A 2018 single-center retrospective study (Kobayashi, Acta Neurologica Belgica) proposed a purely visual criterion for horizontal gaze deviation (HGD) on non-contrast head CT — both eyes unidirectionally deviated by at least 14∘14^{\circ} — reporting a 4.9% false-positive rate in non-stroke controls and strong associations with cardioaortic-embolic middle cerebral artery and posterior inferior cerebellar artery (PICA) strokes (odds ratios 24.2 [9.6, 66.3] and 15.6 [3.2, 76.2]). The unresolved issues that follow from the study's retrospective, single-center, single-rater design, its stated aim of enhancing the availability of a caliper-free visual criterion, and its postulated vestibular mechanism define one coherent open question: the criterion's diagnostic performance (sensitivity, specificity, positive predictive value) has never been validated prospectively or in an external population; inter-rater reliability of the visual 14∘14^{\circ} determination (the reader's present/absent call at the 14∘14^{\circ} threshold), including use by non-radiologists, is unmeasured; and the vestibular tone-imbalance mechanism proposed for PICA stroke remains postulated rather than directly tested. No later work resolves these issues: the only meta-analysis (Li et al. 2022) pooled exclusively retrospective studies with heterogeneous sign definitions and still treats validation as outstanding; the author's own 2024 follow-up remained retrospective and single-rater; a 2025 review and a 2026 external vertebrobasilar cohort advance mechanism and adjacent phenomena but do not validate the criterion. The problem is to produce, for the three open issues that follow, an evidence-gap map of what the current literature establishes and the validation, reliability, and mechanism-testing protocols that would decide whether the 14∘14^{\circ} visual CT sign is a valid, reproducible, and mechanistically grounded indicator of acute ischemic stroke.

Background

Conjugate eye deviation denotes rotation of both eyes in the same horizontal direction; in acute hemispheric ischemic stroke the eyes typically deviate toward the injured hemisphere, and this gaze deviation is a classical bedside localizing sign. Non-contrast computed tomography (CT) is the first imaging test in most acute stroke pathways because it is fast and widely available, but early ischemic changes on CT are subtle, so any CT-visible ocular sign that predicts stroke — and in particular large vessel occlusion (LVO, an artery-occluding stroke that may be treatable by mechanical thrombectomy) — has triage value. 'Radiographic' or 'radiological' conjugate eye deviation means this eye position read off the CT image itself: on an axial slice through the lenses, both globes are visibly rotated in the same direction.

Nishimura et al. (2015) first quantified the phenomenon on CT in acute cerebellar infarction, using a greater-than-10∘10^{\circ} deviation of each eye. Kobayashi (2018) then proposed a purely visual rule requiring no calipers: radiological horizontal gaze deviation (HGD) is judged present when both eyes are unidirectionally deviated by at least 14∘14^{\circ}, a threshold chosen so that only 4.9% of 193 non-stroke controls exceeded it. In 327 stroke patients this sign marked the affected side and the stroke subtype: it was strongly associated with cardioaortic-embolic middle cerebral artery (MCA) stroke (odds ratio 24.2, 95% CI 9.6–66.3) and, unexpectedly, with posterior inferior cerebellar artery (PICA) stroke (odds ratio 15.6, 95% CI 3.2–76.2 — a wide interval, indicating that this association is imprecisely estimated); negative association with daily onset and independent associations with cardioaortic embolism and lenticular lesions were reported. A parallel literature established that a binary visual judgment of 'eyes deviated, yes or no' is reliable between stroke specialists and a neuroradiologist (Spokoyny et al. 2016, kappa 0.88, where kappa is the chance-corrected inter-rater agreement statistic ranging from 0 to 1), but no reliability data exist for the visual 14∘14^{\circ} determination itself.

