Abstract
Errors in radiologic interpretation are largely the result of failures of perception. This remains true despite the increasing use of computer-aided detection and diagnosis. We surveyed the literature on visual illusions during the viewing of radiologic images. Misperception of anatomical structures is a potential cause of error that can lead to patient harm if disease is seen when none is present. However, visual illusions can also help enhance the ability of radiologists to detect and characterize abnormalities. Indeed, radiologists have learned to exploit certain perceptual biases in diagnostic findings and as training tools. We propose that further detailed study of radiologic illusions would help clarify the mechanisms underlying radiologic performance and provide additional heuristics to improve radiologist training and reduce medical error.
Introduction
Most diagnostic errors in radiologic practice are secondary to failures of perception (Renfrew et al., 1992; Slater et al., 2006; ; Waite et al., 2017, 2020). Some of these errors are driven by visual illusions that radiologists encounter as they analyze radiographs. Illusions are hard to rigorously define (), and—while there have been efforts to precisely categorize or describe illusions (Ninio, 2014)—there is no consensus definition among perception scientists. In this paper, we define illusions as mismatches between physical reality and perception (see Westheimer, 2008; Murray and Herrmann, 2013; Shapiro and Todorovic, 2016 for similar approaches). In radiology, such mismatches can potentially interfere with accurate diagnosis.
Radiologists can fail to see pathologies due to perceptual biases—or see pathologies where none exist (Perrin and McBroom, 1987; Renfrew et al., 1992; Waite et al., 2019, 2020). Although missed diagnoses are more commonly discussed in the literature, false positives arising from over-diagnosing normal variations in anatomy as pathological can be harmful too, secondary to complications from unnecessary tests and treatments (Keats and Mark, 2001).
Although variable in the literature, the effective error rate in radiological studies is estimated to be about 4%, unchanged over the last 70 years (Waite et al., 2017). Worldwide, a 4% error rate would translate to approximately 40 million errors per year (Imanzadeh et al., 2020). Computer aided detection and machine learning methods promise to improve diagnostic accuracy, yet these same technologies place new demands on radiologists and can introduce novel sources of perceptual error (Slater et al., 2006; McGurk et al., 2008; Maskell, 2019).
Here, we review some of the more common illusions in radiology and their impact on clinical diagnosis (see Table 1). We discuss both diagnostic errors and the potential benefits of illusions. Understanding the differences between medical images and their perception in the observer can help enhance the ability of radiologists to detect pathology. Thus, radiologists armed with the knowledge of common illusions may not only better avoid misdiagnosis but even use illusions, when present, to help establish diagnosis (). In the future, error rates in radiology could be reduced through a better understanding of the role that illusions play, and radiology residents might be trained to both prevent and exploit such phenomena.
TABLE 1
Common visual illusions observed in radiology.
Brightness and Contrast Illusions
Our brains do not detect the actual brightness of objects in the world, but instead compare an object’s physical luminance to that of nearby surfaces, frequently creating inaccurate representations of the natural world (Martinez-Conde and Macknik, 2017). Sometimes, the brightness and contrast illusions that result from such neural comparisons improve the visibility of structures on medical images, i.e., by enhancing boundary perception. Examples include any objects or surfaces where their physical luminance differs from their perceived brightness or contrast, such as Mach bands and simultaneous contrast effects.
Mach Bands
Mach bands are a form of contrast enhancement, visible as a bandlike line at the edge of almost any shadow and at the borders between adjacent, overlapping objects with different luminance (see Figure 1A). They are commonly encountered in radiology on routine chest radiographs, in places where structures of different image intensities overlap (Lane et al., 1976; ; ), and occur most frequently along the vertebral column (; Raby et al., 2014). Mach bands can be “negative” (dark) or “positive” (bright), but only one type of Mach band is typically visible at each boundary created by most biological shapes (; Papageorges and Sande, 1990). Mach bands are often helpful in demarcating boundaries between anatomic structures—though this is not always the case. Moreover, negative Mach bands and their associated boundaries can be too dark to be seen clearly on radiographs.
FIGURE 1
Negative Mach bands are typically associated with convex (outward-curving) structures, and positive Mach bands with concave (inward-curving) structures (Papageorges, 1991). Papageorges (1991) suggested that these associations could be used to deduce the shape of unknown anatomical structures in radiographs, or to more accurately identify the shape of known structures (see Figure 1B). For instance, radiologists can use Mach bands to better visualize abnormalities that are present in radiographs, but obscured by overlapping structures: when one structure overlaps another, the resulting Mach band from the edge contrast difference can elucidate the shape and position of the occluded structure (; see Figure 1C). Information from Mach bands can be critical in cases where relying on memory to reconstruct the 3-dimensional anatomy would otherwise be difficult, overly complicated, or misleading.
Unfortunately, Mach bands can also hinder accurate diagnosis: Mach bands that overlap with bone can be misperceived as fractures (, ; see Figure 2). Mach bands caused by skin folds can mimic the appearance of pneumothorax (air in the space between the thin pleural covering that surrounds the lungs; (Kattea and Lababede, 2015). Mach bands are also a cause of erroneous diagnosis of cavities (caries) on dental radiographs (Thomson and Johnson, 2012).
