Accuracy of Visual Fields in Localizing MRI Lesions in Posterior Cerebral Artery Infarction

Background: The representation of the visual field in visual cortex was established over a century ago by correlating perimetric defects with the estimated location of war wounds. The availability of high-definition MRI offers the possibility of more precise correlation. Methods: Homonymous hemianopias disclosed on automated visual fields (HVFs) were drawn from an electronic medical record search from 2009 to 2020 at the Michigan Medicine, a tertiary care academic medical center. The patterns of the visual field defects (VFDs) were interpreted by a consensus of 2 authors. The VFDs were correlated with the location of MRI lesions in 92 patients with posterior cerebral artery (PCA) domain ischemic strokes, as determined by the neuroradiologist author, who was masked as to the VFDs. Results: Among the 77 VFDs confined to 1 hemifield, 74 (96%) correctly predicted the side of the visual cortex lesion. In 3 cases, the MRI lesion in the opposite cerebral hemisphere was not foretold. Among the 15 VFDs present in both hemifields, 5 (33.3%) overestimated the MRI lesions, which were evident in only 1 hemisphere. Among the 30 VFDs confined to 1 quadrant, 29 (97%) correctly predicted the lesioned visual cortex quadrant. However, 14 VFDs failed to predict MRI lesions present in both superior and inferior visual cortex quadrants on the same side. Those unpredicted lesions mostly had subtle or indistinct signal abnormalities or were confined to anterior visual cortex, an area that is inaccessible with the HVF test protocol used in this study. Conclusion: In this study of PCA ischemic stroke, VFDs limited to 1 hemifield were accurate in locating the side and quadrant of the MRI visual cortex lesions. However, the quadrantic VFDs sometimes failed to predict that the lesions involved both the superior and inferior quadrants on the same side, largely because those lesions had subtle imaging features that defied accurate radiologic assessment or were out of the reach of the visual field test protocol.

Original Contribution Section Editors: Clare Fraser, MD Susan Mollan, MD Accuracy of Visual Fields in Localizing MRI Lesions in Posterior Cerebral Artery Infarction Juno Cho, MA, Eric Liao, MD, Jonathan D. Trobe, MD Background: The representation of the visual field in visual cortex was established over a century ago by correlating perimetric defects with the estimated location of war wounds. The availability of high-definition MRI offers the possibility of more precise correlation. Methods: Homonymous hemianopias disclosed on automated visual fields (HVFs) were drawn from an electronic medical record search from 2009 to 2020 at the Michigan Medicine, a tertiary care academic medical center. The patterns of the visual field defects (VFDs) were interpreted by a consensus of 2 authors. The VFDs were correlated with the location of MRI lesions in 92 patients with posterior cerebral artery (PCA) domain ischemic strokes, as determined by the neuroradiologist author, who was masked as to the VFDs. Results: Among the 77 VFDs confined to 1 hemifield, 74 (96%) correctly predicted the side of the visual cortex lesion. In 3 cases, the MRI lesion in the opposite cerebral hemisphere was not foretold. Among the 15 VFDs present in both hemifields, 5 (33.3%) overestimated the MRI lesions, which were evident in only 1 hemisphere. Among the 30 VFDs confined to 1 quadrant, 29 (97%) correctly predicted the lesioned visual cortex quadrant. However, 14 VFDs failed to predict MRI lesions present in both superior and inferior visual cortex quadrants on the same side. Those unpredicted lesions mostly had subtle or indistinct signal abnormalities or were confined to anterior visual cortex, an area that is inaccessible with the HVF test protocol used in this study. Conclusion: In this study of PCA ischemic stroke, VFDs limited to 1 hemifield were accurate in locating the side and quadrant of the MRI visual cortex lesions. However, the quadrantic VFDs sometimes failed to predict that the lesions involved both the superior