- Email: [email protected]
Visual Fixation and Scan Patterns of Dentists Viewing Dental Periapical Radiographs: An Eye Tracking Pilot Study
Clinical Research
Visual Fixation and Scan Patterns of Dentists Viewing Dental Periapical Radiographs: An Eye Tracking Pilot Study Brian P. Hermanson, DDS, MSD,* Grant C. Burgdorf, DDS, MSD,† John F. Hatton, DMD,‡ Darrin M. Speegle, PhD,§ and Karl F. Woodmansey, DDS, MA‡ Abstract Introduction: The visual search patterns of dentists and the areas that attract their attention when interpreting dental periapical radiographs are currently unknown. This research identifies areas and patterns of visual fixation when observing dental periapical radiographs. Methods: In an observational study using eye tracking technology and a convenience sample of 44 observers, the interpretations of 4 dental periapical radiographs were recorded using Camtasia Software (TechSmith, Okemos, MI) with a gaze tracking ‘‘bubble’’ denoting where within the radiograph the observers’ eyes gazed. The recorded observations included the scanning pattern, the area of first fixation, and revisits of areas. Also noted was whether the area of first fixation or revisit was radiopaque, radiolucent, or of normal radiodensity and whether it was a coronal or radicular area. Results: The first fixation is more likely to be an area of high contrast that is either radiopaque or radiolucent compared with areas that were normal or of average gray scale. Significantly more revisits occurred on areas that were radiopaque and located in the radicular area. Of the 4 categorized scanning patterns, tooth by tooth scanning predominated. Conclusions: When interpreting dental periapical radiographs, significantly more observers initially fixated on areas of the radiograph that were of high contrast (ie, radiopaque or radiolucent) compared with ‘‘normal areas.’’ A tooth by tooth scanning pattern was most commonly used. (J Endod 2018;-:1–6)
Key Words Dental, eye tracking, perception, periapical radiograph, radiography, x-ray
C
linical tests and periSignificance apical radiographs The visual search patterns of dentists and the areas provide dentists with inthat attract their attention when interpreting dental formation for pulpal and periapical radiographs are currently unknown. This periapical diagnoses. The is the first study to use eye tracking technology for interpretation of dental radental periapical/endodontic radiographic interdiographs is a dual propretation. Using eye tracking technology, this cess that requires both research identifies areas and patterns of visual fixperception and cognition ation of observers viewing dental periapical radio(1). The visual scan of graphs. the radiograph is perceptual, whereas the diagnostic reasoning and decision making are cognitive (2). Dental research has documented some aspects of the cognitive interpretation of dental radiographs; however, minimal research has documented perceptual components of dental radiographic interpretation. Two published studies in dental radiography have used computerized eye tracking technology to evaluate dentists’ perceptions of panoramic and computed tomographic images (3, 4). Computerized eye tracking technology allows researchers to specifically determine where an observer is gazing within an image and illustrates patterns in the scanning process. This is the first study to use eye tracking technology for dental periapical/endodontic radiographic interpretation. Two essential components of perception are termed fixations and saccades. Fixations are locations where the observer’s gaze pauses momentarily. Along with the belief that perception is coincident with fixations is the idea that fixations are where cognition occurs (5). Individual fixations are separated by saccades, which are uniform movements of both eyes at a very high speed directing the gaze onto the next fixation. Some saccades may be voluntary, whereas many are not (6). The purpose of this study was to use computerized eye tracking technology to determine visual fixation and scan patterns of observers when viewing dental periapical radiographs. The following parameters of interest were selected for examination: area of first fixation, revisits, search patterns, effect of radiodensity, effect of coronal or apical location, and differences between observer experience levels. The first null hypothesis of this study was that there would be no differences in fixations or revisits whether the structure was located in a coronal or radicular area or was of normal radiodensity, radiopaque, or radiolucent. The second null hypothesis was that there would be no difference between the scan patterns and the percentage of pathology, coronal, intraradicular, and periapical areas scanned. A third null hypothesis was that there
From the *Private Practice, Pierre, South Dakota; †Private Practice, New York, New York; ‡Center for Advanced Dental Education and §Mathematics and Statistics, Saint Louis University, St Louis, Missouri. Address requests for reprints to Dr Karl F. Woodmansey, Saint Louis University, Department of Endodontics, Center for Advanced Dental Education, 3320 Rutger Street, St Louis, MO 63104. E-mail address: [email protected] 0099-2399/$ - see front matter Copyright ª 2018 American Association of Endodontists. https://doi.org/10.1016/j.joen.2017.12.021
JOE — Volume -, Number -, - 2018
Visual Fixation and Scan Patterns of Dentists
1
Clinical Research would be no difference in scan patterns between observer groups of different types or experience levels.
