Comparison of Intensity- and Polarization-based Contrast in Amyloid-beta Plaques as Observed by Optical Coherence Tomography

Featured Application: The accumulation of pathological amyloid-beta protein is a key hallmark of Alzheimer’s disease. In this study, we evaluated intensity- and polarization-sensitive optical coherence tomography as complementary methods for imaging amyloid-beta plaques as compared to quantitative pathology. Abstract: One key hallmark of Alzheimer’s disease (AD) is the accumulation of extracellular amyloid-beta protein in cortical regions of the brain. For a deﬁnitive diagnosis of AD, post-mortem histological analysis, including sectioning and staining of di ﬀ erent brain regions, is required. Here, we present optical coherence tomography (OCT) as a tissue-preserving imaging modality for the visualization of amyloid-beta plaques and compare their contrast in intensity- and polarization-sensitive (PS) OCT. Human brain samples of eleven patients diagnosed with AD were imaged. Three-dimensional PS-OCT datasets were acquired and plaques were manually segmented in 500 intensity and retardation cross-sections per patient using the freely available ITK-SNAP software. The image contrast of plaques was quantiﬁed. Histological staining of tissue sections from the same specimens was performed to compare OCT ﬁndings against the gold standard. Furthermore, the distribution of plaques was evaluated for intensity-based OCT, PS-OCT and the corresponding histological amyloid-beta staining. Only 5% of plaques were visible in both intensity and retardation segmentations, suggesting that di ﬀ erent types of plaques may be visualized by the two OCT contrast channels. Our results indicate that multicontrast OCT imaging might be a promising approach for a tissue-preserving visualization of amyloid-beta plaques in AD.


Introduction
Alzheimer's disease (AD) is the leading cause of dementia accounting for up to 70% of cases [1]. In 2015, up to 47 million people were affected worldwide [2][3][4]. This number is expected to double every 20 years, resulting in a dramatically increased social as well as financial burden [5]. AD is characterized

OCT and Data Acquisition
A spectral-domain PS-OCT system (TEL220PSC2, Thorlabs, Lübeck, Germany) was utilized for imaging the brain samples. The setup used a multiplexed superluminescent diode in the near infrared (central wavelength 1300 nm, bandwidth 170 nm) and achieved a theoretical axial resolution of 5.5 µm in air. The theoretical lateral resolution was 7 µm (at the focal plane), using a commercial scanning lens with a focal length of 18 mm (OCT-LK2, Thorlabs, Lübeck, Germany). The focus was positioned at the tissue surface for imaging. For the acquisition, a lateral (x,y) field of view of 500 pixels × 500 pixels in a cortical region of the sample was imaged, corresponding to 500 µm × 500 µm. The imaging depth range was 3.5 mm in air. The A-scan rate was 76 kHz and a sensitivity of 109 dB was measured. OCT images displaying backscattered intensity and retardation were computed from the raw spectral data. All volumes were saved as image stacks for further processing.

Histology
After OCT imaging, the whole fixed AD cerebral tissue samples were dehydrated and embedded in paraffin in preparation for histopathological workup. The brain samples were cut into 2.5 µm thick sections on a microtome. One slice was stained immunohistochemically using an anti-amyloid-beta antibody (clone 6F/3D, diluted 1:100, Dako, Santa Clara, CA, United States). With a second tissue slice, Congo red (CI22120, 0.5% dissolved in 50% ethanol, Merck, Kenilworth, NJ, United States) staining was performed to visualize neuritic plaques. Sections stained by Congo red were imaged using both bright-field and polarized light microscopy (BH2-RFCA, Olympus, Hamburg, Germany). In addition, for one specimen, 180 consecutive 3 µm thick serial sections were cut and stained immunohistochemically against amyloid-beta. All sections were further scanned using a slide scanner (NanoZoomer 2.0 HT, Hamamatsu, Shizuoka, Japan). Afterwards, the Fiji [37] plugin StackReg [38] was used to align the serial micrographs in order to generate a three-dimensional (3D) rendering of the stained AD cerebral tissue.

