When measuring whether an eye is nearsighted, farsighted, or perfectly focused, we aim to capture its static, structural state. However, the human eye is an active optical system. It contains a dynamic focusing mechanism called accommodation, driven by the involuntary contraction of the ciliary muscles which alter the shape of the crystalline lens.
When a patient—particularly a child with a highly robust focusing system—looks into a standard vision testing device, their eye frequently accommodates in response to the proximity of the machine or instrument strain. As detailed by Maoqiang Zhuang et al. in Prevalence and influence factors for myopia and high myopia in schoolchildren in Shandong, China, this active muscle contraction temporarily increases the refractive power of the lens. To an automated testing device, this transient optical shift looks identical to structural nearsightedness. This phenomenon is known as pseudo-myopia, and it introduces a systematic bias that significantly overestimates the prevalence and severity of myopia.
The Pharmacological Solution and Its Bottlenecks
To isolate the eye's true physical dimensions from its muscular adjustments, clinical practice relies on cycloplegic refraction. As noted by Yunyun Sun et al. in Prevalence of visual impairment and refractive error-related risk factors in preschool children in beijing, China, this technique involves administering pharmacological eye drops that temporarily paralyze the ciliary muscles. By chemically freezing the eye's internal focusing mechanism, cycloplegic autorefraction ensures that measurements reflect the eye's baseline physical structure rather than its temporary muscular strain (Shengxin Liu et al., Breastfeeding and myopia: A cross-sectional study of children aged 6–12 years in Tianjin, China; Wennan He et al., Effect of wearing peripheral focus-out glasses on emmetropization in Chinese children aged 6–8 years: study protocol for a 2-year randomized controlled intervention trial).
While cycloplegia is the diagnostic gold standard, it introduces severe practical limitations. The drops cause prolonged blurred vision and light sensitivity, resulting in patient discomfort and parental resistance (Haohan Zou et al., Identification of ocular refraction based on deep learning algorithm as a novel retinoscopy method). Furthermore, the workflow is highly resource-intensive and time-consuming, making it difficult to scale to large populations (Rajat S. Chandra et al., Myopia Prediction Using Machine Learning: An External Validation Study).
The Conundrum of Large-Scale Screenings
Because of these clinical bottlenecks, large-scale vision screenings and epidemiological field studies frequently resort to non-cycloplegic refraction. However, bypassing the drops introduces systematic measurement errors. Qinghua Ma et al. demonstrate in The Relationship of Season of Birth with Refractive Error in Very Young Children in Eastern China that relying on non-cycloplegic methods in young children heavily confounds results due to their active ocular accommodation during testing.
The downstream effect of this bias is a substantial distortion of public health data. Meta-analyses and screenings by Li Dong et al. (Screening for novel risk factors related to high myopia using machine learning) and Hui Gao et al. (Prevalence of myopia in Chinese children and adolescents: a systematic review and meta-analysis) emphasize that non-cycloplegic data skews epidemiological conclusions by overestimating how widespread myopia truly is. Even when massive datasets are compiled—such as the 1.2 million children evaluated by Jieying Guan et al. in Prevalence Patterns and Onset Prediction of High Myopia for Children and Adolescents in Southern China via Real-World Screening Data: Retrospective School-Based Study—the foundational non-cycloplegic screening data contains inherent inaccuracies that necessitate downstream clinical verification.
Subtracting Bias via Predictive Modeling
To reconcile the convenience of non-cycloplegic testing with the accuracy of cycloplegic results, researchers are developing computational models designed to mathematically "subtract" accommodation bias.
The core intuition behind this approach is that while accommodation alters the refractive state, it leaves distinct structural and optical signatures that can be modeled. For example, Qiang Su et al. demonstrated in Predictive Modeling of Cycloplegic Refraction Using Non-Cycloplegia Ocular Parameters With Emphasis on Lens-Related Features that machine learning models can leverage non-cycloplegic ocular parameters—specifically focusing on lens-related features—to accurately predict what the true cycloplegic refraction would be. By isolating how the lens behaves under natural accommodation, these models infer the underlying baseline physical structure without requiring chemical paralysis.
[Non-Cycloplegic Data] + [Lens-Related Features] ───► [Machine Learning Model] ───► [Predicted True Cycloplegic Refraction]
│ ▲
└───────────► (Contains Accommodation Bias / Involuntary Strain) ──────────────────────┘
(Model mathematically isolates and removes bias)
Other efforts focus on optimizing screening modalities to minimize or account for this error. In a separate study, Qiang Su et al. designed a smartphone-based subjective refraction platform (Innovative myopic screening platform based on smartphones) that utilizes the human iris as a natural ruler for distance calibration, explicitly accounting for the involuntary eye strain typical of younger users. Concurrently, researchers like Yadi Lei et al. have evaluated advanced binocular wavefront optometers (Comparisons of objective and subjective refraction with and without cycloplegia using binocular wavefront optometer with autorefraction and retinoscopy in school-age children) to map exactly how non-cycloplegic measurements systematically deviate toward myopic values.
By shifting the burden of accommodation control from pharmacology to machine learning and precise optical modeling, this research direction aims to make large-scale, accurate vision screening accessible outside the specialized clinic.
Key Foundational Work
The Relationship of Season of Birth with Refractive Error in Very Young Children in Eastern China (2014) — contributed 2 concept elements to this direction.
Breastfeeding and myopia: A cross-sectional study of children aged 6–12 years in Tianjin, China (2018) — contributed 1 concept elements to this direction.