FIGURE 2
Indeed, trainees sometimes misinterpret Mach bands as fractures, only to be corrected by their mentors (see Reeves, 2004 for the description of one medical student’s experiences). Thus, residents are taught that when a Mach band might be present, they should look for additional findings that suggest a fracture: in the absence of such findings, apparent dark lines are likely indicative or Mach bands, rather than fractures (Samei and Krupinski, 2018). Expert radiologists are more adept at picking up on other subtle cues—or their absence, thereby avoiding diagnostic error (Nielsen, 2001). Thus, while the perceptual expertise of radiologists will not always prevent them from misperceiving images, prior knowledge and experience may improve diagnostic accuracy (). For example, in a case study reported by Panikkath and Panikkath (2014), Mach bands at the lateral margin of the right atrium were initially interpreted as evidence of pneumopericardium in a chest radiograph. Awareness that this might be a perceptual effect—and the discovery that other follow-up imaging did not show signs of air around the heart—revealed that this radiolucent shadow was in fact caused by Mach bands. The effects of context and experience on the interpretation of Mach bands in radiology have also been demonstrated experimentally by Nielsen (2001): dental students frequently misinterpreted a Mach band illusion as a root fracture, but experienced dentists (with three or more years of practice) only tended to provide the same misdiagnosis when they were given a correlative history, such as that the patient suffered trauma during a sporting event (a scenario in which a root fracture might be expected).
Simultaneous Contrast Effect
The simultaneous contrast effect is another brightness/contrast illusion, occurring when differences in luminance between an object and its background, or between one object and another, alter the object’s perceived brightness. In radiological contexts, differences in background density can alter the perceived density of two adjacent objects due to simultaneous contrast (Gordenne and Malchair, 1988; see Figure 3). Importantly, whereas Mach bands usually cover a narrow area (resembling a thin band), simultaneous contrast can cover wide areas.
FIGURE 3
Pareidolia
Pareidolia is the illusion of significance in meaningless sensory inputs. Everyday examples include seeing a face on the moon or finding animal shapes in clouds. We note that pareidolia does not always result in visual illusions and it can occur in other sensory domains: for example, hearing lyrics in music played backwards. Pareidolias result from the same neural processes that extract actual (rather than imagined) meaning from meaningful, real-world objects (Voss et al., 2011).
Pareidolia often serves as an amusing finding that does not hinder or help the radiologist. For example, in one case report describing a man with painful inflammation on his testicles, the testicular mass on the ultrasound image resembled the face of a man in severe pain (Roberts and Touma, 2011; see Figure 4).
FIGURE 4
Yet, because radiologic diagnosis involves the recognition of patterns, pareidolias can be used in similar ways as other mental representations of normal and abnormal conditions exploited by expert radiologists. Importantly, because there is consistency in the perception of pareidolic elements across observers, mentors can share their own perceptual experiences with trainees, and highlight those pareidolia patterns that can aid the diagnostic process. Just as expert radiologists call to mind pre-existing mental representations (such as that of chronic lung disease) while attempting to fit cases to a possible diagnosis (Lesgold et al., 1988), pareidolia may help particular representations to be called to mind or fit to images.
Indeed, pareidolias can be representative of specific conditions, and therefore useful in diagnosis (Maranhão-Filho and Vincent, 2009). Radiologists have described hundreds of such diagnostic “signs”—visual analogies that suggest the presence of a condition or disease (Ridley, 2018; Ridley et al., 2018). Below, we list several pareidolias that serve as effective diagnostic heuristics.
The Snowman Sign
Radiologists often learn that the “snowman” sign in the pituitary region indicates that a pituitary macroadenoma is more likely than a meningioma. The characteristic “snowman” appearance of macroadenomas in that region—a “Figure 8” shape—results from the fact that macroadenomas are softer tumors that become indented where they pass through the sella turcica (the skull bone surrounding the pituitary gland) (Hess and Dillon, 2012).
The Swallow Tail Sign
In some cases, the absence of pareidolia can signal the presence of a disorder (). For example, some linear or comma shapes (resembling the tail of a swallow) are present on normal images of the substantia nigra, but absent in most patients with Parkinson Disease or dementia with Lewy Bodies. Thus, “loss of the swallow tail sign” indicates likely Parkinson Disease or Lewy Body dementia (Shams et al., 2017).
The Molar Tooth Sign
In the “molar tooth sign,” the midbrain resembles a molar or wisdom tooth in axial CT scans (see Figure 5). The molar tooth sign was first observed in a rare condition known as Joubert syndrome, a ciliopathy (a disorder affecting cellular cilia) characterized by an abnormal respiratory pattern, ocular motor apraxia, hypotonia and developmental delay. The syndrome is genetically heterogenous with over 30 causative genes identified, and its characteristic morphology has been reported in 82–100% of Joubert Syndrome patients (Maria et al., 1999a; Poretti et al., 2017). The molar tooth sign is also consistently found in a variety of conditions that share similar features to classic Joubert Syndrome, but with varying causative genes and hence variable involvement of organ systems. Collectively these are referred to as Joubert Syndrome and Related Disorders (JSRD) (Manley and Maertens, 2015). The molar tooth sign is not typically observed on fetal MRI until the 22nd week of gestation, so further identification of the genetic factors causing JSRD could improve early detection (Fluss et al., 2006; Saleem and Zaki, 2010; Romani et al., 2013). In addition, JSRD patients consistently have hypoplasia of the cerebellar vermis, producing an abnormal cleft between the cerebellar hemispheres and another pareidolia, the “batwing appearance” of the fourth ventricle (McGraw, 2003).
FIGURE 5
The Hummingbird Sign
Progressive supranuclear palsy (PSP), a degenerative disease characterized by ataxia and supranuclear vertical gaze palsy (; Leigh and Zee, 2015; ), is associated with the “hummingbird sign,” also called the “penguin sign” (Graber and Staudinger, 2009). On mid-sagittal plain MRI of PSP patients, midbrain atrophy appears to resemble a hummingbird (Kato et al., 2003; see Figure 6A). Because this midbrain atrophy is present only in PSP patients, the hummingbird sign can effectively differentiate PSP from Parkinson’s disease patients with a diagnostic sensitivity of around 100% (Verma and Gupta, 2012).
