846 Medical Informatics in a United and Healthy Europe K.-P. Adlassnig et al. (Eds.) IOS Press, 29 29 European Federation for Medical Informatics. All rights reserved. doi:1.3233/978-1-675-44-5-846 On Signal Analysis of Three-Dimensional Nystagmus Martti JUHOLA a,1, Heikki AALTO b, Timo HIRVONEN b a Department of Computer Sciences, University of Tampere, Finland b Department of Otorhinolaryngology, Helsinki University Central Hospital, Finland Abstract. Nystagmus needs to be stimulated for healthy subjects, but in patients it can also be spontaneous. By recording spontaneous nystagmus it is possible to reveal underlying disorders of the semicircular canals of the inner ear. We developed a signal analysis technique for this purpose and tested it with 28 otoneurological patients who had disorders in their semicircular canals. Our test results support the postulate how nystagmus should appear under these conditions and agree with results presented in the otoneurological literature. Keywords. eye movements, nystagmus, otoneurology, signal analysis, 3D signals 1. Introduction Signal analysis is used to explore eye movements related to otoneurological disorders. Nystagmus [1] (Figures 1 and 2) is an involuntary eye movement stimulated, e.g., by injecting cool or warm air or water to a subject s ear canal. For otoneurological patients nystagmus may spontaneously be induced by a disorder of a semicircular canal located in the inner ear. There is a great demand for technologies of eye movement studies. Video-oculography with small videocameras directed at each eye has become a promising way to measure eye movements. It has extended to three-dimensional (3D) measurements, where torsional eye movements are recorded in addition to horizontal and vertical movements. Compared to older electro-oculography it is fairly noiseless. Its weakness is still a low sampling frequency. We employed slow phases of nystagmus (Figures 1 and 2), which are sufficiently slow for video-oculography. We also studied a hypothesis called Ewald s first law [2] originally from 1892: the trajectory of nystagmus induced by a semicircular canal ought to reflect the anatomic orientation of that semicircular canal [2]. Nystagmus induced by a horizontal semicircular canal is mostly horizontal. Nystagmus from anterior or posterior semicircular canals is assumed to be mixed torsional and vertical, since these canals are oriented diagonally. Direct evidence of Ewald s first law is still restricted to selected patients [2]. We explored spontaneous nystagmus of otoneurological, dizzy patients with signal analysis. 1 Corresponding Author: Martti Juhola, Department of Computer Sciences, 3314 University of Tampere, Finland; E-mail: Martti.Juhola@cs.uta.fi.
M. Juhola et al. / On Signal Analysis of Three-Dimensional Nystagmus 847 2. Signal Data We used nystagmus signals recorded from 28 patients (15 females and 13 males) with a mean age of 49±16 years. The majority of patients suffered from acute, unilateral loss of vestibular function: 17 had vestibular neuritis, 1 had vestibular loss due to intratympanic gentamicin treatment to control Menière s disease, and 6 had labyrinthitis (vestibular loss accompanied with hearing loss). One patient had horizontal semicircular canal fistula due to unoperated cholesteatoma, two patients had dizziness originating from central nervous system disorder, and one patient had bilateral asymmetric loss of vestibular function after chronic mastoiditis. Figure 1 presents 3D nystagmus of a subject in the 1D form. Horizontal eye movements are from left to right or vice versa, vertical eye movements up down or vice versa and torsional movements are clockwise or counterclockwise round the axis in the forward direction. A subject was sitting relaxed and alert, and the head was still. left eye -1-2 -3 5 1 15 4 right eye 2-2 -4 5 1 15 samples Figure 1. Spontaneous nystagmus: the lowest curve of each subplot is the horizontal signal, the middle curve of the left eye and fairly smooth curve of the right eye are the vertical signals, and the upper curve of the left eye and the middle curve (with spikes) of the right eye are the torsional signals. A nystagmus beat includes a slow phase (down in the figure) followed by the fast phase in the opposite direction (Figure 2). fast phase time [s] slow phase (a) Figure 2. (a) A slow phase and fast phase of nystagmus. (b) A measurement eyes covered Before a 3 s recording (Figure 1), it was calibrated by asking a subject to alternately look at nine points located symmetrically on the wall. Recordings were made with covered eyes (Figure 2) by using bilateral video-oculography with cameras (b)
848 M. Juhola et al. / On Signal Analysis of Three-Dimensional Nystagmus embedded in a mask (3D VOG Video-Oculography, Sensomotoric Instruments, Berlin, Germany) in a dim room. The darkness achieved with the eyes covered test condition was preferred, since spontaneous nystagmus is fastest without visual fixation. The sampling frequency was 5 Hz. Horizontal and vertical amplitude resolutions were.5 and the torsional resolution.1. After the calibration, we estimated the resolution to be less than 1. 3. Methods When video-oculography signals were of high quality with infrequent noise, no lowpass filtering was applied. Our Matlab program consists of the following stages: (1) The recording device outputs values between. and 1. depending on the quality of each torsional signal (Figure 3), which was first divided into consecutive segments of 2 s. By choosing the better average quality value of such a segment from the quality signals it was decided from which eye signal triple the next segment was taken. In this way, the whole 3 s was considered. The 2 s length was selected because it included at least one nystagmus beat even with the slowest beats. (2) Angular velocity of three-dimensional signals are computed, signal by signal, using slope values given by linear regression and calculated for every sequential sample (from the signals of the better eye according to step (1)). (3) The direction of nystagmus is defined according to the fast phases. We computed it statistically counting whether the majority of velocity values were either positive or negative. Positive value is an eye movement to the right, up or clockwise and negative to the left, down or counterclockwise depending on their rotational axes. (4) Nystagmus beats are recognized by searching for minima and maxima from the horizontal eye movements, which often have larger amplitudes compared to two others. A minimum or maximum is recognized when the magnitudes of velocity values are less than 1 /s. The slow and fast phases are distinguished, but the former are used. The sampling frequency of 5 Hz (2 ms interval) is too low to detect fast phases accurately enough, since fast phases usually last 7 8 ms [3]. (5) After the recognition of slow phases, their average angular velocities [ /s] are computed with linear regression. (6) Because of noise there may be outliers among slow phase velocities. Thus, 5% of extreme velocities, the smallest and greatest absolute velocity values are pruned. A torsional amplitude greater than 15 obviously as a spike (Figure 1) also produces the rejection of a beat. Altogether, maximally approximately 3% of beats are left out. (7) The velocity values of the accepted nystagmus beats are used as three components of velocity vectors. These are divided by their lengths to form unit velocity vectors. (8) A unit sphere is drawn to model the head, whose gaze direction is that of a subject (not a spectator) (Figure 4). (9) The normal vectors of the planes of the semicircular canals are set on the sphere. Their locations are based on [4]. The unit velocity vectors are also drawn as points on the surface of the unit sphere. Their mean vector is computed and drawn. (1) The angles between the mean vector and the six normal vectors are computed. (11) The least angle is used to predict the disorder of the nearest semicircular canal.
