How to detect flow in a sequence of images
By applying Fourier transformation and weighted high-pass filtering to a series of images, the method enhances the detection of blood vessels in OCTA, addressing sensitivity and imaging depth issues while improving processing speed.
Patent Information
- Application Number
- JP2021524373
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2018-11-08
- Filing Date
- 2019-11-04
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2039-11-04
AI Technical Summary
Optical coherence tomography angiography (OCTA) faces challenges in improving detection sensitivity for small vessels with low flow contrast, expanding imaging depth and field of view, and enhancing image processing speed.
A method involving Fourier transformation of a series of images to identify regions of interest with higher frequency amplitudes, distinguishing blood flow from static regions by analyzing the distribution of frequencies, and generating images with enhanced contrast using weighted high-pass filtering.
The method improves the contrast between blood flow and static tissue, enabling clearer identification of blood vessels and wider tissue areas, with faster processing times compared to traditional OCTA methods.
Smart Images

Figure 0007588287000008 
Figure 0007588287000009 
Figure 0007588287000010
Abstract
Description
[Technical field]
[0001] The present invention relates generally to methods for detecting blood flow, etc., in blood vessels, from a series of images, such as images of biological tissue, and, by way of example and not limitation, to methods for processing optical coherence tomography (OCT) images of tissue to improve vascular contrast. [Background technology]
[0002] Optical coherence tomography angiography (OCTA), an extension of optical coherence tomography (OCT), provides a non-invasive method to image tissue vasculature, including arterioles, capillaries, and venules. While image contrast in OCT is determined by the backscattering level in tissue, in OCTA motion-induced changes in the OCT signal can be used to image microvascular networks. Image resolution and field of view in OCTA typically range from 2 to 20 μm and a few mm to 20 mm, respectively. In human tissue, imaging depth may be limited to less than 1 mm.
[0003] In OCTA, blood vessels are identified by identifying the temporal difference between the OCT signal due to moving scatterers in the blood and that due to the surrounding largely stationary tissue. These flow-induced differences are encoded in both the amplitude and phase of the complex OCT signal, and temporal changes in the OCT amplitude signal can be quantified and detected using speckle variance and / or correlation mapping / speckle decorrelation analysis. Summary of the Invention [Problem to be solved by the invention]
[0004] However, there is a need for improved detection sensitivity for small blood vessels with low flow contrast, and for an expanded imaging depth and field of view to image deeper blood vessels and wider tissue regions. It would also be advantageous if the image processing speed of OCTA could be improved. [Means for solving the problem]
[0005] According to a first aspect of the present invention there is provided a method of detecting flow from a series of images of a material, the method comprising: providing an image including a plurality of voxels or regions of interest as a series of at least three images, such that the at least three images of the region of the material provide an intensity I(t) for each voxel or region of interest as a function of time t at at least three time points; obtaining a distribution I(ω) of frequencies ω by Fourier transforming I(t) for each voxel or region of interest, where I(t) comprises intensities at the at least three time points; High frequency band frequency ω H , which has a larger amplitude I L (ω H ) as a first visual characteristic, and further associating voxels or target regions having a smaller amplitude I S and analyzing I(ω) for each voxel or region of interest, including associating a voxel or region of interest with I(ω) as a second visual property, to generate a processed image of the region of the material containing said voxel or said region of interest.
[0006] The first and second visual characteristics may be, for example, a shade of grey, a color or an intensity. but different Hey The step of analyzing I(ω) may include L (ω H ) voxels or regions of interest associated with I S (ω H ) may be performed to increase the contrast of the processed image between related voxels or regions of interest.
[0007] The inventors have determined that the amplitude I(ω H) is often larger in regions of interest related to flow, such as blood flow in a blood vessel, compared to stationary regions. As a result, embodiments of the present invention have the advantage that the contrast between, for example, blood flow (and associated blood vessels) in biological tissue and stationary regions of biological tissue can be increased, improving discrimination.