and inferior quadrants on the same side, largely because those lesions had subtle imaging features that defied accurate radiologic assess- ment or were out of the reach of the visual field test protocol. Journal of Neuro-Ophthalmology 2022;42:360–366 doi: 10.1097/WNO.0000000000001602 © 2022 by North American Neuro-Ophthalmology Society T he representation of the visual field in primary visual cortex was first recognized by Munk in 1881, based on correlating hemianopic visual field loss with extirpation of one occipital lobe in monkeys (1,2). This discovery was confirmed in humans by Henschen in 1890, based on clinical reports (1,2). In the early 20th century, studies of patients with occipital war wounds by Inouye (3) and Holmes (4–6) correlated homonymous quadrantanopias with lesions above and below the calcarine fissure and suggested that the central visual field was represented in the posterior portion of visual cortex. The availability of high-definition MRI offers the possibility of more accurate in vivo correlation of visual field defects (VFDs) and visual cortex lesions. In this study, we sought to determine the accuracy of automated static perimetry in predicting the location of occipital MRI lesions. To allow more precise imaging assessment of lesion extent, we included only posterior cerebral artery (PCA) ischemic strokes. The lesions were defined by restricted diffusion (acute stroke), postcontrast T1 enhancement (subacute stroke), and precontrast T1 encephalomalacia or gyral hyperintensity (chronic stroke). METHODS Kellogg Eye Center (JC, JDT), Department of Ophthalmology and Visual Sciences, Ann Arbor, MI; Department of Neurology (JDT, Ann Arbor, MI); and Department of Radiology (Neuroradiology) (EL), University of Michigan, Ann Arbor, MI. The authors report no conflicts of interest. Address correspondence to Jonathan D. Trobe, MD, Kellogg Eye Center, 1000 Wall Street, Ann Arbor, MI 48105; E-mail: jdtrobe@ umich.edu 360 We obtained permission from the Michigan Medicine (University of Michigan) Institutional Review Board to conduct a 2009–2020 electronic medical records (Epic) search of patients with “homonymous hemianopia,” “visual fields,” and “MRI” using the Electronic Medical Record Search Engine of the of University of Michigan. We restricted the cohort to patients with clinical and imaging documentation of PCA ischemic stroke. From that cohort, Cho et al: J Neuro-Ophthalmol 2022; 42: 360-366 Copyright © North American Neuro-Ophthalmology Society. Unauthorized reproduction of this article is prohibited. Original Contribution we excerpted patient age, sex, cause of the lesion, pertinent neurologic deficits, date of diagnosis, first ophthalmologic examination that included automated visual field (HVF) examination, and follow-up HVF examinations. HVF examinations were performed by the 24-2 Sita Standard protocol. The HVF results on each patient were interpreted independently and in masked fashion by 2 authors (J.C. and J.D.T.). The interpretations were repeated by each reader until there was consistency and consensus. In their interpretations, the readers relied on the gray scale in combination with the pattern deviation. The gray scale was included because it is commonly used by clinicians. We excluded 3 patients whose visual field results were unreliable on the basis of excessive false-positive errors, and 3 patients whose visual fields were uninterpretable on the basis of massive defects. There were 2 VFD patterns (Fig. 1): 1) Homonymous hemianopias, defined as defects involving superior and inferior quadrants in the same affected hemifields of both eyes; 2) homonymous quadrantanopias, defined as hemianopic defects confined to the superior or inferior quadrants in the same affected hemifields of both eyes. We drew on the 1.5T or 3T brain MRIs performed closest to the time of the first HVF examinations. The studies had to disclose an ischemic stroke in the distribution of the PCA on one or both sides. We measured lesion extent based on restricted diffusion (acute stroke), postcontrast T1 enhancement (subacute stroke), and precontrast T1 encephalomalacia