Materials and Methods This study was performed in accordance with a protocol approved by the Institutional Review Board of Saint Louis University, St Louis, MO. A convenience sample of 44 observers with differing years of experience and levels of training were recruited. These observers consisted of third- and fourth-year dental students (n = 12), advanced education in general dentistry residents (n = 8), general dentists (n = 12), and endodontists (n = 12). Observers were excluded if they were unable to visually calibrate the eye tracking device. Digital periapical radiographs were obtained from patients treated at the endodontic clinic at the Saint Louis University Center for Advanced Dental Education. A Delphi panel consisting of 4 endodontists selected the radiographs to include a variety of radiographic findings including normal teeth, direct or indirect restorations, coronal caries, periapical radiolucencies, posts, and endodontic treatment. All radiographs had patient identifiers removed to maintain confidentiality. Four radiographs were selected for analysis: 2 from the maxillary arch and 2 from the mandibular arch (Fig. 1). Two were of the left side and 2 of the right side. Each periapical radiograph contained a first molar, second molar, and at least the second premolar. The full extent of a periapical radiolucency, if present, was visible in the radiograph. Six other radiographs were selected as ‘‘filler’’ or ‘‘practice’’ radiographs. The radiographs were displayed on a 21.5-inch Apple iMac Desktop computer/monitor (Apple Inc, Cupertino, CA) using an automated PowerPoint presentation (Microsoft Corporation, Redmond, WA) with each of the 10 radiographs displayed in sequence for 10 seconds. The experimental radiographs were sequenced at positions 4, 5,
6, and 10. Data were collected using an optical video-based Tobii EyeX eye tracking device (Tobii Technology, Danderyd, Sweden), which illustrates gaze paths via an integrated 1.5-inch diameter gaze tracking ‘‘bubble’’ (Fig. 2A and B). A Camtasia Studio Software (TechSmith, Okemos, MI) recording of the display captured the path of the gaze tracking bubble. A digital video file preserved each observer’s gazes of the 4 radiographs. Observers first viewed introductory slides, which included a calibration exercise that ensured the Tobii EyeX could accurately track the observers’ gaze. Observers were instructed to interpret the radiograph as if a patient was waiting in the chair, that the radiographs would cycle automatically after 10 seconds, and to disregard the gaze tracking bubble. The Camtasia digital video files of the 4 experimental radiographs were blinded by an unassociated individual using a random number generator. Two blinded researchers then performed evaluations of the Camtasia recordings. When there was disagreement between the 2 evaluators in categorizing data, a consensus was reached after a discussion. In all instances, a consensus was able to be reached. Data from the other 6 radiographs were not evaluated. The observers’ area of first fixation was noted, defined as the first region of a definitive stop of the eye along the scan path. This area was categorized by location and radiodensity as coronal or radicular/periapical and normal, radiopaque, or radiolucent. The Delphi panel defined these categorizations. Radiopaque was defined as an area on the radiograph with radiodensity significantly greater than the average gray scale of the radiograph. These areas included metal restorations or root canal obturating materials. Radiolucent was defined as an area on the radiograph with radiodensity significantly lower than the average gray scale of the radiograph. These areas included periapical radiolucencies or carious lesions. Normal was defined as areas on the radiograph that were within the expected gray scale of the radiograph.
Figure 1. The 4 experimental radiographs.
2
Hermanson et al.
JOE — Volume -, Number -, - 2018
Clinical Research radiograph until all areas were scanned. Scan patterns that could not be categorized into those 3 categories were classified as random.
Statistical Analysis Data were analyzed using SPSS Statistics Software Version 24 (IBM Corp, Armonk, NY). Data from all participants across all 4 radiographs were pooled for the analysis. Comparisons between the observer groups are published elsewhere (7). For analysis of first fixations and revisits, intraradicular and periapical areas were pooled together because on many occasions the gaze tracking bubble would cover both the intraradicular and periapical areas simultaneously. Chi-square analysis was used to determine if significant differences existed for first fixations and revisits between coronal and radicular/ periapical areas and for differences among normal, radiopaque, and radiolucent areas. The significance level was set at P < .05. Post hoc pair-wise chi-square analysis was used to compare groups for significant differences between first fixation of normal and radiopaque, normal and radiolucent, and radiopaque and radiolucent areas. The significance level was set at P < .05. The Bonferroni correction for multiple comparisons was used. Using chi-square analysis, the scan patterns were compared for differences in the percentages of the pathology, coronal, intraradicular, and periapical areas scanned. The significance level was set at P < .05. Post hoc pair-wise chi-square analysis was used to compare for significant differences between each scan pattern and the specific area on the radiograph. The Bonferroni correction for multiple comparisons was used.
Results
Figure 2. (A) Study simulation with the Tobi EyeX eye tracking device (red arrow). (B) The gaze tracking bubble (yellow arrow).