[0008] In one embodiment, the step of analyzing I(ω) comprises analyzing the amplitude I(ω) at frequency ω in the lower frequency band. L ) by I L (ω H ) and I S (ω H ) may be included.
[0009] I L (ω H ) and I S (ω H ) may be the respective average values of the amplitude in a given frequency band, for example the frequency bands above 0.5, 1.2 or 3 Hz.
[0010] I(ω L ) may be the amplitude at a frequency of substantially 0 Hz (DC).
[0011] The step of providing at least three series of images may include providing a series of at least three depth images, which may be OCT images, such as OCT B-scans consisting of a series of OCT A-scans.
[0012] Each OCT B-scan may be obtained by detecting a series of optical spectra (as for the OCT A-scan) and then converting the spectral intensity distribution into a spatial intensity distribution to form an OCT image (OCT B-scan) by applying an inverse Fourier transform to each acquired optical spectrum.
[0013] Additionally, an OCT image may include multiple OCT B-scans from different locations within the material, which are combined to form an OCT volumetric image.
[0014] The material may be biological tissue such as intraocular tissue, such as the human eye, skin, or brain. The method is typically performed in vivo, but may alternatively be performed ex vivo.
[0015] The present invention will be more fully understood from the following description of specific embodiments thereof, which description is made with reference to the accompanying drawings. [Brief description of the drawings]
[0016] [Figure 1] FIG. 1 is a flow chart of a method for detecting flow in a sequence of images of a material according to one aspect of the present invention. [Diagram 2] Figure 2(a) is a plot of the OCT signal frequency magnitude versus frequency for the capillary flow region and static matrix in the fabricated phantom, and Figure 2(b) is a plot of the OCT signal frequency magnitude versus frequency for human skin blood vessels and static tissue. [Diagram 3] Figure 3(a) is a plot of average high-pass frequency magnitude and contrast as a function of time sample number (i.e., number of B-scans at the same location) for flow field and stationary tissue in a phantom according to an embodiment of the present invention, and Figure 3(b) is a plot of average high-pass frequency magnitude and contrast as a function of time sample number (i.e., number of B-scans at the same location) for flow field and stationary tissue in human skin according to an embodiment of the present invention. [Figure 4] 4(a) and 4(b) are cross-sectional images of blood vessels in a short time series OCT angiogram (OCTA) before and after weighting by the inverse of the mean of the OCT intensity signal, according to an embodiment of the present invention. [Diagram 5]5(a) and 5(b) are OCTA images showing vessel projections obtained by short time series OCTA based on OCT intensity signals before and after weighting with the inverse of the mean of the OCT intensity signals, respectively, according to an embodiment of the present invention, and FIG. 5(c) is a vessel projection by short time series OCTA based on complex OCT signals with weighting. [Figure 6] Figures 6(a), (c), and (e) are OCTA images showing projections of blood vessels obtained by speckle decorrelation, short time series, and speckle dispersion, respectively, according to certain embodiments of the present invention, while Figures 6(b), (d), and (f) are magnified views of the areas outlined in Figures 6(a), (c), and (e), respectively. [Figure 7] Figures 7(a), (c) and (e) are further OCTA images showing projections of blood vessels obtained by speckle decorrelation, short time series and speckle dispersion, respectively, according to certain embodiments of the present invention, while Figures 7(b), (d) and (f) are enlargements of the areas outlined in Figures 7(a), (c) and (f), respectively. [Figure 8] FIG. 8 is a plot of normalized OCTA signal as a function of velocity in the flow region of a phantom for a short time series, speckle decorrelation and speckle dispersion, according to an embodiment of the present invention. [Figure 9] 9(a) is an OCTA image acquired with a short time series showing the projection of blood vessels located between the surface and a depth of about 300 μm in a region of laser-treated skin of a subject, and FIG. 9(b) is an OCTA image acquired with a short time series showing the projection of blood vessels located between the surface and a depth of about 300 μm in a region of normal skin adjacent to the region of laser-treated skin of a subject. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0017] While other embodiments may fall within the scope of the present disclosure as described in the Summary, specific embodiments will be described by way of example with reference to the accompanying drawings.
