AstigMate (2026)

A desktop astigmatism monitoring device utilizing computer vision and machine learning techniques for high-screen-time professionals.

Project Overview

Project Summary

Project Summary Placeholder

Tech Stack

  • Machine Learning: LoRA fine-tuning, EED-Astig dataset + self-collected adult dataset

  • Frontend: React + JS

Acknowledgements

  • This project is a personal, non-commercial exploration. No part of this work is intended for commercial use, clinical application, or professional medical diagnosis.

  • I have read and agree to the terms of the EED-Astig Data Use Agreement (DUA) 1.0.0 (DOI: 10.5281/zenodo.17231120). I will use the data solely for non-commercial research purposes and will not attempt to re-identify any participant.

  • All self-collected image data were obtained with explicit informed consent from each individual. Sensitive or personally identifiable information of participants will not be displayed.

Problem Framing

Personal Story

In the span of 1 year working as a UX designer after my graduation, my daily screen time rapidly climbed past 10 hours, with at least 8 of those spent staring at Claude sessions, Figma frames, Slack threads, and endless browser tabs. Consequently, my astigmatism went from zero to -0.5D (left) and -1.25D (right).

What’s worse than my own prescription is that eye strain has become silently accepted as an “occupational hazard” among the professionals in the tech industry. And the AI era hasn’t eased it at all. What people stare at has shifted from GUIs to agentic Web UIs, IDEs, or CLIs; yet even as AI reduces much of the tedious cognitive load, the strain on their eyes has never diminished.

Problem Scope

Globally, astigmatism is affecting an estimated 40% of adults. In China, the prevalence is particularly striking—up to 62% in certain regions—representing some of the highest prevalence worldwide.

Globally, astigmatism is affecting an estimated 40% of adults across all WHO regions. China reports the most striking prevalence range worldwide—from 8% in rural areas to 62% in urban cohorts—driven by demographic, environmental, and methodological differences. Locally, the Shanghai Eye Study found that nearly half (49.3%) of adults in Shanghai were affected by astigmatism, and the prevalence of astigmatism increased with age.

Astigmatism Prevalence in Adults

Estimated share of adults with astigmatism

Global

40%

China

62%

Shanghai

49.3%

References:

Zhang J, Wu Y, Sharma B, Gupta R, Jawla S, Bullimore MA. Epidemiology and Burden of Astigmatism: A Systematic Literature Review. Optom Vis Sci. 2023 Mar 1;100(3):218-231. doi: 10.1097/OPX.0000000000001998. Epub 2023 Feb 7. PMID: 36749017; PMCID: PMC10045990.

The Relationship between Crystalline Lens Power and Refractive Error in Older Chinese Adults: The Shanghai Eye Study
He J, Lu L, He X, Xu X, Du X, et al. (2017) The Relationship between Crystalline Lens Power and Refractive Error in Older Chinese Adults: The Shanghai Eye Study. PLOS ONE 12(1): e0170030. https://doi.org/10.1371/journal.pone.0170030

Technical Solution

Design Requirements

Given the device is a non-clinical, desk-based, self-operated monitoring solution for high-screen-time professionals, the technical approach must meet the following requirements based on the scenario:

Technical Research

Selecting the right technical approach requires balancing accuracy and reliability with design constraints.

Modal Training

Data Collection

Data collection in the original EED-Astig dataset was performed using an iPhone 12 camera at a standardized distance of 33 cm, with five images captured per subject. The additional self-collected data maintains methodological consistency by utilizing similar imaging equipment, working distance, and acquisition angles as the original dataset.

Back

Back

Problem Framing

Technical Research & Execution

Development & Prototyping

User Testing & Results