Revisits were similarly recorded. These were defined as the first area returned to and fixated on after completion of the initial scanning pattern. For example, the first area that was fixated upon after a circular pattern was completed was categorized as a revisit. Revisits were not counted when a scan pattern was incomplete or for observers with a random pattern. Similar to areas of first fixation, revisits were categorized by location and radiodensity. Seven areas of pathology or abnormalities were identified across the 4 radiographs by the Delphi panel (Fig. 3). The number of areas of pathology or abnormalities scanned was tabulated to calculate the percent of pathology that the observer scanned in all 4 radiographs. Each tooth was divided into 3 regions: coronal, intraradicular, and apical. The total of the regions on each radiograph that were covered by the tracking ‘‘bubble’’ was calculated as a percentage of all possible regions that could possibly be viewed. Scan patterns of the entire radiograph over the 10-second period were subjectively classified as circular, tooth by tooth, or horizontal scanning (Fig. 4). Circular patterns were defined as a pattern that traced the periphery of the radiograph in either a clockwise or counterclockwise direction to complete a circle. A tooth by tooth pattern occurred when the observer began on the left or right side of the radiograph and scanned an entire tooth before moving to the adjacent tooth in the direction of the most distant tooth. A horizontal scanning pattern occurred when the observer began in either the coronal or apical region of the radiograph and scanned back and forth from the left to right of the JOE — Volume -, Number -, - 2018
A total of 44 screen recordings were obtained from the 44 participants; however, 2 participants were excluded because of an inability to calibrate the eye tracking device (1 general dentist and 1 endodontist). The resulting data yielded 42 recordings per radiograph for a total of 168 10-second Camtasia recordings. These included 168 scan patterns, 168 first fixations, and 75 revisits. The first null hypothesis of this study was that there would be no differences in fixations or revisits whether the structure was located in a coronal or radicular area or was of normal radiodensity, radiopaque, or radiolucent. Chi-square analysis showed no significant difference (P < .643) between first fixations on coronal or radicular/periapical areas, with 48.2% of first fixations occurring on coronal areas and 51.8% on radicular/periapical areas (Fig. 5). When compared with areas of normal radiodensity, a significantly greater (P # .001) number of first fixations were made on radiopaque or radiolucent areas (Fig. 5). When comparing these 3 individual categories, a statistically significant difference was found between normal and radiopaque and normal and radiolucent areas but not radiopaque and radiolucent areas with significance values of P < .001, P < .003, and P < .609, respectively (Table 1). Significantly more revisits occurred on radicular/periapical areas compared with coronal areas (P < .002), with 32% of revisits occurring on coronal areas and 68% on radicular/periapical areas (Fig. 5). When compared with areas of normal radiodensity, a significantly greater (P # .001) number of revisits were made on radiopaque or radiolucent areas. When comparing these 3 individual categories, a statistically significant difference was found between normal and radiopaque and normal and radiolucent areas but not radiopaque and radiolucent areas, with significance values of P < .001, P < .027, and P < .081, respectively (Table 1). The second null hypothesis was that there would be no difference between the scan patterns and the percentage of the pathology, coronal,
Visual Fixation and Scan Patterns of Dentists
3
Clinical Research
Figure 3. The 7 areas of pathology identified by the Delphi panel on the (A–D) 4 experimental radiographs.
intraradicular, and periapical areas scanned. The 168 observations were categorized into the following scanning patterns: 54.2% tooth by tooth, 31.5% circular, 12.5% random, and 1.8% horizontal. For statistical analysis, horizontal scanning was not included because its use was too low to be included. Chi-square analysis showed no significant difference in the percentage of pathology scanned or in periapical areas scanned by all scan patterns with P < .687 and P < .327, respectively. A significant difference was found in the percentage of coronal and intraradicular areas scanned by all scan patterns with P < .018 and P < .001, respectively. When comparing the different scan patterns, a significant difference was found in the percentage of the coronal areas scanned between tooth by tooth and circular scanning (P < .004). No significant difference was found between tooth by tooth and random or circular and random scanning with significance values of P < 1 and P < .146, respectively. A significant difference was found in the percentage of intraradicular areas scanned between tooth by tooth and circular and tooth by tooth and random scanning with significance values of P < .000 and P < .014, respectively. No significant difference was noted in the percentage of the intraradicular areas scanned between circular and random scanning with a significance value of P < 1.
Discussion The null hypotheses of this study were that there would be no differences in first fixation or revisits whether the structure was located in a coronal or radicular area or was of normal radiodensity, radiopaque, or radiolucent. The second was that there would be no difference between the scan patterns and the percentages of the pathology, coronal, intraradicular, and periapical areas scanned. These findings generally agree with those of Davis and Palladino (8) who reported that observers’ attention tends to ‘‘focus on objects 4
Hermanson et al.
that are smaller, brighter, more centrally located in an image.’’ A study by Berbaum et al (9) in 1996 reported that when areas of contrast were present, less time was spent scanning normal, low-contrast areas. Although the current study did not measure scanning times, it found that high-contrast areas were more likely to be the areas of first fixation. This study has numerous limitations and is therefore termed a pilot study. The most notable limitation is that the findings can only be related to these specific 4 radiographs and to this limited cohort of observers. Furthermore, these observers varied greatly in radiographic interpretation expertise. Fixations are typically measured in milliseconds. In their 1996 article, Berbaum et al (9) defined a fixation as gazing in a location within 1 degree for a minimum of 100 milliseconds with eye position measured 60 times per second. In the current study, gaze times could not be precisely measured. The first definitive stop along the scan pattern was considered the first fixation. This was subjective and complicated by the varying speeds at which the observers scanned the radiograph. The tracking ‘‘bubble’’ was inexact because of its size (Fig. 4B) but provided an adequate impression of where the observer was gazing. Observers were only given 10 seconds to evaluate each radiograph, which in most instances appeared to be an adequate amount of time to visualize all elements. However, some very systematic observers seemed slower and may not have had enough time for complete visualization, which may have affected the results. Many participants were observed to increase their speed throughout the study in order to view the entire radiograph. Although observers were instructed to interpret the radiograph as if a patient was waiting in the chair, a 10-second observation time was used to standardize the experimental protocol and to prevent observers from consciously trying to fixate on 100% of the radiograph. Because the 4 experimental radiographs were embedded within the 10-radiograph PowerPoint presentation JOE — Volume -, Number -, - 2018
Clinical Research 24
radiolucent
62 42
radiopaque
9
normal
77
29
revisits first fixations
51
Radicular
24
Coronal 0
87 81
50
100
Figure 5. First fixations and revisits, displayed by radiodensity and by radicular/coronal area. (first fixations n = 168; revisits n = 75)
In this study, the observers may have modified their visual searches to ensure that all areas of the radiograph were scanned in order to be a ‘‘good participant.’’ Although 2 observers fixated on the same areas, they may have actually interpreted the radiographs very differently. Goldman et al (12, 13) classically explored the cognitive aspects of dental radiographic interpretation, noting significant inter- and intraexaminer variability. Their findings show that radiographs are not ‘‘read’’ as if they were words on a page but are cognitively interpreted and not just visually processed. Consequently, the interpretation of radiographs is a human-centric process that has subjective variability between examiners. The present study reported where observers’ eyes gazed within these radiographs but did not demonstrate their cognitive interpretations. Observers’ interpretations may not necessarily be associated with their gaze fixations (14). Eye tracking research alone cannot fully explain radiographic interpretation; both perceptual and cognitive processes are necessary. Revisiting an area of potential abnormality was suggested by Brocklebank (15), recommending that a medical radiograph be reviewed systematically in a way that will cover the entire radiograph. She suggested noting any possible abnormalities and returning for closer examination once the initial scan of the radiograph was completed. The findings of this study support that concept. Only 12% of revisits occurred on areas of normal radiodensity or of average gray scale. Within the limitations of this study, when interpreting periapical radiographs, the first fixation is more likely to be an area of high contrast that is either radiopaque or radiolucent compared
TABLE 1. First fixation & revisit Chi-square analysis Figure 4. Scan patterns: (A) circular, (B) tooth by tooth, and (C) horizontal scanning.