[0018] In a first aspect, the present invention provides a method for detecting flow from a series of images of a material. With reference to FIG. 1, the method comprises in a first step providing a series of at least three images of a region of the material, each image comprising a plurality of voxels or regions of interest, such that the at least three images of the region of the material provide intensities I(t) at at least three time points. The method comprises in a second step obtaining a frequency distribution I(ω) by Fourier transforming I(t) for each voxel or region of interest, I(t) comprising intensities at at least three time points. The method then comprises in a third step analysing I(ω) for each voxel or region of interest to generate a processed image of the region of the material comprising each voxel or region of interest, wherein the higher frequency bands have larger amplitudes I(t). L (ω H ) as the first visual feature, and associates voxels or regions of interest with smaller amplitudes I S (ω H ) as a second visual characteristic.
[0019] The material is biological tissue, more specifically human skin tissue, in a specific embodiment, and the method is performed in vivo, the method being a method for detecting blood flow in blood vessels in human skin. In a further embodiment, the material is a fabricated flow phantom, such as a capillary region or a static matrix region, which respectively model blood vessels and static tissue in human tissue. The method more specifically includes providing at least three successive depth OCT images.
[0020] However, it will be understood that other biological tissues are contemplated, such as tissue within the human eye. Also, materials other than biological tissue are contemplated and are within the scope of the present invention. Additionally, the method may be performed ex vivo.
[0021] As mentioned above, the inventors have determined that the amplitude I(ωH ) is often large. This finding can be exploited, for example, to image blood vessels with higher contrast.
[0022] Here, we describe improving vascular contrast for flow phantoms and in vivo human skin tissue. However, those skilled in the art will appreciate that the invention has broader application. Furthermore, those skilled in the art will appreciate that the invention is not limited to OCT, but can be used in (for example) magnetic resonance imaging (MRI) and ultrasound imaging.
[0023] One embodiment of the present invention involves obtaining a frequency spectrum of the detected OCT signal by multiple imaging of specific voxels analyzed in the flow phantom and human skin tissue, respectively, and the method for detecting blood flow in capillaries and intravascular vessels, respectively, is referred to herein as short time-series OCT angiography (OCTA). The short time-series OCTA method is also compared with commonly used intensity-based OCTA methods, including speckle decorrelation (correlation mapping) and speckle dispersion. The results generally demonstrate improved contrast and visibility, especially in small vessels, with a small increase in imaging time at a given OCT A-scan rate in human tissue. Moreover, the method is relatively simple and can be rapidly implemented. These advantages suggest potential future applications.
[0024] method Short time series OCTA angiography The basic premise of the method according to an embodiment of the present invention is that blood flow induces stronger non-zero frequency components in the OCT signal than those induced by surrounding stationary tissue. Similar to other OCTA methods, the method first requires the acquisition of co-located OCT B-scans (i.e., B-scans from the same lateral position) at multiple time points throughout the acquisition volume. The OCT intensity signal at the same voxel location (i.e., the modulus of the complex amplitude of the OCT signal) consists of a discrete time series with the nth sample at location (x,y,z) as follows:
[0025]
number
[0026] The time series at each voxel in equation (1) is subjected to a discrete Fourier transform to obtain a complex frequency signal with frequency components F. The frequency F can be expressed as:
[0027]
number
[0028]
number
[0029] After Fourier transformation, even in the presence of flow, voxels with low OCT signal intensity have correspondingly small complex non-zero frequency components. To improve the detectability of flow at low OCT signal levels, a weighting by the inverse of the OCT signal intensity (i.e., the zero frequency component scaled by the number of B-scans at the same location) is introduced, given by:
[0030]
number
[0031]
number
[0032]
number
[0033]
number
[0034] Flow phantom and OCT scans of human skin To demonstrate the short time series OCTA technique on both flow phantoms and normal human skin, OCT scans were taken using a commercial spectral domain scanner (upgraded TELESTO II, Thorlabs Inc., USA). The system has a central wavelength of 1300 nm and provides an image resolution of 5.5 μm (in air) and 13 μm in the axial and lateral directions (as defined by the vendor), respectively. The scanner was operated at 76 kHz (A-scans / s), below a maximum of 146 kHz. Scans were taken in one of two modes: 2D scanning, which involved taking 200 isotopic B-scans from one lateral position with a FOV of 6×3.6 mm (1024×1024 pixels) in x and z, respectively, and 3D scanning with a FOV of 6×1.8×3.6 mm in x, y, and z directions. In the 3D scan mode, 240 locations in the transverse direction (y) were scanned with a set of 5 co-located B-scans taken from each location, using the same pixel size in x and z as in the 2D mode. It took about 4 and 21 seconds to acquire the scans in the 2D and 3D scan modes, respectively. Furthermore, the time interval between B-scans was 17.8 ms (~56 B-scans / s) for both 2D and 3D modes, resulting in a discrete frequency spectrum with components up to 28 Hz.
