My primary project at the Datta lab involved studying a mouse social olfactory behavior called the social transmission of food preference. In this paradigm, a simple odor input from the breath or excrement of another mouse induces a change the exposed mouse’s food preference: an interesting type of learning. We used various techniques to study this including many behavioral recordings with fiber photometry, optogenetics, and behavioral videos as odors and stimuli were being presented. I was responsible for helping build the experimental rigs, running experiments, and writing python code to scale up our analysis pipelines from handling single sessions, to TBs of data. Shown are photos of some of the computer vision algorithms I implemented and some example data visualizations.

Conducted experiments studying human movement and complex object manipulation in healthy and stroke populations. My primary personal project at the Action Lab was called the Bell Task, an experiment where we investigated how humans control a bell, which is a complex object requiring precise force control. I also helped collect data for and rewrite a python-based computer vision system for an experiment called Magic Table. In this experiment, we record stroke and healthy patients manipulating a 3D printed cup with a ball inside (a complex object again) and analyze their movments. I upgraded the code to be compatible with a 3D depth camera instead of a 2D webcam, for future 3D experiments, and made it easier for the user to use.

Stroke often results in dexterity loss, so having a way to progressively and personally train this is useful for rehabilitation. Incorporating gross arm movements helps improve whole body coordination training, and varying between rhythmic and discrete movements engages more of the motor system.

~70 million Americans are over 65 — and most will lose hand dexterity as they age. This loss signals cognitive decline and reduced independence. Yet clinicians have no way to accurately quantify dexterity under real-world conditions. We are developing a series of devices to solve this problem by bringing low-cost, highly accurate force sensors to an accessible form factor for impaired populations with a neurology and rehabilitation informed software analysis and game suite.