14
Christina Vong 24244058 1

Christina Vong 24244058 - Psychophysics of Vision in Medical Imaging Essay.docx

Embed Size (px)

Citation preview

Christina Vong 24244058

1

RAD4160 Psychophysics of Vision in Medical Imaging

INTRODUCTION

Visual perception is a one of the fundamental facets of image interpretation in the medical imaging field. While increasing technological developments in modern medicine today have lead to advanced imaging techniques, the interpretation of radiographic images is still heavily reliant upon the visual system and associated perception and cognitive processes (Krupinski 2010).

THE EYE AND VISUAL PATHWAY

The visual pathway begins at the retina of the eye, where information from the environment is received in the form of light and converted into a neural signal via phototransduction. This transformation process occurs in the photoreceptors of the retina, which comprises of two types of cells; rods and cones. Visual pigments contained in the photoreceptors alter slightly depending on the photons that are absorbed (McCaa 1982). Rods and cones are active in different levels of light, with rods more active in dim lighting (scotopic vision) and cones more active in higher light levels (Remington 2012a). Cone cells are further designated into three different types that are excited by different wavelengths of light, allowing colour perception. On the other hand, scotopic vision is colourblind, as rods lack the mechanism to retain wavelength information (Cornsweet 1970).

Once phototransduction is complete, the corresponding rod and cone bipolar cells are excited and act as either indirect or direct pathways to the third order neurons known as the ganglion cells. Axons of the ganglion cells converge at the optic disc to form the optic nerve as they exit the eye (McCaa 1982). The signal continues along until the two optic nerves cross at the optic chiasm to synapse at the lateral geniculate nucleus (LGN) via the optic tract (Alfano 1961). The LGN serves as a complex processing centre in the thalamus that is responsible for regulating the visual information (Remington 2012b).

The final neurons in the visual pathway that pass the signal to the visual cortex are the optic radiations. The visual cortex itself, is sub-divided into several regions known as the Brodmann areas 17 (primary or striate cortex), 18, 19 as well as the extrastriate cortex (Remington 2012b).

Figure 1. The visual pathway, from the eye to the visual cortex.

(Remington 2012b)

VISUAL PERCEPTION AND IMAGE INTERPRETATION

The visual pathway comprises the perceptual processing component of image interpretation by obtaining information from the external environment. The second element in visual perception involves cognitive processes (Morita et al. 2008).

There are two techniques our visual cortex employs during visual object recognition; the bottom-up process and top-down process (Taddei-Ferretti et al. 2008). In bottom-up processing, perceptual factors drive the cognitive counterpart, where sensory stimulation from individual features first occurs before they are combined to form a complete image in the brain (Morita et al. 2008). Conversely, top-down processing involves initial recognition of the whole image, followed by segmentation into its distinctive features. Epshtein et al (2008) proposed a single bottom-up top-down cycle, where the bottom-up and top-down processes were performed consecutively to correct for visual errors that were overlooked in the first bottom-up method.

Visual search and attention is also required due to the restricted foveal vision of the human eye. Selective attention is executed in order to find an object or analyse a fixed target. The eye moves around the image to scrutinize relevant details so that the brain can better distinguish differences between the target object and its surroundings (Krupinski 2010). Recognition is then possible, by connecting the sensory input with stored data in memory. However, visual search is influenced by context; objects in highly familiar settings are recognized more accurately and more information is retrieved (Epstein 2005).

Furthermore, conceptual knowledge associated with corresponding visual stimuli facilitates visual perception and interpretation. Features of the target object such as shape, dimensions and colour, can assist in more efficient detection when used in conjunction with the relevant background (Cheung & Gauthier 2014). In situations where scenic material is not provided, structure of the background and previous knowledge can be applied as additional information to aid in accurate object recognition and localization (Oliva & Torralba 2007).

INTERPRETATION IN MEDICAL IMAGING

Image interpretation is especially crucial in radiology, where the images seen and reported on have a significant impact on the patients health outcomes. Several elements of interpretation such as contrast, signal-to-noise ratio (SNR), colour and grey-scale play important parts in radiologic image interpretation (Sabih et al. 2011).

Depending on the lesion of interest, different contrast levels are required for target identification. Low-contrast pathology, for example, isoechoic lesions on ultrasound, are the most difficult to detect as they are only distinguishable by other features like border irregularities (Sabih et al. 2011). In such situations, it is natural to assume that moving closer or making the target larger will assist in localisation, however SNRs role in perception indicates otherwise. In imaging, SNR is related to the amount of photons that are actually detected at the detector plate. It is a parameter that affects image quality, which in turn, has an impact on image interpretation (Krupinski 2010). In fact, increasing viewing distance has shown to improve perception of many lesions (Sabih et al. 2011).