[0035] For comparison with short time series OCTA, speckle variance of the same 3D scans was calculated for five co-located B-scans using the method described by Mariampillai et al. in "Speckle variance detection of microvasculature using swept-source optical coherence tomography," Opt. Lett. 33(13), 1530-1532 (2008). Speckle decorrelation was determined for each adjacent pair of co-located B-scans with a 3x3 pixel window in the fast scan and depth axes using the formula described in P. Gong, S. Es'haghian, KA Harms, A. Murray, S. Rea, BF Kennedy, FM Wood, DD Sampson, and RA McLaughlin et al. in "Optical coherence tomography for longitudinal monitoring of vascular scars treated by laser fracturing," J. Biophotonics 9(6), 626-636 (2016). This resulted in four decorrelated B-scans from each lateral position, which were averaged to produce one amplified decorrelated B-scan. Furthermore, the speckle decorrelation and speckle variance were weighted by the mean value and threshold of the OCT signal at the corresponding pixels to reduce noise (J. Enfield, E. Jonathan, and M. Leahy, "In vivo imaging of the microcirculation in the volar forearm using correlation mapping optical coherence tomography (cmOCT)," Biomed. Opt. Express 2(5), 1184-1193 (2011)). The threshold used in the short time series OCTA was used here. To make a fair comparison, the same lateral averaging window (3 × 3 pixels) was applied to the short time series and speckle variance images.
[0036] Blood vessels were compared primarily in a depth range of 300 μm from the skin tissue surface (determined from OCT depth scans assuming an average refractive index of 1.4) to ensure that sufficiently strong signals were obtained from all three methods. For each method, the maximum OCTA signal of each A-scan in this depth range was used to generate a projection image of the blood vessels. For visualization, the same color map was used for the projection and cross-sectional OCTA images. The lower and upper thresholds were set at the 50% and 99.5% points, respectively, of the cumulative distribution function of the OCTA signal in the image. These thresholds were chosen empirically to maximize the vascular contrast without losing low-signal vessels. For quantification, each projection image was processed to measure the vascular area density, defined as the ratio of the total tissue area to the total blood vessel area in the thresholded blood vessel image. The threshold was set for each image using the method of Otsu (N. Otsu, "A method for selecting a threshold from a gray level histogram," IEEE Trans. Syst. Man Cybern. 9(1), 62-66 (1979)).