FIGURE 6
The Double Panda Sign
The “double panda sign” is associated with Wilson’s disease, characterized by copper accumulation in the body leading to psychiatric symptoms (Jacobs et al., 2003). It includes two separate panda faces: a “face of the giant panda” on the midbrain and a “face of the miniature panda” on the tegmentum region of the pons (see Figures 6B,C). Other disorders, such Methyl alcohol poisoning and Leigh disease, can also produce the double panda sign; thus, its presence does not result in a definitive diagnosis without additional findings ().
The Scottie Dog Sign
Pars interarticularis fractures are common sports injuries in young athletes (Syrmou et al., 2010). The “Scottie dog sign” helps radiology students to rapidly orient themselves to the different parts of the vertebrae, and then recognize this injury in oblique radiographs of the spine (Foye et al., 2014). Different parts of the vertebrae can be visualized as different parts of a dog. If the dog’s neck appears to have a collar or break, this represents a fracture or defect in the pars interarticularis (see Figure 7A).
FIGURE 7
The Winking Owl Sign
The “winking owl sign” is the most common finding in plain spinal x-rays in patients with symptomatic extradural metastasis (Livingston and Perrin, 1978). The cancer might not be recognized if the sign is not detected. Thus, the presence or absence of the “winking owl sign” sign can aid diagnosis, as the sign is not seen when the metastasis is intradural or extramedullary (Perrin et al., 1982). Foye et al. (2014) argued that teaching students the “winking owl sign” facilitates their detection of missing pedicles, allowing them to determine if any destruction is symmetrical (see Figure 7B).
The examples above indicate the value of pareidolia illusions as educational and training tools in radiology, easy for trainees to remember and apply quickly to improve diagnostic accuracy (Maranhão-Filho and Vincent, 2009; Foye et al., 2014; Manley and Maertens, 2015). Many radiologists use pareidolia in the practice of their profession, even if they are unfamiliar with the meaning of the term as an illusion involving pattern recognition (Maranhão-Filho and Vincent, 2009).
Illusions Due to Viewpoint in Space
When only 2D radiographic images are used, the limited viewpoints involved can prevent radiologists from seeing important anatomical structures. Except for cases in which contact between an object and local structures causes changes in opacity (thus providing a cue to the object’s relative location, called the “silhouette sign”; Kumaresh et al., 2015), it can be difficult or impossible to judge the anteroposterior location of an object from a single frontal image. In addition, the apparent position of structures can change with changes in line of sight, an effect called the “parallax phenomenon” (see Figures 8A–F).
FIGURE 8
Illusions from parallax phenomena or overlap of structures can be resolved by taking additional images with oblique viewpoints (as opposed to only two 90° views) or by using different imaging methods (such as fluoroscopy) to view the structures from different angles as needed (Volz and Martin, 1977; ; see Figure 9).
FIGURE 9
Expectancy Effects
It is difficult to draw a hard line between pure vision vs. visual cognition: some biases typically considered to be cognitive can prevent observers from perceiving an image in a way that matches reality. For example, radiologists change the way they view images based on how likely they believe an abnormality will be—a cognitive phenomenon with important perceptual and diagnostic implications. When abnormalities are rare, as is typically the case in radiology, there can be a higher frequency of false-negative readings (Reed et al., 2011; ; Wolfe et al., 2016). This “prevalence effect” also causes radiologists to increase the amount of viewing time for individual images when they expect the incidence of abnormalities to be high (i.e., when reading the chest radiographs of known smokers)—however, one study found that increased expectation of abnormalities was not linked to false positives (). Conversely, radiologists can fail to perceive unexpected findings even when seemingly obvious. In an illustration of the “inattentional blindness” phenomenon, radiologists may even look directly at something unexpected without perceiving it. In one study, a large image of a gorilla—48 times the size of a 5-mm lung nodule—was inserted into the last case examined during a nodule detection task. Twenty of twenty-four (83%) radiologists failed to perceive the gorilla, and the majority failed to notice it even after setting their eyes on it (). Despite this high failure rate, expert radiologists performed better than non-experts: every naïve observer tested failed to notice the gorilla. Knowing the patient history—and thus having some idea of what to look for—can additionally improve accuracy (McNeil et al., 1983; , ): providing image readers with pertinent clinical history increased the accuracy of chest radiolograph interpretations from 16 to 72% for trainees and from 38 to 84% for experienced radiologists in one study ().
Distinguishing Illusion From Reality
Perception begins with the visual input itself; thus, medical image optimization can help prevent certain types of radiological illusions—especially those caused by artifacts of image formation (see Krupinski, 2006; Society for Imaging Informatics in Medicine, 2020). Light conditions, screen resolution, and room luminosity can all contribute to illusions. Thus, optimization of each of these factors can reduce uncertainty and decrease the prevalence of perceptual illusions (Sabih et al., 2011).
Radiologists have begun exploring the possibility of using smartphones for interpreting some radiographic images (), though some caution is advisable in these and similar approaches, as the incidence of radiologic illusions on (the smaller) smartphone displays is yet to be assessed. The development of new medical displays and techniques in radiology practice (i.e., those involving augmented or mixed reality) may moreover create new visual contexts for illusions to occur. For example, when head-mounted displays or operating equipment are augmented to overlay neuroradiological images on a patient during surgery, the surfaces, contours, and other visual attributes of the image may overlap with the patient’s form in ways that distort perception. Thus, it is critical to explore and be aware of potential illusions that may arise as a direct result of advances in medical technologies.
