2.1.

OCT Otoscope System

Conceptually, the hand-held OCT (HHOCT) clinical system design (see Fig. 1) follows our previous study that reported the otoscopic OCT system.²² It is comprised of several components: the hand-held imaging probe, an endoscopy cart, and a wireless foot pedal serving as an input assist device. The system employed a swept laser with a center wavelength of 1310 nm and a 39 nm bandwidth, yielding an axial resolution of $33.4\mu \mathrm{m}$ (in tissue, $n=1.3$) for OCT imaging, which is sufficient for resolving the human middle ear structures.²⁴

Fig. 1

System configuration (a) schematic diagram. Abbreviations: 90:10 fc, 90:10 fiber fused coupler; Circ, fiber optic circulator; cL, collimating lens; MEMS, MEMS scan mirror; DM, dichroic mirror; ILL, illuminator; OBJL, objective lens; MESPL, middle ear speculum; EC, ear canal; ML, malleus; IN, incus; ST, stapes; aL, achromatic lens; CM, video camera; PC, polarization controller; 50:50 fc, 50:50 fiber fused coupler; BPD, balanced photo-detector; GPU, graphic processing unit; DGT, digitizer; DAQ, data acquisition board, MEMS CTRL, MEMS scan mirror controller; FP, foot pedal. (b) Entire system setup in a regular clinic, (c) close-up of the hand-held imaging probe. (d) Ray tracing diagram in Zemax. (e) Lateral point-spread function (PSF) of the $x$ and $y$ axis measured at the focal plane and (f) size comparison of a standard otoscope and the HHOCT imaging probe.

The essential optical system design is shown in Fig. 1(a). The swept laser output was directed into a 90:10 optical fiber coupler (fC). The 10% side illuminated the reference arm of the interferometer. After passing through a fiber optic patch cable sized to an optical path length matched with the sample arm, the reference light passed into a 50:50 fC where it was combined with the sample arm. The 90% side illuminated the sample arm, passing through a circulator (Circ) and then onto the hand-held imaging probe. Inside the hand-held imaging probe, the light was guided by a dual axes MEMS mirror (Mirrocle Technology, Northamptonshire, United Kingdom) through the objective lens and onto the sample for scanning. The back-reflected light passed back through the circulator and was combined with the reference arm via the 50:50 optical fC. The two outputs of the fC were directed into the balanced photodetector (BPD, WL-BPD600MA, Wieserlabs, Penzberg, Germany).

The objective yielded a $38\mu \mathrm{m}$ (FWHM) scanning beam spot at the focal plane measured by a commercial scanning-slit optical beam profiler (BP104-IR, Thorlabs, Newton, New Jersey, United States) and 5.03 mm depth of focus (DOF) calculated from the waist ($1/e^2$) of a Gaussian beam, as shown in Fig. 1(e). A ray tracing diagram for the objective lens is shown in Fig. 1(d). As can be seen, the rays cross at $\sim 16\mathrm{mm}$ from the final lens surface at a point coinciding approximately with the tip of the speculum. From that point, the rays fan out and come to a focus at an arc about 27 mm from the crossing. As a result, the image is sampled in a spherical polar coordinate system, unlike our prior telecentric system, which naturally samples in the more convenient Cartesian space. Below, we describe our approach to resampling in Cartesian space and aberration correction. The advantage of this imaging pattern is a larger field of view, 7.4 mm at the focal plane through the limited aperture of the middle ear speculum. This represents an approximately two-fold increase over our prior telecentric system.

A high-pass dichroic mirror (DM) positioned between the MEMS mirror and an achromatic objective lens (aL) directed the back-reflected visible light onto a small CMOS sensor (MU9PC-MH, XIMEA, Münster, Germany) within the hand-held imaging probe to generate a video otoscope image. Visible light illumination of the ear canal was accomplished by a set of visible LEDs mounted around the objective lens. Real-time video output from the CMOS was shown on the monitor and used by the clinician to guide the hand-held unit to the proper position for OCT imaging. As a size comparison of our HHOCT imaging probe with a typical otoscope, we photographed the two side-by-side as shown in Fig. 1(f). The outer shell of the HHOCT device is designed to resemble the grip angle and size of a standard otoscope. This design choice aims to make it familiar for the clinician and ensure patient comfort during examinations.