[0037] Silicone flow phantoms were created in-house by mixing Elastosil® P7676A and P7676B fluids (Wacker Chemie AG, Germany) with titanium dioxide in a 3D-printed plastic container (see S. Es'haghian, KM Kennedy, P. Gong, DD Sampson, RA McLaughlin, and BF Kennedy, "In vivo imaging of human skin lesions using optical palpation and mechanical contrast," J. Biomed. Opt. 20(1), 016013 (2015)). The container was customized with two holes in the sidewall to accommodate miniature glass capillaries (outer diameter: 80 μm, inner diameter: 50 μm) that mimic blood vessels. After curing, the capillaries were embedded in static tissue-mimicking silicone. The capillary was then connected to a syringe filled with a suspension of polystyrene granules (nominal diameter 0.5 μm) to mimic blood flow. The syringe was connected to a pump (Fusion 200, Chemyx Inc., USA) for introduction and flow rate control. The scattering properties of the phantom were adjusted by adjusting the ratio of titanium dioxide to Elastosil® P7676A and P7676B so that the signal attenuation of the phantom closely matched that of normal human skin.
[0038] Human subjects (n=4) were recruited for in vivo OCT scanning with ethical approval from the Human Research Ethics Committee of the University of Western Australia. Written informed consent was obtained from all subjects, including one who had undergone laser treatment for wart removal, prior to OCT scanning of the volar forearm. For this subject, one area from the treatment site and one area from adjacent normal skin were selected for OCT imaging. To reduce bulk tissue motion during data acquisition, a spacer was attached to the skin surface to bring the OCT probe into close contact with the skin tissue. In addition, a thin metal strip with a central hole (5 mm diameter) was attached to the skin during imaging as a fiducial marker to check for motion artifacts. As a result, we observed mostly good vascular contrast and negligible vascular distortion, so no motion correction was performed. Further details of the imaging probe spacer and scan setup are described in P. Gong et al., "Optical coherence tomography for longitudinal monitoring of vascular scars treated by laser fracturing," J. Biophotonics 9(6), 626-636 (2016). Scans taken from the phantom and skin tissue were then processed and compared using the three OCTA techniques.
[0039] result In this section, we first consider the contrast in short time-series OCTA and its optimization by selecting the number of samples and weighting the signals. Also, the differences between short time-series implemented in OCT intensity and complex signals are presented. Then, the results from the optimized short time-series OCTA are compared with those obtained from speckle decorrelation and speckle dispersion.
[0040] Vascular contrast Vascular contrast in short time-series OCTA is due to the increase in non-zero frequency components induced by scatterers moving through blood. An example of such vascular contrast, obtained from an extended time series of 200 B-scans, is shown in Figure 2(a), which plots the Fourier transform (magnitude) of the time series of OCT intensity signals in a phantom scan across a flow of polystyrene beads in a capillary tube.
[0041] The bilateral spectral densities of the signals obtained from 200 B-scans in the flow and stationary regions are shown in FIG. 2(a) for the phantom and in FIG. 2(b) for human skin tissue.
[0042] The frequency spectrum of the OCT signal from the static matrix region, shown in Figure 2(a), is approximately 20 dB lower than the peak value at approximately 1.1 Hz and remains consistently lower above that frequency. A similar sharp drop is seen in the capillary flow region (flow velocity 3 mm / s), but the magnitude of the frequency spectrum above 1.1 Hz is much higher than in the static matrix region (Figure 2(a)).
[0043] A similar plot was obtained from skin in vivo, with the spectral density shown in Figure 2(b), determined from 200 isotopic B-scans. The spectral density shows that there is consistent contrast between blood vessels and stationary tissue, which can be parameterized as the average magnitude of non-zero (high-pass) frequencies using Eq. (3). Figure 2(b) further shows that contrast between flow and stationary tissue exists at frequencies higher than about 2 Hz, which was chosen as the cutoff for calculating the average magnitude at high frequencies in Eq. (3). Note that the peak centered at zero frequency for stationary tissue in the skin (Figure 2(b)) is wider than the peak for the stationary matrix in the phantom (Figure 2(a)). This could be due to residual motion in the skin tissue, pulsation (approximately 60-100 beats per minute), or other causes. In the following sections, if only five isotopic B-scans are taken for analysis, the frequency interval is very large (i.e., 11.2 Hz) and only the removal of the zero frequency component as per equation (3) is required to achieve the desired high pass filter.