Data Processing and Statistical Analysis
Using ITK-SNAP, the plaques were manually segmented in the intensity and the retardation data [39]. The resulting volumetric segmentation data were saved as binary files. Using the 3D Objects Counter plugin [40] in Fiji and the segmentation data of the plaques, their diameter, area and plaque density (in plaques per mm 3 ) were quantified over the whole volume [37]. The mean plaque diameter in µm and plaque density were also extracted by the 3D Objects Counter tool. For plaque diameters, the average of the horizontal and vertical diameter was calculated.
The mean intensity and retardation signals (µ) of OCT image data were calculated for the plaques (µ P ) and the surrounding brain parenchyma (µ BP ) using the retardation averaging method 1 [41]. To estimate the signal levels in the brain parenchyma, the same amount of pixels used for the evaluation of the signal in the plaques was used. Values adjacent to the plaques at the same depth (located 10 µm to the left in OCT tomograms) were chosen. Using a custom Matlab (Version R2015b, The MathWorks, Inc., Natick, MA, United States) code, the Weber contrast (WC) [42] was calculated in both contrast modalities for the plaques compared to the surrounding cortical gray matter: WC was evaluated for the intensity data (Int) and for the retardation data (Ret) in linear scale based on the plaque segmentations performed in both the intensity data (Int Seg ) and the retardation data (Ret Seg ), resulting in four different WC analyses. The same evaluation was performed for the signal to noise ratio (SNR), where σ P and σ BP are the standard deviation of the signals in plaques and the brain parenchyma, respectively [42]: Furthermore, the plaque overlap (i.e., the percentage of plaques segmented in both intensity and retardation images) was evaluated using Fiji. Finally, histograms and box plots were generated for visual representation. Student's t-tests were used to analyze statistical differences for equal mean values and Bonferroni correction was applied when evaluations on multiple groups were performed. The data are presented with their associated standard deviations, unless otherwise stated.
For a direct comparison to histology, the number of amyloid-beta plaques in fields of view (FOV) of 500 µm × 500 µm × 2.5 µm (corresponding to the thickness of a histological section) were quantified for OCT intensity images, PS-OCT retardation images and the corresponding histological section of one specimen. The dense neuritic and diffuse plaques in one histological slice were manually counted in five randomly chosen regions and the mean plaque diameter and the areal plaque density (in plaques per mm 2 ) were evaluated.

Appearance of Plaques in Intensity and Retardation OCT Images
Samples from eleven brains affected by AD were imaged. Figure 1a,b show the intensity en-face projection over a depth of 10 µm and a representative B-scan, respectively. Figure 1d,e show the corresponding retardation en-face projection and B-scan. Plaques appeared as highly scattering features in the intensity OCT images. In retardation data sets, plaques exhibited increased phase retardation compared to the polarization preserving brain parenchyma. As cumulative phase retardation effects occur, some plaques exhibit retardation trails. This effect can be observed in Figure 1e in the plaque on the right. Plaques visible in both OCT and PS-OCT data are highlighted with yellow arrows. Reconstructions of the acquired 3D volumes are displayed in Figure 1c,f. corresponding retardation en-face projection and B-scan. Plaques appeared as highly scattering features in the intensity OCT images. In retardation data sets, plaques exhibited increased phase retardation compared to the polarization preserving brain parenchyma. As cumulative phase retardation effects occur, some plaques exhibit retardation trails. This effect can be observed in Figure  1e in the plaque on the right. Plaques visible in both OCT and PS-OCT data are highlighted with yellow arrows. Reconstructions of the acquired 3D volumes are displayed in Figure 1c,f.    Figure 2e,f, respectively. In Figure 2h, an overlay of both segmentation data sets is shown. Only few plaques appear in both data sets. The amount of plaques overlapping in both data sets was quantified as shown in Figure 2g. On average, 5% of the amyloid-beta plaques in all data sets were visualized by both contrast channels.

