The schematic diagram of the DOCT system is shown in Fig. 1. The detector of the DOCT system was a 12-bit CMOS line scan camera (Sprint spL2048-140 km; Basler, Ahrensburg, Germany) with an effective scan rate of 70,000 lines/sec at 2048-pixel mode. The transmission-type diffraction grating (spatial frequency, 1,800 lines/mm; nominal angle of incidence and angle of diffraction, 46.05°; Wasatch Photonics, Durham, NC, USA) was adapted to enhance light efficiency in the detection path. A fiberbased interferometer was combined with a superluminescent diode (SLD; λ0=850 nm, Δλ=55 nm; Exalos, Schlieren, Switzerland) as the light source. We used SLDs for the optical source used in the OCT system. The source power is 14 mW and the final sample arm power is 7 mW which is exposed to the skin. This is by far less than the maximum permissible exposure of 500 mW, as stated in American National Standards Institute [15]. The light source was split into sample and reference paths, with the latter terminated by a stationary mirror. The sample path delivered light to a sample and collected light backscattered from different depths in the sample. B-mode scanning was performed using a galvanometer scanning mirror (GVS002; Thorlabs, Newton, NJ, USA) at the back focal plane of the objective lens in the sample path. In this paper, a GPU accelerated mode using a graphics card (GeForce GTX 480, 700-MHz clock rate, 480 CUDA processors; NVIDIA, Santa Clara, CA, USA) was used for real-time DOCT signal processing [14,16,17]. The axial and lateral resolutions of the system were approximately 4 μm and 12 μm, respectively, and the measured depth range was 2 mm. A pure tone was generated by a loudspeaker with the sound ranging in frequency from 20 Hz to 20 kHz at an output of 15 W.

The Doppler frequency shift generated by vibration of the sample was detected by a phase signal and the effect was calculated for each corresponding pixel of the M-mode image. We computed the phase difference between adjacent A-lines, which is proportional to the modulation of the difference in optical path induced by the acoustic stimulation. The phase information was extracted from real and imaginary values. The Doppler phase shift f_d induced by a moving sample is expressed as

where f_a is the camera line period between successive A-lines and tan^–1{(Q_m,nI_m,n+1–I_m,nQ_m,n+1)/Q_m,nQ_m,n+1+I_m,nI_m,n+1)} is the phase difference between successive A-lines [18]. M and N are the horizontal and vertical sizes of the averaging mask, respectively. In addition, I (in-phase) and Q (quadrature phase) are the result elements of the fast Fourier transform operation. The Doppler phase shift is continuously observed while the TM vibrates. The velocity v of the vibrating TM is expressed as

Where λ₀ is the center wavelength, n_t is the refractive index, and θ is the angle between the TM and the direction of the sample arm beam.

The Doppler signal was measured using the Kasai autocorrelation algorithm [19,20]. This method is efficient at calculating the phase change when there is noise and its fast processing makes real-time processing possible. The phase difference was marked in quadrants from –π to π by calculating the arctangent, and the extent of the phase difference was expressed as a color variation. Restoration of the phase signal was possible by segregating the red (R)/green (G)/blue (B) information in the region of interest of the image. Thus, the vibration velocity was calculated using the phase difference. Red, green, and blue are represented by 8 bit color, respectively. In the first quadrant, the color between B and Black is mapped. In the second quadrant, R changes from fixed B to map between B and magenta (M), color between R and M on the third quadrant, and color between R and black on the fourth quadrant. When the direction of Doppler effects is inversed, the color also becomes inverse color. In the post-processing, 3D-demodulation for pattern analysis was performed. Horizontal and vertical resolution information was obtained from the region of interest of the phase shift image, and R/G/B information was separated from each pixel. And the phase shift was calculated by the inverse color mapping.

To test the sensitivity of the DOCT system, a latex drum was constructed by stretching a 100-μm-thick latex membrane over and attaching it to a 50-mm-diameter glass tube. Sound stimuli comprising sinusoidal signals of 1, 2, and 4 kHz from a functional generator were broadcast through the loudspeaker, causing the latex drum to vibrate. To minimize the effect of reflected sound, all procedures were performed in a 250×200×120-mm sound-proof chamber. A calibrated probe microphone was placed near the latex drum to record the sound pressure level at the surface of the latex drum. The OCT beam scanned the entire drum and detected the Doppler frequency shift caused by vibration.