M. Juhola et al. / On Signal Analysis of Three-Dimensional Nystagmus 849 4. Results Table 1 presents average numbers of nystagmus beats detected from the signals of the patients and average velocities computed from them in the three-dimensional coordinate system. Table 2 includes average angles between the mean unit vectors of the nystagmus beats accepted and their angles related to the normal unit vectors, which denote the planes of the semicircular canals. The nystagmus beats were mostly in the horizontal direction, but vertical and torsional nystagmus also existed. This shows the advantage of three-dimensional measurements to acquire as precise measurement as possible from nystagmus phenomenon. Table 2 shed light on the actual modelling..8 left eye.6.4.2 5 1 15.8 right eye.6.4.2 5 1 15 samples Figure 3. The torsional quality signals of Figure 1. Since the quality signal of the left eye was higher for the most of time, the eye movement signals of the left eye were used. Compare to Figure 1, where the torsional signal of the lower subplot contains noisy spikes. Table 1. Means and standard deviations of the slow phases of the nystagmus beats recognized when three velocity values are computed from the accepted beats Number of nystagmus beats before pruning Number of nystagmus beats after pruning Horizontal velocity [ /s] Vertical velocity [ /s] Torsional velocity [ /s] 45.9±17.3 33.±13.4 7.6±5.1 2.4±2.3 3.4±2.7 Table 2. Means and standard deviations of the angles between the mean vectors and the nearest normal vectors representing semicircular canals Normal anterior horizontal posterior anterior horizontal posterior Number 1 18 Angle [ ] - 19.6±8.8 - - 16.5±1.2 -
85 M. Juhola et al. / On Signal Analysis of Three-Dimensional Nystagmus 5. Discussion and Conclusion Horizontal nystagmus component is often prominent in clinical examination of dizzy patients with peripheral cause, and our current 3D analysis results confirm this finding. The mean unit vectors clustered closest to those of horizontal canals, and dominant side was that with stronger input. In unilateral weakness, the normal side drives the nystagmus, but in unilateral stimulation, stimulated side determines the vector orientation, as in one patient with horizontal semicircular canal fistula due to unoperated cholesteatoma. He had the horizontal canal vector of the same side at distance of 4 from the mean vector. Clear unilateral, peripheral weakness was clinically noted in 25 of our patients, and the mean unit vector was closest to the horizontal canal of opposite side. In two patients the underlying pathology was undefined, but the mean unit vector was still closest to that of horizontal canal. Our results confirm in the largest sample published that in unilateral peripheral vestibular loss horizontal canal function loss is prominent and the other canals are affected in lesser degree. Selective defects of one canal only were rare, but they also existed and confirmed the importance of Ewald s first law. RH Z axis: horizontal SPV: right<-->left 1 LA.5 -.5-1 -.5.5 Y axis: vertical SPV: up<-->down LP LH RA.5 RP -.5 X axis: torsional SPV: CW<-->CCW Figure 4. The unit sphere model for the spontaneous nystagmus of a patient: positive X axis is parallel with the nose, positive Y axis comes out from the left ear and positive Z axis from the top of the head. Semicircular canals are represented by their normals. For the mean vector of the spots of the velocity vectors, the angles were computed with the normals. RH is the nearest with 4 angle (L=left and R=right inner ear; A=anterior, P=posterior and H=horizontal semicircular canal; SPV=slow phase velocity [ /s]). References [1] Furman, J., Whitney, S.L. (Eds.) (25) Otoneurology. Neurologic Clinics 23(3). [2] Cremer, P.D., Migliaccio, A.A., Pohl, D.V., Curthoys, I.S., Davies, L., Yavor, R.A., Halmagyi, G.M. (2) Posterior semicircular canal nystagmus is conjugate and its axis is parallel to that of the canal. Neurology 54:216 22. [3] Juhola, M., Aalto, H., Hirvonen, T. (27) Using results of eye movement signal analysis in the neural network recognition of otoneurological patients. Computer Methods and Programs in Biomedicine 86:216 226. [4] Della Santina, C.C., Potyagaylo, V., Migliaccio, A.A., Minor, L.B., Carey, J.P. (25) Orientation of human semicircular canals measured by three-dimensional multiplanar CT reconstruction. Journal of the Association for Research in Otolaryngology 6:191 26.