# Design and Implementation of an Automated Drosophila Locomotor Assay Using Computer Vision Tracking

## TL;DR

This paper introduces an automated system using computer vision to study fruit fly movement, enabling faster and more detailed analysis of locomotion behaviors.

## Contribution

The novel contribution is an automated, high-throughput platform for analyzing Drosophila locomotion with sex-specific and age-related precision.

## Key findings

- The automated system processes data 2.8 times faster and with 800 times higher data density than manual scoring.
- Sex-specific differences in locomotor performance were observed in circadian ClockOut mutations and PoIG knockdowns.
- The system reveals reproducible phenotypes across multiple genotypes, including sex-dimorphic aging patterns.

## Abstract

Drosophila has long served as a powerful model for investigating locomotor behavior, using geotaxis assays to generate valuable insights into genetics, aging, and neurobiology. Nonetheless, mostly their use can be constrained by subjective scoring, modest thought, and challenges in reproducibility. We developed and validated an integrated hardware–software platform that enables automated, high-resolution locomotor analysis across 12 vials in parallel. The system integrates 3D-printed mechanical components, Raspberry Pi–based video acquisition, and programmable environmental controls to ensure standardized conditions. A deep learning pipeline segments vials with near-perfect accuracy (IoU > 0.95), while computer vision algorithms quantify climbing trajectories, velocity, and positional zone occupancy at 60 frames per second. The end-to-end workflow converts raw video into time-resolved metrics, supports sex-specific aggregation, and incorporates advanced statistical analyses, including Linear Mixed Effects regression, harmonic mean p-values, and Mann–Whitney U tests. Relative to manual scoring, this automated pipeline yields 2.8-fold faster processing and about 800-fold higher data density. Application of the platform uncovered reproducible phenotypes of multiple genotypes. For example, in circadian ClockOut mutation, males displayed progressive climbing deficits with age, whereas females-maintained age-resilient trajectories. Moreover, male ClockOut exhibited a reduced locomotor performance compared to age-matched control (w1118) males, however, female ClockOut showed subtle reduction in locomotor performance. Additionally, glial-specific knockdown of PoIG, encoding the DNA polymerase gamma catalytic subunit, revealed striking sex-dimorphic aging patterns: females outperformed controls at older age in Glaz driven and at younger age in Elav driven, while males exhibited some marked decline. To promote broad adoption, a user-friendly Python interface (Tkinter GUI) enables accessibility independent of computational expertise. Collectively, this standardized, high-throughput framework advances the resolution of genotype-, age-, and sex-dependent locomotor dynamics, offering new opportunities in aging, circadian biology, and neurodegeneration research.