A judicious strategy to help distinguish reality from illusion is to rely on multiple sources of information for diagnosis, as opposed to a single finding. For example, the absence of secondary signs of trauma can help distinguish apparent from true fractures, where the presence of other findings can provide confirmatory evidence of a lesion. Similarly, second readings of uncertain radiographic findings (Sabih et al., 2011) are known to improve accuracy (Lauritzen et al., 2016). Thus, we recommend that referral clinicians look at their patients’ studies rather than relying solely on the radiology report. While radiologists’ perceptual abilities to detect abnormalities are more developed than those of referral clinicians (Reinus, 1995; Waite et al., 2019, 2020; ), a second viewer who is more attuned to the patient’s history may notice details that the first viewer missed.
Ultimately, if a radiologist has difficulty distinguishing an illusion from a true lesion, repeating the image (ideally on a different axis) or using a different imaging modality could help. If needed, radiologists should rely on additional diagnostic tests and information from the patient history. If an image contradicts all other available evidence, a prudent radiologist should consider that the image interpretation may be wrong.
Conclusion
Perceptual errors in radiology, including the illusions described above, are a significant contributor to patient harm (Waite et al., 2017, 2019). Yet, the perceptual training of radiologists relies on the informal teaching of some “tips and tricks” and “the techniques taught, while valid, do not result from a systematic review of the perceptual literature or understanding of the human eye-brain system” (, p. 472). Visual illusions can mimic lesions, causing radiologists to report pathology where there is none—leading to unnecessary workups or more invasive procedures. Thus, knowledge of how to detect and handle these illusions may help prevent premature or incorrect diagnoses.
Extending the existing knowledge about illusory perception from well-controlled lab studies to radiological practice is far from straightforward. Searching for abnormalities in radiologic images is likely to differ in many ways from searches conducted in non-radiological settings. Notably, radiologists can arrive to the correct diagnoses even when given very little time to search images and forced to guess (, ). However, prior research has found that radiologists are no better than non-specialists at finding hidden images in line drawings and “Where’s Waldo?” illustrations, suggesting that radiology training does not result in any cognitive or perceptual improvement that generalizes across search domains. Further, increased practice with “Where’s Waldo?” images does not enhance radiologic search accuracy (though it could improve overall search speed) (Sahraian et al., 2020). Instead, radiologic expertise is specific to radiologic images, suggesting critical differences between radiologic and non-radiological perceptual tasks (Nodine and Krupinski, 1998; see also Kelly et al., 2017).
Medical image perception has many qualities that are known to increase task difficulty in other contexts: the displays are complex and cluttered (Neider and Zelinsky, 2011), targets are unknown and vary widely in appearance (), are often low in salience (), are often similar to distractors (, ), and are sometimes occluded by other structures (). In addition, images may include more than one target or no targets (; Waite et al., 2017). The intrinsic difficulty of the task may produce a set of behaviors that facilitate rapid and accurate performance in most cases but may increase the likelihood of particular illusions. A better understanding of how radiologists might be trained to avoid such illusions, and/or use them to their advantage, could enhance patient safety and save lives.
Statements
Author contributions
RA, FY, SW, SLM, and SM-C conceptualized, planned, and supervised the project. FY, RA, and SM-C reviewed the literature. FY, RA, SW, and SM-C wrote the main manuscript text. RA, FY, ZC, and SW prepared the figures and tables. All authors reviewed the manuscript.
Funding
This study was supported by the National Science Foundation (Award 1734887 to SLM and SM-C) and the National Institute of Health (Awards R01EY031971 and R01CA258021 to SM-C and SLM).
Conflict of interest
The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
References
1
AlexanderR. G.MacknikS. L.Martinez-CondeS. (2018). Microsaccade characteristics in neurological and ophthalmic disease [Review].Front. Neurol.9:144. 10.3389/fneur.2018.00144
2
AlexanderR. G.NahviR. J.ZelinskyG. J. (2019). Specifying the precision of guiding features for visual search.J. Exp. Psychol. Hum. Percept. Perform.451248–1264. 10.1037/xhp0000668
3
AlexanderR. G.WaiteS.MacknikS. L.Martinez-CondeS. (2020). What do radiologists look for? Advances and limitations of perceptual learning in radiologic search.J. Vis.201–13. 10.1167/jov.20.10.17
4
AlexanderR. G.ZelinskyG. J. (2011). Visual similarity effects in categorical search.J. Vis.11:9. 10.1167/11.8.9
5
AlexanderR. G.ZelinskyG. J. (2012). Effects of part-based similarity on visual search: the Frankenbear experiment.Vis. Res.5420–30. 10.1016/j.visres.2011.12.004
6
AlexanderR. G.ZelinskyG. J. (2018). Occluded information is restored at preview but not during visual search.J. Vis.111–16. 10.1167/18.11.4
7
AnbariM. M.WestO. C. (1997). Cervical spine trauma radiography: sources of false-negative diagnoses.Emerg. Radiol.4218–224. 10.1007/bf01508174
8
AuffermannW. F.MazurowskiM. (2018). “Perception and Training,” in The Handbook of Medical Image Perception and Techniques, edsSameiE.KrupinskiE. A. (Cambridge: Cambridge University Press).
9
BerbaumK. S.FrankenE.Jr.El-KhouryG. (1989). Impact of clinical history on radiographic detection of fractures: a comparison of radiologists and orthopedists.Am. J. Roentgenol.1531221–1224. 10.2214/ajr.153.6.1221
10
BerbaumK. S.FrankenE. A.Jr.AndersonK. L.DorfmanD. D.ErkonenW. E.FarrarG. P.et al (1993). The influence of clinical history on visual search with single and multiple abnormalities.Invest. Radiol.28191–201. 10.1097/00004424-199303000-00001
11
BerryJ. H. (1983). Cervical burnout and Mach band: two shadows of doubt in radiologic interpretation of carious lesions.J. Am. Dent. Assoc. (1939)106622–625. 10.14219/jada.archive.1983.0109
12
BiggsA. T.MitroffS. R. (2015). Improving the efficacy of security screening tasks: a review of visual search challenges and ways to mitigate their adverse effects.Appl. Cogn. Psychol.29142–148. 10.1002/acp.3083
13
BradyA.LaoideR. O.McCarthyP.McDermottR. (2012). Discrepancy and error in radiology: concepts, causes and consequences.Ulster Med. J.813–9.