or gyral hyperintensity (chronic stroke). We considered the superior primary visual cortex region to lie above the calcarine fissure and the inferior primary visual cortex to lie below the calcarine fissure. To define the superior and inferior extent of the lesion, we relied on sagittal images. Where imaging in this plane was unsatisfactory or unavailable, we reformatted the relevant sequences into the sagittal plane. The final cohort consisted of 92 patients with 105 MRI hemispheric lesions (13 patients had bilateral hemispheric lesions). We divided the 92 patients into 3 groups according to the anteroposterior extent of the MRI-defined infarctions (Fig. 2): 1) Group A: infarctions limited to gray matter of primary visual cortex, defined as extending from the occipital tip to the parieto-occipital fissure; 2) Group B: infarctions involving gray matter of primary visual cortex and white matter of posterior optic radiations lying posteromedial to the calcar avis (atrium); 3) Group C: infarctions involving gray matter of primary visual cortex, posterior optic radiations, and anterior optic radiations. In correlating the VFDs with the MRI lesions, we separately analyzed the unilateral VFDs (restricted to one hemifield) and the bilateral VFDs (involved both hemifields). We then correlated the VFDs in each affected quadrant with the location of the MRI lesions in the visual cortex quadrants. We performed a subgroup analysis on the quadrantic VFDs to determine if the correlations were affected by whether the infarctions were limited to visual cortex (Group A), extended to involve posterior optic radiations (Group B), or extended to involve anterior optic radiations (Group C). The coding and data analysis were conducted with R Studio (R Studio Team, Boston, MA) using ggplot2 package (Wickham, New York, NY). In a subsequent report, we will describe the results of correlating macular sparing and homonymous paracentral scotomas with the extent of lesions along the anteroposterior extent of visual cortex. RESULTS FIG. 1. The visual field defect patterns. A. Homonymous hemianopia. B. Homonymous quadrantanopia. Cho et al: J Neuro-Ophthalmol 2022; 42: 360-366 Correlation of all VFDs with the side of visual cortex lesions (Table 1). Among the 77 VFDs confined to one hemifield, the side of the lesion was correctly predicted in 74 cases (96%). In the 3 remaining VFDs, HVFs underestimated the number of MRI lesions, which were detected in the visual cortex of both hemispheres. The MRI lesions in those 3 cases were obvious (Fig. 3A). In 2 of them, retrospective review of the HVFs suggested that we might have failed to identify a limited VFD corresponding to the second MRI lesion (Fig. 3B). Among the 15 VFDs that were present in both hemifields, 10 (66.6%) predicted the MRI lesions in both hemispheres. The remaining 5 (33.3%) HVFs overestimated the number of MRI lesions, in that there was a lesion in only 1 hemisphere. The VFDs not represented on MRI were unequivocal (Fig. 3C). Imaging disclosed a chronic lesion in 361 Copyright © North American Neuro-Ophthalmology Society. Unauthorized reproduction of this article is prohibited. Original Contribution FIG. 2. The 3 stroke groups according to the extent of infarction identified by restricted diffusion (left) and apparent diffusion coefficient (right). A. Infarction limited to visual cortex (Group A). B. Infarction involves visual cortex and posterior optic radiations (Group B). C. Infarction involves visual cortex, posterior optic radiations, and anterior optic radiations (Group C). TABLE 1. Correlation of side of hemianopic visual field defects (VFDs) with side of cerebral hemispheric lesions defined by MRI Left hemianopic VFDs (n = 41) Right hemianopic VFDs (n = 36) Bilateral hemianopic VFDs (n = 15) 362 Lesions in Both Cerebral Hemispheres Lesions in Right Hemisphere Only Lesions in Left Hemispheric Only 0 41 0 3 0 33 10 3 2 Cho et al: J Neuro-Ophthalmol 2022; 42: 360-366 Copyright © North American Neuro-Ophthalmology Society. Unauthorized reproduction of this article is prohibited. Original Contribution FIG. 3. Mismatches between