Académique Documents
Professionnel Documents
Culture Documents
Abstract
Introduction: The influence of the radiographic appear-
ances of the coronal and intraradicular areas on periap-
ical radiographic interpretation has been minimally
R adiography is an inte-
gral part of dentistry
and endodontics. Radio-
Signicance
This study asserts that the radiographic appear-
ances of coronal and intraradicular areas inuence
evaluated in dentistry and endodontics. The purpose graphic images, in conjunc-
the interpretation of periapical areas. This raises
of this study was to evaluate the effects that the coronal tion with clinical testing, are
concerns regarding the validity of radiographic
and intraradicular radiographic appearance has on end- routinely used for end-
outcome assessments of the periapical area that
odontists radiographic interpretations of periapical odontic diagnosis, treat-
have been used as outcome measures for decades
areas. Methods: In a split-group study design using ment planning, treatment,
to assess periapical healing. This effect may be sig-
an online survey format, 2 pairs of digital periapical and as an outcome measure
nicant in previously endodontically treated cases.
radiographic images were evaluated by 2 groups for evaluating treatment
Both routine radiographic recall assessments and
(A and B) of endodontist readers for the presence of a efficacy. Interpretation of
radiographic outcome research should consider
periapical finding. The images in each pair were iden- radiographic images in-
the potential inuences of coronal and intraradicu-
tical except that 1 image of each image pairs had coro- volves both perceptual and
lar areas.
nal restorations and/or root canal fillings altered using cognitive skills. Visual in-
Adobe Photoshop software (Adobe Systems, San Jose, spection of the image is a
CA). The periapical areas were not altered. Using a 5- perceptual skill, whereas reasoning through possible diagnoses and inferring decisions
point Likert scale, the endodontist readers were asked is a cognitive skill.
to Please evaluate the periapical area(s). A Mann- Shortly after World War II, a medical study reported inter- and intraobserver var-
Whitney U test was used to statistically evaluate the dif- iations among radiologists of more than 30% (1). Since then, considerable research
ference between the groups. Significance was set at has been performed in medical radiography of the various influences on the perception
P < .01. Results: There were 417 readers in group A of radiographic findings; however, far fewer studies have been performed in dentistry
and 442 readers in group B. The Mann-Whitney U test (213). Most endodontic radiographic research has focused on diagnostic accuracy,
showed a significant difference in the responses overlooking the directly related but more fundamental perceptual and interpretive
between the groups for both image pairs (P < .01). issues (1417). Goldman et als classic study (6) showed that radiographic interpre-
Conclusions: Because the periapical areas of the image tation is unreliable and that success depends on who is reading the radiograph. Two
pairs were unaltered, the differing coronal and intrara- previous studies have shown the effects of the coronal and intraradicular areas on the
dicular areas of the radiographs appear to have influ- radiographic assessment of periapical areas (13, 18).
enced endodontists interpretations of the periapical This study examined the effects the coronal and intraradicular areas of dental
areas. This finding has implications for all radiographic radiographic images have on endodontists interpretations of periapical areas. The
outcome assessments. (J Endod 2017;43:723727) null hypothesis was that the coronal and intraradicular areas would not affect the peri-
apical interpretation.
Key Words
Dental radiography, diagnosis, endodontics, interpreta- Materials and Methods
tion, outcomes, perception This study was performed in accordance with a protocol approved by the Institu-
tional Review Board of Saint Louis University, St Louis, MO (Institutional Review Board
#26444). All patient identifiers were removed from all radiographic images before their
use in the study.
A list of e-mail addresses of potential study participants was compiled for a total of
4386 potential participants. Using a random number technique, the potential partici-
From the *Private Practice, Butte, Montana; Graduate Endodontics and Center for Advanced Dental Education, Saint Louis University, St Louis, Missouri; Illinois Air
National Guard and kPrivate Practice, Durango, Colorado.