While the cone cells in the retina make the human eye sensitive to colour, medical images are instead displayed in grey scale, as individuals perceive colours in different ways. Hence in radiology, colour is not beneficial in terms of identifying normal versus abnormal anatomy. A broad range of grey shades is required to avoid the brain seeing adjacent shades as one level of grey (Kimpe & Tuytschaever 2007).

VISUAL ILLUSIONS IN RADIOLOGIC IMAGE INTERPRETATION

Visual illusions occur when reality is represented in a distorted or altered way (Figure 2). When imaged, the human body is a complex combination of overlapping shadows and differing radiographic densities. Sensory and perceptual aspects of our visual system can give rise to artefacts that can be mistaken as pathology. Illusions can be categorized into either sensory or perceptual illusions, where sensation illusions occur during the phototransduction process and perceptual illusions emerge from the interpretation phase after sensory input has been analysed (Buckle et al. 2013). One of the most common illusions of sensation in radiology is the Mach band effect (Panikkath & Panikkath 2014).

(Buckle et al. 2013)

Figure 2. (a) Typical ambiguous image in the form of a visual illusion; Mother or Wife? which can either be interpreted as (b) a young woman looking away (blue circle) or (c) as an old woman looking downwards (red circle)

MACH BAND SIGNS

Mach bands appear as bright and dark lines at the borders of two objects with different contrast levels or optical densities (Daffner 1989). The Mach band effect is a result of lateral inhibition of the bipolar neurons in the retina by the horizontal cells in the eye. Once a phototransduced impulse excites a bipolar cell, the adjacent bipolar neurons are inhibited (Buckle et al. 2013). This either increases or decreases their response to light signals, depending on the distance between the two cells (Chasen 2001).

The edge enhancement effect that results is commonly seen in radiography, such as the borders that demarcate the lungs and mediastinum. Mach bands can also provide diagnostic information; the radiologic halo-sign that is indicative of a benign breast mass describes a Mach band sign where a dark outline surrounds a smooth-bordered breast lesion (Buckle et al. 2013).

However, Mach bands can also simulate pathology that is not actually present (Figure 3). Skin folds or the posterior arch of the atlas projected over the base of the dens is a common illusion created by a Mach band (Figure 3a) (Daffner 1989). Superimposition of surrounding bony structures can also give rise to pseudofractures.

Figure 3. (a) Mach band seen (arrows) where the posterior arch of the atlas is projected over C2 and mimics a dens fracture. Lateral view was normal. (b) AP chest x-ray demonstrating a Mach band between the lungs and left heart border simulating pneumomediastinum.

(Daffner 1989)

(a) (b)

MEASURING DETECTION ACCURACY IN RADIOLOGY

Taking into account these issues with identification and detection of pathology, radiologists diagnostic accuracy can be measured by first determining the sensitivity and specificity (by calculating from the decision matrix in Figure 4), followed by the overall accuracy of their interpretations (Zhou et al. 2011).

Figure 4. Decision matrix illustrating how TPs, FPs, FNs and TNs are determined by comparing the radiologists interpretations with the real situation. Sensitivity gives the proportion of true positives and is calculated by TP/(TP + FN), while specificity gives the proportion of true negatives and is given by TN/(TN + FP). Accuracy can then be determined by TN + TP/(TN + TP + FP + FN).

REDUCING ERROR

To reduce perception errors in radiologic reporting, continued education and training of radiologists is important (Sabih et al. 2011). Understanding the physiology of visual illusions such as Mach bands is a vital aspect of image interpretation, not only because they can aid in making diagnoses, but also because they can be mistaken for pathology (Buckle et al. 2013). If radiologists are aware of its existence, misinterpretations due to visual illusions are less likely to occur.

CONCLUSION

The human visual system is a crucial component of the radiologic image interpretation process; ultimately, the radiologists findings are based on their visual impression of the images. Images first interact with the eye, before being processed through the visual pathway to the brain, where the signals are interpreted. In order to provide accurate diagnostic reports, the basic principles of sensory and perceptual vision should be understood. Errors associated with visual perception such as optical illusions do exist; hence the detection accuracy of radiologists should be measured regularly to ensure that correct diagnoses are being made.

