Abstract
Objective
Using a well-established measure of attention, we aimed to objectively identify medical students’ gaze patterns when assessing children with/without unicoronal craniosynostosis (UCS) and identify potential diagnostic and educational gaps in their assessment.
Design
Medical student participants viewed a series of images of children with/without UCS. Eye movements were recorded using a table-mounted eye-tracking device. Dwell times for 8 interest areas (forehead, brow, eyes, nose, lower face, mouth, left ear, right ear) were compared.
Participants
Thirty medical students (21 males, 9 females, mean age = 29.9 years old) were recruited from the local medical school.
Main Outcome Measure
The main outcome measure was the cumulative dwell times (milliseconds) participants spent within the 8 facial regions (forehead, brow, eyes, nose, lower face, mouth, left ear, right ear).
Results
Participants spent significantly more time on the brow region (P < 0.001) and less on the nose region (P < 0.05) when viewing UCS images. Furthermore, there were significantly longer dwell times to the side of the forehead contralateral to the fused suture in UCS images (P < 0.05). In control images, participants focused more on the right side of the mouth (P < 0.05). No other significant differences in dwell times were observed.
Conclusion
This study demonstrates that medical students exhibit attentional biases when viewing children with UCS and focus on only some of the asymmetries seen in UCS. These findings highlight potential educational gaps emphasizing the need for targeted training in craniofacial assessment.
Introduction
Unicoronal craniosynostosis (UCS) is among the most common congenital craniofacial differences, affecting approximately 1 in 2000 live births. 1 UCS restricts craniofacial growth leading to ipsilateral fronto-orbital retrusion and elevation with compensatory contralateral frontal bossing. In some cases, UCS can lead to increased intracranial pressure, developmental delays, and psychosocial challenges. 2 While surgical intervention is held to mitigate the risk of these sequelae, timely differentiation of UCS from benign craniofacial conditions can allow for less invasive surgical approaches such as endoscopic strip craniectomy and helmeting compared to traditional open fronto-orbital remodeling. A common craniofacial difference mistaken for UCS is deformational plagiocephaly (DP). 3 DP results in similar oblique asymmetries of the head, caused by external forces (eg, supine sleeping). Unlike UCS, DP requires no surgical intervention and is generally self-limiting with conservative management. The ability to distinguish between UCS and DP depends heavily on anatomical knowledge, clinical experience, and training in craniofacial anomalies. 3
Misdiagnosis of craniosynostosis is common in primary care due to a lack of familiarity with abnormal craniofacial structures. 4 One possible explanation for this may be reduced exposure to basic anatomy among medical trainees. A national survey found that Canadian medical students spend significantly less time on anatomical dissections than their U.S. counterparts, potentially impacting their foundational knowledge. 5 Reduced gross anatomy education has been shown to hinder diagnostic accuracy and clinical reasoning, both of which are crucial for identifying conditions with subtle anatomical variations, such as UCS. 6 Furthermore, most Canadian medical schools provide minimal exposure to plastic surgery during medical education. 7 This is particularly concerning for craniofacial conditions like UCS, where accurate identification requires a deeper level of anatomical knowledge along with clinical exposure. To address gaps in medical students’ understanding of craniofacial anatomy, three-dimensional (3D) skull models have become a widely used teaching tool. A previous study by Lane & Black 8 demonstrated that 3D printed skulls improved medical students’ comprehension of craniosynostosis pathology and surgical approaches compared to lecture-based modules. While these models capture the skeletal changes observed in UCS, they do not reflect how patients with UCS present to clinicians. Eye-tracking technology provides a complementary approach by identifying perceptual biases in the assessment of facial features. Typically, facial gaze patterns concentrate within the “central triangle of fixation,” defined by the eyes, nose, and lips. However, gaze deviations increase proportionally with the severity of craniofacial deformities.9–11 Linz et al 12 observed, through eye-tracking, that university students without medical training deviated from the central triangle when viewing pre-operative UCS patients, and their gaze normalized to the central region following corrective surgery. By analyzing gaze patterns, eye-tracking technology can help educators determine what features trainees are focusing on when observing a patient with a craniofacial difference, offering valuable insights into what information medical learners use to correctly or incorrectly come to a given diagnosis.
In this study, an eye-tracking device was used to examine how medical students direct their attention to facial areas of interest when viewing images of children with UCS compared to children without a craniofacial difference.
