- Email: [email protected]
Validating a threshold of ocular gaze deviation for the prediction of acute ischaemic stroke
Clinical Radiology 69 (2014) 1244e1248
Contents lists available at ScienceDirect
Clinical Radiology journal homepage: www.clinicalradiologyonline.net
Validating a threshold of ocular gaze deviation for the prediction of acute ischaemic stroke D. McKean a, *, M. Kudari a, M. Landells b, D. Grant a, S. Johnson a, pez de Heredia c, S. Yanny c, E.K. Woo c L. Lo a
Oxford University Hospitals National Health Service Trust, Radiology Department, John Radcliffe Hospital, Headley Way, Oxford, UK b New College, University of Oxford, Holywell St, Oxford OX1 3BN, UK c Buckinghamshire Healthcare National Health Service Trust, Radiology Department, Aylesbury, UK
article in formation Article history: Received 22 January 2014 Received in revised form 1 May 2014 Accepted 9 July 2014
AIM: To determine a threshold at which the degree of ocular gaze deviation (OGD) on axial imaging is highly specific for the prediction of acute ischaemic stroke. MATERIALS AND METHODS: A retrospective analysis of 517 patients who had received MRI with diffusion-weighted imaging (DWI) for suspected acute stroke was performed. The degree of OGD was measured in all patients and the presence and location of infarction determined. The difference in OGD between groups was compared using the independent t-test for normally distributed data and the ManneWhitney test for non-normal data. The sensitivity and specificity for degrees of OGD in the prediction of acute infarction was calculated using a receiver operating curve (ROC) analysis. RESULTS: The imaging of 448 patients meeting the inclusion criteria was reviewed. Acute infarct was demonstrated in 34.8% (n¼156). There was a significant difference in the degree of OGD between patients with an acute infarct and those without evidence of acute ischaemia (p<0.001). ROC curve analysis for OGD demonstrated area under the curve (AUC) ¼ 0.619 with increasing degrees of OGD more specific for acute infarct. OGD >11.95 had a sensitivity of 17% and specificity of 95.9% in predicting acute infarction. CONCLUSION: Significant OGD>11.95 has a high specificity for acute infarct. This threshold may provide a helpful additional sign in the detection of subtle acute infarct, particularly on axial CT brain imaging. Ó 2014 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
Introduction Ocular gaze deviation (OGD) was first described by Jean Louis Prevost in 18651 when he observed that patients with
* Guarantor and correspondent: D. McKean, Radiology Department, Oxford University Hospitals NHS Trust, John Radcliffe Hospital, Headley Way, Oxford, UK. Tel.: þ44 (0) 1865 231750; fax: þ44 (0) 1865 231750. E-mail addresses: [email protected], drdavidmckean@gmail. com (D. McKean).
hemiplegia had eyes that were deviated towards the damaged cerebral hemisphere. Subsequent studies have examined the frequency, clinical features,2e5 and lesion locations6e9 associated with OGD in acute stroke, with the incidence of OGD reported in approximately 20% of acute hemispheric infarcts.9,10,13,14 Previous studies looking at the clinical significance of OGD have shown that the observation of eye deviation on axial CT has been shown to improve the detection of acute ischaemic stroke12 and to be associated with severe clinical symptoms, poor outcome, and increased mortality.13
http://dx.doi.org/10.1016/j.crad.2014.07.011 0009-9260/Ó 2014 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
D. McKean et al. / Clinical Radiology 69 (2014) 1244e1248
Supratentorial lesions causing OGD typically reflect dysfunction in cortical areas involved in the control of the frontal eye fields, spatial attention, and eye movements.9,10 These lesions classically cause ipsilateral gaze deviation, and has been more commonly reported in right hemispheric infarcts.9 Infratentorial infarcts involving the cerebellum and pons may also induce OGD but have been less frequently studied.11 Contraversive shift of eye deviation from the lesional side to the contralateral side in hemispheric stroke may indicate emerging mass effect on thalamic mediated circuits and the corticopontine projection within the internal capsule.14 The assessment of OGD may be confounded by incidental variations in gaze direction15 and the threshold at which OGD becomes a significant observation has not yet been defined in patients with suspected acute infarction. The purpose of the present study was, therefore, to determine whether a general threshold of radiologically determined gaze deviation on axial brain imaging for suspected acute stroke could be defined, which may provide a useful additional sign in the prediction of acute infarction.
Materials and methods Case identification Five hundred and seventeen consecutive patients with suspected acute stroke patients presenting at our institution were identified between 1 December 2011 and 31 March 2012 were selected (Fig 1). The present inclusion criteria included1: retrievable stroke protocol MRI brain on admission acquired within 24 h of presentation2; no previously demonstrated intracranial disease3; no nonischaemic intracranial disease identified on the study examination4; axial imaging of the orbits unobscured by movement artefact or patient position. Sixty-nine patients were excluded due to newly identified non-ischaemic disease or image degradation by artefact. Patients were not given any instructions as to where to direct their gaze during the examination and no features of the layout of the MRI suite were identified which would direct a patients gaze either leftwards to rightwards. Both supratentorial and infratentorial lesions were included, as infarct in both these regions may result in OGD.9e11,14
Figure 1 Study cohort.