with 3 unscored radiographs being viewed before the first scored radiograph, the observers had an opportunity to learn the flow of the automated 10-second timing. Another limitation of this study is that the observers were aware that their eye movements were being recorded. This may have led them to behave in ways that were different than if they were unaware their eye movements were being tracked. The term demand characteristic describes how experimental subjects attempt to identify the purpose of the study and accordingly modify their behavior (10, 11). JOE — Volume -, Number -, - 2018
Normal/ Normal/ Radiopaque/ Coronal/ Radiopaque Radiolucent Radiolucent Radicular First Fixations Chi21.736 square df 1 Asymp .000 Sig. Revisits Chi21.353 square df 1 Asymp .000 Sig.
11.967
1.619
.214
1 .003
1 .609
1 .643
6.818
4.909
9.720
1 .027
1 .081
1 .002
Visual Fixation and Scan Patterns of Dentists
5
Clinical Research with areas that were normal or of average gray scale. Significantly more revisits occurred on areas that were radiopaque (56%) and located in the radicular area (68%). Of the 4 categorized scanning patterns, tooth by tooth scanning predominated. There was no difference between scan patterns in covering the pathologic areas or periapical areas. Significant differences between scan patterns in covering coronal and intraradicular areas were observed. Circular scan patterns covered significantly more coronal areas than tooth by tooth or random scan patterns. Tooth by tooth scan patterns scanned significantly more intraradicular areas than circular or random scan patterns. As a pilot study, this work has numerous limitations. Even so, it documents previously unknown aspects of dental periapical radiograph interpretation. Further studies are needed to better explore the visual search and cognitive processing of dental periapical radiographs. Future research in this domain may have revolutionary implications on the psychological understanding of radiographic image perception and cognition. The authors of this study are now advancing these findings with more sensitive advanced eye tracking equipment.
Acknowledgments This work was supported by an AAE Foundation/DENTSPLY Innovation/Start-up in Research Grant. The authors wish to thank the participants of this study and Mr. Dan Kilfoy of Saint Louis University for his technical support.
6
Hermanson et al.
References 1. Nodine CF, Kundel HL. The cognitive side of visual search. In: O’Regan J, LevySchoen A, eds. Eye movements: from physiology to cognition. Amsterdam, The Netherlands: Elsevier; 1987:573–82. 2. Donovan T, Manning DJ. The radiology task: Bayesian theory and perception. Br J Radiol 2007;80:389–91. 3. Suwa K, Furukawa A, Matsumoto T, et al. Analyzing the eye movements of dentists during their reading of CT images. Odontology 2001;89:54–61. 4. Turgeon D, Lam E. Influence of experience and training on dental students’ examination performance regarding panoramic images. J Dent Educ 2016;80:156–64. 5. Just MA, Carpenter PA. A theory of reading: from eye fixations to comprehension. Psychol Rev 1980;87:329–54. 6. Yarbus AL. Eye Movements and Vision. New York: Plenum Press; 1967. 7. Burgdorf GC. Periapical image scanning between levels of expertise (master’s thesis]. St Louis: Saint Louis University; 2017. 8. Davis SF, Palladino JJ. Psychology: Media and Research Update, 3rd ed. Upper Saddle River: Prentice Hall; 2002. 9. Berbaum K, Franken EA, Dorfman D, et al. Cause of satisfaction of search effects in contrast studies of the abdomen. Acad Radiol 1996;3:815–26. 10. Orne MT. Demand characteristics and the concept of quasi-controls. In: Rosenthal R, Rosnow R, eds. Artifact in Behavioral Research. New York: Academic Press; 1969. 11. Nichols AL, Maner JK. The good-subject effect: investigating participant demand characteristics. J Gen Psychol 2008;135:151–65. 12. Goldman M, Pearson A, Darzenta N. Endodontic success – who’s reading the radiograph? Oral Surg Oral Med Oral Pathol 1972;33:432–7. 13. Goldman M, Pearson A, Darzenta N. Reliability of radiographic interpretations. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 1974;38:287–93. 14. Drew T, Vo M, Wolfe J. The invisible gorilla strikes again: sustained inattentional blindness in expert observers. Psychol Sci 2013;24:1848–53. 15. Brocklebank LM. Assessment of the radiographic image: recognition of abnormal features. Dent Update 1999;26:51–6, 58.
