Thin-section CT of the lungs: Eye-tracking analysis of the visual approach to reading tiled and stacked display formats

doi:10.1016/j.ejrad.2006.05.006

European Journal of Radiology

Volume 59, Issue 2, August 2006, Pages 257-264

https://doi.org/10.1016/j.ejrad.2006.05.006 Get rights and content

Abstract

Objective

To use eye-tracking analysis to identify the differences in approach to and efficiency of reading thin-section CT of the lungs presented tiled and stacked soft-copy displays.

Materials and methods

Four chest radiologists read 16 thin-section CT examinations displayed in either a tiled (four images at once) or stacked (full screen cine) format. Eye-movements were recorded and analysed in terms of movement type; saccade distance (classified by the calculated range of useful peripheral vision), number of fixations, duration and direction of gaze—comparison of the areas of the images viewed.

Results

Cases presented in stacked format were read quicker than when presented in tiled format with a greater fixation frequency (5 fixations versus 4.5 fixations points per 100 data points; p < 0.001) and a greater proportion of short saccades (97% versus 94%; p < 0.005). The consistency with which the observers viewed equivalent areas of the scan images in different cases was greater when viewing in stacked format (mean kappa 0.45 versus 0.36; p < 0.05) suggesting a more systematic approach to reading.

Conclusion

Eye-tracking data demonstrates why thin-section CT examinations of the lungs are read more efficiently when displayed in a stack as opposed to a tiled format.

Introduction

The proportion of radiologic image reading that is performed directly from a computer screen (‘soft copy’ reporting) as opposed to that performed from film (‘hard copy’ reporting) has dramatically increased [1]. Computing power and monitor quality have developed to the extent that the comparison of reporting efficacy of CT images between viewing hard and soft copy images [2] is no longer an issue, and attention has now turned towards the optimal way of viewing images in terms of workstation design [3], including the size and number of images that should be displayed at any one time [4]. An inherent advantage of a stacked, as opposed to a tiled, display format for reading contiguous images of a 3-D volume, has been demonstrated in terms of reporting accuracy and viewing speed [5]. However, the relevance of this observation to non-contiguous thin-section CT imaging is uncertain as features that traverse multiple images may present on adjacent images in distinctly different positions.

Modern eye-trackers enable the non-obtrusive assessment of workstation presentation ergonomics to be analysed in terms of eye-movements [6]. The characterisation of eye-movements involves assumptions as to the cognitive and subliminal processes underlying them including the influence of peripheral vision. In order to identify eye-movements that may be guided by peripheral vision, the range of peripheral vision for a given task needs to be assessed.

It is assumed that a more ordered and therefore efficient approach to reading an image will be reflected by a greater proportion of short saccadic eye movements either directed by a system of search or by useful peripheral vision. A more chaotic approach to reading would result in a greater number of large saccadic movements between fixations.

A structured approach to reading CT images will be reflected in the consistency with which the most important areas of the image are viewed. Such areas may relate to specific abnormalities found in that case but others will relate to the anatomical structures that tend to be affected by the disease process being hypothesised by the reader. This may alter according to the characteristics of a given disease. The direct comparison of spatial data between different cases is impossible due to varying anatomy. The mapping of spatial data onto a calculated ‘normal’ template, retaining the anatomical features present, allowed a more accurate comparison of spatial data between cases from different patients. No attempt was made to identify the areas the reader considered most important as the experiment was designed around comparing reader approach rather than identifying the features each reader found significant.

1.
An initial experiment to determine to what extent the readers used in this study could discern contrast and fine detail entirely from their peripheral vision.
2.
Recording of the eye-movements of the readers whilst reading thin-section CT scans in one of two different display formats to explain differences in reader approach due to display format
3.
Comparison of where on the CT images the readers looked by mapping the eye tracking data onto a standardized stack of CT images generated by combining 24 normal thin-section CT scans. To determine whether the areas viewed are consistent and whether display format has an impact on this consistency.