Comparison of Segmentation and Image Contrast
Weber contrast was assessed in the intensity and retardation data sets based on the segmentation of both the intensity data (Int Seg ) and retardation data (Ret Seg ). As shown in Figure 3a, the image contrast of plaques was higher in intensity data and Int Seg compared to intensity data and Ret Seg ; however, no significance was found (p = 0.8). The same analysis for the retardation values showed a significant contrast difference when comparing plaques segmented for retardation and intensity data (p = 0.02), see Figure 3b. Additionally, the SNR was assessed in the intensity and retardation data sets based on the segmentation of both the intensity data (Int Seg ) and retardation data (Ret Seg ). A slightly higher difference for intensity images and Int Seg was observed compared to intensity images and Ret Seg (p = 0.6). For retardation data and Ret Seg, a higher SNR was calculated when compared to retardation images and Int Seg (p = 0.06). A detailed overview of plaques segmented in all data sets is shown in Figure 4a.
respectively. Hyperscattering or turquoise pixels indicate amyloid-beta plaques. B-scans with manually segmented plaques highlighted as red and yellow dots are displayed in Figure 2c,d. Representative 3D renderings of the segmentations based on the intensity and phase retardation images are shown in Figure 2e,f, respectively. In Figure 2h, an overlay of both segmentation data sets is shown. Only few plaques appear in both data sets. The amount of plaques overlapping in both data sets was quantified as shown in Figure 2g. On average, 5% of the amyloid-beta plaques in all data sets were visualized by both contrast channels. Weber contrast was assessed in the intensity and retardation data sets based on the segmentation of both the intensity data (IntSeg) and retardation data (RetSeg). As shown in Figure 3a, the image contrast of plaques was higher in intensity data and IntSeg compared to intensity data and RetSeg; however, no significance was found (p = 0.8). The same analysis for the retardation values showed a significant contrast difference when comparing plaques segmented for retardation and intensity data (p = 0.02), see Figure 3b. Additionally, the SNR was assessed in the intensity and retardation data sets based on the segmentation of both the intensity data (IntSeg) and retardation data (RetSeg). A slightly higher difference for intensity images and IntSeg was observed compared to intensity images and RetSeg (p = 0.6). For retardation data and RetSeg, a higher SNR was calculated when compared to retardation images and IntSeg (p = 0.06). A detailed overview of plaques segmented in all data sets is shown in Figure 4a. The Weber contrast (WC) was calculated for all intensity data sets in linear scale. WC for the plaques segmented in the intensity data (IntSeg) compared to those segmented in the retardation data (RetSeg) was evaluated. A slightly higher but not statistically significant difference for intensity data and IntSeg was observed compared to intensity data and RetSeg. (b) WC for the retardation data. A significant WC difference for retardation data and RetSeg over retardation data and IntSeg was calculated (p-value 0.02). (c) The signal to noise ratio (SNR) was calculated for all intensity data sets in linear scale. SNR for the plaques segmented in the intensity data (IntSeg) compared to those segmented in the retardation data (RetSeg) was evaluated. A slightly higher but not statistically significant difference for intensity data and IntSeg was observed compared to intensity data and RetSeg.
(d) The SNR for the retardation data. A slightly higher but not statistically significant difference for retardation data and RetSeg was observed compared to retardation data and IntSeg. The Weber contrast (WC) was calculated for all intensity data sets in linear scale. WC for the plaques segmented in the intensity data (Int Seg ) compared to those segmented in the retardation data (Ret Seg ) was evaluated. A slightly higher but not statistically significant difference for intensity data and Int Seg was observed compared to intensity data and Ret Seg . (b) WC for the retardation data.
Appl. Sci. 2019, 9, 2100 7 of 13 A significant WC difference for retardation data and Ret Seg over retardation data and Int Seg was calculated (p-value 0.02). (c) The signal to noise ratio (SNR) was calculated for all intensity data sets in linear scale. SNR for the plaques segmented in the intensity data (Int Seg ) compared to those segmented in the retardation data (Ret Seg ) was evaluated. A slightly higher but not statistically significant difference for intensity data and Int Seg was observed compared to intensity data and Ret Seg . (d) The SNR for the retardation data. A slightly higher but not statistically significant difference for retardation data and Ret Seg was observed compared to retardation data and Int Seg . Figure 3. (a) The Weber contrast (WC) was calculated for all intensity data sets in linear scale. WC for the plaques segmented in the intensity data (IntSeg) compared to those segmented in the retardation data (RetSeg) was evaluated. A slightly higher but not statistically significant difference for intensity data and IntSeg was observed compared to intensity data and RetSeg. (b) WC for the retardation data. A significant WC difference for retardation data and RetSeg over retardation data and IntSeg was calculated (p-value 0.02). (c) The signal to noise ratio (SNR) was calculated for all intensity data sets in linear scale. SNR for the plaques segmented in the intensity data (IntSeg) compared to those segmented in the retardation data (RetSeg) was evaluated. A slightly higher but not statistically significant difference for intensity data and IntSeg was observed compared to intensity data and RetSeg.
(d) The SNR for the retardation data. A slightly higher but not statistically significant difference for retardation data and RetSeg was observed compared to retardation data and IntSeg.