JOE — Volume -, Number -, - 2018
Visual Fixation and Scan Patterns of Dentists Viewing Dental Periapical Radiographs: An Eye Tracking Pilot Study Brian P. Hermanson, DDS, MSD,* Grant C. Burgdorf, DDS, MSD,† John F. Hatton, DMD,‡ Darrin M. Speegle, PhD,§ and Karl F. Woodmansey, DDS, MA‡ Abstract Introduction: The visual search patterns of dentists and the areas that attract their attention when interpreting dental periapical radiographs are currently unknown. This research identifies areas and patterns of visual fixation when observing dental periapical radiographs. Methods: In an observational study using eye tracking technology and a convenience sample of 44 observers, the interpretations of 4 dental periapical radiographs were recorded using Camtasia Software (TechSmith, Okemos, MI) with a gaze tracking ‘‘bubble’’ denoting where within the radiograph the observers’ eyes gazed. The recorded observations included the scanning pattern, the area of first fixation, and revisits of areas. Also noted was whether the area of first fixation or revisit was radiopaque, radiolucent, or of normal radiodensity and whether it was a coronal or radicular area. Results: The first fixation is more likely to be an area of high contrast that is either radiopaque or radiolucent compared with areas that were normal or of average gray scale. Significantly more revisits occurred on areas that were radiopaque and located in the radicular area. Of the 4 categorized scanning patterns, tooth by tooth scanning predominated. Conclusions: When interpreting dental periapical radiographs, significantly more observers initially fixated on areas of the radiograph that were of high contrast (ie, radiopaque or radiolucent) compared with ‘‘normal areas.’’ A tooth by tooth scanning pattern was most commonly used. (J Endod 2018;-:1–6)
Key Words Dental, eye tracking, perception, periapical radiograph, radiography, x-ray
C
linical tests and periSignificance apical radiographs The visual search patterns of dentists and the areas provide dentists with inthat attract their attention when interpreting dental formation for pulpal and periapical radiographs are currently unknown. This periapical diagnoses. The is the first study to use eye tracking technology for interpretation of dental radental periapical/endodontic radiographic interdiographs is a dual propretation. Using eye tracking technology, this cess that requires both research identifies areas and patterns of visual fixperception and cognition ation of observers viewing dental periapical radio(1). The visual scan of graphs. the radiograph is perceptual, whereas the diagnostic reasoning and decision making are cognitive (2). Dental research has documented some aspects of the cognitive interpretation of dental radiographs; however, minimal research has documented perceptual components of dental radiographic interpretation. Two published studies in dental radiography have used computerized eye tracking technology to evaluate dentists’ perceptions of panoramic and computed tomographic images (3, 4). Computerized eye tracking technology allows researchers to specifically determine where an observer is gazing within an image and illustrates patterns in the scanning process. This is the first study to use eye tracking technology for dental periapical/endodontic radiographic interpretation. Two essential components of perception are termed fixations and saccades. Fixations are locations where the observer’s gaze pauses momentarily. Along with the belief that perception is coincident with fixations is the idea that fixations are where cognition occurs (5). Individual fixations are separated by saccades, which are uniform movements of both eyes at a very high speed directing the gaze onto the next fixation. Some saccades may be voluntary, whereas many are not (6). The purpose of this study was to use computerized eye tracking technology to determine visual fixation and scan patterns of observers when viewing dental periapical radiographs. The following parameters of interest were selected for examination: area of first fixation, revisits, search patterns, effect of radiodensity, effect of coronal or apical location, and differences between observer experience levels. The first null hypothesis of this study was that there would be no differences in fixations or revisits whether the structure was located in a coronal or radicular area or was of normal radiodensity, radiopaque, or radiolucent. The second null hypothesis was that there would be no difference between the scan patterns and the percentage of pathology, coronal, intraradicular, and periapical areas scanned. A third null hypothesis was that there
From the *Private Practice, Pierre, South Dakota; †Private Practice, New York, New York; ‡Center for Advanced Dental Education and §Mathematics and Statistics, Saint Louis University, St Louis, Missouri. Address requests for reprints to Dr Karl F. Woodmansey, Saint Louis University, Department of Endodontics, Center for Advanced Dental Education, 3320 Rutger Street, St Louis, MO 63104. E-mail address: [email protected] 0099-2399/$ - see front matter Copyright ª 2018 American Association of Endodontists. https://doi.org/10.1016/j.joen.2017.12.021
JOE — Volume -, Number -, - 2018
Visual Fixation and Scan Patterns of Dentists
1
Clinical Research would be no difference in scan patterns between observer groups of different types or experience levels.
