In what order should you read a bitewing for caries?

Read a bitewing for caries in a consistent sequence, tooth-by-tooth. Start with a quick look at the whole radiograph, then examine the proximal surfaces one arch at a time, followed by the occlusal surfaces, restoration margins and cervical regions. After every finding, reset and continue until the entire image has been covered.

There is no good evidence that one particular bitewing reading sequence is more accurate than another. The value of a fixed route is simpler than that: it stops you skipping things.

Two habits probably matter more than the exact path your eyes take:

Look wide before you look close.

Do not stop searching after the first finding.

A practical bitewing reading sequence

If you want a fixed route, use this:

1. Look at the whole bitewing first. Orient yourself, check whether the image is diagnostic, look for overlapped contacts and notice anything that immediately stands out.
2. Read the proximal surfaces tooth by tooth, one arch at a time. Start with the most distal tooth and work mesially. At every contact, deliberately assess both proximal surfaces.
3. Make a separate pass over the occlusal surfaces.
4. Check every restoration margin.
5. Make a separate pass over the cervical regions.
6. After every finding, reset and continue until you have covered the whole image.

This sequence is not proven to be better than another systematic route.

Its value is coverage.

Does the exact order matter?

Less than you might expect from how firmly particular sequences get taught.

Early on, a fixed route is useful because it stops you wandering around the image and forgetting the last molar, a restoration margin or an entire contact. But the order itself is not what separates a strong reader from a struggling one.

Start with one of the closest things we have to a test of an "order-like" intervention.

A vocalized checklist did not reduce or eliminate satisfaction of search, the tendency to miss a second finding after finding a first, across 20 radiologists reading 64 chest radiographs. The authors still noted that a checklist may help organise search and reporting, particularly among students, but it did not prevent the errors it was intended to prevent.¹

The same group found that computer-aided detection prompts that always pointed correctly at an added lesion also failed to counteract satisfaction of search.²

That is chest radiography, not dentistry. The distinction matters, and much of the visual-search evidence in this article comes from general radiology.

The evidence usually offered in support of systematic searching is real, but weaker than it sometimes sounds.

More systematic scanning patterns were observed in experienced observers than in inexperienced observers searching hand and wrist radiographs for fractures.³ In volumetric chest imaging, greater image coverage was a strong predictor of accuracy.⁴

Both findings are useful.

Neither proves that adopting a particular route makes you accurate.

Experienced readers searching systematically may simply be what expertise looks like from the outside. It does not necessarily tell us what produced that expertise.

Dental evidence points to another part of the problem.

In a randomised study of 73 second-year dental and dental hygiene students, those asked to give a diagnosis first and then identify the supporting features were more accurate than those who listed features first and then diagnosed.⁵

That study is about reasoning rather than visual coverage, but the implication is important:

Getting your eyes to the right place is only one part of interpretation.

The fairest summary is this:

No retrieved study compares one bitewing reading sequence with another, or with unstructured looking, for diagnostic accuracy.

What the dental error literature does support is teaching systematic search. In a mixed-methods study of fifth- and sixth-year dental students, false-negative errors were the most frequent interpretation error, and overlooking, inattentional blindness and satisfaction of search were all linked to ineffective visual scanning.⁶

So a route is worth having.

Just do not mistake the route for the skill.

Look wide before you look close

Before you settle on a single tooth, spend a few seconds looking at the whole radiograph.

There is a mechanism behind this, and it has been studied for decades.

Kundel and Nodine showed in 1975 that radiologists shown a chest radiograph for only 0.2 seconds, far too short for deliberate visual search, still achieved 70% true positives. That rose to 97% with unlimited free search. They described interpretation as beginning with a global first impression that helps organise the detailed checking fixations that follow.⁷

The wide look is not extra work.

It helps stop you starting in the wrong place.

Ask four questions.

Which side, which teeth?

Orient yourself before anything else.

Identify the quadrant and tooth range so you know what anatomy should be present.

Is the radiograph diagnostic?

The FGDP two-tier grading used for intraoral radiographs provides a useful gate. A radiograph is diagnostically acceptable when it has no or minimal errors in patient preparation, exposure, positioning, receptor processing or image reconstruction, and when it is of sufficient image quality to answer the clinical question.⁸

That second part matters.

A radiograph can look technically acceptable and still fail to answer the caries question you took it for.

In an audit of 300 intraoral periapical radiographs taken by dental assistants at one dental school, 65.67% were graded diagnostically acceptable and 34.33% were not. Contrast problems were the most common fault at 28.3%, followed by incorrect receptor positioning at 22%, incorrect vertical angulation at 12.3%, processing errors at 5% and incorrect horizontal angulation at 3.7%.⁸

That was periapical imaging at one institution.

Treat those numbers as examples of failure modes, not rates to expect on bitewings.

Are the contacts open?