The endoscopy cart housed the major electronic components. Detailed information about the electronic devices and signal processing procedures can be found in previous studies.^22,23 Briefly, the interferometric signal from the BPD was connected to a digitizer and processed using custom software written in Python, C++, and Cuda C. The 400 MHz clock from the swept laser served as the master clock. The data acquisition (DAQ) board generated an enable trigger that worked as a logical AND so that the digitizer only acquired data during the desired portion of the mirror scan. The DAQ also generated a synchronized analog signal for controlling the MEMS mirror. In addition, the foot pedal facilitated hands-free data acquisition, minimizing motion artifacts caused by the operator. The foot pedal, operated through the Bluetooth Secure Personal Network (PAN), was paired with the system software for image capture that allowed a single operator hands-free operation for the standalone and meticulous measurements in clinical procedures.

As noted above, the new imaging device has a much larger field of view than our prior telecentric systems. As a consequence, we sample in spherical polar space rather than the typical Cartesian space as well as introduce some image aberration. To correct this and enable accurate image dimensions, we developed a two-part algorithm that includes coordinate transformation and aberration correction.

2.2.

Model for Coordinate Transformation

Although our OCT system nominally scans in spherical polar coordinates, we chose to treat it as two polar coordinate systems, in part because each axis of the MEMs mirror has its own hysteresis and center of rotation. Hence, OCT volume scanning involves two steps: $x\text{-}z$ plane scanning (fast axis) and $y\text{-}z$ plane scanning (slow axis). As shown in Fig. 2, we establish two Cartesian coordinate systems to represent the real sample space; one follows the slow axis $R_{C1}(O,X,Y,Z)$, and the other is based on the fast axis $R_{C2}(O^{\prime },X^{\prime },Y^{\prime },Z^{\prime })$, with $O(O^{\prime })$ representing the mirror rotation center. A polar coordinate system $R_{\text{raw}}(O_{\text{raw}},x_{\text{raw}},y_{\text{raw}},z_{\text{raw}})$ is introduced to represent the measured image space.

Fig. 2

Model of the coordinate transformation for the otoscopic OCT system. The green 3D shape represents the collected data. The fast-axis and slow-axis coordinate systems are shown in blue and red, respectively. The gray plane represents an arbitrary plane represented in both spaces.

Assuming a point $P$ in the plane is denoted as $P_1(x_1,y_1,z_1)$ in $R_{C1}$ and $P_2(x_2,y_2,z_2)$ in $R_{C2}$, we establish the relationship between them as

According to Eqs. (1) and (2), we derive

In the polar space, we know point $P$ is represented as

2.3.

Model for Aberration Correction

After performing a coordinate transformation, the OCT image exhibited distortion attributed to optical aberrations, as illustrated in Fig. 3(a). We imaged a checkerboard calibration target (Edmund Optics Inc., Barrington, New Jersey, United States), $0.2\times 0.2\mathrm{mm}$ square pattern, at various imaging depths while controlling the spatial offset to assess the image quality and the extent of the aberrations. We estimated the residual aberration present in depth slices of an OCT volume by constructing a polynomial model based on the coordinate correspondence between the OCT enface images with the standard checkerboard pattern image serving as the reference.

2.4.

Clinical Imaging Protocol

This study was approved by the institutional review board at the University of Southern California (HS-21-00338). Over 100 patients were imaged after providing informed consent in a tertiary care otology clinic at Keck Hospital of USC from October 2022 to June 2023. Patients’ ears were cleaned of wax with a curette prior to imaging. Inclusion criteria included anyone with a patent ear canal with visibility of the TM on the otoscopic view. Exclusion criteria included patients with active bleeding, discharge, or open wound in the external ear canal; a non-visible TM; or a non-patent ear canal. Patients were imaged while sitting upright in an exam chair. Both ears were imaged unless a patient had abnormal anatomy or intolerance to the plastic ear speculum that made imaging one ear not possible within a clinic setting.