automated visual fields (HVFs) and the number of MRI visual pathway lesions. A. HVFs underestimate the number of MRI lesions. Right superior homonymous quadrantanopia has corresponding left acute visual cortex lesion (arrowheads) visible on diffusion-weighted axial (left) and apparent diffusion coefficient (right) images; however, right acute visual cortex lesion (arrow) has no corresponding visual field defect (VFD). B. HVFs underestimate the number of MRI lesions. Right superior homonymous quadrantanopia has corresponding left inferior visual cortex chronic lesion visible as encephalomalacia on precontrast T1 sagittal (top left) and axial (top right) images; right superior visual cortex chronic lesion, visible as encephalomalacia on precontrast T1 sagittal (bottom left) and axial (bottom right) images, has no definite corresponding left inferior VFD; however, the lesion lies anteriorly in visual cortex, and we may have overlooked the corresponding VFD, which has wide macular sparing. C. HVFs overestimate the number of MRI lesions. Left homonymous hemianopia has corresponding right visual cortex chronic lesion visible as encephalomalacia in precontrast T1 sagittal (left) and axial (center) images (arrows); right inferior quadrantanopia has no corresponding lesion visible in left visual cortex on precontrast T1 axial (center) and sagittal (right) images. TABLE 2. Correlation of quadrant visual field defects (VFDs) and quadrant visual cortex lesions defined by MRI Lesions in Superior and Inferior Visual Cortex Quadrants Lesions in Inferior Visual Cortex Quadrant Only Lesions in Superior Visual Cortex Quadrant Only 5 4 0 9 1 11 59 7 6 Groups A, B, and C combined Superior quadrant VFDs only (n = 9) Inferior quadrant VFDs only (n = 21) Superior and inferior quadrant VFDs (n = 72) Cho et al: J Neuro-Ophthalmol 2022; 42: 360-366 363 Copyright © North American Neuro-Ophthalmology Society. Unauthorized reproduction of this article is prohibited. Original Contribution FIG. 4. HVFs underestimate the number of lesioned visual cortex quadrants. A. VFD limited to right inferior homonymous quadrantanopia, yet diffusion-weighted sagittal image shows lesions in both the left superior (arrow) and inferior (black arrow) visual cortex quadrants; calcarine fissure is indicated by arrowheads. B. VFD limited to right inferior homonymous quadrantanopia, yet precontrast T1 sagittal (left) and axial (right) images show gyral high signal lesions (laminar necrosis, arrows) that are subtle enough to defy accurate localization in relation to the estimated position of the calcarine fissure (arrowhead). C. VFD limited to left inferior homonymous paracentral scotomas, yet postcontrast T1 sagittal image shows a lesion that appears to straddle the calcarine fissure (arrowheads), involving inferior as well as superior visual cortex. D. VFD limited to right inferior homonymous quadrantanopia, yet there is an enhancing left visual cortex lesion on axial postcontrast T1 image (left) with indistinct borders on reconstructed postcontrast T1 sagittal image (right), suggesting lesions (arrows) lying inferior as well as superior to the calcarine fissure (arrowhead). E. VFD limited to inferior quadrants, yet precontrast T1 sagittal image shows encephalomalacia above and below the calcarine fissure (arrowhead) that involves the superior (black arrow) and inferior (arrow) visual cortex; however, the inferior visual cortex lesion is confined to the far anterior visual cortex, which is not accessible by the HVF protocols used in this study. HVF, automated visual fields; VFD, visual field defect. 