Address requests for reprints to Dr John F. Hatton, Graduate Endodontics, Center for Advanced Dental Education, Saint Louis University, 3320 Rutger Street, St Louis,
MO 63104. E-mail address: Hattonjf@slu.edu
0099-2399/$ - see front matter
Copyright 2017 American Association of Endodontists.
http://dx.doi.org/10.1016/j.joen.2017.01.003
JOE Volume 43, Number 5, May 2017 Radiographic Perception of the Periapical Region 723
Clinical Research
pants were randomly assigned to 1 of 2 groups referred to as group A images than group B. Each group viewed only 1 image from each image
and group B. This provided a total of 2193 potential participants in test pair (Fig. 2). Time was not limited for completion of the question-
group A and 2193 potential participants in group B. naire; however, participants were prevented from returning to previous
An invitation to participate in the study was e-mailed to the 4386 pages.
potential participants on March 7, 2016. The invitations were hyper- Each image had a 5-point Likert scale (referred to as the confi-
linked to 2 similar, but slightly different, online questionnaires using dence scale) associated with it. Using this scale, participants re-
an online survey tool (SurveyMonkey, Palo Alto, CA). One week after sponded to the following question: Please evaluate the periradicular
the initial invitation, a second invitation was e-mailed to individuals area(s). The scale used in this study was as follows: 1, definitely,
who had not opened the hyperlink. The study was closed on March or almost definitely, an abnormal feature; 2, probably abnormal;
21, 2016, after 2 weeks of availability. All participants were volunteers. 3, possibly abnormal; 4, an anomalous feature present but probably
No incentives were offered. normal; and 5, definitely, or almost definitely, normal. Although Lik-
A total of 6 digital periapical radiographic images were selected for ert scales typically have a neutral option, the selected scale used in the
use in this study. Each image was cropped to consist of a single tooth study did not. The wording for this scale was based on a similar one
with a clearly visible crown, root, and periapical area. Two of the 6 im- used in medical radiography perception research (19).
ages served as control images, and 2 served as test images. The 2 test In addition, the participant was asked the following question for an
images were of single-rooted maxillary premolar teeth with ambiguous action category: What is your recommendation to the patient? The
periapical areas. Each of the 2 test images were then digitally modified choices of the 3-point Likert scale were Treatment, Future re-evalu-
using Adobe Photoshop software (Adobe Systems, San Jose, CA) to ation, or No endodontic treatment/no future re-evaluation.
create 2 test image pairs. Image pair 1 was modified to suggest an For result analysis, the 2 control images consisting of obvious
adequate coronal restoration with an inadequate (short) obturation. normal and abnormal periapical areas were used to assess partic-
Image pair 2 was digitally modified to suggest an adequate coronal ipants for eligibility. The intent was to ensure that the participant was
restoration with an adequate obturation (Fig. 1). The periapical areas focused on the questionnaire and capable of identifying normal and
remained unaltered in all images and were identical in each image pair. abnormal periapical areas. Data were excluded from participants
The study used a split design to allow for comparison of the re- who failed to correctly score the normal and abnormal control images.
sponses between each pair of test images. Each respondent was asked Data were excluded if participants identified the normal image as
his or her number of years of experience as an endodontist. Both ver- definitely, or almost definitely, an abnormal feature, probably
sions of the online questionnaire included 6 pages with a single radio- abnormal, or possibly abnormal. Data were excluded if participants
graphic image and the following Likert scale question: Please evaluate identified the abnormal image as an anomalous feature present but
the periapical areas. The 6 images shown on these pages included 2 probably normal or definitely, or almost definitely, normal. Data
test images, 1 from each image pair; 2 control images; and 2 addi- were also excluded from participants who failed to score both test im-
tional images that were not scored. Both groups viewed the same con- ages. These screenings effectively were exclusion/inclusion criteria,
trol images and nonscored images, but group A viewed different test such that the remaining data set was considered for analysis (Fig. 2).
Statistical analysis was performed using SPSS for Windows Version
Images Viewed Images Viewed by 14.0 (SPSS Inc, Chicago, IL). The median and mode were calculated for
by Group A Group B the confidence scales of each test image. The medians of the confidence
Image scales of each image pair were compared using the Mann-Whitney U
Pair 1 test. Significance was set at P < .01 to reject the null hypothesis. Using
the Mann-Whitney U test results, the Cliffs delta effect size of the data
was calculated to evaluate the magnitude of difference between the orig-
inal and modified images of each image pair. The null hypothesis was
that there would be no differences between the images of each image
pair.
