DentAI – Liquid on the Tooth Surface and Intraoral Scanner Accuracy

Liquid significantly reduces accuracy

The wet condition had the worst trueness and precision in every case, and the dry condition was the best. This was significant for both trueness and precision. The practical point is that the liquid-induced error is an optical-physical phenomenon, not a chemical one. For this reason no significant difference was seen between pure water and artificial saliva. That is, within the scope of this study the chemical composition of the liquid is not decisive; the mere presence of a liquid film on the surface produces the error.

Note that this equivalence was measured only between two transparent liquids that refract light (water and saliva). Blood behaves differently, because instead of simply refracting light it absorbs it and is opaque, so it should not be assumed to belong to the same category, and bleeding control is a separate story. In practice, saliva, gingival crevicular fluid, and rinsing water must all be controlled, not just one of them.

Where the error concentrates

The deviations caused by liquid were almost always positive, meaning the digital surface was recorded higher and thicker than reality. These deviations were not spread uniformly but were concentrated in three regions, all three of which are clinically sensitive:

• The pits and fissures of the occlusal surface of premolars and molars
• The interproximal area
• The margins of the abutments, especially around the shoulder, where more liquid is trapped

This distribution pattern is what determines the clinical importance of the study, because these regions are exactly the ones the final fit of the restoration depends on.

The 120-micron threshold

The positive deviations caused by liquid could exceed 120 microns. This number is not arbitrary, because 120 microns is the clinically acceptable marginal discrepancy. In other words, the presence of liquid at the margin alone can produce an error at or beyond the entire marginal error budget and create a virtual high point that ultimately leads to restoration misfit. Note that this is an inference based on comparison with a threshold, because no restoration was fabricated and seated in this study. But the direction and magnitude of the error are large enough that ignoring it would not be reasonable.

Drying is effective but not complete

Drying with an air syringe (10 seconds of compressed air) removed most of the liquid-induced error, and the deviations accumulated in the sensitive regions almost disappeared. This finding confirms the common clinical recommendation with direct evidence for the first time. However, the compensation was not complete. In some of the conditions, accuracy after drying was still statistically different from the initial dry state. The precise interpretation is that drying greatly reduces the error but does not always return it to the original zero level. The practical message is that drying is necessary but does not replace careful field-moisture control.

A second error source that has nothing to do with liquid

This part is clinically important because it is easily confused with liquid error. In the molar region, especially the second molars on both sides, a separate error was seen that was horizontal and in the buccolingual direction. This error has no optical or moisture origin; rather it arises from the stitching algorithm and error accumulation when stitching frames during a full-arch scan. The most important point is that this error is not removed by drying, because it does not come from liquid at all. In practice this means that if you see misfit in the posterior region of the arch, more drying does not necessarily help, and you should think about scanning strategy and keeping the stitching path short — not just moisture.

The difference between the two scanners

Both scanners were affected by liquid, but precision was mainly determined by the device itself rather than the moisture condition. Primescan had higher precision than Trios 3 under all conditions. The subtle point is that Trios 3 was more scattered from the start, and liquid did not make that scatter statistically significantly worse, whereas Primescan had a much better baseline accuracy that dropped more in the wet state but still remained better. When choosing a device for precise full-arch work, this baseline precision difference is a more decisive factor than sensitivity to moisture.

Clinical takeaway

Liquid on the tooth surface is a real, measurable obstacle to recording an accurate digital impression, not a marginal detail. Moisture control before scanning, with emphasis on the margin and abutment shoulder, can be the difference between an acceptable fit and a clinical misfit. Drying with an air syringe is an effective and recommended step, but we must realize that part of the full-arch error, especially in the posterior region, has an algorithmic origin and is not solved by drying.

A temporal caveat is also needed: these results pertain to the 2021 generation of these two devices and their software. The article itself notes at the end that scanners able to detect saliva and blood still have room for development. Newer generations with better processing may change the magnitude of this error, although the physical logic of the liquid's effect still holds.