying cellular abnormalities that often result in changes in neuronal morphology. Neuronal morphology refers to the shape and structure of a neuron and is directly linked to proper function. Changes in morphology can be quantitatively assessed using features such as soma size, dendritic length and branching, also known as arborization, and spine density, providing objective outputs of these alterations. While there have been attempts to automate certain parts of this process, there is currently no fully automated pipeline that takes images as inputs, requires little to no user input, and performs the entire analysis. My project focuses on creating an end-to-end pipeline that performs morphological analysis with minimal user input. Initially, the project achieves high throughput through the use of a box microscope, which trades lower resolution for increased imaging scalability. These images are then segmented using a machine learning (ML) model with a dual-head system, in which PSPNet segments neurites and a Swin Transformer segments soma. Downstream analysis is then performed using existing Python toolboxes to extract specific morphological features and conduct Sholl analysis, a fundamental method for characterizing dendritic branching complexity. This system not only provides accurate quantitative characterization but also makes the overall process high-throughput, scalable, and more objective, allowing for comparisons across different datasets. Achieving this will strengthen ongoing studies and lay the groundwork for broader adoption of morphological biomarkers in certain NDD models, mechanistic studies, and future therapeutic screening or target identification.
STS Project
My thesis project focused on investigating the intersection between social media and the normalization of steroid use across different age groups of men. Since the 1980s, anabolic-androgenic steroid (AAS) use has become increasingly popular among the general public. This creates concern not only as a societal issue through the shifting of male body ideals, but also as a public health concern because steroids can have serious health consequences, including increased cardiovascular disease risk, reproductive health issues, and mental health problems. This relationship was studied using two methods: discourse analysis and a meta-review. Discourse analysis helped provide insight into the varying opinions of social media users on anabolic steroids, including factors such as results, side effects, health concerns, and social media validation. The meta-review provided research on how social media shapes body ideals, how individuals view themselves in relation to social media use, the relationship between steroids and social media, and associated mental health effects. The findings showed that social media intensifies male body comparison, promotes hypermuscular ideals, and makes steroid-enhanced physiques appear more standard and desirable. Furthermore, younger individuals and less experienced lifters appeared to be more influenced by external pressures, while older individuals tended to frame steroid use as a performance-based or goal-oriented decision rather than one motivated primarily by aesthetics. Actor-Network Theory (ANT) allows steroid use to be viewed as the result of a network of actors rather than just an individual choice. This framework helps explain AAS normalization as the result of interactions among multiple actors, including users, influencers, online forums, social media platforms, algorithms, peer comparisons, and idealized body images.