Evaluation of Plaque Load and Size Using Multicontrast OCT
The diameters of the plaques segmented in intensity and retardation data was evaluated. The size distributions of all data sets revealed by the two OCT contrast channels are shown in the histogram in Figure 5a. In the intensity images, the diameters ranged from 11 to 85 µm with a mean value of 42 µm. The plaque size observed in the retardation data ranged from 14 to 148 µm in diameter (mean 54 µm). Next, the observed plaque volume density (plaques per mm 3 ) was analyzed and is plotted in Figure 5b. The plaque density in intensity data (median of 224 [Q1-Q3: 112-332 p/mm 3 ]) and retardation data (median of 156 [Q1-Q3: 124-208 p/mm 3 ]) was rather similar. Lastly, the mean plaque area (in µm 2 ) was calculated for each data set. Results are displayed in box plots in Figure 5c. In intensity data, plaque areas ranged between 1250 and 1450 µm 2 (median of 1394 [Q1-Q3: 1256-1458 µm 2 ]). This is significantly smaller than those measured in retardation images, where the plaque area was clustered between 1400 and 1750 µm 2 (median of 1518 [Q1-Q3: 1430-1727 µm 2 ]). Q1 and Q3 represent the first and third quartile.

Evaluation of Plaque Load and Size Using Multicontrast OCT
The diameters of the plaques segmented in intensity and retardation data was evaluated. The size distributions of all data sets revealed by the two OCT contrast channels are shown in the histogram in Figure 5a. In the intensity images, the diameters ranged from 11 to 85 µm with a mean value of 42 µm. The plaque size observed in the retardation data ranged from 14 to 148 µm in diameter (mean 54 µm). Next, the observed plaque volume density (plaques per mm³) was analyzed and is plotted in Figure 5b. The plaque density in intensity data (median of 224 [Q1-Q3: 112-332 p/mm 3 ]) and retardation data (median of 156 [Q1-Q3: 124-208 p/mm³]) was rather similar. Lastly, the mean plaque area (in µm²) was calculated for each data set. Results are displayed in box plots in Figure 5c. In intensity data, plaque areas ranged between 1250 and 1450 µm² (median of 1394 [Q1-Q3: 1256-1458 µm²]). This is significantly smaller than those measured in retardation images, where the plaque area was clustered between 1400 and 1750 µm² (median of 1518 [Q1-Q3: 1430-1727 µm²]). Q1 and Q3 represent the first and third quartile.

Comparison to Histology
Immunohistochemical stainings against amyloid-beta were performed in order to confirm OCT findings. In Figure 6a, a 3D reconstruction and rendering of 180 stained serial sections is shown. Figure 6b displays a zoom-in into the cortical region of the brain sample. Immunohistochemically stained amyloid-beta plaques appear as brownish deposits. Neuritic plaques appear as red