The PICA finding raised a mechanistic puzzle: the PICA supplies the cerebellum and the lateral medulla, not the hemispheric gaze centers, and Kobayashi postulated that gaze deviation in PICA stroke arises from imbalanced vestibular tone due to impaired vestibulocerebellar projections. The flocculonodular lobe (the vestibulocerebellum) normally calibrates the medial vestibular nucleus (MVN), a brainstem nucleus that holds horizontal eye position, so loss of its Purkinje cells could leave the ipsilesional MVN overactive and drive the eyes contralesionally, whereas lateral medullary lesions could render the MVN underactive and drive the eyes ipsilesionally. Later work enlarged the PICA experience (Jiang et al. 2021: HGD in 43.5% of 46 angiographically proven acute PICA occlusions, at a greater mean deviation angle than in LVO stroke — 30.0∘30.0^{\circ} versus 22.9∘22.9^{\circ}, p < 0.001 — making PICA deviation a potential LVO mimic on triage CT) and the same author refined the circuit hypothesis while dissociating radiographic from physical gaze deviation (Kobayashi 2024); a review by Badihian et al. (2025) synthesized supratentorial, infratentorial, and peripheral vestibular causes of radiographic gaze deviation and noted the sign is rarely reported in routine radiology reports and cannot by itself distinguish supratentorial from infratentorial lesions. None of this work directly tested vestibular function against CT gaze angle in PICA stroke.

The source study was retrospective, single-center, and single-rater, and the open issues that follow from that design, from the study's stated aim of enhancing the availability of a caliper-free visual criterion, and from its postulated vestibular mechanism are: the diagnostic performance of the 14∘14^{\circ} criterion has not been validated in a prospective cohort or in an external population; inter-rater reliability of the visual 14∘14^{\circ} determination has not been assessed and its utility as a rapid screening tool by non-radiologists has not been evaluated; and the vestibular mechanism was not directly tested. These components jointly serve one objective — establishing whether the 14∘14^{\circ} visual CT sign is a valid, reproducible, and mechanistically grounded indicator of acute ischemic stroke — so the problem is stated at that full generality, and the most recent synthesis (Li et al. 2022, eleven retrospective studies pooled) still describes the sign as expected to be 'extensively used and validated in acute stroke imaging'.

Problem Statement

Determine, for the three linked open issues that follow from the retrospective, single-center, single-rater design of Kobayashi (2018) and its postulated vestibular mechanism, both what the existing literature establishes and what a decisive study must test about whether the visual criterion for horizontal gaze deviation on non-contrast head CT — both eyes unidirectionally deviated by at least 14∘14^{\circ} — is a valid, reproducible, and mechanistically grounded sign of acute ischemic stroke:

(1) Diagnostic performance: prospective and external or multi-center measurement of the criterion's sensitivity, specificity, and positive predictive value in consecutive patients on an acute stroke pathway, with the visual reading blinded to the final diagnosis and independent adjudication of stroke versus mimics, quantified separately for the subtypes where the sign was strongest (cardioaortic-embolic middle cerebral artery stroke, PICA stroke) and for the large-vessel-occlusion setting relevant to thrombectomy triage.

(2) Reproducibility and usability: the inter-rater reliability of the visual 14∘14^{\circ} determination — defined once here as the reader's binary call of whether both eyes are unidirectionally deviated by at least 14∘14^{\circ} (present versus absent), scored by the chance-corrected agreement statistic kappa — among readers of different training, explicitly including non-radiologists, and whether the sign can serve as a rapid screening tool in clinical workflows.

(3) Mechanism: direct testing, rather than postulation, of the vestibular tone-imbalance explanation of gaze deviation in PICA stroke, quantitatively relating vestibular function and lesion location (vestibulocerebellar versus lateral medullary) to the presence, direction, and degree of radiographic gaze deviation.

For each of (1)-(3), the question asks for an evidence-gap map of what the existing literature establishes and where it is silent or insufficient, and a fully specified study protocol, giving design, target measures, acceptance thresholds, and expected precision, that would decide the component. Because the prospective cohort, multi-reader, and vestibular-function data called for in (1)-(3) do not yet exist, the decisive study for each issue is specified as a protocol rather than a primary measurement. The criterion being validated is the published visual 14∘14^{\circ} rule, not a caliper-measured, automated, or binary any-deviation criterion.