14
BrancatiF.DallapiccolaB.ValenteE. M. (2010). Joubert Syndrome and related disorders.Orphanet J. Rare Dis.5:20.
15
BrunoM. A.WalkerE. A.AbujudehH. H. (2015). Understanding and confronting our mistakes: the epidemiology of error in radiology and strategies for error reduction.Radiographics351668–1676. 10.1148/rg.2015150023
16
BuckleC. E.UdawattaV.StrausC. M. (2013). Now you see it, now you don’t: visual illusions in radiology.Radiographics332087–2102. 10.1148/rg.337125204
17
ChasenM. H. (2001). Practical applications of Mach band theory in thoracic analysis.Radiology219596–610. 10.1148/radiology.219.3.r01jn37596
18
ChenA. L.RileyD. E.KingS. A.JoshiA. C.SerraA.LiaoK.et al (2010). The disturbance of gaze in progressive supranuclear palsy: implications for pathogenesis.Front. Neurol.1:147. 10.3389/fneur.2010.00147
19
ClarkK.CainM. S.AdamoS. H.MitroffS. R. (2012). “Overcoming hurdles in translating visual search research between the lab and the field,” in The Influence of Attention, Learning, and Motivation on Visual Search, edsDoddM.FlowersJ. (New York, NY: Springer), 147–181. 10.1007/978-1-4614-4794-8_7
20
CruzA. D.CastroM. C.AguiarM. F.GuimaraesL. S.GomesC. C. (2018). Impact of room lighting and image display device in the radiographic appearances of the endodontic treatments.Dentomaxillofac. Radiol.47:20170372. 10.1259/dmfr.20170372
21
CupplesT. E.EklundG.CardenosaG. (1996). Mammographic halo sign revisited.Radiology199105–108. 10.1148/radiology.199.1.8633130
22
DaffnerR. H. (1977). Pseudofracture of the dens: mach bands.Am. J. Roentgenol.128607–612. 10.2214/ajr.128.4.607
23
DaffnerR. H. (1980). Visual illusions in computed tomography: phenomena related to Mach effect.Am. J. Roentgenol.134261–264. 10.2214/ajr.134.2.261
24
DaffnerR. H. (1989). Visual illusions in the interpretation of the radiographic image.Curr. Probl. Diagn. Radiol.1862–87. 10.1016/0363-0188(89)90030-3
25
DaffnerR. H. (1995). Skeletal pseudofractures.Emerg. Radiol.296–104. 10.1007/bf02628786
26
DaffnerR. H.DeebZ. L.RothfusW. E. (1986). Pseudofractures of the cervical vertebral body.Skeletal Radiol.15295–298. 10.1007/bf00349818
27
DaffnerR. H.RosenbloomS. A.RodanB. A.MoylanJ. A. (1982). Are two views adequate for foreign object localization?J. Trauma Acute Care Surg.2266–67. 10.1097/00005373-198201000-00014
28
DasS. K.RayK. (2006). Wilson’s disease: an update.Nat. Rev. Neurol.2:482.
29
De MarziR.SeppiK.HöglB.MüllerC.ScherflerC.StefaniA.et al (2016). Loss of dorsolateral nigral hyperintensity on 3.0 tesla susceptibility-weighted imaging in idiopathic rapid eye movement sleep behavior disorder.Ann. Neurol.791026–1030. 10.1002/ana.24646
30
DrewT.VõM. L.-H.WolfeJ. M. (2013). The invisible gorilla strikes again: sustained inattentional blindness in expert observers.Psychol. Sci.241848–1853. 10.1177/0956797613479386
31
EaglemanD. M. (2001). Visual illusions and neurobiology.Nat. Rev. Neurosci.2920–926. 10.1038/35104092
32
EdholmP. (1981). Boundaries in the radiographic Image:I. general principles for perception of boundaries and their application to the image.Acta Radiol. Diagn.22457–473. 10.1177/028418518102200410
33
EvansK. K.BirdwellR. L.WolfeJ. M. (2013a). If you don’t find it often, you often don’t find it: why some cancers are missed in breast cancer screening.PLoS One8:e64366. 10.1371/journal.pone.0064366
34
EvansK. K.Georgian-SmithD.TambouretR.BirdwellR. L.WolfeJ. M. (2013b). The gist of the abnormal: above-chance medical decision making in the blink of an eye.Psychon. Bull. Rev.201170–1175. 10.3758/s13423-013-0459-3
35
EvansK. K.HaygoodT. M.CooperJ.CulpanA.-M.WolfeJ. M. (2016). A half-second glimpse often lets radiologists identify breast cancer cases even when viewing the mammogram of the opposite breast.Proc. Natl. Acad. Sci. U.S.A.11310292–10297. 10.1073/pnas.1606187113
36
FlussJ.BlaserS.ChitayatD.AkouryH.GlancP.SkidmoreM.et al (2006). Molar tooth sign in fetal brain magnetic resonance imaging leading to the prenatal diagnosis of Joubert syndrome and related disorders.J. Child Neurol.21320–324. 10.1177/08830738060210041001
37
FoyeP.AbdelshahedD.PatelS. (2014). Musculoskeletal pareidolia in medical education.Clin. Teach.11251–253. 10.1111/tct.12143
38