Comparison to Histology
Immunohistochemical stainings against amyloid-beta were performed in order to confirm OCT findings. In Figure 6a, a 3D reconstruction and rendering of 180 stained serial sections is shown. Figure 6b displays a zoom-in into the cortical region of the brain sample. Immunohistochemically stained amyloid-beta plaques appear as brownish deposits. Neuritic plaques appear as red accumulations when labeled with Congo red and were investigated with brightfield microscopy (see Figure 6c). This staining was performed for each brain sample as well. Evaluating the same Congo red stained section using polarization-contrast microscopy, neuritic plaques exhibit an apple-green color with a characteristic cross shape as displayed in Figure 6d. For a direct comparison of OCT and state-of-the-art histology, the number of amyloid-beta plaques was evaluated for intensity-based OCT, PS-OCT and the corresponding histological amyloid-beta stainings in five regions of one patient. In the immunohistochemically stained histological section, 240 ± 64 plaques/mm 2 were counted. Of these plaques, 33.5% (≙ 79.2 ± 20.8 plaques/mm 2 ) were classified as cored/dense and 66.5% (≙ 160.8 ± 42.8 plaques/mm 2 ) as diffuse. In the corresponding OCT intensity and retardation images, plaque densities of 34 ± 10 plaques/mm 2 and 148 ± 32 plaques/mm 2 were measured, as shown in Figure 7a. The diameters of the diffuse and dense plaques annotated in histological sections were evaluated and plotted in the histogram shown in Figure 7b. Representative images of a dense plaque (diameters ranged from 4.8 µm to 25.5 µm) and a diffuse plaque (diameters ranged from 6.9 µm to 83.6 µm) are shown in Figure 7c,d, respectively. The histology data showed that significantly fewer plaques are visualized by OCT (p < 0.01), and they appeared significantly bigger (p < 0.01). For a direct comparison of OCT and state-of-the-art histology, the number of amyloid-beta plaques was evaluated for intensity-based OCT, PS-OCT and the corresponding histological amyloid-beta stainings in five regions of one patient. In the immunohistochemically stained histological section, 240 ± 64 plaques/mm 2 were counted. Of these plaques, 33.5% (= 79.2 ± 20.8 plaques/mm 2 ) were classified as cored/dense and 66.5% (= 160.8 ± 42.8 plaques/mm 2 ) as diffuse. In the corresponding OCT intensity and retardation images, plaque densities of 34 ± 10 plaques/mm 2 and 148 ± 32 plaques/mm 2 were measured, as shown in Figure 7a. The diameters of the diffuse and dense plaques annotated in histological sections were evaluated and plotted in the histogram shown in Figure 7b. Representative images of a dense plaque (diameters ranged from 4.8 µm to 25.5 µm) and a diffuse plaque (diameters ranged from 6.9 µm to 83.6 µm) are shown in Figure 7c,d, respectively. The histology data showed that significantly fewer plaques are visualized by OCT (p < 0.01), and they appeared significantly bigger (p < 0.01). and 148 ± 32 plaques/mm were measured, as shown in Figure 7a. The diameters of the diffuse and dense plaques annotated in histological sections were evaluated and plotted in the histogram shown in Figure 7b. Representative images of a dense plaque (diameters ranged from 4.8 µm to 25.5 µm) and a diffuse plaque (diameters ranged from 6.9 µm to 83.6 µm) are shown in Figure 7c,d, respectively. The histology data showed that significantly fewer plaques are visualized by OCT (p < 0.01), and they appeared significantly bigger (p < 0.01).