Materials and Methods This study was performed in accordance with a protocol approved by the Institutional Review Board of Saint Louis University, St Louis, MO. A convenience sample of 44 observers with differing years of experience and levels of training were recruited. These observers consisted of third- and fourth-year dental students (n = 12), advanced education in general dentistry residents (n = 8), general dentists (n = 12), and endodontists (n = 12). Observers were excluded if they were unable to visually calibrate the eye tracking device. Digital periapical radiographs were obtained from patients treated at the endodontic clinic at the Saint Louis University Center for Advanced Dental Education. A Delphi panel consisting of 4 endodontists selected the radiographs to include a variety of radiographic findings including normal teeth, direct or indirect restorations, coronal caries, periapical radiolucencies, posts, and endodontic treatment. All radiographs had patient identifiers removed to maintain confidentiality. Four radiographs were selected for analysis: 2 from the maxillary arch and 2 from the mandibular arch (Fig. 1). Two were of the left side and 2 of the right side. Each periapical radiograph contained a first molar, second molar, and at least the second premolar. The full extent of a periapical radiolucency, if present, was visible in the radiograph. Six other radiographs were selected as ‘‘filler’’ or ‘‘practice’’ radiographs. The radiographs were displayed on a 21.5-inch Apple iMac Desktop computer/monitor (Apple Inc, Cupertino, CA) using an automated PowerPoint presentation (Microsoft Corporation, Redmond, WA) with each of the 10 radiographs displayed in sequence for 10 seconds. The experimental radiographs were sequenced at positions 4, 5,
6, and 10. Data were collected using an optical video-based Tobii EyeX eye tracking device (Tobii Technology, Danderyd, Sweden), which illustrates gaze paths via an integrated 1.5-inch diameter gaze tracking ‘‘bubble’’ (Fig. 2A and B). A Camtasia Studio Software (TechSmith, Okemos, MI) recording of the display captured the path of the gaze tracking bubble. A digital video file preserved each observer’s gazes of the 4 radiographs. Observers first viewed introductory slides, which included a calibration exercise that ensured the Tobii EyeX could accurately track the observers’ gaze. Observers were instructed to interpret the radiograph as if a patient was waiting in the chair, that the radiographs would cycle automatically after 10 seconds, and to disregard the gaze tracking bubble. The Camtasia digital video files of the 4 experimental radiographs were blinded by an unassociated individual using a random number generator. Two blinded researchers then performed evaluations of the Camtasia recordings. When there was disagreement between the 2 evaluators in categorizing data, a consensus was reached after a discussion. In all instances, a consensus was able to be reached. Data from the other 6 radiographs were not evaluated. The observers’ area of first fixation was noted, defined as the first region of a definitive stop of the eye along the scan path. This area was categorized by location and radiodensity as coronal or radicular/periapical and normal, radiopaque, or radiolucent. The Delphi panel defined these categorizations. Radiopaque was defined as an area on the radiograph with radiodensity significantly greater than the average gray scale of the radiograph. These areas included metal restorations or root canal obturating materials. Radiolucent was defined as an area on the radiograph with radiodensity significantly lower than the average gray scale of the radiograph. These areas included periapical radiolucencies or carious lesions. Normal was defined as areas on the radiograph that were within the expected gray scale of the radiograph.
Figure 1. The 4 experimental radiographs.
2
Hermanson et al.
JOE — Volume -, Number -, - 2018
Clinical Research radiograph until all areas were scanned. Scan patterns that could not be categorized into those 3 categories were classified as random.
Statistical Analysis Data were analyzed using SPSS Statistics Software Version 24 (IBM Corp, Armonk, NY). Data from all participants across all 4 radiographs were pooled for the analysis. Comparisons between the observer groups are published elsewhere (7). For analysis of first fixations and revisits, intraradicular and periapical areas were pooled together because on many occasions the gaze tracking bubble would cover both the intraradicular and periapical areas simultaneously. Chi-square analysis was used to determine if significant differences existed for first fixations and revisits between coronal and radicular/ periapical areas and for differences among normal, radiopaque, and radiolucent areas. The significance level was set at P < .05. Post hoc pair-wise chi-square analysis was used to compare groups for significant differences between first fixation of normal and radiopaque, normal and radiolucent, and radiopaque and radiolucent areas. The significance level was set at P < .05. The Bonferroni correction for multiple comparisons was used. Using chi-square analysis, the scan patterns were compared for differences in the percentages of the pathology, coronal, intraradicular, and periapical areas scanned. The significance level was set at P < .05. Post hoc pair-wise chi-square analysis was used to compare for significant differences between each scan pattern and the specific area on the radiograph. The Bonferroni correction for multiple comparisons was used.
Results
Figure 2. (A) Study simulation with the Tobi EyeX eye tracking device (red arrow). (B) The gaze tracking bubble (yellow arrow).