GleesonJ. G.KeelerL. C.ParisiM. A.MarshS. E.ChanceP. F.GlassI. A.et al (2004). Molar tooth sign of the midbrain–hindbrain junction: occurrence in multiple distinct syndromes.Am. J. Med. Genet. Part A125125–134. 10.1002/ajmg.a.20437
39
GordenneW.MalchairF. (1988). Mach bands in mammography.Radiology16955–58. 10.1148/radiology.169.1.2843941
40
GraberJ. J.StaudingerR. (2009). Teaching NeuroImages: “Penguin” or “hummingbird” sign and midbrain atrophy in progressive supranuclear palsy.Neurology72:e81. 10.1212/WNL.0b013e3181a2e815
41
GriffinF. M.InsallJ. N.ScuderiG. R. (1998). The posterior condylar angle in osteoarthritic knees.J. Arthroplasty13812–815. 10.1016/s0883-5403(98)90036-5
42
HardyS. M. (2003). The sandwich sign.Radiology226651–652. 10.1148/radiol.2263020109
43
HessC. P.DillonW. P. (2012). Imaging the pituitary and parasellar region.Neurosurg. Clin.23529–542. 10.1016/j.nec.2012.06.002
44
ImanzadehA.PourjabbarS.MezrichJ. (2020). Medicolegal training in radiology; an overlooked component of the non-interpretive skills curriculum.Clin. Imaging65138–142. 10.1016/j.clinimag.2020.04.002
45
JacobsD. A.MarkowitzC. E.LiebeskindD. S.GalettaS. L. (2003). The “double panda sign” in Wilson’s disease.Neurology61:969. 10.1212/01.wnl.0000085871.98174.4e
46
KatoN.AraiK.HattoriT. (2003). Study of the rostral midbrain atrophy in progressive supranuclear palsy.J. Neurol. Sci.21057–60. 10.1016/s0022-510x(03)00014-5
47
KatteaM. O.LababedeO. (2015). Differentiating pneumothorax from the common radiographic skinfold artifact.Ann. Am. Thorac. Soc.12928–931. 10.1513/annalsats.201412-576as
48
KeatsT. E.MarkW. (2001). Atlas of Normal Roentgen Variants That May Simulate Disease.St. Louis: Mosby, 888–895. Year Book.
49
KellyB.RainfordL. A.McEnteeM. F.KavanaghE. C. (2017). Influence of radiology expertise on the perception of nonmedical images.J. Med. Imaging5:031402. 10.1117/1.JMI.5.3.031402
50
KrupinskiE. A. (2006). Technology and perception in the 21st-century reading room.J. Am. Coll. Radiol.3433–440. 10.1016/j.jacr.2006.02.022
51
KumareshA.KumarM.DevB.GorantlaR.SaiP. V.ThanasekaraanV. (2015). Back to basics–‘must know’classical signs in thoracic radiology.J. Clin. Imaging Sci.5:43. 10.4103/2156-7514.161977
52
LaneE. J.ProtoA. V.PhillipsT. W. (1976). Mach bands and density perception.Radiology1219–17. 10.1148/121.1.9
53
LauritzenP. M.AndersenJ. G.StokkeM. V.TennstrandA. L.AamodtR.HeggelundT.et al (2016). Radiologist-initiated double reading of abdominal CT: retrospective analysis of the clinical importance of changes to radiology reports.BMJ Qual. Saf.25595–603. 10.1136/bmjqs-2015-004536
54
LeighR. J.ZeeD. S. (2015). The Neurology of Eye Movements, 5th Edn. Oxford: Oxford University Press.
55
LesgoldA.RubinsonH.FeltovichP.GlaserR.KlopferD.WangY. (1988). “Expertise in a complex skill: diagnosing x-ray pictures,” in The Nature of Expertise, edsChiM. T. H.GlaserR.FarrM. J. (Hillsdale, NJ: Lawrence Erlbaum Associates, Inc), 311–342.
56
LienW. C.HuangS. P.LiuK. L.ChangJ. H.LaiT. I.LiuY. P.et al (2009). The sandwich sign of non-lymphomatous origin.J. Clin. Ultrasound37212–214. 10.1002/jcu.20540
57
LivingstonK. E.PerrinR. G. (1978). The neurosurgical management of spinal metastases causing cord and cauda equina compression.J. Neurosurg.49839–843. 10.3171/jns.1978.49.6.0839
58
ManleyA. T.MaertensP. M. (2015). The Shepherd’s crook sign: a new neuroimaging pareidolia in joubert syndrome.J. Neuroimaging25510–512. 10.1111/jon.12159
59
Maranhão-FilhoP.VincentM. B. (2009). Neuropareidolia: diagnostic clues apropos of visual illusions.Arq. Neuropsiquiatr.671117–1123. 10.1590/s0004-282x2009000600033
60
MariaB. L.BoltshauserE.PalmerS. C.TranT. X. (1999a). Clinical features and revised diagnostic criteria in Joubert syndrome.J. Child Neurol.14583–590. 10.1177/088307389901400906
61
MariaB. L.QuislingR. G.RosainzL. C.YachnisA. T.GittenJ.DedeD.et al (1999b). Molar tooth sign in Joubert syndrome: clinical, radiologic, and pathologic significance.J. Child Neurol.14368–376. 10.1177/088307389901400605
62
Martinez-CondeS.MacknikS. (2017). Champions of Illusion: The Science Behind Mind-boggling Images and Mystifying Brain Puzzles.New York, NY: Scientific American/Farrar, Straus and Giroux.