Multiple OCT volumes were obtained from each of the four anatomical quadrants of the TM using superficially identified anatomical landmarks (e.g., Umbo, pars flaccida) on live otoscopic video feed to orient the scanning volume location and corroborate ME structures seen in real time OCT B scans. A single volume scan took 0.4 s. The relatively high speed helped to reduce motion artifacts that may arise from patient or operator movement. The region of interest was selected with the center of the OCT beam positioned directly over the umbo. This was done to encompass the entire tympanic membrane (TM) and cover the area with suspected pathology, thereby maximizing the likelihood of capturing intriguing pathological features within the OCT volumes. The total imaging time, including consenting the patient, reviewing the images, and explaining to patients what they were seeing, was $<5\min$. OCT volumes were post-processed and analyzed using MatLab (MathWorks), Amira (Thermofisher Scientific), and Fiji (NIH).

3.1.

Coordinate Transformation and Aberration Correction

We found the following parameters in Table 1 for mapping our volumetric images into Cartesian space. Examples of images at the focal plane and $\pm 2\mathrm{mm}$ before and after coordinate transformation are shown in Fig. 3(a) columns 1 and 2. The imaging target has $200\mu \mathrm{m}$ squares with an asymmetric cross near the center of the image to aid in registration. As can be seen, although the image quality is improved, there is still aberration, resembling barrel distortion, that can be corrected.

Table 1

Parameters of coordinate transformation.

Fig. 3

(a) Calibration target images at the focal plane (0 mm) and $\pm 2\mathrm{mm}$. Enface images at the depth of the target, above a B-scan through the middle of the target. (b) and (c) Representative linear fits of corresponding $x$ and $y$ pixel positions as a function of depth.

We performed aberration correction using the same checkerboard calibration target at six different depths ranging from $-5$ to 5 mm. Between these depths, we assumed a linear relationship. As demonstrated in Figs. 3(b) and 3(c), the linear fit among corresponding points has an $R^2$ of 0.99 and 0.97, respectively. Therefore, we can readily realize the aberration correction of 3D OCT images by linear fitting. The result following this algorithm is the third column in Fig. 3(a), i.e., after coordinate transformation and aberration correction.

3.2.

Normal TM and Middle Ear

Here, we present several example cases from the clinical study. Our first example is a patient with a healthy middle ear as confirmed by an expert otologist through clinical assessment and otoscopic examination. From the otoscopic view in Fig. 4(a), the tympanic membrane can be seen to be fairly transparent, which allows us to see the incus, malleus, and cochlear promontory through the membrane. Other anatomical landmarks include the umbo and cone-of-light. Figure 4(b) is a summed voxel projection (SVP) of the OCT volume in which the 2D image is generated by summing along the depth dimension. This generates an image that can be directly compared with the standard otoscope view to verify orientation. Figures 4(c) and 4(d) are representative cross-sectional B-scans that were extracted along the dashed colored lines, as shown in panels (a) and (b). The cross-sections show a normal thin tympanic membrane along with sections of the chorda tympani and stapes in addition to the anatomy that was also visible in the otoscopic view. Figure 4(e) displays the 3D volume rendering that visualizes major anatomical structures presented in the otoscopic view.

Fig. 4

Otoscope and OCT image acquired from normal ear. Abbreviations: MA, malleus; CT, chorda tympani; TM, tympanic membrane; IN, incus; UM, umbo; CL, cone of light; P, promontory. (a) A snapshot of the otoscopic video camera. Dashed colored lines indicate the positions where the linear B-scans are shown in panels (c) and (d). (b) En face image generated by SVP from an OCT volume. Dashed colored lines indicate the positions where the linear B-scans are shown in panels (c) and (d). (c) B-scan extracted along the green-colored dashed line from the OCT volume. (d) B-scan extracted along the red-colored dashed line from the OCT volume. (e) 3D volume rendering.

3.3.

Abnormal Tissue Growth

This case is a patient presented with 2 weeks of ear pain and aural fullness. An otoscopic evaluation by an expert otologist revealed what appeared to be abnormal tissue growth superficially on bilateral TMs. The level of involvement with the TM and if the tissue was directly against the TM were not discernible with a routine physical exam. The physician used OCT imaging to determine if tissue debridement was possible in the clinic without damage to the TM. Representative images are shown in Fig. 5.

