How Accurate Is Smartphone Interpupillary Distance Measurement?

01 · STUDY DESIGN

Forty adults and four measurement methods

Forty participants with a mean age of about 25.7 years underwent distance and near binocular IPD measurement using a pupillometer, a ruler, an autorefractor, and Eye Measure 1.22 on an iPhone. The autorefractor contributed distance data only.

The pupillometer was selected as the reference. Measurements were repeated to examine consistency, but the sample excluded manifest eye misalignment and does not automatically represent children or complex cases.

02 · RESULTS

Similar group averages can hide individual measurement differences

Mean distance IPDs were 64.2 mm with the pupillometer, 64.8 mm with the ruler, 63.6 mm with the app, and 64.4 mm with the autorefractor. The app averaged about 0.6 mm below the reference. In pairwise comparisons of distance IPD, the app gave statistically significantly smaller readings than the pupillometer, ruler, and autorefractor. For near IPD, the app did not differ significantly from the pupillometer, whereas the ruler gave higher readings than both.

The limits of agreement for app versus pupillometer distance IPD were approximately −2.89 to +1.70 mm. A modest mean difference can therefore conceal larger individual discrepancies. Good repeatability is not the same as agreement with the reference.

03 · OPTICAL EFFECT

The optical effect of a PD error depends on lens power

Using Prentice’s rule, the paper illustrates that 0.6 mm of decentration with a −3.00 D lens may induce approximately 0.18 prism diopters. The authors also note that this example is below the cited 0.33-prism-diopter horizontal induced-prism tolerance. It is a conditional calculation, not a universal safety conclusion for all prescriptions.

High prescriptions, facial asymmetry, monocular PD, and multifocal fitting heights require additional attention. The study assessed binocular IPD and did not validate monocular PD accuracy.

04 · LIMITATIONS

Useful screening is not the same as final dispensing

The authors saw potential in screening where services are limited, while urging caution for complex prescriptions and misaligned eyes. Only one application was tested, so the findings cannot validate other apps or devices.

A developer needs separate validation for camera distance, head rotation, gaze, and device model. Clear communication about what a displayed measurement can and cannot be used for matters as much as the algorithm.

05 · MEASUREMENT CONDITIONS

Why the measurement conditions matter

The study measured binocular distance and near interpupillary distance twice in 40 adults with a mean age of 25.65 years. Ruler-based distance measurements used a target approximately 6 m away; near fixation was around 40 cm. The phone app acquired facial measurements with the phone about 40 cm away.

Pupillometers, rulers, autorefractors and camera apps do not use identical measurement principles. The study excluded participants with abnormal ocular alignment and assigned trained examiners to the instruments. Unsupervised measurements taken at home should not be assumed to replicate these conditions.

Near PD can differ from distance PD because the eyes converge when fixating a close target. A recorded number needs the target distance and measurement method to remain interpretable.

06 · STATISTICAL READING

A small average difference is not individual agreement

For distance binocular PD, the study reported a mean of 64.2 mm with a pupillometer and 63.6 mm with the phone app. The average difference was only 0.6 mm, but agreement analysis showed that individual discrepancies could be larger.

Bland–Altman limits of agreement describe the distribution of differences between methods. The reported app-versus-pupillometer distance limits of approximately −2.89 to +1.70 mm show why one average correction cannot eliminate error for every person.

Repeatability—getting a similar number on repeated attempts—and accuracy relative to a reference instrument are different properties. Responsible reporting of mobile measurement should separate them.

07 · OPTICAL APPLICATION

The validation steps before using app data for dispensing

For higher-power lenses, small centration differences can generate meaningful induced prism. Prentice's rule links lens power and decentration, illustrating why the same geometric error need not have the same optical consequence across prescriptions.

This paper primarily evaluated binocular PD. It did not comprehensively validate monocular PD, fitting height, frame positioning, children or patients with ocular misalignment. The result of one iPhone app cannot certify the accuracy of all facial-measurement services.

A deployable system should separately test device models, camera distance, head rotation, gaze, lighting, error distributions and failure rates—and explicitly disclose conditions under which measurement is not possible. A number without a defined scope is not a complete measurement product.