Materials and Methods
Participants
Institutional ethical approval and written informed consent from all participants were obtained for this study. Medical students in their second, third, or fourth (final) year of undergraduate medical training were recruited at the study institution by word-of-mouth and online notices. The target recruitment was set at a minimum of 30 participants based on previous reports which outlined recommendations for effective eye-tracking studies and demonstrated that at a minimum of 30 participants dwell times and fixation patterns were more reliable.13,14 Other eye-tracking studies exploring attentional biases have had a comparable number of participants to the current study, which all showed statistically significant results following data analysis.10–12 All participants reported normal or corrected-to-normal vision. All recruited participants completed the study in full; no data was withdrawn owing to inability to complete tasks, or visual strain during the eye-tracking process. All participants were naïve to the true intent of the study but were informed once the tasks of the data collection were completed. This was done to minimize attentional biases when stimulus images were presented.
Stimuli
A total of 8 images were presented, including four images of children aged 6 to 9 months with UCS obtained with parental consent from the local hospital's craniofacial patient population with an equal distribution by sex. Four control images of children without UCS or other craniofacial difference were obtained from a public domain photo collection (iStock; July 8, 2024). UCS and control images represented a diverse patient population. All photos were cropped and resized to display only the face of each child. Additionally, photos were edited to remove potential distractors (eg, earrings, embellishments on their clothing) and were given the same background color to minimize attentional biases (Figure 1). Participants were presented the images in a randomized order.
Representative Images of Experimental Stimuli Depicting a Child Without UCS (A) and a Child With UCS (B). UCS, unicoronal craniosynostosis.
Apparatus
Eye movements were recorded using the Eyelink 1000 (SR Research Ltd, Mississauga, Ontario, Canada). This device is composed of a camera which records eye movements in real-time, a monitor displaying the study's images, and an isolated head mount, used to minimize head movement throughout the study (Figure 2).
A Participant Using the EyeLink 1000 Device.
Procedure
Once written consent was obtained, participants were seated approximately 60 cm from a 19-inch monitor displaying the images. After the head mount was adjusted for participant comfort, pupillary calibration was performed to ensure eye-tracking was accurate. Following calibration, participants re-entered the study's display with a set of instructions. The participant was prompted to press the keyboard when they were ready to commence the experiment. Each image was shown to the participant for 5000 milliseconds in a random sequential order. In eye-tracking studies, it is common to become less focused to the stimuli as time progresses, a phenomenon known as habituation. 13 To minimize this, following the display of each picture, participants were asked to approximate the age of the child on a 2-month interval scale from ages 2-10 months. The age approximation acted as a distractor task not only to hide the true intent of the study and thereby minimize attentional biases, but also to maintain a participant's focus. Participants were given 10 s to record their age estimation following each photo. Once the study was completed, participants were debriefed and told the true intent of the study.
Data Analysis
The Eyelink 1000 software (SR Research Ltd.) was used to record the amount of time (milliseconds) participants spent looking at eight facial areas of interest (Figure 3). The dwell times of these regions of the face were compared between images of children with UCS and control images. After the initial comparison, interest areas were then divided into halves to analyze the distribution of attention. UCS images were divided into “ipsilateral” and “contralateral” sides to the affected suture. Analogously, control images were divided into left and right sides (Figure 4).
Images Displaying the Interest Areas Used to Measure Dwell Times for Children Without (A)/with (B) UCS: (1) Forehead Region, (2) Brow Region, (3) Eye Region, (4) Nose Region, (5) Lower Face Region, (6) Mouth Region, (7) Right Ear, (8) Left Ear. UCS, unicoronal craniosynostosis.
UCS Image Displaying the Interest Areas Being Divided Into Ipsilateral and Contralateral Sides Relative to the Affected Suture. Control Images were Divided Analogously Into Halves. UCS, unicoronal craniosynostosis.
Statistical Analysis
Data was analyzed using the SPSS software version 20.0.0 (IBM SPSS Statistics, Armonk, New York). The dwell times for each interest area for UCS and control images were analyzed with descriptive statistics (median, range, and interquartile range). Kolmogorov-Smirnov testing was performed to assess normality. For normally distributed data, a two-sided t-test was used, while for non-normally distributed data, a paired Wilcoxon signed-rank test was used to determine if dwell times differed between UCS and control images and further between “ipsilateral” and “contralateral” sides. Significance was set at P < 0.05 when comparing each interest area (forehead, eyes, nose, mouth, left and right ear).