63
MaskellG. (2019). Error in radiology—where are we now?Br. J. Radiol.92:20180845. 10.1259/bjr.20180845
64
MassoudT. F.KalninsA. (2016). Glioblastoma invoking “killer” rabbits of the middle ages.World Neurosurg.92140–141. 10.1016/j.wneu.2016.04.116
65
McGrawP. (2003). The molar tooth sign.Radiology229671–672. 10.1148/radiol.2293020764
66
McGurkS.BrauerK.MacfarlaneT.DuncanK. (2008). The effect of voice recognition software on comparative error rates in radiology reports.Br. J. Radiol.81767–770. 10.1259/bjr/20698753
67
McNeilB.HanleyJ.FunkensteinH.WallmanJ. (1983). Paired receiver operating characteristic curves and the effect of history on radiographic interpretation. CT of the head as a case study.Radiology14975–77. 10.1148/radiology.149.1.6611955
68
MurrayM. M.HerrmannC. S. (2013). Illusory contours: a window onto the neurophysiology of constructing perception.Trends Cogn. Sci.17471–481. 10.1016/j.tics.2013.07.004
69
NeiderM. B.ZelinskyG. J. (2011). Cutting through the clutter: searching for targets in evolving complex scenes.J. Vis.11:7. 10.1167/11.14.7
70
NielsenC. J. (2001). Effect of scenario and experience on interpretation of mach bands.J. Endod.27687–691. 10.1097/00004770-200111000-00009
71
NinioJ. (2014). Geometrical illusions are not always where you think they are: a review of some classical and less classical illusions, and ways to describe them.Front. Hum. Neurosci.8:856. 10.3389/fnhum.2014.00856
72
NodineC. F.KrupinskiE. A. (1998). Perceptual skill, radiology expertise, and visual test performance with NINA and WALDO.Acad. Radiol.5603–612. 10.1016/s1076-6332(98)80295-x
73
OhmoriT.KabataT.KajinoY.TagaT.InoueD.YamamotoT.et al (2018). Usefulness of the “grand-piano sign” for determining femoral rotational alignment in total knee arthroplasty.Knee2515–24. 10.1016/j.knee.2017.11.008
74
PanikkathR.PanikkathD. (2014). Mach band sign: an optical illusion.Proc. Bayl. Univ. Med. Cent.27364–365. 10.1080/08998280.2014.11929161
75
PapageorgesM. (1991). How the mach phenomenon and shape affect the radiographic appearance of skeletal structures.Vet. Radiol.32191–195. 10.1111/j.1740-8261.1991.tb00106.x
76
PapageorgesM.SandeR. D. (1990). The mach phenomenon.Vet. Radiol.31274–280. 10.1111/j.1740-8261.1990.tb00801.x
77
PerrinR. G.LivingstonK. E.AarabiB. (1982). Intradural extramedullary spinal metastasis: a report of 10 cases.J. Neurosurg.56835–837. 10.3171/jns.1982.56.6.0835
78
PerrinR. G.McBroomR. (1987). Anterior versus posterior decompression for symptomatic spinal metastasis.Can. J. Neurol. Sci.1475–80. 10.1017/s0317167100026871
79
PorettiA.SnowJ.SummersA. C.TekesA.HuismanT. A.AygunN.et al (2017). Joubert syndrome: neuroimaging findings in 110 patients in correlation with cognitive function and genetic cause.J. Med. Genet.54521–529. 10.1136/jmedgenet-2016-104425
80
PropperR. A.SkolnickM. L.WeinsteinB. J.DekkerA. (1980). The nonspecificity of the thyroid halo sign.J. Clin. Ultrasound8129–132. 10.1002/jcu.1870080206
81
RabyN.BermanL.MorleyS.De LaceyG. (2014). Accident and Emergency Radiology: A Survival Guide E-Book.Philadelphia, PA: Elsevier Health Sciences.
82
ReedW. M.RyanJ. T.McEnteeM. F.EvanoffM. G.BrennanP. C. (2011). The effect of abnormality-prevalence expectation on expert observer performance and visual search.Radiology258938–943. 10.1148/radiol.10101090
83
ReevesP. (2004). Visions of normality? Early experiences of radiographic reporting.Radiography10213–216. 10.1016/j.radi.2004.05.002
84
ReinusW. R. (1995). Emergency physician error rates for interpretation of plain radiographs and utilization of radiologist consultation.Emerg. Radiol.2207–213. 10.1007/bf02615821
85
RenfrewD. L.FrankenE. A.Jr.BerbaumK. S.WeigeltF. H.Abu-YousefM. M. (1992). Error in radiology: classification and lessons in 182 cases presented at a problem case conference.Radiology183145–150. 10.1148/radiology.183.1.1549661
86
RidleyL. J. (2018). The use of animal signs in Radiology: lessons in image interpretation from art theory, patternicity and analogy.J. Med. Imaging Radiat. Oncol.623–6. 10.1111/1754-9485.12782
87
RidleyL. J.XiangH.HanJ.RidleyW. E. (2018). Animal signs in Radiology: method of creating a compendium.J. Med. Imaging Radiat. Oncol.627–11. 10.1111/1754-9485.12783
88
RobertsG. G.ToumaN. J. (2011). The face of testicular pain: a surprising ultrasound finding.Urology78:565. 10.1016/j.urology.2010.11.017
89
RomaniM.MicalizziA.ValenteE. M. (2013). Joubert syndrome: congenital cerebellar ataxia with the molar tooth.Lancet Neurol.12894–905. 10.1016/S1474-4422(13)70136-4
90
SabihA.SabihQ.KhanA. N. (2011). Image perception and interpretation of abnormalities; can we believe our eyes? Can we do something about it?Insights Imaging247–55. 10.1007/s13244-010-0048-1
91
SahraianS.YousemD.BeheshtianE.JalilianhasanpourR.MoralesR. E.KrupinskiE. A.et al (2020). Improving radiology trainees’ perception using where’s waldo?Acad. Radiol.10.1016/j.acra.2020.10.023[Epub ahead of print].