Section snippets

Materials and methods

Four experienced chest radiologists with experience ranging from 10 to 27 years read a selection of 16 thin-section CT examinations of the chest on a 22 in. diagonal high contrast 100 Hz multisync computer monitor. The analysis of the observers’ eye-movements required that the examinations be read in an environment designed around an eye-tracking camera that kept natural light to a minimum. A chin rest 70 cm from the computer monitor was used to stabilize the head position to aid the eye-tracking

Results

The range of useful peripheral vision for the four observers were 40, 140, 160 and 200 screen pixels (corresponding to 3, 11.2, 12.8 and 16 cm) or a visual angle of 3°, 9.1°, 10.4° and 13° respectively. This did not translate into a significant difference in reading efficiency between observers. The lowest calculated useful peripheral vision reading resulted from the observers’ difficulty in keeping their direction of gaze in the centre circle. Two of the observers used spectacles in the

Discussion

The results from this experiment show that the vast majority of saccadic movements, regardless of display format, occur within the ambit of useful peripheral vision. The extent of useful peripheral vision for the identification of fine detail, in the case of airway and vascular morphology, varied between observers but for one individual it effectively encompassed about a quarter of the screen, a finding previously noted in an observer performance study that evaluated the detection of nodules on

Conclusion

Most CT reading is now performed from computer workstations (soft copy) and increasing pressure to improve reporting efficiency necessitates the viewing of CT cases in the most efficient display format. It is intuitively clear that contiguous CT images would be read more efficiently in a stacked display format, our study introduces eye-tracking analysis to determine why non-contiguous CT imaging should also be viewed in a stacked format.

Reference (11)

J.E. van der Heyden et al.
Exploring presentation methods for tomographic medical image viewing
Artif Intell Med
(2001)
R.L. Arenson et al.
Computers in imaging and health care: now and in the future
J Digit Imaging
(2000)
D.V. Beard et al.
Interpretation of CT studies: single-screen workstation versus film alternator
Radiology
(1993)
D.V. Beard et al.
A study of radiologists viewing multiple computed tomography examinations using an eyetracking device
J Digit Imaging
(1990)
N.H. Strickland et al.
Default display arrangements of images on PACS monitors
Br J Radiol
(1995)

There are more references available in the full text version of this article.

Cited by (33)