1245
MRI interpretation MRI studies were obtained on a 1.5 T Symphony MRI machine (Siemens, Erlangen, Germany) or 1.5 T Achieva MRI system (Phillips, Andover, USA), using an eight-channel neurovascular phased-array coil. The standardized stroke protocol included T2 axial turbo gradient and spin-echo and diffusion-weighted axial (DWI) sequences. DWI was performed with a single-shot echo planar imaging (EPI) spinecho sequence: 108 ms echo time, 90 flip angle, 230 230 mm field of view, 192 192 matrix, 19 sections, 5 mm sections thickness, 1.5 mm gap, scan acquired in three directions at B ¼ 0, B ¼ 500, and B ¼ 1000. Axial T2weighted brain MRI sequences were used for the detection of ocular gaze orientation. OGD was then calculated for each globe by drawing three intersecting lines (Fig 2); line A was drawn anteroposteriorly through the midline, line B was drawn 90 perpendicular to line A, and further lines were drawn through the long axis of each lens. OGD, leftwards or rightwards, was recorded for both globes by measuring the angle formed by the intersection of these lines. The average OGD was calculated for each patient (right globe OGD þ left globe OGD)/2).10 Patients were excluded if movement artefact precluded the identification of the midline or long axis of the ocular lenses.
Statistical analysis Statistical analysis was performed using SPSS for Windows, version 20. Normality was assessed visually with histograms and using the KolmogoroveSmirnov and ShapiroeWilk tests. The difference in OGD between groups was also compared using the independent t-test for normally distributed data and the ManneWhitney test for nonnormal data. (These are different statistical tests, both of which were performed.) A p-value of <0.05 was considered statistically significant.
Figure 2 Measurement of OGD. Line A is drawn through the midline. Line B, is drawn perpendicular to line A. Lines are then drawn through the long axes of both lenses to create angles with line B from which the average OGD may be calculated.
1246
D. McKean et al. / Clinical Radiology 69 (2014) 1244e1248
An empirical receiver operating characteristic (ROC) curve analysis was performed to demonstrate the sensitivity and specificity of increasing degrees of OGDin the prediction of acute infarction. Interobserver reliability in measuring OGD was characterized with the k statistic based on a randomized selection of 25 patients. The cohort used to calculate interobserver reliability including a range from 32 rightward to 27 leftwards. K over 0.75 was characterized as excellent, 0.40 to 0.75 as fair to good, and below 0.40 as poor correlation.
Results The present study cohort consisted of 448 patients of whom 48.5% were male and 51.5% female (Table 1). The mean patient age was 70 years (range 15e88 years). Thirtyfour point eight percent (n¼156) of the patients had lesions demonstrated on DWI in keeping with acute infarction. Of these, 92.3% were supratentorial and 7.7% in the posterior circulation. Fifty seven point seven percent of infarcts occurred in the right or left middle cerebral artery (MCA) territories (Table 2). Of the posterior circulation infarcts (n¼12), six were cerebellar (two right hemispheric, one left hemispheric, and three bilateral), four were in the territory of the vertebral artery perforators and two in the basilar artery perforators.
Presence of OGD OGD in both case and control groups was not normally distributed (p<0.001 for KolomogroveSmirnov and ShapiroeWilk tests, respectively; Fig 3). OGD was not normally distributed (p<0.001) and a significant difference in the distribution of OGDwas seen between the non-infarct and acute infarct groups (p<0.001) with the acute infarct group demonstrating increased degrees of OGD compared to the control group (Figs 3 and 4).
Performance of OGD ROC curve analysis of the sensitivity and 1especificity of OGD in the prediction of acute infarct produced area under the curve (AUC) ¼ 0.619 (Fig 5). The diagnostic performance of an OGD >11.95 in predicting acute infarction is shown in Table 2. An illustrative example of significant OGD >11.95 is shown in Fig 2. The interobserver reliability in measuring OGD was excellent with k score of 0.9530 and interclass correlation coefficient of 0.9597. Table 1 Demographics of study cohort. Patient characteristic
Number of patients % (n)
Male Female Mean age, years Infarct on DWI
48.5 (217) 51.5 (231) 70 (range 15e88) 34.8 (156)
N ¼ 448.
Table 2 Vascular territory of acute infarction identified on DWI. Location of infarct
Number of patients % (n)
R- MCA R- LSA R- AChA R- ACA R- PCA L- MCA L- LSA L- AChA L- ACA L- PCA R- Cerebellar infarct L- Cerebellar infarct Bilateral cerebellar Vertebral branches Basilar branches Multi territory infarcts
33.3 1.9 2.6 0.6 6.4 24.4 3.2 3.2 1.3 7.7 1.3 0.6 1.9 2.6 1.3 7.7
(52) (3) (4) (1) (10) (38) (5) (5) (2) (12) (2) (1) (3) (4) (2) (12)
N ¼ 156. R, right; L, left; MCA, middle cerebral artery; LSA, lenticulo-striate arteries; AChA, anterior choroideal artery; ACA, anterior cerebral artery; PCA, posterior cerebral artery.