2 cases, an acute lesion in 2 cases, and a subacute lesion in 1 case. In 4 of these cases, the lesser hemianopia corresponded to the normal-appearing visual cortex on MRI. 364 Correlation of visual field quadrant defects and visual cortex lesions (Table 2). Among the 30 VFDs limited to either the superior or inferior visual field quadrants, 29 Cho et al: J Neuro-Ophthalmol 2022; 42: 360-366 Copyright © North American Neuro-Ophthalmology Society. Unauthorized reproduction of this article is prohibited. Original Contribution FIG. 5. HVFs overestimate number of lesioned visual cortex quadrants. A. Left homonymous hemianopia, yet diffusionweighted sagittal (left) and axial (right) images show that the lesion (arrow) is confined to right inferior visual cortex below the calcarine fissure (arrowheads); in this case, the lesion extended into anterior optic radiations, where axons destined for right superior visual cortex might have been damaged. B. Left homonymous hemianopia, yet diffusion-weighted sagittal (left) and axial (right) images show that the lesion (arrow) is confined to right inferior visual cortex, below the calcarine fissure (arrowheads); in this case, the lesion extended into posterior optic radiations, where axons destined for right superior visual cortex might have been damaged. HVF, automated visual fields. (97%) correctly identified the lesioned quadrant. There were 14 VFDs that underestimated the number of MRI lesions, which were present in superior and inferior visual cortex quadrants on the same side. In 1 of those 14 VFDs, we may have failed to notice a limited homonymous quadrantanopia. In 4 of the 13 VFDs that failed to predict all the MRI lesions, the overlooked lesion was obvious (Fig. 4A). In the remaining 9 VFDs, imaging features explained the predictive errors: 1) 4 lesions had subtle signal abnormalities and may have been too slight to cause VFDs (Fig. 4B); 2) 3 lesions straddled the calcarine fissure (Fig. 4C) or had indistinct borders (Fig. 4D), making it challenging to decide if they had damaged both the superior and inferior visual cortex; 3) 2 lesions were confined to anterior visual cortex, a region not accessible on the HVF examination protocol we used (Fig. 4E). Among the 72 VFDs that involved both the superior and inferior quadrants in the same hemifield, the fields correctly predicted the visual cortex quadrant lesions in 59 cases (82%). In the remaining 13 cases, MRIs showed lesions in only 1 TABLE 3. Correlation of quadrant visual field defects (VFDs) and quadrant visual cortex lesions according to the anteroposterior extent of the infarction Lesions in Superior and Inferior Quadrants Lesions in Inferior Quadrant Only Lesions in Superior Quadrant Only 1 4 13 2 1 1 0 3 1 1 4 27 2 0 3 0 6 0 3 1 19 0 0 3 0 2 5 Group A (infarction limited to visual cortex) Superior quadrant VFDs only (n = 3) Inferior quadrant VFDs only (n = 8) Superior and inferior quadrant VFDs (n = 15) Group B (infarction extends to posterior optic radiations) Superior quadrant VFDs only (n = 3) Inferior quadrant VFDs only (n = 10) Superior and inferior quadrant VFDs (n = 30) Group C (infarction extends to anterior optic radiations) Superior quadrant VFDs only (n = 3) Inferior quadrant VFDs only (n = 3) Superior and inferior quadrant VFDs (n = 27) Cho et al: J Neuro-Ophthalmol 2022; 42: 360-366 365 Copyright © North American Neuro-Ophthalmology Society. Unauthorized reproduction of this article is prohibited. Original Contribution visual cortex quadrant. Among 11 (85%) of those 13 cases, the lesion extended into the anterior optic radiations (Fig. 5A) or posterior optic radiations (Fig. 5B). When lesions were confined to visual cortex, VFDs overestimated the MRI visual cortex lesions in only 2 cases (15%) (Table 3). DISCUSSION In this study, VFDs confined to 1 hemifield predicted the side of the visual cortex lesion in 74 (96%) of 77 cases. The remaining 3 VFDs failed to predict obvious MRI lesions in the other cerebral hemisphere. Among 15 VFDs present in both hemifields, 5 (33.3%) overestimated the number of MRI lesions, which appeared in only 1 hemisphere. The VFDs in those 5 cases were unequivocal, suggesting that the visual dysfunction was real and that the MRIs simply did not show corresponding lesions. This lack of sensitivity of MRI to ischemic stroke, albeit infrequent, could be clinically meaningful. Among the 30 VFDs that were limited to either the superior or inferior visual field quadrants, only 1 predicted the wrong visual cortex quadrant. Most of the remaining errors were in failing to predict MRI lesions in both superior and inferior visual cortex quadrants on