Sprague Dawley rats (5 weeks old, 60 g) were anesthetized by intramuscular injection of a mixture of tiletamine-zolazepam (1.8 mg/100 g) and xylazine hydrochlorate (0.7 mg/100 g). After local injection of 2% lidocaine HCl and epinephrine (1:100,000), a postauricular skin incision was made, the auricle and cartilaginous external auditory canal were removed, and the bony external auditory canal was widened by drilling to expose the entire TM.

During in vivo experimental measurements of the TM and ossicles under normal conditions, the phase or displacement of the umbo and contact area between the TM and the ossicles was acquired using DOCT and 1-, 2-, and 4-kHz pure tone stimuli at 15 W in the sound-proof chamber. To simulate middle ear effusion (MEE), a pick was used to make a pinhole in the bulla, viscous liquid was injected through the pinhole, and the hole was sealed with bone wax. The same experiment was performed to evaluate the in vivo vibration of the TM and ossicles in the simulated effusion condition.

Our study was permitted by Institutional Animal Care and Use Committee of Kyungpook National University Hospital (No. 2014-12). All tests were performed within the guidelines of the Animal Care Ethics Committee of Kyungpook National University and the National Institutes of Health guidelines.

Fig. 2A-C shows the 2D images and the 2D phase signals from the latex drum of pure tone 1, 2, and 4 kHz stimuli at 15 W. The phase signal was acquired from analysis of the R/G/B information from the lateral line using the phase shift for demodulation processing and 2D phase mapping. Because the time gap between the A-lines was 14.2 μs, the oscillation cycles of the latex drum were 1.006, 2.012, and 3.912 kHz for the 1-, 2-, and 4-kHz stimuli, respectively. Thus, the oscillation cycle of the drum was shortened in proportion to the frequency of the stimulus. On the basis of these data, this DOCT could be able to analyze TM vibration shapes.

Fig. 2D-F shows the vibration pattern of the TM stimulated by the same frequencies used for the latex drum (1, 2, and 4 kHz) obtained by the DOCT. The umbo is most sensitive to vibration of the TM caused by sound stimuli [21,22], so the vibration pattern of the umbo accurately reflects the status of the middle ear. The oscillation cycles of the umbo were 0.893, 1.929, and 4.176 kHz for the 1-, 2-, and 4-kHz stimuli, respectively. The vibration shapes had an integer ratio similar to that of the latex drum.

Fig. 3A-C shows the shape of vibration in the contact area between the TM and the malleus. The oscillation cycles of the contact area were 0.999, 1.472, and 4.176 kHz for the 1-, 2-, and 4-kHz stimuli, respectively, which showed similar pattern to those of the umbo. The results presented in Figs. 2 and 3 show that the 2D demodulation of the phase signal provided a change in phase with time.

The oscillation pattern of the contact area in the simulated effusion condition is shown in Fig. 3D-F. Compared with Fig. 3A-C, the TM and ossicles appear thick and the contact area between the malleus and the TM is not clear because of effusion. The Doppler signal tended toward attenuation at 4 kHz frequency but was prominent at 1 and 2 kHz.

The errors from the mean of the results for the five samples were as follows: 0.006% (latex drum), –0.107% (umbo), and –0.001% (contact area) at 1 kHz, 0.006% (latex drum), –0.036% (umbo), and –0.264% (contact area) at 2 kHz, 0.022% (latex drum), 0.044% (umbo), and 0.044% (contact area) at 4 kHz, each.

To gain a rigorous evaluation of the proposed method, we obtained 3D DOCT images of the area in the white boxes with arrows in Figs. 2 and 3. Fig. 4 shows the 3D reconstruction images for the latex drum and the in vivo normal TM of a rat. The 3D images revealed the changes of the oscillation pattern according to the frequencies of signals.

Our system performance was verified through the acquired 3D reconstruction phase shift images of the contact area between TM and malleus in normal TM and simulated effusion condition in vivo (Fig. 5). Compared with normal TM in vivo (Fig. 5A), the 3D Doppler signal tended toward attenuation at all frequencies but was prominent at 1 and 2 kHz in TM with simulated effusion condition (Fig. 5B).

The oscillating frequencies were compared among latex drum, umbo, and contact area in normal TM at 1, 2, and 4 kHz pure tone stimuli, each (Fig. 6). The results show the restoration pattern according to the stimulated frequency, which means that oscillating frequency follows the stimulated frequency.

These results suggest that the oscillation pattern provided more detailed information on the Doppler shift that the 3D phase pattern. The 3D phase pattern provides a more comprehensive view of the condition of the TM for identifying pathological features and making a diagnosis.