Discussion The present study is the first to validate a threshold of OGD in the prediction of acute infarct as demonstrated on DWI. In the present cohort, the performance of OGD >11.95 in predicting infarction demonstrates high specificity but relatively low sensitivity (Table 3), similar to that of the hyperdense middle cerebral artery (HMCA) sign or dot sign.16e22 Although OGD is not pathognomic of cerebral hemispheric disease, as patients may incidentally divert their gaze away from the midline, in the context of clinically suspected acute infarction increasing degrees of deviation are highly specific for acute stroke. In the present cohort, 100% positive predictive value (PPV) was seen with OGD of >20 . A lower threshold of >11.95 also demonstrates a high degree of specificity and improved sensitivity. This observation may be of greatest use in the interpretation of brain CT. Early detection of stroke is essential for the initiation of effective treatment and is associated with improved patient outcome.23 Although some institutions routinely perform either MRI or CT perfusion, in many centres the selection of appropriate patients for urgent stroke intervention is still often dependent on the presenting clinical examination and an initial unenhanced head CT examination.24 Unenhanced CT identification of hyperacute stroke is often limited by the absence of measurable objective criteria. Other signs that have been demonstrated to improve the detection of acute stroke, such as the HMCA, insular ribbon sign, or dot sign, vary in conspicuity and are dependent on appropriate window levels.16 Observation of these subtle CT signs of infarction is important as these have diagnostic as well as prognostic value in the acute stroke setting.25,26 Previous studies have demonstrated that observation of eye deviation is useful in identifying cerebral parenchymal
D. McKean et al. / Clinical Radiology 69 (2014) 1244e1248
1247
Figure 3 Histograms demonstrating the distribution of gaze deviation rightwards (negative values) and leftwards (positive values) in (a) noninfarct patients and (b) acute infarct patients.
hypoattenuation indicative of stroke, even in the absence of additional patient clinical data.13 Any objective sign that may improve the sensitivity of CT in the detection of acute infarction warrants further investigation. Although significant OGD will not been seen in all patients with acute infarction, detection of significant OGD should, in a similar manner to the HMCA sign or dot sign,16e22 prompt the reporting radiologist to look for further evidence of subtle acute infarction in the most commonly affected anatomical locations. The lesion pattern of OGD, seen in the present cohort is similar to that demonstrated in previous studies,9 using a clinical definition of conjugate eye deviation, with OGD more commonly seen in right hemispheric than left hemispheric lesions. The supratentorial lesions most often associated with OGD include the ipsilateral basal ganglia, frontal lobes, parietal eye field, and temporoparietal cortical regions involved in spatial attention. The number of posterior circulation infarcts in the present cohort was small (n¼12), but the results confirm the previous observations that infratentorial lesions causing OGD may be seen in the cerebellum and contralateral pons11,27 and that these lesions typically result in
Figure 5 Empirical ROC curve illustrating the sensitivity and 1especificity of gaze deviation in predicting acute infarct.
contralateral gaze deviation, away from the side of the lesion and towards the hemiplegic side. The present study has the following limitations. DWI in acute ischaemia only partially reflects penumbral tissue and underestimates the area of functionally impaired tissue.9 Neither correlation with the findings of clinical
Table 3 Performance of the ocular gaze deviation (>11.95 ) in predicting acute infarction.
Figure 4 Boxplots to show the distribution of gaze deviation rightwards (negative values) and leftwards (positive values) in (a) noninfarct patients and (b) acute infarct patients.
Parameter
Value
Specificity, % Sensitivity, % NPV, % PPV, % Positive LR Negative LR
95.9 17.3 68.2 64.3 3.37 0.87
N ¼ 448. NPV, negative predictive value; PPV, positive predictive value; LR, likelihood ratio.
1248
D. McKean et al. / Clinical Radiology 69 (2014) 1244e1248
examination, specifically with respect to conjugate eye deviation, nor patient follow-up was performed. Previous studies have shown that clinically detected OGD is associated with worse patient prognosis,15 however, the observation of radiologically identified OGD was not correlated with long-term patient outcome. The study reviewed patients with suspect acute infarction; however, OGD is not specific to stroke and may also be seen secondary to tumour, seizures, or traumatic brain injury. In summary, the present study shows that the presence of gaze deviation >11.95 in patients with suspected acute infarction provides a useful threshold for the interpretation of OGD, and should prompt detailed review of the locations commonly associated with OGD for evidence of subtle ischaemic lesions. This radiological equivalent of the clinically detected Prevost’s sign represents a useful addition to the catalogue of signs of acute cerebral ischaemia that may be evident on early CT studies.