Discussion
The post-mortem assessment of amyloid-beta plaques is an essential component of neuro-pathological AD diagnostics [16]. In this article, we investigated the imaging capabilities of OCT intensity and polarization contrast as alternative techniques for the visualization and quantitative assessment of amyloid-beta plaques in ex-vivo cerebral tissue of AD patients.
PS-OCT visualizes amyloid-beta plaques based on their optical scattering and birefringent properties [33][34][35][36]. Previous studies mainly exploited the hyperscattering characteristics of plaques and only one report investigated the birefringent characteristics of neuritic plaques using PS-OCT [19]. Here, we presented a comparative analysis of the plaque visualization capabilities of intensity-and PS-OCT contrast. Surprisingly, most plaques appeared in only one contrast channel and were in most of the cases not visible in the other. While for each patient, we observed different fractions of plaques in intensity and retardation respectively (Figure 4), only very few plaques-on average 5%-did actually overlap (Figure 2). This finding may indicate different underlying contrast mechanisms, suggesting that distinct subgroups of plaques provide independent optical contrast. Figure 7 may suggest that intensity and retardation contrasts visualize different plaque morphologies. However, further investigation and a one-on-one correlation to histology would be required in order to elucidate the relationship between the two OCT contrast channels with the different subgroups of plaques.
In our earlier work, we simulated the birefringent [19] and scattering signals [36] produced by neuritic plaques assuming a geometry of radiating fibrils [43]. These simulations suggested that neuritic plaques are birefringent, resulting in increased phase retardation observed in PS-OCT [19]. As cumulative phase retardation effects occur, some plaques exhibit retardation trails. In our recently published simulations we found that low numerical aperture (NA) values resulted in straight retardation shadows in z-direction whereas higher NA values produced a lateral spread of those shadows [19]. The intensity of directly backscattered light was found to be rather low for this plaque geometry, such that strong neuritic plaque signals were only observed using a dark-field detection scheme for OCT [36].
By analyzing the signal levels in plaques compared to the surrounding gray matter, most birefringent plaques (i.e., those visualized in retardation images) were invisible in co-localized OCT intensity data. However, higher retardation was observed only in these birefringent plaques as compared to those highlighted in the intensity images. This finding, which is also supported by previous research reporting the visualization of plaques by OCT without polarization contrast [31][32][33][34][35], reinforces our assumption that intensity and retardation-based contrasts are caused by different families of amyloid-beta plaques. To further evaluate this, WC and SNR calculations were performed. As Figure 3 shows, there was no significant difference either in WC or SNR when comparing intensity data. This slight difference may be explained by a depth-dependent contrast increase between plaques and surrounding brain parenchyma. However, note that plaques vary not only in size but also in shape and density. Therefore, in intensity images, many plaques appeared as strong hyperreflective spots and only some expressed a vague hyperreflectivity, resulting in weaker WC and SNR values. For retardation images, WC differences were significant. These findings indicate that plaques in retardation images overall express a higher contrast than plaques visualized in intensity data when compared to surrounding neural tissue.
The plaque sizes as measured by histology and the two different OCT contrasts differed from each other. The diameters of the plaques segmented in the retardation images showed a trend to be bigger than those segmented in the intensity data (Figure 5a). Plaques segmented in histology data, the gold standard, had a smaller plaque diameter when compared to plaques segmented in retardation and intensity images (Figure 5a versus Figure 7b). The minimum plaque size detected in histology was around 3 µm, whereas the smallest plaques detected in OCT were around 12 µm. While the detected size of plaques in all modalities did agree with earlier literature reports [44][45][46], the direct comparison between histology and OCT (Figures 5a and 7) showed that OCT seems to severely overestimate plaque size. Potential reasons for this could be the expansion (convolution) of imaged objects by the imaging system's point spread function (PSF), the inherently different contrast mechanisms and tissue shrinkage due to dehydration and fixation processes. Previous studies reported tissue shrinkage due to histologic workup up to 17% [47,48]. Most importantly, the image resolution of OCT is rather poor when compared to histology. The observed size of a point-like object is described by the PSF of the imaging system. In an image, any object will be smeared out (convolved) with the PSF in x, y and z directions. The PSF in OCT is described by the coherence length in z-direction (which is determined by the covered spectral bandwidth in wavenumber space) and by the beam spot size in x-y direction [49]. Moreover, the beam spot size depends on the distance to the focal plane. Due to the limited resolution, multiple single plaques in close vicinity detected by histology might also be mistaken for a single, large plaque in OCT. In contrast, immunostaining visualizes amyloid-beta deposits of (almost) all shapes and sizes very specifically, and is expected to pick up much fainter signals and structures. OCT seems to visualize much fewer plaques (Figure 7a).
Several methodological modifications may improve the performance of OCT for visualizing amyloid-beta plaques. In order to enhance the axial image resolution, a light source providing a broader bandwidth in wavenumber space (e.g., with a shorter central wavelength and/or broader wavelength coverage than the system used here) could be employed [32,33]. Furthermore, objective lenses with higher magnification could be used for improving the lateral resolution. Note, however, that the use of objectives with a shorter focal length usually comes at the cost of a reduction of the FOV.
As a logical next step, tissues of patients with different stages of the disease could be investigated to evaluate the time point at which plaques can be visualized with OCT. Further elucidation in the three-dimensional shape could be obtained by reconstructed histological serial sections similar to that shown in Figure 6a.
As an advantage over histology, OCT can investigate ex-vivo cerebral tissue in a tissue-preserving manner, without any fixation, sectioning or staining. However, at the same time, OCT relies on intrinsic contrast and therefore does not require additional tissue processing steps. Hence, for some applications, multimodal OCT imaging might be a promising high throughput approach for the investigation of pathological alterations of the human brain.

Conclusions
A commercial polarization-sensitive optical coherence tomography (PS-OCT) setup was used to investigate ex-vivo human brain samples of patients diagnosed with Alzheimer's disease (AD). The performance of multicontrast OCT for the imaging of amyloid-beta plaques, one hallmark of the disease, was investigated. Following OCT imaging, histological staining was performed for a comparison to the gold-standard technique for AD diagnosis. Plaques were segmented in the OCT intensity and retardation data sets and the plaque contrast was analyzed. Notably, complementary distinct plaque patterns were observed in both contrast channels, only 5% of plaques were visualized in both intensity-and retardation-based contrast. The segmentation data were further used to evaluate the plaque diameter, area and the number of plaques per mm 3 . The same parameters were evaluated for histology data, showing that significantly fewer plaques are visualized by OCT and that they appeared significantly bigger (p < 0.01). In the OCT intensity contrast, a similar amount of plaques was counted compared to the retardation data. Plaques in retardation images appeared significantly larger. In conclusion, the results suggest that a multicontrast OCT is a promising tissue-preserving imaging technique to investigate amyloid-beta plaques.