An overlapped contact can hide a lesion or make a normal surface look suspicious.

You need to know which surfaces are compromised before you begin calling them sound or carious.

Among 74 dental students reading 50 bitewing radiographs for enamel-only proximal caries, 38% of proximal surfaces were overlapped. Both positive and negative agreement with the reference standard were significantly lower on overlapped surfaces than on non-overlapped ones, with p < .001.⁹

Overlapped contacts may simply be unreadable for the question you are asking.

Unreadable is not the same as sound.

Does anything immediately stand out?

Take in the image before focusing tightly on individual teeth.

Notice the obvious findings.

Do not chase them yet.

The first thing you see should not be allowed to choose the rest of your search.

Then go close.

The three ways to miss a lesion

When something gets missed, eye-tracking work in radiology separates the failure into three broad categories.¹⁰

Sampling errors: the finding is never fixated at all.

Recognition errors: the finding is looked at, but not recognised as abnormal.

Decision errors: the finding receives enough attention, but is ultimately called normal anyway.

These are different problems.

"I never looked at the distal surface of the last molar" is not the same as "I looked straight at the lesion and decided it was normal."

They improve in different ways.

This is also why a fixed route can only do so much.

A systematic sequence mainly helps with sampling errors, the places your eyes never reached.

It does much less for recognition and decision errors.

That is not an argument against having a route.

It is an argument against expecting the route to carry you.

Eye-position recordings in chest radiograph interpretation found that many missed nodules had actually been fixated. Obvious abnormalities also captured attention and reduced vigilance for subtler findings.¹¹

Dental eye-tracking research points in a similar direction.

In a study of 65 final-year dental undergraduates reading panoramic radiographs, participants frequently failed to reach the correct diagnosis despite having fixated the relevant regions. Their searches showed no clear sequential viewing pattern, and they spent substantially more time examining the dentition than the surrounding structures.¹²

That was panoramic imaging rather than bitewings.

But the shape of the failure is the same:

The eye arrived, and the diagnosis did not.

The fourth miss: you found one thing and stopped looking

There is another way to miss something, and it is harder to notice in yourself.

You find one abnormality and quietly stop searching.

This is called satisfaction of search.

It has been studied in radiology for more than fifty years. The modern experimental version was established by Berbaum and colleagues in 1990, whose readers of chest and abdominal radiographs showed reduced accuracy for original lesions after another abnormality had been added to the image.¹³ ¹⁰

Dental-specific evidence exists, although it is limited.

In a pilot study, 16 examiners, including endodontic and periodontic residents, alumni and faculty, read six periapical radiographs. Four contained a second abnormality inserted digitally.

A true satisfaction-of-search error occurred in 20.31% of interpretation sets, 13 of 64. When a second problem was present, the original lesions were reported 25 times out of 64, compared with 30 out of 64 when no second abnormality was present.¹⁴

That was a small study using periapical radiographs.

But it is among the closest dental evidence available.

The severity of the first finding may also matter.

In simulated multitrauma interpretation, adding a low-morbidity fracture substantially reduced detection of subtle test fractures, while adding a high-morbidity fracture did not produce the same effect.¹⁵

That result should not be transferred directly to bitewings.

But it raises an uncomfortable possibility: the thing that ends your search does not have to be dramatic.

It could be a small overhang, a bit of calculus or an obvious restoration.

Something noticeable enough to capture your attention without you consciously deciding that the search is finished.

You cannot reliably feel this happening in the moment.

So build a habit against it.

The stop rule: how you know the read is over

After every finding, reset and continue.

Every remaining contact.

Every restoration margin.

Every surface you intended to examine.

The read is finished when you have covered the entire radiograph.

Not when you have found something.

A simple prompt can help:

Found one. What else?

One finding is a reason to keep looking, never a reason to stop.

That is the stop rule.

Three places to go back to

Some findings are particularly easy to skate past once your attention has already been captured elsewhere.

Early proximal lesions

Radiographic detection is imperfect, particularly for early lesions.

In a meta-analysis of 117 studies covering 13,375 teeth and 19,108 surfaces, radiographic detection of lesions including initial caries showed low sensitivities, reported between 0.24 and 0.42, with moderate to high specificities between 0.70 and 0.97. For dentine lesions, sensitivity ranged from 0.36 for proximal lesions to 0.56 for occlusal lesions.¹⁶

Much of the evidence was at high risk of bias and combined in-vitro with clinical studies.

The practical point is:

A radiographically sound-looking proximal surface does not rule out early caries.

Radiographic appearance has to be interpreted in the context of lesion stage and the rest of the clinical picture.

How early a proximal lesion can be seen is a question in its own right, and caries depth classification is how you record what you find.