Revisits were similarly recorded. These were defined as the first area returned to and fixated on after completion of the initial scanning pattern. For example, the first area that was fixated upon after a circular pattern was completed was categorized as a revisit. Revisits were not counted when a scan pattern was incomplete or for observers with a random pattern. Similar to areas of first fixation, revisits were categorized by location and radiodensity. Seven areas of pathology or abnormalities were identified across the 4 radiographs by the Delphi panel (Fig. 3). The number of areas of pathology or abnormalities scanned was tabulated to calculate the percent of pathology that the observer scanned in all 4 radiographs. Each tooth was divided into 3 regions: coronal, intraradicular, and apical. The total of the regions on each radiograph that were covered by the tracking ‘‘bubble’’ was calculated as a percentage of all possible regions that could possibly be viewed. Scan patterns of the entire radiograph over the 10-second period were subjectively classified as circular, tooth by tooth, or horizontal scanning (Fig. 4). Circular patterns were defined as a pattern that traced the periphery of the radiograph in either a clockwise or counterclockwise direction to complete a circle. A tooth by tooth pattern occurred when the observer began on the left or right side of the radiograph and scanned an entire tooth before moving to the adjacent tooth in the direction of the most distant tooth. A horizontal scanning pattern occurred when the observer began in either the coronal or apical region of the radiograph and scanned back and forth from the left to right of the JOE — Volume -, Number -, - 2018
A total of 44 screen recordings were obtained from the 44 participants; however, 2 participants were excluded because of an inability to calibrate the eye tracking device (1 general dentist and 1 endodontist). The resulting data yielded 42 recordings per radiograph for a total of 168 10-second Camtasia recordings. These included 168 scan patterns, 168 first fixations, and 75 revisits. The first null hypothesis of this study was that there would be no differences in fixations or revisits whether the structure was located in a coronal or radicular area or was of normal radiodensity, radiopaque, or radiolucent. Chi-square analysis showed no significant difference (P < .643) between first fixations on coronal or radicular/periapical areas, with 48.2% of first fixations occurring on coronal areas and 51.8% on radicular/periapical areas (Fig. 5). When compared with areas of normal radiodensity, a significantly greater (P # .001) number of first fixations were made on radiopaque or radiolucent areas (Fig. 5). When comparing these 3 individual categories, a statistically significant difference was found between normal and radiopaque and normal and radiolucent areas but not radiopaque and radiolucent areas with significance values of P < .001, P < .003, and P < .609, respectively (Table 1). Significantly more revisits occurred on radicular/periapical areas compared with coronal areas (P < .002), with 32% of revisits occurring on coronal areas and 68% on radicular/periapical areas (Fig. 5). When compared with areas of normal radiodensity, a significantly greater (P # .001) number of revisits were made on radiopaque or radiolucent areas. When comparing these 3 individual categories, a statistically significant difference was found between normal and radiopaque and normal and radiolucent areas but not radiopaque and radiolucent areas, with significance values of P < .001, P < .027, and P < .081, respectively (Table 1). The second null hypothesis was that there would be no difference between the scan patterns and the percentage of the pathology, coronal,
Visual Fixation and Scan Patterns of Dentists
3
Clinical Research
Figure 3. The 7 areas of pathology identified by the Delphi panel on the (A–D) 4 experimental radiographs.
intraradicular, and periapical areas scanned. The 168 observations were categorized into the following scanning patterns: 54.2% tooth by tooth, 31.5% circular, 12.5% random, and 1.8% horizontal. For statistical analysis, horizontal scanning was not included because its use was too low to be included. Chi-square analysis showed no significant difference in the percentage of pathology scanned or in periapical areas scanned by all scan patterns with P < .687 and P < .327, respectively. A significant difference was found in the percentage of coronal and intraradicular areas scanned by all scan patterns with P < .018 and P < .001, respectively. When comparing the different scan patterns, a significant difference was found in the percentage of the coronal areas scanned between tooth by tooth and circular scanning (P < .004). No significant difference was found between tooth by tooth and random or circular and random scanning with significance values of P < 1 and P < .146, respectively. A significant difference was found in the percentage of intraradicular areas scanned between tooth by tooth and circular and tooth by tooth and random scanning with significance values of P < .000 and P < .014, respectively. No significant difference was noted in the percentage of the intraradicular areas scanned between circular and random scanning with a significance value of P < 1.
Discussion The null hypotheses of this study were that there would be no differences in first fixation or revisits whether the structure was located in a coronal or radicular area or was of normal radiodensity, radiopaque, or radiolucent. The second was that there would be no difference between the scan patterns and the percentages of the pathology, coronal, intraradicular, and periapical areas scanned. These findings generally agree with those of Davis and Palladino (8) who reported that observers’ attention tends to ‘‘focus on objects 4
Hermanson et al.
that are smaller, brighter, more centrally located in an image.’’ A study by Berbaum et al (9) in 1996 reported that when areas of contrast were present, less time was spent scanning normal, low-contrast areas. Although the current study did not measure scanning times, it found that high-contrast areas were more likely to be the areas of first fixation. This study has numerous limitations and is therefore termed a pilot study. The most notable limitation is that the findings can only be related to these specific 4 radiographs and to this limited cohort of observers. Furthermore, these observers varied greatly in radiographic interpretation expertise. Fixations are typically measured in milliseconds. In their 1996 article, Berbaum et al (9) defined a fixation as gazing in a location within 1 degree for a minimum of 100 milliseconds with eye position measured 60 times per second. In the current study, gaze times could not be precisely measured. The first definitive stop along the scan pattern was considered the first fixation. This was subjective and complicated by the varying speeds at which the observers scanned the radiograph. The tracking ‘‘bubble’’ was inexact because of its size (Fig. 4B) but provided an adequate impression of where the observer was gazing. Observers were only given 10 seconds to evaluate each radiograph, which in most instances appeared to be an adequate amount of time to visualize all elements. However, some very systematic observers seemed slower and may not have had enough time for complete visualization, which may have affected the results. Many participants were observed to increase their speed throughout the study in order to view the entire radiograph. Although observers were instructed to interpret the radiograph as if a patient was waiting in the chair, a 10-second observation time was used to standardize the experimental protocol and to prevent observers from consciously trying to fixate on 100% of the radiograph. Because the 4 experimental radiographs were embedded within the 10-radiograph PowerPoint presentation JOE — Volume -, Number -, - 2018
Clinical Research 24
radiolucent
62 42
radiopaque
9
normal
77
29
revisits first fixations
51
Radicular
24
Coronal 0
87 81
50
100
Figure 5. First fixations and revisits, displayed by radiodensity and by radicular/coronal area. (first fixations n = 168; revisits n = 75)
In this study, the observers may have modified their visual searches to ensure that all areas of the radiograph were scanned in order to be a ‘‘good participant.’’ Although 2 observers fixated on the same areas, they may have actually interpreted the radiographs very differently. Goldman et al (12, 13) classically explored the cognitive aspects of dental radiographic interpretation, noting significant inter- and intraexaminer variability. Their findings show that radiographs are not ‘‘read’’ as if they were words on a page but are cognitively interpreted and not just visually processed. Consequently, the interpretation of radiographs is a human-centric process that has subjective variability between examiners. The present study reported where observers’ eyes gazed within these radiographs but did not demonstrate their cognitive interpretations. Observers’ interpretations may not necessarily be associated with their gaze fixations (14). Eye tracking research alone cannot fully explain radiographic interpretation; both perceptual and cognitive processes are necessary. Revisiting an area of potential abnormality was suggested by Brocklebank (15), recommending that a medical radiograph be reviewed systematically in a way that will cover the entire radiograph. She suggested noting any possible abnormalities and returning for closer examination once the initial scan of the radiograph was completed. The findings of this study support that concept. Only 12% of revisits occurred on areas of normal radiodensity or of average gray scale. Within the limitations of this study, when interpreting periapical radiographs, the first fixation is more likely to be an area of high contrast that is either radiopaque or radiolucent compared
TABLE 1. First fixation & revisit Chi-square analysis Figure 4. Scan patterns: (A) circular, (B) tooth by tooth, and (C) horizontal scanning.