the same side. In a minority of those errors, the missed lesions were obvious. In most cases, the imaging displayed subtle or indistinct signal abnormalities or the lesion straddled the calcarine fissure. HVFs also failed to predict lesions located in anterior visual cortex, which is inaccessible with the 24-2 test protocol that we used. In the few cases where the quadrantic VFDs overestimated the number of MRI lesions, the lesions usually extended into the anterior or posterior optic radiations. We posit that this overestimation occurred because proximal damage to optic radiation axons produced visual dysfunction not reflected in the visual cortex lesions themselves. Thus, the MRI portrayal of visual cortex lesions in proximal PCA stroke would likely underestimate the extent of the VFDs. Our estimate of the accuracy of static perimetry in locating visual cortex lesions is moderated by several considerations. It applies only to PCA ischemic strokes, in which lesion definition can be relatively well defined by diffusion-weighted, precontrast, and postcontrast T1 imaging, as applied in this study. Brain tumors, hemorrhages, trauma, malformations, abscesses, and demyelination are much less amenable to correlations with VFDs because lesion extent cannot be well delineated with those MRI pulse sequences. However, even with high-definition MRI, 366 visualizing the course of the calcarine fissure is difficult because the fissure undulates, often disappearing out of the plane of imaging. As a result, determining whether the lesion lies above or below the fissure is challenging, especially when small lesions appear to straddle the fissure. Finally, locating lesion borders in relation to visual cortex quadrants depends on high-definition sagittal sequences, which were not always available in this study. Reconstructions from other imaging planes were often indistinct. The localizing accuracy of visual fields also depends on reliable patient performance and skilled provider interpretation of the test results. We did exclude patients who generated unreliable and uninterpretable patterns. Even so, visual field test taking can be difficult, especially for a patient who has had a recent stroke. The result may be suboptimal. Interpretation of the test results is equally challenging. Of the 2 authors who interpreted the VFDs, one (J.C.) had no prior experience and was trained by the senior author (J.D.T.). Despite a shared point of view, reaching consensus required several iterations on at least half of the cases. Decisions about whether high-threshold clusters constituted a relevant defect were subjective. We relied on a combination of gray scale and pattern deviation, a decision that could be questioned. Despite these qualifications, we believe that our results attest to the accuracy of HVFs in locating lesions associated with PCA infarcts. STATEMENT OF AUTHORSHIP Conception and design: J. Cho, E. Liao, J. D. Trobe; Acquisition of data: J. Cho, E. Liao, J. D. Trobe; Analysis and interpretation of data: J. Cho, E. Liao, J. D. Trobe. Drafting the manuscript: J. Cho, E. Liao, J. D. Trobe; Revising the manuscript for intellectual content: J. Cho, E. Liao, J. D. Trobe. Final approval of the completed manuscript: J. Cho, E. Liao, J. D. Trobe. REFERENCES 1. Glickstein M, Whitteridge D. Inouye and the mapping of the visual fields on the human cerebral cortex. Trends Neurosci. 1987;10:350–353. 2. Glickstein M. The discovery of the visual cortex. Sci Am. 1988;259:118–127. 3. Inouye T. Die Sehstorungen bei Schussverletzungender kortikalen Sehsphare. Leipzig, Germany: W Engelmann, 1909. 4. Holmes G, Lister WT. Disturbances of vision from cerebral lesions with special reference to the cortical representation of the macula. Brain. 1916;39:34–73. 5. Holmes G. Disturbances of vision by cerebral lesions. Br J Ophthalmol. 1917;2:353–384. 6. Holmes G. The organization of the visual cortex in man. Proc R Soc Lond Ser B (Biol). 1945;132:348–361. Cho et al: J Neuro-Ophthalmol 2022; 42: 360-366 Copyright © North American Neuro-Ophthalmology Society. Unauthorized reproduction of this article is prohibited.