Restoration margins

Secondary caries at restoration margins had reported radiographic sensitivity in the range of 0.50 to 0.59 and specificity from 0.78 to 0.83 in a body of evidence that was mostly in vitro, focused on permanent posterior teeth and judged at high risk of bias.¹⁷

A restoration is also a high-contrast object sitting in the middle of your search.

That makes every restoration margin worth a deliberate pass rather than something you hope to notice incidentally.

Residual caries and restorative materials that are themselves radiolucent may not announce themselves during a quick coverage pass.

You have to be asking the question.

The places you stared at and dismissed

This one is counterintuitive.

In fracture-search eye-tracking research, gaze durations were significantly longer for false-negative decisions than for true-negative decisions.³

The place you stared at, hesitated over and eventually called normal may deserve another look.

And name the surfaces you cannot interpret.

An overlapped contact recorded as unreadable is an honest result.

Recorded as sound, it becomes a false-negative risk.

Whether feedback on your own search behaviour can improve detection has been tested in radiology, although the evidence is small.

A randomised trial of ten radiology residents reported a 38.9% absolute improvement in detecting suspicious pulmonary nodules among residents given automated gaze feedback, compared with 5.6% in the control group.¹⁸

Ten residents.

One centre.

Chest radiographs.

Interesting, but not enough to build a curriculum around.

So what should the sequence look like in practice?

If you want a route, here is a workable one.

It is scaffolding, not the point.

Any fixed order that gives you complete coverage could work. The advantage of making it fixed is that your path is not chosen by whatever happens to be brightest or most obvious on the image.

The one piece of evidence describing systematic search on bitewings comes from an eye-tracking trial of 22 dentists reading approximately 18 bitewing radiographs each, randomised with and without AI assistance.

A more systematic search pattern was characterised by a higher proportion of lateral, tooth-wise transitions, moving tooth by tooth along the arch.¹⁹

That study measured how AI changed dentists' gaze.

It did not test whether adopting a fixed sequence improves diagnostic accuracy.

With that caveat, here is the route in more detail.

1. Take the wide look

Orient yourself.

Check image quality.

Check the contacts.

Notice obvious findings without allowing the first one to take over the search.

2. Read the proximal surfaces tooth by tooth

Choose one arch.

Start with the most distal tooth and work mesially.

At every contact, assess both proximal surfaces before moving on.

Then repeat the same route on the opposing arch.

Proximal caries deserves its own deliberate pass on a bitewing caries read.

3. Make a separate pass over the occlusal surfaces

Use the same tooth-by-tooth route back.

Occlusal caries has its own limitations on this modality and is better treated as a separate question than something you hope to pick up while concentrating on proximal contacts.

4. Check every restoration margin

Again, tooth by tooth.

Do not let the restoration itself become the finding that ends your search.

5. Check the cervical regions

Normal anatomy and radiographic effects can mimic disease here.

Cervical burnout and the Mach band effect can create radiolucent appearances that are not carious lesions.

The distinction causes enough trouble that it has its own article.

6. Apply the stop rule

After every finding:

Found one. What else?

Continue until the image has been covered.

Then stop.

Do not memorise the route harder than you learn the disease

A systematic route is useful because it helps you cover the image.

That matters particularly while you are learning.

But the exact direction your eyes travel is not the whole skill.

Spend your effort on three things:

Coverage: Did you actually inspect the whole image?

Recognition: When your eyes reached an abnormality, did you recognise it?

Decision: Once something looked suspicious, did you classify it appropriately?

And keep the two habits from the beginning:

Look wide before you look close.

After every finding, keep searching.

The route becomes much less important once you know what you are looking for.

Where the evidence is unsettled

Most of the visual-search literature discussed here comes from general radiology rather than dentistry.

That matters.

Satisfaction of search

Dental-specific evidence exists but is thin: one pilot using periapical radiographs¹⁴ and one student study.⁶

Most of the evidence comes from chest and trauma imaging.¹³ ¹⁰ ¹¹ ¹⁵

Search behaviour on bitewings

There is one descriptive eye-tracking trial.¹⁹

No dental trial comparing one fixed bitewing reading sequence with another was found.

Has the exact order been tested?

No retrieved study directly tested whether one bitewing reading sequence is more accurate than another.

The wider dental literature suggests this is a genuine evidence gap.

A systematic review of interpretation errors in dentistry identified six contributing factors: clinical experience, clinical knowledge and technical ability, case complexity, time pressure, the location and duration of education and training, and cognitive load. It also concluded that interpretation errors have not been widely investigated in dentistry.²⁰

Even in radiology, where visual-search research is much deeper, a review of perceptual learning concluded that the precise visual features expert radiologists use in clinical practice remain incompletely understood.²¹

So use a route if it helps you cover the image.

Hold the exact route loosely.

The habits are the part worth drilling.

Reading about the stop rule and using it are different skills. Try three real bitewings and find out whether you keep looking after the first finding: take the 2-minute reading test.