References 1. Prevost MJL. Deviation des yeux et de la tete dans quelques cas d’hemiplegie. Gazette Hebdomadaire de Medecine et de Chirurgie 1865;41:649e50. 2. Chung CS, Caplan LR, Yamamoto Y, et al. Striatocapsular haemorrhage. Brain 2000;123:1850e62. 3. Chung CS, Caplan LR, Han W, et al. Thalamic haemorrhage. Brain 1996;119:1873e86. 4. Karnath HO, Himmelbach M, Rorden C. The subcortical anatomy of human spatial neglect: putamen, caudate nucleus and pulvinar. Brain 2002;125:350e60. 5. Bartanusz V, Daniel RT, Villemure JG. Conjugate eye deviation due to traumatic striatalesubthalamic lesion. J Clin Neurosci 2005;12:92e3. 6. De Renzi E, Colombo A, Faglioni P, et al. Conjugate gaze paresis in stroke patients with unilateral damage. An unexpected instance of hemispheric asymmetry. Arch Neurol 1982;39:482e6. 7. Mohr J, Rubinstein LV, Kase C. Gaze palsy in hemispheral stroke: the NINCDS stroke data bank. Neurology 1984;34(Suppl. 1):199. 8. Okinaka S, Toyokura Y, Nakamura H, et al. A contribution to the study of pathogenesis of conjugate deviation of eyes in cerebral apoplexy. Folia Psychiatr Neurol Jpn 1952;6:125e37. 9. Singer OC, Humpich MC, Laufs H, et al. Conjugate eye deviation in acute stroke. Incidence, hemispheric asymmetry and lesion pattern. Stroke 2006;37:2726e32.
10. Becker E, Karnath H-O. Neuroimaging of eye position reveals spatial neglect. Brain 2010;133:909e14. 11. Fruhmann-Berger M, Karnath HO. Spontaneous eye and head position in patients with spatial neglect. J Neurol 2005;252:1194e200. 12. Mahajan V, Minshew PT, Khoury J, et al. Eye position information on CT increases the identification of acute ischemic hypoattenuation. AJNR Am J Neuroradiol 2008;29:1144e6. 13. Tijssen CC, Schulte BP, Leyten AC. Prognostic significance of conjugate eye deviation in stroke patients. Stroke 1991;22:200e2. 14. Wu YJ, Tsai TT, Huang CW. Contraversive shift of conjugate eye deviation in hemispheric stroke indicates emerging mass effect. J Clin Neurosci 2011;18:1276e8. 15. Fruhmann Berger M, Pross RD, Ilg UJ, et al. Deviation of eyes and head in acute cerebral stroke. BMC Neurology 2006;6:23. 16. Leary MC, Kidwell CS, Villablanca JP, et al. Validation of computed tomographic middle cerebral artery “dot” sign: an angiographic correlation study. Stroke 2003;34:2636e40. 17. Kidwell CS, Villablanca JP, Saver JL. Advances in neuroimaging of acute stroke. Curr Atheroscler Rep 2000;2:126e35. 18. von Kummer R, Nolte PN, Schnittger H, et al. Detectability of hemispheric ischemic infarction by computed tomography within 6 hours after stroke. Neuroradiology 1996;38:31e3. 19. Truwit CL, Barkovitch AJ, Gean-Marton A, et al. Loss of the insular ribbon: another early CT sign of acute middle cerebral artery infarction. Radiology 1990;176:801e6. 20. von Kummar R, Meyding-Lamade U, Forsting M, et al. Sensitivity and prognostic value of early computed tomography in middle cerebral artery trunk occlusion. AJNR Am J Neuroradiol 1994;15:9e15. 21. Tomsick TA, Brott TG, Olinger CP, et al. Hyperdense middle cerebral artery: incidence and quantitative significance. Neuroradiology 1989;31:312e5. 22. Kalafut MA, Schriger DL, Saver JL, et al. Detection of early CT signs of 1/3 middle cerebral artery infarctions. Stroke 2000;31:1667e71. 23. Jauch EC, Saver JL, Adams HP, et al. Guidelines for the early management of patients with acute ischemic stroke: a guideline for healthcare professionals from the American Heart Association/American Stroke Association. Stroke 2013;44:870e947. 24. Tissue plasminogen activator for acute ischemic stroke. The National Institute of Neurological Disorders and Stroke rt-PA Stroke Study Group. N Engl J Med 1995;333:1581e7. 25. Scott JN, Buchan AM, Sevick RJ. Correlation of neurologic dysfunction with early CT findings in acute ischemic stroke. Can J Neurol Sci 1999;26:182e9. 26. Kucinski T, Koch C, Grzyska U, et al. The predictive value of early CT and angiography for fatal hemispheric swelling in acute stroke. AJNR Am J Neuroradiol 1998;19:839e46. 27. Simon JE, Morgan SC, Pexman JHW, et al. CT assessment of conjugate eye deviation in acute stroke. Neurology 2003;60(1):135e7.