92
SaleemS.ZakiM. (2010). Role of MR imaging in prenatal diagnosis of pregnancies at risk for Joubert syndrome and related cerebellar disorders.Am. J. Neuroradiol.31424–429. 10.3174/ajnr.a1867
93
SameiE.KrupinskiE. A. (2018). The Handbook of Medical Image Perception and Techniques.Cambridge: Cambridge University Press.
94
SecginC. K.GulsahiA.ArhunN. (2016). Diagnostic challenge: instances mimicking a proximal carious lesion detected by bitewing radiography.OHMD151–5.
95
ShamsS.FällmarD.SchwarzS.WahlundL.-O.Van WestenD.HanssonO.et al (2017). MRI of the swallow tail sign: a useful marker in the diagnosis of Lewy body dementia?Am. J. Neuroradiol.381737–1741. 10.3174/ajnr.a5274
96
ShapiroA. G.TodorovicD. (2016). The Oxford Compendium of Visual Illusions.Oxford: Oxford University Press.
97
SlaterA.TaylorS. A.TamE.GartnerL.ScarthJ.PeirisC.et al (2006). Reader error during CT colonography: causes and implications for training.Eur. Radiol.162275–2283. 10.1007/s00330-006-0299-x
98
Society for Imaging Informatics in Medicine (2020). ACR-AAPM-SIIM Practice Guidelines and Technical Standards. Available online at: https://siim.org/page/practice_guidelines(accessed February 2, 2020).
99
SogurE.BaksiB. G.MertA. (2012). The effect of delayed scanning of storage phosphor plates on occlusal caries detection.Dentomaxillofac. Radiol.41309–315. 10.1259/dmfr/12935491
100
SonamK.BinduP. S.GayathriN.KhanN. A.GovindarajuC.ArvindaH. R.et al (2014). The “double panda” sign in Leigh disease.J. Child Neurol.29980–982. 10.1177/0883073813484968
101
SwannC. A.KopansD. B.KoernerF. C.McCarthyK. A.WhiteG.HallD. A. (1987). The halo sign and malignant breast lesions.Am. J. Roentgenol.1491145–1147. 10.2214/ajr.149.6.1145
102
SyrmouE.TsitsopoulosP.MarinopoulosD.TsonidisC.AnagnostopoulosI.TsitsopoulosP. (2010). Spondylolysis: a review and reappraisal.Hippokratia14:17.
103
ThomsonE. M.JohnsonO. N. (2012). Essentials of Dental Radiography for Dental Assistants and Hygienists.Upper Saddle River, NJ: Pearson.
104
VermaR.GuptaM. (2012). Hummingbird sign in progressive supranuclear palsy.Ann. Saudi Med.32663–664. 10.5144/0256-4947.2012.663
105
VolzR. G.MartinM. D. (1977). Illusory biplane radiographic images.Radiology122695–697. 10.1148/122.3.695
106
VossJ. L.FedermeierK. D.PallerK. A. (2011). The potato chip really does look like Elvis! Neural hallmarks of conceptual processing associated with finding novel shapes subjectively meaningful.Cereb. Cortex222354–2364. 10.1093/cercor/bhr315
107
WaiteS.FarooqZ.GrigorianA.SistromC.KollaS.MancusoA.et al (2020). A review of perceptual expertise in radiology-how it develops, how we can test it, and why humans still matter in the era of Artificial Intelligence. [Special Review].Acad. Radiol.2726–38. 10.1016/j.acra.2019.08.018
108
WaiteS.GrigorianA.AlexanderR. G.MacknikS. L.CarrascoM.HeegerD. J.et al (2019). Analysis of perceptual expertise in radiology – current knowledge and a new perspective [Review].Front. Hum. Neurosci.13:213. 10.3389/fnhum.2019.00213
109
WaiteS.ScottJ.GaleB.FuchsT.KollaS.ReedeD. (2017). Interpretive error in radiology.Am. J. Roentgenol.208739–749. 10.2214/ajr.16.16963
110
WestheimerG. (2008). Illusions in the spatial sense of the eye: geometrical–optical illusions and the neural representation of space.Vis. Res.482128–2142. 10.1016/j.visres.2008.05.016
111
WolfeJ. M.EvansK. K.DrewT.AizenmanA.JosephsE. (2016). How do radiologists use the human search engine?Radiat. Protect. dosimetry16924–31. 10.1093/rpd/ncv501
Summary
Keywords
radiological error, illusions, false positives, perceptual expertise, image quality, medical image perception, medical images, false negatives
Citation
Alexander RG, Yazdanie F, Waite S, Chaudhry ZA, Kolla S, Macknik SL and Martinez-Conde S (2021) Visual Illusions in Radiology: Untrue Perceptions in Medical Images and Their Implications for Diagnostic Accuracy. Front. Neurosci. 15:629469. doi: 10.3389/fnins.2021.629469
Received
17 November 2020
Accepted
19 April 2021
Published
11 June 2021
Volume
15 - 2021
Edited by
Britt Anderson, University of Waterloo, Canada
Reviewed by
Elizabeth Krupinski, Emory University, United States; Philip Tseng, Taipei Medical University, Taiwan
Updates
Copyright
© 2021 Alexander, Yazdanie, Waite, Chaudhry, Kolla, Macknik and Martinez-Conde.
This is an open-access article distributed under the terms of the Creative Commons Attribution License (CC BY). The use, distribution or reproduction in other forums is permitted, provided the original author(s) and the copyright owner(s) are credited and that the original publication in this journal is cited, in accordance with accepted academic practice. No use, distribution or reproduction is permitted which does not comply with these terms.
*Correspondence: Robert G. Alexander, robert.alexander@downstate.eduStephen Waite, stephen.waite@downstate.edu
†These authors have contributed equally to this work
This article was submitted to Perception Science, a section of the journal Frontiers in Neuroscience
Disclaimer
All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article or claim that may be made by its manufacturer is not guaranteed or endorsed by the publisher.