[0044] Choosing the length of a time series The frequency spectrum shown in FIG. 2 is from 200 isotopic B-scans selected to allow detailed analysis at a single lateral location. Such long acquisition times are not practical for clinical use, and a trade-off is required to reduce the number of time samples while maintaining high vascular contrast. The inventors have examined this trade-off, as presented in FIG. 3. FIG. 3 shows the average magnitude and contrast obtained by calculating the average magnitude of frequencies above 2 Hz for flow regions (solid lines) and stationary tissue (dashed lines) in a 3×3 pixel area of a phantom (FIG. 3(a)) and skin tissue (FIG. 3(b)). Indeed, FIG. 3 shows the vascular contrast in time-series OCTA versus the number of time samples (i.e., the number of isotopic B-scans) for the phantom (FIG. 3(a)) and skin tissue (FIG. 3(b)). With reference to the left vertical axis, the average high-pass frequency magnitude versus the number of time samples is shown for the flow regions (solid lines) and stationary tissue regions (dashed lines). The ratios are plotted as dotted lines against the right-hand vertical axis, with the average value shown by the dashed-dotted line. The inset shows a magnified trace of the contrast for samples 3-9. The circles indicate the ratios for 5 samples.
[0045] In both the flow regions of the capillaries / vessels and the static matrix / stationary tissue, the average magnitude increases with the number of isotopic B-scans taken, as does the difference in average magnitude between the flow and static regions. In particular, the two ratios (plotted by dotted lines) peak around 5-10 isotopic B-scans and then stabilize (albeit with local variations). In Fig. 3(a) and Fig. 3(b), the dashed and dotted lines show the average of all ratios (n=3-200), which are in the range of 7-8. The circles show the ratios over 5 B-scans. From this figure, it can be seen that generating about 5 isotopic B-scans is sufficient to obtain good vascular contrast with the short time series method. Generating more isotopic B-scans may be beneficial in some situations, but will increase acquisition time. On the other hand, generating only 3-4 isotopic B-scans, while showing clear contrast for the vessels analyzed in Fig. 3(b), may cause problems for vessels with lower contrast. Therefore, we decided to use five isotopic B-scans from the same tissue location for the 3D imaging of the vascular network presented below.
[0046] Signal Amplification by Weighting To amplify the vascular signal in flow regions with low OCT signal, we further weight the average magnitude of non-zero frequencies by the inverse of the linear OCT signal intensity, as described in Eq. (4). The weighted image shown in Fig. 4(b) shows an amplified vascular signal (e.g., blood vessels marked by two arrows) compared to Fig. 4(a) before weighting. The improvement level is better evaluated in a projection image containing blood vessels from the skin surface to a depth of 300 μm in the forearm, as shown in Fig. 5(b). Compared to the unweighted corresponding image shown in Fig. 5(a), the weighted image shows improved visibility in terms of vascular connectivity and number of visible blood vessels. Moreover, this weighting significantly suppresses spurious vascular signals caused by very strong surface reflections, as shown by the arrows in Fig. 4(a). Therefore, we incorporate this weighting in our method as a key step to optimize vascular image quality.
[0047] Figure 4 shows pre-(a) and post-(b) time-series OCTA vessel cross-sectional images weighted by the inverse of the mean OCT intensity signal. The arrows and arrowheads indicate corresponding pixels in the two images, at the tissue surface (arrow) and the vessel (arrowhead), respectively. The scale bar corresponds to 500 μm.
[0048] Figure 5 shows the time-series OCTA projections of blood vessels based on the unweighted OCT intensity signal (a), the weighted OCT intensity signal (b), and the weighted complex OCT signal (c). The projections show blood vessels from the skin surface to a depth of 300 μm. The scale bar corresponds to 500 μm.
[0049] Intensity vs. Complex Signal Based Processing As with other OCTA methods, it is possible to analyze either the intensity or the full complex OCT signal (i.e. intensity and phase). The short time series method was applied in both cases on the same skin scan. A representative example is shown in the vessel projection image in Fig. 5. A comparison of Fig. 5(b) and Fig. 5(c) shows that the processing based on the intensity and complex signals leads to very comparable vessel detection. However, using the complex signal results in more motion artifacts (horizontal lines), as is evident in Fig. 5(c).