Contents lists available at ScienceDirect
Clinical Radiology journal homepage: www.clinicalradiologyonline.net
Validating a threshold of ocular gaze deviation for the prediction of acute ischaemic stroke D. McKean a, *, M. Kudari a, M. Landells b, D. Grant a, S. Johnson a, pez de Heredia c, S. Yanny c, E.K. Woo c L. Lo a
Oxford University Hospitals National Health Service Trust, Radiology Department, John Radcliffe Hospital, Headley Way, Oxford, UK b New College, University of Oxford, Holywell St, Oxford OX1 3BN, UK c Buckinghamshire Healthcare National Health Service Trust, Radiology Department, Aylesbury, UK
article in formation Article history: Received 22 January 2014 Received in revised form 1 May 2014 Accepted 9 July 2014
AIM: To determine a threshold at which the degree of ocular gaze deviation (OGD) on axial imaging is highly specific for the prediction of acute ischaemic stroke. MATERIALS AND METHODS: A retrospective analysis of 517 patients who had received MRI with diffusion-weighted imaging (DWI) for suspected acute stroke was performed. The degree of OGD was measured in all patients and the presence and location of infarction determined. The difference in OGD between groups was compared using the independent t-test for normally distributed data and the ManneWhitney test for non-normal data. The sensitivity and specificity for degrees of OGD in the prediction of acute infarction was calculated using a receiver operating curve (ROC) analysis. RESULTS: The imaging of 448 patients meeting the inclusion criteria was reviewed. Acute infarct was demonstrated in 34.8% (n¼156). There was a significant difference in the degree of OGD between patients with an acute infarct and those without evidence of acute ischaemia (p<0.001). ROC curve analysis for OGD demonstrated area under the curve (AUC) ¼ 0.619 with increasing degrees of OGD more specific for acute infarct. OGD >11.95 had a sensitivity of 17% and specificity of 95.9% in predicting acute infarction. CONCLUSION: Significant OGD>11.95 has a high specificity for acute infarct. This threshold may provide a helpful additional sign in the detection of subtle acute infarct, particularly on axial CT brain imaging. Ó 2014 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
Introduction Ocular gaze deviation (OGD) was first described by Jean Louis Prevost in 18651 when he observed that patients with
* Guarantor and correspondent: D. McKean, Radiology Department, Oxford University Hospitals NHS Trust, John Radcliffe Hospital, Headley Way, Oxford, UK. Tel.: þ44 (0) 1865 231750; fax: þ44 (0) 1865 231750. E-mail addresses: [email protected], drdavidmckean@gmail. com (D. McKean).
hemiplegia had eyes that were deviated towards the damaged cerebral hemisphere. Subsequent studies have examined the frequency, clinical features,2e5 and lesion locations6e9 associated with OGD in acute stroke, with the incidence of OGD reported in approximately 20% of acute hemispheric infarcts.9,10,13,14 Previous studies looking at the clinical significance of OGD have shown that the observation of eye deviation on axial CT has been shown to improve the detection of acute ischaemic stroke12 and to be associated with severe clinical symptoms, poor outcome, and increased mortality.13
http://dx.doi.org/10.1016/j.crad.2014.07.011 0009-9260/Ó 2014 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.
D. McKean et al. / Clinical Radiology 69 (2014) 1244e1248
Supratentorial lesions causing OGD typically reflect dysfunction in cortical areas involved in the control of the frontal eye fields, spatial attention, and eye movements.9,10 These lesions classically cause ipsilateral gaze deviation, and has been more commonly reported in right hemispheric infarcts.9 Infratentorial infarcts involving the cerebellum and pons may also induce OGD but have been less frequently studied.11 Contraversive shift of eye deviation from the lesional side to the contralateral side in hemispheric stroke may indicate emerging mass effect on thalamic mediated circuits and the corticopontine projection within the internal capsule.14 The assessment of OGD may be confounded by incidental variations in gaze direction15 and the threshold at which OGD becomes a significant observation has not yet been defined in patients with suspected acute infarction. The purpose of the present study was, therefore, to determine whether a general threshold of radiologically determined gaze deviation on axial brain imaging for suspected acute stroke could be defined, which may provide a useful additional sign in the prediction of acute infarction.
Materials and methods Case identification Five hundred and seventeen consecutive patients with suspected acute stroke patients presenting at our institution were identified between 1 December 2011 and 31 March 2012 were selected (Fig 1). The present inclusion criteria included1: retrievable stroke protocol MRI brain on admission acquired within 24 h of presentation2; no previously demonstrated intracranial disease3; no nonischaemic intracranial disease identified on the study examination4; axial imaging of the orbits unobscured by movement artefact or patient position. Sixty-nine patients were excluded due to newly identified non-ischaemic disease or image degradation by artefact. Patients were not given any instructions as to where to direct their gaze during the examination and no features of the layout of the MRI suite were identified which would direct a patients gaze either leftwards to rightwards. Both supratentorial and infratentorial lesions were included, as infarct in both these regions may result in OGD.9e11,14
Figure 1 Study cohort.