How to Read an Abdominal CT: Insights from the Visual and Cognitive Sciences Translated for Clinical Practice
2022, Current Problems in Diagnostic Radiology
Citation Excerpt :
Likewise, missed cancers in the unprepped colon only confirms what all abdominal radiologists know: the unprepped colon is five twisty feet of hazardous road.33 Eye gaze and tracking techniques have been applied to understand how radiologists examine an image or set of images.3,12,37 Studies of eye gaze for detection of lung nodules on CT have identified at least 2 general patterns of search: drilling and scanning.3,14,38
When first learning abdominal CT studies, residents are often given little concrete, practical direction. There is, however, a large literature from the visual and cognitive sciences that can provide guidance towards search strategies that maximize efficiency and comprehensiveness. This literature has not penetrated radiology teaching to any great extent. In this article, we will examine the current pedagogy (and why that falls short), why untutored search fails, where misses occur in abdomen/pelvis CT, why these misses occur where they do, how expert radiologists search 3d image stacks, and how novices might expedite the acquisition of expertise.
Medical students' cognitive load in volumetric image interpretation: Insights from human-computer interaction and eye movements
2016, Computers in Human Behavior
Citation Excerpt :
Conversely, volumetric image interpretation may also decrease cognitive load. The possibility of examining the anatomical structure and its relative position from multiple angles can arguably provide the student with additional contextual information, i.e. the student does not need to infer the shape, size and position of a structure based on one 2D image (Ellis et al., 2006; Hegarty, Keehner, Cohen, Montello, & Lippa, 2007; van der Land et al., 2013). This contextual information allows for less specific prior knowledge needed for image comprehension (van Merriënboer & Sweller, 2010).
Medical image interpretation is moving from using 2D- to volumetric images, thereby changing the cognitive and perceptual processes involved. This is expected to affect medical students' experienced cognitive load, while learning image interpretation skills. With two studies this explorative research investigated whether measures inherent to image interpretation, i.e. human-computer interaction and eye tracking, relate to cognitive load. Subsequently, it investigated effects of volumetric image interpretation on second-year medical students' cognitive load. Study 1 measured human-computer interactions of participants during two volumetric image interpretation tasks. Using structural equation modelling, the latent variable ‘volumetric image information’ was identified from the data, which significantly predicted self-reported mental effort as a measure of cognitive load. Study 2 measured participants' eye movements during multiple 2D and volumetric image interpretation tasks. Multilevel analysis showed that time to locate a relevant structure in an image was significantly related to pupil dilation, as a proxy for cognitive load. It is discussed how combining human-computer interaction and eye tracking allows for comprehensive measurement of cognitive load. Combining such measures in a single model would allow for disentangling unique sources of cognitive load, leading to recommendations for implementation of volumetric image interpretation in the medical education curriculum.
A Passing Glance? Differences in Eye Tracking and Gaze Patterns Between Trainees and Experts Reading Plain Film Bunion Radiographs
2015, Journal of Foot and Ankle Surgery
Eye tracking and gaze pattern studies have been used to evaluate human behavior for decades. This is because of its ability to reveal conscious and subconscious behaviors when subjects are tasked with observation, decision making, and surgical performance. Many have popularized the use of this technology for radiographic assessment while evaluating radiologist behaviors, but little has been described for surgeon behavior patterns when evaluating preoperative deformities by radiograph. Because the radiographic assessment strongly influences surgical selection, the present study was designed to evaluate the differences between groups of novice and experienced surgeons' gaze patterns when tasked to describe hallux valgus deformities. The subjects were asked to rate the deformity as “none,” “mild,” “moderate,” or “severe.” Using an externally mounted eye tracking system, our study assessed saccades, fixations, overall time spent per radiograph, and the subjects' chosen bunion rating. Both the novice and advanced groups of foot and ankle surgeons were tasked to evaluate 25 total anteroposterior radiographs from patients who presented with a primary complaint of bunion pain. These patients were chosen at random, such that all participating surgeons had no previous patient familiarization. Statistically significant differences were observed with regard to the activity and rating of the moderate bunion films. The experience of surgeons does appear to modify gaze behavior with respect to time and attention, such that less overall time spent per image is needed by the advanced group, with improved efficiency. Future academic curriculum and training techniques could be developed to reflect these potential technical differences in search behavior, diagnostic technique, and surgical selection strategy.
Volumetric and two-dimensional image interpretation show different cognitive processes in learners
2015, Academic Radiology
Citation Excerpt :
We did not find other studies reporting differences in cognitive processes in volumetric versus 2D image interpretation in the context of radiology education. Previously reported differences in visual search patterns and error patterns (5–7) only suggest a difference in cognitive processes, although the actual cognitive processes during image interpretation, in terms of the thinking process of the image interpreter, was not investigated. Most verbal protocol studies investigating cognitive processes in radiologic image interpretation research are related to differences in expertise levels.
In current practice, radiologists interpret digital images, including a substantial amount of volumetric images. We hypothesized that interpretation of a stack of a volumetric data set demands different skills than interpretation of two-dimensional (2D) cross-sectional images. This study aimed to investigate and compare knowledge and skills used for interpretation of volumetric versus 2D images.
Twenty radiology clerks were asked to think out loud while reading four or five volumetric computed tomography (CT) images in stack mode and four or five 2D CT images. Cases were presented in a digital testing program allowing stack viewing of volumetric data sets and changing views and window settings. Thoughts verbalized by the participants were registered and coded by a framework of knowledge and skills concerning three components: perception, analysis, and synthesis. The components were subdivided into 16 discrete knowledge and skill elements. A within-subject analysis was performed to compare cognitive processes during volumetric image readings versus 2D cross-sectional image readings.
Most utterances contained knowledge and skills concerning perception (46%). A smaller part involved synthesis (31%) and analysis (23%). More utterances regarded perception in volumetric image interpretation than in 2D image interpretation (Median 48% vs 35%; z = −3.9; P < .001). Synthesis was less prominent in volumetric than in 2D image interpretation (Median 28% vs 42%; z = −3.9; P < .001). No differences were found in analysis utterances.
Cognitive processes in volumetric and 2D cross-sectional image interpretation differ substantially. Volumetric image interpretation draws predominantly on perceptual processes, whereas 2D image interpretation is mainly characterized by synthesis. The results encourage the use of volumetric images for teaching and testing perceptual skills.
Lesion detection in digital breast tomosynthesis: human reader experiments indicate no benefit from the integration of information from multiple planes
2023, Journal of Medical Imaging
The Radiologist’s Gaze: Mapping Three-Dimensional Visual Search in Computed Tomography of the Abdomen and Pelvis
2019, Journal of Digital Imaging

View all citing articles on Scopus

View full text

Thin-section CT of the lungs: Eye-tracking analysis of the visual approach to reading tiled and stacked display formats

Abstract

Objective

Materials and methods

Results

Conclusion

Introduction

Section snippets

Materials and methods

Results

Discussion

Conclusion

Artif Intell Med

Computers in imaging and health care: now and in the future

J Digit Imaging

Interpretation of CT studies: single-screen workstation versus film alternator

Radiology

A study of radiologists viewing multiple computed tomography examinations using an eyetracking device

J Digit Imaging

Default display arrangements of images on PACS monitors

Br J Radiol