[0050] A customized imaging spacer and setup is used to minimize movement during data acquisition, which has previously been shown to be effective when combined with a speckle decorrelation technique (P. Gong et al., "Optical coherence tomography for longitudinal monitoring of vascular scars treated by laser fracturing," J. Biophotonics 9(6), 626-636 (2016)).
[0051] In Figures 5(a) and 5(b), which are based on intensity alone, the residual tissue motion is almost invisible, whereas in Figure 5(c), due to the incorporation of motion-sensitive phase information, the residual tissue motion is detected as multiple horizontal lines. Additional motion correction algorithms would be required to mitigate such artifacts if the method were based on complex signals. To avoid this requirement, the time-series OCTA results presented in the following sections are calculated using only the OCT intensity signal.
[0052] Speckle decorrelation versus speckle dispersion. To further evaluate the performance of the OCTA method according to an embodiment of the present invention, the short time series OCTA method was compared with two commonly used intensity-based OCTA methods, speckle decorrelation (correlation mapping) and speckle dispersion, applied to a set of five concentric B-scans in a 3D scan.
[0053] Referring to Figure 6, the projection of blood vessels with speckle decorrelation is shown in Figure 6(a), the projection of blood vessels with a short time series in Figure 6(c), and the projection of blood vessels with speckle dispersion in Figure 6(e). The areas outlined in Figures 6(a), 6(c), and 6(e) are enlarged in Figures 6(b), 6(d), and 6(f), respectively.
[0054] The blood vessels of the forearm skin are projected from the surface to a depth of 300 μm. The arrows in Fig. 6 indicate the same vascular section. The scale bars correspond to 500 μm in (a), (c) and (e) and 200 μm in (b), (d) and (f).
[0055] Thus, FIG. 6 shows an example of skin from the forearm. This example projects blood vessels up to 300 μm from the surface of the skin. The vascular images generated by the short time series method are placed in the middle row for easy comparison with the images generated by the speckle decorrelation method (top row) and the speckle dispersion method (bottom row). In FIG. 6(c), the short time series method provides a visualization of the vascular network comparable to that of the speckle decorrelation (FIG. 6(a)) and speckle dispersion (FIG. 6(e)). Further investigation shows that the short time series projection images provide improved vascular contrast, improving vascular connectivity and visibility. One such example taken from the outlined tissue area in FIG. 6(c) is enlarged in FIG. 6(d). Comparing FIG. 6(b) and FIG. 6(g), obtained using speckle decorrelation and speckle dispersion, respectively, some vascular parts are more clearly observed, with a representative example indicated by the arrow.
[0056] This improvement is further quantified by measuring the vessel area density with an estimated accuracy of about 1%. This resulted in a superior area density of 28% for the short time series image in Figure 6(c), compared to 21% for the speckle decorrelation method and 20% for the speckle dispersion method. The relatively higher density for the short time series method is due to the improved vessel contrast, as shown in Figure 6.