1245
MRI interpretation MRI studies were obtained on a 1.5 T Symphony MRI machine (Siemens, Erlangen, Germany) or 1.5 T Achieva MRI system (Phillips, Andover, USA), using an eight-channel neurovascular phased-array coil. The standardized stroke protocol included T2 axial turbo gradient and spin-echo and diffusion-weighted axial (DWI) sequences. DWI was performed with a single-shot echo planar imaging (EPI) spinecho sequence: 108 ms echo time, 90 flip angle, 230 230 mm field of view, 192 192 matrix, 19 sections, 5 mm sections thickness, 1.5 mm gap, scan acquired in three directions at B ¼ 0, B ¼ 500, and B ¼ 1000. Axial T2weighted brain MRI sequences were used for the detection of ocular gaze orientation. OGD was then calculated for each globe by drawing three intersecting lines (Fig 2); line A was drawn anteroposteriorly through the midline, line B was drawn 90 perpendicular to line A, and further lines were drawn through the long axis of each lens. OGD, leftwards or rightwards, was recorded for both globes by measuring the angle formed by the intersection of these lines. The average OGD was calculated for each patient (right globe OGD þ left globe OGD)/2).10 Patients were excluded if movement artefact precluded the identification of the midline or long axis of the ocular lenses.
Statistical analysis Statistical analysis was performed using SPSS for Windows, version 20. Normality was assessed visually with histograms and using the KolmogoroveSmirnov and ShapiroeWilk tests. The difference in OGD between groups was also compared using the independent t-test for normally distributed data and the ManneWhitney test for nonnormal data. (These are different statistical tests, both of which were performed.) A p-value of <0.05 was considered statistically significant.
Figure 2 Measurement of OGD. Line A is drawn through the midline. Line B, is drawn perpendicular to line A. Lines are then drawn through the long axes of both lenses to create angles with line B from which the average OGD may be calculated.
1246
D. McKean et al. / Clinical Radiology 69 (2014) 1244e1248
An empirical receiver operating characteristic (ROC) curve analysis was performed to demonstrate the sensitivity and specificity of increasing degrees of OGDin the prediction of acute infarction. Interobserver reliability in measuring OGD was characterized with the k statistic based on a randomized selection of 25 patients. The cohort used to calculate interobserver reliability including a range from 32 rightward to 27 leftwards. K over 0.75 was characterized as excellent, 0.40 to 0.75 as fair to good, and below 0.40 as poor correlation.
Results The present study cohort consisted of 448 patients of whom 48.5% were male and 51.5% female (Table 1). The mean patient age was 70 years (range 15e88 years). Thirtyfour point eight percent (n¼156) of the patients had lesions demonstrated on DWI in keeping with acute infarction. Of these, 92.3% were supratentorial and 7.7% in the posterior circulation. Fifty seven point seven percent of infarcts occurred in the right or left middle cerebral artery (MCA) territories (Table 2). Of the posterior circulation infarcts (n¼12), six were cerebellar (two right hemispheric, one left hemispheric, and three bilateral), four were in the territory of the vertebral artery perforators and two in the basilar artery perforators.
Presence of OGD OGD in both case and control groups was not normally distributed (p<0.001 for KolomogroveSmirnov and ShapiroeWilk tests, respectively; Fig 3). OGD was not normally distributed (p<0.001) and a significant difference in the distribution of OGDwas seen between the non-infarct and acute infarct groups (p<0.001) with the acute infarct group demonstrating increased degrees of OGD compared to the control group (Figs 3 and 4).
Performance of OGD ROC curve analysis of the sensitivity and 1especificity of OGD in the prediction of acute infarct produced area under the curve (AUC) ¼ 0.619 (Fig 5). The diagnostic performance of an OGD >11.95 in predicting acute infarction is shown in Table 2. An illustrative example of significant OGD >11.95 is shown in Fig 2. The interobserver reliability in measuring OGD was excellent with k score of 0.9530 and interclass correlation coefficient of 0.9597. Table 1 Demographics of study cohort. Patient characteristic
Number of patients % (n)
Male Female Mean age, years Infarct on DWI
48.5 (217) 51.5 (231) 70 (range 15e88) 34.8 (156)
N ¼ 448.
Table 2 Vascular territory of acute infarction identified on DWI. Location of infarct
Number of patients % (n)
R- MCA R- LSA R- AChA R- ACA R- PCA L- MCA L- LSA L- AChA L- ACA L- PCA R- Cerebellar infarct L- Cerebellar infarct Bilateral cerebellar Vertebral branches Basilar branches Multi territory infarcts
33.3 1.9 2.6 0.6 6.4 24.4 3.2 3.2 1.3 7.7 1.3 0.6 1.9 2.6 1.3 7.7
(52) (3) (4) (1) (10) (38) (5) (5) (2) (12) (2) (1) (3) (4) (2) (12)
N ¼ 156. R, right; L, left; MCA, middle cerebral artery; LSA, lenticulo-striate arteries; AChA, anterior choroideal artery; ACA, anterior cerebral artery; PCA, posterior cerebral artery.