References

  1. Kobayashi M. Horizontal gaze deviation on computed tomography: the visual criterion and lesion characteristics in ischemic stroke. Acta Neurologica Belgica 2018;118(4):581-587. DOI: 10.1007/s13760-018-0949-1. https://doi.org/10.1007/s13760-018-0949-1
  2. Li M, Liang W, Yue P, Jiang X, Zhao Z, Zhao B, Xu Z, Mang J. Does radiological conjugate eye deviation sign play a role in acute stroke imaging? A meta-analysis. Journal of Neurology 2022;269(3):1142-1153. DOI: 10.1007/s00415-021-10540-7. https://doi.org/10.1007/s00415-021-10540-7
  3. Jiang NN, Sahlas DJ, Fong C, Wu W, Monteiro S, Larrazabal R. Radiographic horizontal gaze deviation in the setting of acute PICA territory ischemia: A potential mimic of large vessel occlusion. Journal of the Neurological Sciences 2021;420:117226. DOI: 10.1016/j.jns.2020.117226. https://doi.org/10.1016/j.jns.2020.117226
  4. Kobayashi M. Clinical Differences between Radiographic and Physical Horizontal Gaze Deviation: Utility of Using the Two in Patients with Ischemic Stroke. The Neurohospitalist 2024;14(3):264-272. DOI: 10.1177/19418744241245748. https://doi.org/10.1177/19418744241245748
  5. Badihian S, Samaha E, Newman-Toker DE, Zee DS, Kattah JC. Radiographic Horizontal Conjugate Gaze Deviation: Clinical Correlates. Neurology Clinical Practice 2025;15(1):e200375. DOI: 10.1212/CPJ.0000000000200375. https://doi.org/10.1212/CPJ.0000000000200375
  6. Eremina IO, Kulesh AA, Khalo NV, Astanin PA. The role of neuroradiological clues in the diagnosis of vertebrobasilar stroke. Consilium Medicum 2026;28(2):72-75. DOI: 10.26442/20751753.2026.2.203534. https://doi.org/10.26442/20751753.2026.2.203534
  7. Spokoyny I, Chen JY, Raman R, Ernstrom K, Agrawal K, Modir RF, Meyer DM, Meyer BC. Visual Determination of Conjugate Eye Deviation on Computed Tomography Scan Predicts Diagnosis of Stroke Code Patients. Journal of Stroke and Cerebrovascular Diseases 2016;25(12):2809-2813. DOI: 10.1016/j.jstrokecerebrovasdis.2016.07.039. https://doi.org/10.1016/j.jstrokecerebrovasdis.2016.07.039
  8. Nishimura K, Ohara T, Nagatsuka K, Minematsu K, Toyoda K. Radiographic conjugate horizontal eye deviation in patients with acute cerebellar infarction. Journal of the Neurological Sciences 2015;355(1-2):68-71. DOI: 10.1016/j.jns.2015.05.018. https://doi.org/10.1016/j.jns.2015.05.018
  9. Farhat R, Abu Awad A, Abu Shaheen W, Alwily D, Avraham Y, Najjar R, Merchavy S, Massoud S. The 'Vestibular Eye Sign'-'VES': a new radiological sign of vestibular neuronitis can help to determine the affected vestibule and support the diagnosis. Journal of Neurology 2023;270(9):4360-4367. DOI: 10.1007/s00415-023-11771-6. https://doi.org/10.1007/s00415-023-11771-6
  10. Maresky HS, Rootman JM, Klar MM, Levitt M, Kossar AP, Zucker D, Glazier M, Kalmanovich-Avnery S, Aviv R, Ertl-Wagner B, Tal S. Bringing Prevost's sign into the third dimension: Artificial intelligence estimation of conjugate gaze adjusted length (CGAL) and correlation with acute ischemic stroke. Medicine 2020;99(49):e23330. DOI: 10.1097/MD.0000000000023330. https://doi.org/10.1097/MD.0000000000023330