with 3 unscored radiographs being viewed before the first scored radiograph, the observers had an opportunity to learn the flow of the automated 10-second timing. Another limitation of this study is that the observers were aware that their eye movements were being recorded. This may have led them to behave in ways that were different than if they were unaware their eye movements were being tracked. The term demand characteristic describes how experimental subjects attempt to identify the purpose of the study and accordingly modify their behavior (10, 11). JOE — Volume -, Number -, - 2018
Normal/ Normal/ Radiopaque/ Coronal/ Radiopaque Radiolucent Radiolucent Radicular First Fixations Chi21.736 square df 1 Asymp .000 Sig. Revisits Chi21.353 square df 1 Asymp .000 Sig.
11.967
1.619
.214
1 .003
1 .609
1 .643
6.818
4.909
9.720
1 .027
1 .081
1 .002
Visual Fixation and Scan Patterns of Dentists
5
Clinical Research with areas that were normal or of average gray scale. Significantly more revisits occurred on areas that were radiopaque (56%) and located in the radicular area (68%). Of the 4 categorized scanning patterns, tooth by tooth scanning predominated. There was no difference between scan patterns in covering the pathologic areas or periapical areas. Significant differences between scan patterns in covering coronal and intraradicular areas were observed. Circular scan patterns covered significantly more coronal areas than tooth by tooth or random scan patterns. Tooth by tooth scan patterns scanned significantly more intraradicular areas than circular or random scan patterns. As a pilot study, this work has numerous limitations. Even so, it documents previously unknown aspects of dental periapical radiograph interpretation. Further studies are needed to better explore the visual search and cognitive processing of dental periapical radiographs. Future research in this domain may have revolutionary implications on the psychological understanding of radiographic image perception and cognition. The authors of this study are now advancing these findings with more sensitive advanced eye tracking equipment.
Acknowledgments This work was supported by an AAE Foundation/DENTSPLY Innovation/Start-up in Research Grant. The authors wish to thank the participants of this study and Mr. Dan Kilfoy of Saint Louis University for his technical support.
6
Hermanson et al.
References 1. Nodine CF, Kundel HL. The cognitive side of visual search. In: O’Regan J, LevySchoen A, eds. Eye movements: from physiology to cognition. Amsterdam, The Netherlands: Elsevier; 1987:573–82. 2. Donovan T, Manning DJ. The radiology task: Bayesian theory and perception. Br J Radiol 2007;80:389–91. 3. Suwa K, Furukawa A, Matsumoto T, et al. Analyzing the eye movements of dentists during their reading of CT images. Odontology 2001;89:54–61. 4. Turgeon D, Lam E. Influence of experience and training on dental students’ examination performance regarding panoramic images. J Dent Educ 2016;80:156–64. 5. Just MA, Carpenter PA. A theory of reading: from eye fixations to comprehension. Psychol Rev 1980;87:329–54. 6. Yarbus AL. Eye Movements and Vision. New York: Plenum Press; 1967. 7. Burgdorf GC. Periapical image scanning between levels of expertise (master’s thesis]. St Louis: Saint Louis University; 2017. 8. Davis SF, Palladino JJ. Psychology: Media and Research Update, 3rd ed. Upper Saddle River: Prentice Hall; 2002. 9. Berbaum K, Franken EA, Dorfman D, et al. Cause of satisfaction of search effects in contrast studies of the abdomen. Acad Radiol 1996;3:815–26. 10. Orne MT. Demand characteristics and the concept of quasi-controls. In: Rosenthal R, Rosnow R, eds. Artifact in Behavioral Research. New York: Academic Press; 1969. 11. Nichols AL, Maner JK. The good-subject effect: investigating participant demand characteristics. J Gen Psychol 2008;135:151–65. 12. Goldman M, Pearson A, Darzenta N. Endodontic success – who’s reading the radiograph? Oral Surg Oral Med Oral Pathol 1972;33:432–7. 13. Goldman M, Pearson A, Darzenta N. Reliability of radiographic interpretations. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 1974;38:287–93. 14. Drew T, Vo M, Wolfe J. The invisible gorilla strikes again: sustained inattentional blindness in expert observers. Psychol Sci 2013;24:1848–53. 15. Brocklebank LM. Assessment of the radiographic image: recognition of abnormal features. Dent Update 1999;26:51–6, 58.
JOE — Volume -, Number -, - 2018