Discussion The present study is the first to validate a threshold of OGD in the prediction of acute infarct as demonstrated on DWI. In the present cohort, the performance of OGD >11.95 in predicting infarction demonstrates high specificity but relatively low sensitivity (Table 3), similar to that of the hyperdense middle cerebral artery (HMCA) sign or dot sign.16e22 Although OGD is not pathognomic of cerebral hemispheric disease, as patients may incidentally divert their gaze away from the midline, in the context of clinically suspected acute infarction increasing degrees of deviation are highly specific for acute stroke. In the present cohort, 100% positive predictive value (PPV) was seen with OGD of >20 . A lower threshold of >11.95 also demonstrates a high degree of specificity and improved sensitivity. This observation may be of greatest use in the interpretation of brain CT. Early detection of stroke is essential for the initiation of effective treatment and is associated with improved patient outcome.23 Although some institutions routinely perform either MRI or CT perfusion, in many centres the selection of appropriate patients for urgent stroke intervention is still often dependent on the presenting clinical examination and an initial unenhanced head CT examination.24 Unenhanced CT identification of hyperacute stroke is often limited by the absence of measurable objective criteria. Other signs that have been demonstrated to improve the detection of acute stroke, such as the HMCA, insular ribbon sign, or dot sign, vary in conspicuity and are dependent on appropriate window levels.16 Observation of these subtle CT signs of infarction is important as these have diagnostic as well as prognostic value in the acute stroke setting.25,26 Previous studies have demonstrated that observation of eye deviation is useful in identifying cerebral parenchymal
D. McKean et al. / Clinical Radiology 69 (2014) 1244e1248
1247
Figure 3 Histograms demonstrating the distribution of gaze deviation rightwards (negative values) and leftwards (positive values) in (a) noninfarct patients and (b) acute infarct patients.
hypoattenuation indicative of stroke, even in the absence of additional patient clinical data.13 Any objective sign that may improve the sensitivity of CT in the detection of acute infarction warrants further investigation. Although significant OGD will not been seen in all patients with acute infarction, detection of significant OGD should, in a similar manner to the HMCA sign or dot sign,16e22 prompt the reporting radiologist to look for further evidence of subtle acute infarction in the most commonly affected anatomical locations. The lesion pattern of OGD, seen in the present cohort is similar to that demonstrated in previous studies,9 using a clinical definition of conjugate eye deviation, with OGD more commonly seen in right hemispheric than left hemispheric lesions. The supratentorial lesions most often associated with OGD include the ipsilateral basal ganglia, frontal lobes, parietal eye field, and temporoparietal cortical regions involved in spatial attention. The number of posterior circulation infarcts in the present cohort was small (n¼12), but the results confirm the previous observations that infratentorial lesions causing OGD may be seen in the cerebellum and contralateral pons11,27 and that these lesions typically result in
Figure 5 Empirical ROC curve illustrating the sensitivity and 1especificity of gaze deviation in predicting acute infarct.
contralateral gaze deviation, away from the side of the lesion and towards the hemiplegic side. The present study has the following limitations. DWI in acute ischaemia only partially reflects penumbral tissue and underestimates the area of functionally impaired tissue.9 Neither correlation with the findings of clinical
Table 3 Performance of the ocular gaze deviation (>11.95 ) in predicting acute infarction.
Figure 4 Boxplots to show the distribution of gaze deviation rightwards (negative values) and leftwards (positive values) in (a) noninfarct patients and (b) acute infarct patients.
Parameter
Value
Specificity, % Sensitivity, % NPV, % PPV, % Positive LR Negative LR
95.9 17.3 68.2 64.3 3.37 0.87
N ¼ 448. NPV, negative predictive value; PPV, positive predictive value; LR, likelihood ratio.
1248
D. McKean et al. / Clinical Radiology 69 (2014) 1244e1248
examination, specifically with respect to conjugate eye deviation, nor patient follow-up was performed. Previous studies have shown that clinically detected OGD is associated with worse patient prognosis,15 however, the observation of radiologically identified OGD was not correlated with long-term patient outcome. The study reviewed patients with suspect acute infarction; however, OGD is not specific to stroke and may also be seen secondary to tumour, seizures, or traumatic brain injury. In summary, the present study shows that the presence of gaze deviation >11.95 in patients with suspected acute infarction provides a useful threshold for the interpretation of OGD, and should prompt detailed review of the locations commonly associated with OGD for evidence of subtle ischaemic lesions. This radiological equivalent of the clinically detected Prevost’s sign represents a useful addition to the catalogue of signs of acute cerebral ischaemia that may be evident on early CT studies.