Results
There were 417 responses that met the inclusion criteria for group
A for a response rate of 19% and 442 responses that met the inclusion
Original Modified criteria for group B for a response rate of 20%. The mean/median num-
Image ber of years of endodontic experience was 14.4/12 for group A and
Pair 2 14.2/11 for group B. Fifty-two (12%) of the readers in group A and
50 (11%) of the readers in group B were endodontic residents.
For image pair 1, the results of the confidence scale resulted in
both a mode and median for group A of possibly abnormal; for group
B, the mode was definitely abnormal, and the median was probably
abnormal (Fig. 3). For image pair 2, the mode and median were
possibly abnormal for group A, and the mode and median were defi-
nitely abnormal for group B (Fig. 4).
Mann-Whitney U statistics compared the median Likert scores of
both image pairs and showed significant differences with actual P values
Modified Original <.001 (Table 1). The null hypothesis was rejected for both image pairs
because the calculated P value was less than the preselected significance
Figure 1. The 4 test images or 2 image pairs. level (P < .01).
Randomization and
Allocation
Test
Images
Using the Cliffs delta, effect sizes were calculated to evaluate the ranges from 1.0 to +1.0, with a value of 0.0 designating zero effect
magnitude of the differences between the original and modified images (21). The calculated effect sizes (+0.51 and +0.55) strongly suggest
for each image pair. The Cliffs delta for image pair 1 was +0.51, and for that radiographic interpretations of the periapical areas by endodontists
image pair 2 it was +0.55. can be biased by the radiographic appearances of the coronal and intra-
radicular areas. Because the radiographic images had ambiguous periap-
ical areas, the judgments regarding these areas were not straightforward.
Discussion Endodontists (and residents) who participated in this study appear to
It should be emphasized that the periapical areas were identical in have interpreted a greater prevalence of apical disease when associated
both image pairs and that participants were instructed to Please eval- with questionable coronal and intraradicular radiographic findings.
uate the periradicular area(s). The results suggest that because the Morgental et al (13) used an Adobe Photoshop image alteration
periapical areas of the image pairs were identical, the differences be- technique to compare periapical evaluations. They had 10 Brazilian
tween the interpretations of the image pairs can be directly attributed general dentists compare original radiographic images with images
to the coronal and intraradicular differences of the radiographic images. digitally modified to either fully mask the root area or the periapical re-
An effect size is a quantitative measure of the strength of a phenom- gion. Although their experimental technique differed substantially from
enon (20). The Cliffs delta is a nonparametric effect size measure that the Photoshop alteration technique used in this study, they had remark-
quantifies the amount of difference between 2 groups of observation ably similar findings. They concluded that root filling quality substan-
beyond P value interpretation (21). The effect size for the Cliffs delta tially affects periapical diagnoses.
JOE Volume 43, Number 5, May 2017 Radiographic Perception of the Periapical Region 725
Clinical Research
In medical radiology interpretation studies, action categories are responses as opposed to the amount of responses, we believe asking
often used instead of a Likert scale to assess confidence in the call and for action categories based on radiographic findings in isolation may
make recommendations for a future course of action such as earlier not be appropriate, and, therefore, the data are not reported.
follow-ups or biopsy (22). For this study, we piloted asking for an action It must be recognized that this study design has limitations and var-
category, which required making a decision as to a course of action, iables. The respondents to the questionnaire were a fraction of the prac-
including treatment. Although the results of the action category strongly ticing endodontists in the United States, such that those who responded
supported the studied effect, we received multiple e-mails expressing a to the questionnaire introduced a self-selection bias into the results. Of
general objection to making a decision as to a course of action without the respondents who completed the survey, 369 (30%) were excluded
subjective and objective patient information. Based on the level of because of incorrect responses to the normal and abnormal
JOE Volume 43, Number 5, May 2017 Radiographic Perception of the Periapical Region 727