Fig. 5

Otoscope and OCT image acquired from myringitis over the subsequent visits. (a) A snapshot of the otoscopic video camera at first visit. Dashed colored lines indicate the positions where the linear B-scans are shown in panels (c) and (d). (b) En face image generated by SVP from an OCT volume at first visit. Dashed colored lines indicate the positions where the linear B-scans are shown in panels (c) and (d). (c) B-scan extracted along the green-colored dashed line from the OCT volume at first visit. (d) B-scan extracted along the red-colored dashed line from the OCT volume. White arrows with circled numbers from 1 to 4 denote the layered plaques of abnormal tissue partially adhere to the TM. (e) Frame from (Video 1) showing 3D volume rendering and B-scan fly through at first visit. (f) A snapshot of the otoscopic video camera at the visit 28 days after the first visit. Dashed colored lines indicate the positions where the linear B-scans are shown in panels (h) and (i). (g) En face image generated by SVP from an OCT volume at the visit 28 days after the first visit. Dashed colored lines indicate the positions where the linear B-scans are shown in panels (h) and (i). (h) B-scan extracted along the green-colored dashed line from the OCT volume at the visit 28 days after the first visit. (i) B-scan extracted along the red-colored dashed line from the OCT volume. White arrows with circled numbers from 5 to 7 denote the symbolic middle ear structures. The circled number 5 indicates the incus. The circled number 6 shows the umbo. The circled number 7 is from a part of a cochlear promontory. (j) Frame from (Video 2) showing 3D volume rendering and B-scan fly through at 28 days after the first visit. (Video 1, MP4, 2.50 MB [URL: https://doi.org/10.1117/1.JBO.29.8.086005.s1]; Video 2, MP4, 2.36 MB [URL: https://doi.org/10.1117/1.JBO.29.8.086005.s2]).

Panels (a) and (b) are the otoscopic image and SVP. Panels (c) and (d) are representative cross-sectional OCT B-scans that show the layered plaques of abnormal tissue partially adherent to the surface of the TM with partial elevation of the TM. The plaques are strongly scattering and therefore strongly shadow the tissue below. This leads to the sharp edges labeled (1) and (2) in panel (d) that look like discontinuities in the TM. Panel (e) displays a still image from a movie showing the 3D volume rendering. The volume rendering helps visualize the plaques.

These are denoted with white arrows and circled numbers. Given the findings, a conservative management approach with Debrox drops (Prestige Consumer Healthcare Inc., Irvington, New York, United States) was chosen. The patient was imaged every week over four subsequent visits. Images from the final visit are shown in Figs. 5(f)–5(j). The images clearly show the resolution and clearance of abnormal tissue. The images revealed no residual tissue plaques, a thin normal-looking TM in cross-section, and middle ear anatomy such as the incus and cochlear promontory. Panel (j) displays a still image from a movie showing the 3D volume rendering. The volume rendering helps show how the plaques have largely been resolved.

3.4.

Chronic Otitis Media with Effusion

Images from a patient with left ear chronic otitis media with effusion are shown in Fig. 6. An otoscopic evaluation revealed opacification of the TM potentially consistent with fluid in the middle ear with no apparent air-fluid levels, bubbles, or active otorrhea. OCT imaging demonstrated speckled signal intensity underlying a thickened TM distinct from the dark middle ear space evident in the OCT images of the right ear. The signal intensity drops off uniformly deeper into the middle ear space along the length of the TM, consistent with OCT imaging through a mucoid effusion with extensive scattering. A representative area with mucoid effusion adjacent to the TM is labeled by white arrows in Fig. 6. Following imaging, subsequent in-office myringotomy expressed thick yellow opaque mucoid fluid from the patients left ear, confirming the presence of fluid.

Fig. 6

Otoscope and OCT image acquired from effusion. (a) A snapshot of the otoscopic video camera at first visit. Dashed colored lines indicate the positions where the linear B-scans are shown in panels (c) and (d). (b) En face image generated by SVP from an OCT volume at first visit. Dashed colored lines indicate the positions where the linear B-scans are shown in panels (c) and (d). (c) B-scan extracted along the green-colored dashed line from the OCT volume at first visit. (d) B-scan extracted along the red-colored dashed line from the OCT volume. White arrows denote the mucoid effusion adjacent to the TM in panels (c) and (d).