Word count: 1441

REFERENCES

Alfano, J 1961, 'ANATOMY AND LESIONS OF THE VISUAL PATHWAY', International Ophthalmology Clinics, vol. 1, no. 3, pp. 731-7

Buckle, CE, Udawatta, V & Straus, CM 2013, 'Now You See It, Now You Dont: Visual Illusions in Radiology', Radiographics, vol. 33, no. 7, pp. 2087-102, doi:10.1148/rg.337125204

Chasen, MH 2001, 'Practical Applications of Mach Band Theory in Thoracic Analysis', Radiology, vol. 219, no. 3, pp. 596-610, doi:10.1148/radiology.219.3.r01jn37596

Cheung, OS & Gauthier, I 2014, 'Visual appearance interacts with conceptual knowledge in object recognition', Frontiers in Psychology, vol. 5, p. 793, doi:10.3389/fpsyg.2014.00793

Cornsweet, TN 1970, 'X - COLOR VISION IIITHE PERCEPTION OF COLOR', in TN Cornsweet (ed.), Visual Perception, Academic Press, pp. 224-67, DOI http://dx.doi.org/10.1016/B978-0-12-189750-5.50014-2, http://www.sciencedirect.com/science/article/pii/B9780121897505500142.

Daffner, RH 1989, 'Visual illusions in the interpretation of the radiographic image', Current Problems in Diagnostic Radiology, vol. 18, no. 2, pp. 62-87, doi:http://dx.doi.org/10.1016/0363-0188(89)90030-3

Epshtein, B, Lifshitz, I & Ullman, S 2008, 'Image interpretation by a single bottom-up top-down cycle', Proceedings of the National Academy of Sciences, vol. 105, no. 38, pp. 14298-303, doi:10.1073/pnas.0800968105

Epstein, R 2005, 'The cortical basis of visual scene processing', Visual Cognition, vol. 12, no. 6, pp. 954-78, doi:10.1080/13506280444000607

Kimpe, T & Tuytschaever, T 2007, 'Increasing the Number of Gray Shades in Medical Display SystemsHow Much is Enough?', Journal of Digital Imaging, vol. 20, no. 4, pp. 422-32, doi:10.1007/s10278-006-1052-3

Krupinski, EA 2010, 'Current perspectives in medical image perception', Attention, Perception, and Psychophysics, vol. 72, no. 5, pp. 1205-17, doi:10.3758/APP.72.5.1205

McCaa, CS 1982, 'The Eye and Visual Nervous System: Anatomy, Physiology and Toxicology', Environmental Health Perspectives, vol. 44, pp. 1-8, doi:10.2307/3429469

Morita, J, Miwa, K, Kitasaka, T, Mori, K, Suenaga, Y, Iwano, S, Ikeda, M & Ishigaki, T 2008, 'Interactions of perceptual and conceptual processing: Expertise in medical image diagnosis', International Journal of Human-Computer Studies, vol. 66, no. 5, pp. 370-90, doi:http://dx.doi.org/10.1016/j.ijhcs.2007.11.004

Oliva, A & Torralba, A 2007, 'The role of context in object recognition', Trends in Cognitive Sciences, vol. 11, no. 12, pp. 520-7, doi:http://dx.doi.org/10.1016/j.tics.2007.09.009

Panikkath, R & Panikkath, D 2014, 'Mach band sign: an optical illusion', Baylor University Medical Center Proceedings, vol. 27, p. 364+

Remington, LA 2012a, 'Chapter 1 - Visual System', in LA Remington (ed.), Clinical Anatomy and Physiology of the Visual System (Third Edition), Butterworth-Heinemann, Saint Louis, pp. 1-9, DOI http://dx.doi.org/10.1016/B978-1-4377-1926-0.10001-3, http://www.sciencedirect.com/science/article/pii/B9781437719260100013.

2012b, 'Chapter 13 - Visual Pathway', in LA Remington (ed.), Clinical Anatomy and Physiology of the Visual System (Third Edition), Butterworth-Heinemann, Saint Louis, pp. 233-52, DOI http://dx.doi.org/10.1016/B978-1-4377-1926-0.10013-X, http://www.sciencedirect.com/science/article/pii/B978143771926010013X.

Sabih, D-e, Sabih, A, Sabih, Q & Khan, A 2011, 'Image perception and interpretation of abnormalities; can we believe our eyes? Can we do something about it?', Insights into Imaging, vol. 2, no. 1, pp. 47-55, doi:10.1007/s13244-010-0048-1

Taddei-Ferretti, C, Radilova, J, Musio, C, Santillo, S, Cibelli, E, Cotugno, A & Radil, T 2008, 'The effects of pattern shape, subliminal stimulation, and voluntary control on multistable visual perception', Brain Research, vol. 1225, no. 0, pp. 163-70, doi:http://dx.doi.org/10.1016/j.brainres.2008.04.064

Zhou, X-H, Obuchowski, NA & McClish, DK 2011, 'Measures of Diagnostic Accuracy', in Statistical Methods in Diagnostic Medicine, John Wiley & Sons, Inc., pp. 13-55, DOI 10.1002/9780470906514.ch2, http://dx.doi.org/10.1002/9780470906514.ch2.