Results
Thirty medical students (21 males, 9 females, mean age of 29.9 years old) participated in this study. Dwell times when participants viewed images of children with or without UCS are summarized in Table 1. A significant difference in nose and brow region dwell times was noted when comparing UCS and control images. More specifically, participants’ gaze was maintained at the brow region significantly longer (P < 0.001) in UCS images than control images. Additionally, it was found that participants dwelled significantly less (P < 0.05) at the nose region when looking at UCS images than their control image counterparts. In the other interest areas (forehead, eyes, mouth, left and right ear), there were no significant differences in dwell times between the two groups. Further analyses were performed to investigate the dwell time distribution across interest areas. This was done to examine if participants dwelled on a certain side of the image in the presence of a craniofacial difference (UCS) versus an image with no difference (control). Table 2 summarizes the dwell times for interest areas “ipsilateral” and “contralateral” to the affected suture in the UCS images, and the corresponding left or right side of the control images. This revealed that when viewing UCS images, participants dwelled on the contralateral forehead region of UCS images at the forehead region only (P < 0.05). No such differences between sides were noted in the control images save for the tendency of participants to dwell on the right side of the mouth (P < 0.05). For each group, distribution of attention did not differ significantly at the other interest areas.
Summary of Dwell Times When Perceiving Images of Children with/Without UC.
IQR, interquartile range; UCS, unicoronal craniosynostosis.
*Two-sided, paired Wilcoxon signed-rank test.
Comparison of Dwell Time Distribution on Different Sides of the Face.
IQR, interquartile range; UCS, unicoronal craniosynostosis.
UCS (Ipsi.), the half of the face that is ipsilateral to the fused coronal suture.
UCS (Contra.), the half of the face that is contralateral to the prematurely fused coronal suture.
*Paired, two-sided Wilcoxon signed-ranked test.
All statistical tests were two sided with P < 0.05 determined to be statistically significant.
Discussion
The current study demonstrates that the craniofacial differences associated with UCS cause a deviation from the expected facial gaze pattern. Overall, participants spent less time looking at the nose within the central triangle of fixation when viewing images of children with UCS. In contrast, they dwelled longer on the brow region, which was distorted by UCS, compared to control images. This suggests that asymmetries caused by UCS disrupt typical patterns of facial attention, redirecting focus to visibly altered features. This aligns with previous eye-tracking studies that found that perception of people with craniosynostosis leads to attentional biases that deviate from the central triangle of fixation, depending on the severity of the craniofacial difference.10,12
Our study expands upon the previous study of Linz et al 12 by incorporating analysis of facial attention based on laterality. While Linz and colleagues demonstrated deviations from the central triangle of fixation in craniosynostosis, they did not assess how asymmetry drives perceptual differences across facial hemispheres. By dividing faces into “ipsilateral” and “contralateral” sides, we revealed that participants disproportionately fixated on the forehead contralateral to the affected suture in UCS images. This would suggest that the principal abnormality being identified by medical trainees is the compensatory contralateral frontal bossing observed in UCS and not the retruded, constricted ipsilateral forehead. Participants also showed a tendency to focus on the right side of the mouth in control images, but no such bias was observed in images of children with UCS. This might be attributed to some degree of non-neutrality in the expressions of children in control images, which captured attention and led to prolonged focus on specific areas of the mouth. In contrast, UCS images did not evoke the same attentional bias, possibly due to the overall appearance affecting how participants distributed their attention across facial features.
Despite UCS leading to distortions in the position and shape of the orbits, participants did not show a significant difference in dwell times at these facial interest areas when viewing children with or without UCS or when comparing facial halves in UCS images. One potential reason could be a confirmation bias that led participants to prioritize areas they believe are diagnostically relevant, such as the forehead deformity. Another possibility is that participants assume asymmetrical eye position in UCS were secondary to the forehead deformity and therefore did not exhibit a perceptual bias to this region. Additionally, these gaze patterns could be associated with visual salience, a cognitive phenomenon in which attention is disproportionately drawn to prominent, high-contrast features. 15 This effect likely contributed to medical student participants’ fixation on exaggerated asymmetries such as contralateral frontal bossing, rather than subtler orbital differences. While visual salience can be beneficial in identifying gross deformities, it can also skew attention away from diagnostically important, but less visually arresting regions. Given that the study population consisted of exclusively medical students, these unique gaze patterns suggest that medical students lack the perceptual framework for a comprehensive craniofacial exam for conditions such as UCS.