3.5.

TM Perforation

A patient with a history of chronic otitis media presented with a small elliptical TM perforation. On otoscopic examination during a 6-week follow-up exam, Fig. 7(a), the perforation appeared to be non-healing and patent into the middle ear cavity. However, the OCT cross-sectional images in Figs. 7(c) and 7(d) clearly show a thin layer of TM $\sim 68\text{-}\mu \mathrm{m}$ thick, filling the space where the perforation had been. The OCT volume showed a healing elliptical perforation with minor and major axes of 300 and $400\mu \mathrm{m}$, respectively, and a thin layer of tissue across the entire area.

Fig. 7

Otoscope and OCT image acquired from a healed thin TM. (a) A snapshot of the otoscopic video camera. Dashed colored lines indicate the positions where the linear B-scans are shown in panels (c) and (d). Dashed colored box indicates where the OCT volume was acquired. (b) En face image generated by SVP from an OCT volume inside of the dashed colored box on the otoscopic video image (a). Dashed colored lines indicate the positions where the linear B-scans are shown in panels (c) and (d). (c) B-scan extracted along the green-colored dashed line from the OCT volume. (d) B-scan extracted along the red-colored dashed line from the OCT volume. White arrows denote the healed TM in panels (c) and (d).

3.6.

TM Retraction

A patient with apparent TM retraction identified on otoscopic evaluation by an expert otologist was imaged. In the OCT cross-sections, Figs. 8(c) and 8(d), the TM is shown to be severely retracted. A co-registered normal TM is shown for reference in grayscale and denoted with a white arrow labeled 3. Inspection of the OCT volume showed that the TM was touching the incus and was very close but not touching the promontory. These are labeled in representative cross-sections by the circled numbers 1 and 2, respectively. TM contact with the incus brings an increased risk of bone necrosis and cholesteatoma formation.

Fig. 8

Otoscope and OCT image acquired from TM retraction patient. (a) A snapshot of the otoscopic video camera. Dashed colored lines indicate the positions where the B-scans are shown in panels (c) and (d). (b) En face image generated by SVP from an OCT volume. Dashed colored lines indicate the positions where the linear B-scans are shown in panels (c) and (d). (c) B-scan extracted along the green-colored dashed line from the OCT volume and overlaid with the gray-shaded OCT image (labeled 3) acquired from a normal condition. (d) B-scan extracted along the red-colored dashed line from the OCT volume. White arrows with circled numbers 1 and 2 denote that the TM touched the incus and was close to the cochlear promontory.

3.7.

Tympanic Membrane Protrusion with Air Pocket

A patient with a history of bilateral otosclerosis, stapedectomies (ear surgery to remove the stapes bone), and unknown pathology affecting the right ear TM presenting with decreased hearing and ear fullness was examined. They had seen several prior otologists with no reported clear diagnosis other than bulging of the TM or post-surgical changes. The otoscopic view appeared to show a partially white opaque TM possibly consistent with post-surgical scarring or middle ear mass of the TM posteriorly in Fig. 9(a). TM continuity with the superior-anterior quadrant of the ear canal was indeterminate with potential perforation on the otoscopic view. In Figs. 9(c) and 9(d), further OCT imaging demonstrated the presence of a $\sim 1.3\text{-}\mathrm{mm}$ diameter air pocket within a body of thick scar tissue. The scar tissue, $\sim 420\mu \mathrm{m}$ thick, was continuous with the interior layer of the TM near the center, annotated with white arrows in Figs. 9(c) and 9(d), demonstrating no perforation on OCT imaging.

Fig. 9

Otoscope and OCT image acquired from TM protrusion with an air pocket. (a) A snapshot of the otoscopic video camera. Dashed colored lines indicate the positions where the linear B-scans are shown in panels (c) and (d). (b) En face image generated by SVP from an OCT volume. Dashed colored lines indicate the positions where the linear B-scans are shown in panels (c) and (d). (c) B-scan extracted along the green-colored dashed line from the OCT volume. (d) B-scan extracted along the red-colored dashed line from the OCT volume. White arrows denote the TM protrusion with an air pocket.