References 1. Prevost MJL. Deviation des yeux et de la tete dans quelques cas d’hemiplegie. Gazette Hebdomadaire de Medecine et de Chirurgie 1865;41:649e50. 2. Chung CS, Caplan LR, Yamamoto Y, et al. Striatocapsular haemorrhage. Brain 2000;123:1850e62. 3. Chung CS, Caplan LR, Han W, et al. Thalamic haemorrhage. Brain 1996;119:1873e86. 4. Karnath HO, Himmelbach M, Rorden C. The subcortical anatomy of human spatial neglect: putamen, caudate nucleus and pulvinar. Brain 2002;125:350e60. 5. Bartanusz V, Daniel RT, Villemure JG. Conjugate eye deviation due to traumatic striatalesubthalamic lesion. J Clin Neurosci 2005;12:92e3. 6. De Renzi E, Colombo A, Faglioni P, et al. Conjugate gaze paresis in stroke patients with unilateral damage. An unexpected instance of hemispheric asymmetry. Arch Neurol 1982;39:482e6. 7. Mohr J, Rubinstein LV, Kase C. Gaze palsy in hemispheral stroke: the NINCDS stroke data bank. Neurology 1984;34(Suppl. 1):199. 8. Okinaka S, Toyokura Y, Nakamura H, et al. A contribution to the study of pathogenesis of conjugate deviation of eyes in cerebral apoplexy. Folia Psychiatr Neurol Jpn 1952;6:125e37. 9. Singer OC, Humpich MC, Laufs H, et al. Conjugate eye deviation in acute stroke. Incidence, hemispheric asymmetry and lesion pattern. Stroke 2006;37:2726e32.
10. Becker E, Karnath H-O. Neuroimaging of eye position reveals spatial neglect. Brain 2010;133:909e14. 11. Fruhmann-Berger M, Karnath HO. Spontaneous eye and head position in patients with spatial neglect. J Neurol 2005;252:1194e200. 12. Mahajan V, Minshew PT, Khoury J, et al. Eye position information on CT increases the identification of acute ischemic hypoattenuation. AJNR Am J Neuroradiol 2008;29:1144e6. 13. Tijssen CC, Schulte BP, Leyten AC. Prognostic significance of conjugate eye deviation in stroke patients. Stroke 1991;22:200e2. 14. Wu YJ, Tsai TT, Huang CW. Contraversive shift of conjugate eye deviation in hemispheric stroke indicates emerging mass effect. J Clin Neurosci 2011;18:1276e8. 15. Fruhmann Berger M, Pross RD, Ilg UJ, et al. Deviation of eyes and head in acute cerebral stroke. BMC Neurology 2006;6:23. 16. Leary MC, Kidwell CS, Villablanca JP, et al. Validation of computed tomographic middle cerebral artery “dot” sign: an angiographic correlation study. Stroke 2003;34:2636e40. 17. Kidwell CS, Villablanca JP, Saver JL. Advances in neuroimaging of acute stroke. Curr Atheroscler Rep 2000;2:126e35. 18. von Kummer R, Nolte PN, Schnittger H, et al. Detectability of hemispheric ischemic infarction by computed tomography within 6 hours after stroke. Neuroradiology 1996;38:31e3. 19. Truwit CL, Barkovitch AJ, Gean-Marton A, et al. Loss of the insular ribbon: another early CT sign of acute middle cerebral artery infarction. Radiology 1990;176:801e6. 20. von Kummar R, Meyding-Lamade U, Forsting M, et al. Sensitivity and prognostic value of early computed tomography in middle cerebral artery trunk occlusion. AJNR Am J Neuroradiol 1994;15:9e15. 21. Tomsick TA, Brott TG, Olinger CP, et al. Hyperdense middle cerebral artery: incidence and quantitative significance. Neuroradiology 1989;31:312e5. 22. Kalafut MA, Schriger DL, Saver JL, et al. Detection of early CT signs of 1/3 middle cerebral artery infarctions. Stroke 2000;31:1667e71. 23. Jauch EC, Saver JL, Adams HP, et al. Guidelines for the early management of patients with acute ischemic stroke: a guideline for healthcare professionals from the American Heart Association/American Stroke Association. Stroke 2013;44:870e947. 24. Tissue plasminogen activator for acute ischemic stroke. The National Institute of Neurological Disorders and Stroke rt-PA Stroke Study Group. N Engl J Med 1995;333:1581e7. 25. Scott JN, Buchan AM, Sevick RJ. Correlation of neurologic dysfunction with early CT findings in acute ischemic stroke. Can J Neurol Sci 1999;26:182e9. 26. Kucinski T, Koch C, Grzyska U, et al. The predictive value of early CT and angiography for fatal hemispheric swelling in acute stroke. AJNR Am J Neuroradiol 1998;19:839e46. 27. Simon JE, Morgan SC, Pexman JHW, et al. CT assessment of conjugate eye deviation in acute stroke. Neurology 2003;60(1):135e7.