Several evidence-based strategies have been discussed in current literature to improve craniofacial assessments in medical trainees. Virtual simulation platforms have emerged at the forefront of medical education tools, allowing for early exposure to cleft and craniofacial conditions, like UCS, that medical students may not be accustomed to. 16 Though current simulations are catered toward surgical residents, these tools can be adapted to align with the scope of undergraduate medical education. This would involve emphasizing foundational craniofacial assessment skills such as pattern recognition and facial asymmetry assessments. A recent program developed by the American Academy of Otolaryngology-Head and Neck Surgery, displayed that virtual simulations improved learner confidence and clinical familiarity through high-fidelity case presentations and structured feedback. 17 Virtual simulations are particularly relevant in Canadian medical education, where clinical exposure to patients with complex craniofacial anomalies may be limited due to the emphasis on early training in primary care and community-based settings. 5 If medical students underwent virtual simulations for craniofacial assessment, clinical competence could be assessed through the Objective Structured Clinical Examinations (OSCE). 18 An OSCE station could present a medical student with a 3D rendering of a craniofacial condition, similar to the effective educational models discussed by Lane & Black. 8 Students could be tasked to identify key features of asymmetry, differentiate from benign variants, such as DP, and communicate appropriate management and referral plans. Integrating these educational strategies could assist in standardizing exposure, and ultimately improve early recognition, management and knowledge of craniofacial conditions, like UCS.
Strengthening early educational exposure to craniofacial conditions could impact how future clinicians evaluate surgical needs. Additional knowledge of facial areas that draw attentional bias in children with UCS can help prioritize these regions when determining the functional and esthetic outcomes of surgery. The use of bilateral fronto-orbital advancement techniques is supported by our findings as these approaches address both the ipsilateral retruded fronto-orbital skeletal as well as the contralateral frontal bossing which attracted the gaze of our study participants. By creating a more symmetrical appearance and reducing the prominence of visually arresting areas, bilateral approaches minimize visual salience.
Further investigations should aim to address some of the limitations of the current study. First, the images that were shown to the participants were derived from two different sources. One being the local hospital's craniofacial clinic for UCS images, in addition to a public domain photo bank (iStock; July 8, 2024) for control images. Although the images for both groups were similar in quality and angle, the differences in facial expression could have played a part in influencing participants’ visual attention throughout the experiment. Additionally, since there is a spectrum of UCS presentations, future work could benefit from selecting UCS images with a gradation of severity. 2 Another limitation is the study population. This study targeted medical students as they represent a group with foundational knowledge but limited clinical experience. This study provided insight into how individuals without specialized training perceive craniofacial differences, but it would be informative to repeat the current methodology with late-stage trainees (resident physicians, craniofacial fellows) and established craniofacial surgeons to see how attention changes with training and experience. Lastly, though a distractor task was used to reduce attentional bias prior to each picture being displayed, the increase in cognitive load to estimate each child's age could have led to changes in gaze patterns that shifted attentional biases to key facial interest areas that were measured. 18 Despite these limitations, the current study identified deviations of visual patterns when comparing participant perceptions of children with and without UCS.
Conclusion
Eye-tracking technology, a well-established measure of attention, was used to demonstrate attentional biases in perceiving children with and without UCS. Longer fixation times on the brows were noted when viewing images of children with UCS. Furthermore, participants tended to show an attentional bias to the forehead contralateral to the affected suture. This study provides evidence for the existence of attentional biases when viewing children with UCS with implications for initial diagnosis and surgical correction. To mitigate perceptual biases in craniofacial assessments, the results of the study support the integration of educational strategies for craniofacial differences in medical school curricula. With adaptation of current standardized exams with novel virtual resources for craniofacial assessment, undergraduate medical education can have a significant impact on early recognition and interpretation of craniofacial differences, such as UCS.
Footnotes
Acknowledgements
The authors would like to thank Drs. Diane MacKenzie, Darren Oystreck, and William Brixius, and Mr. Eric Lee for their technical assistance with the eye-tracking system.
Ethical Considerations & Informed Consent
Ethical approval for this study was obtained from the Research Ethics Board of the IWK Health Centre. All clinical images used were obtained through written informed consent. All participants provided written informed consent prior to the study.
Funding
The authors received no financial support for the research, authorship, or publication of this article.
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.