[0057] In this study, the consistent superiority of vascular contrast with short time series OCTA in all subjects (n=4) is confirmed by further examples shown in Fig. 7. Fig. 7(a) shows the projection of the vessels with speckle decorrelation, Fig. 7(c) shows the projection of the vessels with short time series, and Fig. 7(e) shows the projection of the vessels with speckle dispersion. The areas outlined in Fig. 7(a), Fig. 7(c) and Fig. 7(e) are enlarged in Fig. 7(b), Fig. 7(d) and Fig. 7(f), respectively. The vessels of the skin of the forearm are projected from the surface to a depth of 300 μm. The arrowheads in (a), (c) and (e) indicate the same vessel portion. The arrows in (b), (d) and (f) indicate the same vessel portion. The scale bars correspond to 500 μm in (a), (c) and (e) and 200 μm in (b), (d) and (f). In Fig. 7(a), Fig. 7(c) and Fig. 7(e) the maximum intensity projections of blood vessels from the skin surface to a depth of 300 μm are visualized by the three methods. This case shows the superior visibility of blood vessels provided by the short time series method in Fig. 7c than by speckle decorrelation (Fig. 7(a)) and speckle dispersion (Fig. 7(e)). In the corresponding enlarged areas in Fig. 7(b), Fig. 7(d) and Fig. 7(f), the specific case indicated by the arrows highlights the difference in blood vessel contrast, indicating the superior blood vessel signal intensity and connectivity in the short time series images. In this case, the measured blood vessel area density (27%) is higher than that of speckle decorrelation (21%) and speckle dispersion (19%), which is consistent with the analysis results in Fig. 6. Interestingly, in this case, several examples of parallel blood vessels are seen in localized areas (e.g. blood vessels indicated by arrowheads), which are easier to evaluate in the short time series images than those obtained by speckle decorrelation and speckle dispersion.Other projection methods used in the literature use the average OCTA vessel signal instead of the maximum (see, for example, CL Chen, and RK Wang, "Optical Coherence Tomography Angiography [Invited Lecture]," Biomed. Opt. Express 8(2), 1056-1082 (2017), and A. Zhang et al., "Minimizing Projection Artifacts for Accurate Display of Choroidal Neovascularization in OCT Microangiography," Biomed. Opt. Express 6(10), 4130-4143 (2015)). In this study, consistent vessel contrast differences between the three techniques were observed in the average projections as well (not shown).
[0058] To further clarify the contrast differences between the three techniques, experiments were performed to examine the OCTA signal in a phantom versus flow velocity (nine values ranging from 0 to 2 mm / s). Figure 8 shows the normalized OCTA signal versus flow velocity in the flow region of the phantom for short time series (squares), speckle decorrelation (triangles), and speckle dispersion (circles). Specifically, Figure 8 shows the resulting signal intensity in the flow region, determined by normalizing the original flow signal by its maximum value after subtracting noise in the static region from the original flow signal. All three methods show an increase in signal intensity with increasing flow velocity of fine grains (2 μm in diameter) from a baseline signal due to Brownian motion at zero flow velocity, and the signal all saturates between 0.8 and 1.2 mm / s. Compared to speckle decorrelation and speckle dispersion, the short time series method shows a higher normalized signal at low velocities. This observation is consistent with the improved contrast for small blood vessels seen in Figures 6 and 7.
[0059] Moreover, in addition to visualizing the normal vascular network, the short time-series OCTA also shows good vascular contrast for the subject with treated warts. The resulting images are shown in Fig. 9(a) as compared to adjacent normal skin of the same subject shown in Fig. 9(b). Fig. 9 specifically shows short time-series OCTA images of a subject with a laser-treated wart, where Fig. 9(a) shows the projection of blood vessels from the surface of the laser-treated area to a depth of 300 μm, and Fig. 9(b) shows the projection of blood vessels from the surface of the adjacent normal skin (b) to a depth of 300 μm. The scale bar corresponds to 500 μm.
[0060] The wart was removed by laser approximately 16 years prior to the OCT scan. Comparison of images obtained with the three OCTA techniques consistently shows improved visualization with the short time series technique for both normal and treated skin areas (not shown). Although the treated areas show a skin color very similar to normal skin, the underlying microvasculature visualized by the short time series technique clearly shows morphological differences. For example, the treated areas show a network with more branches and a clear honeycomb-like pattern (i.e., localized loops) that are not present in normal skin. Also, the quantified vascular area density in the treated areas (34%) is significantly higher than that in normal skin (29%). Such visualization and associated contrast demonstrate the potential of short time series OCTA for the study of various skin diseases in the future.
[0061] Another important factor is the computation time, which may be a limitation in applications requiring near-real-time or real-time imaging. Overall, speckle decorrelation takes longer than short time series or speckle dispersion due to the need for windowing (not simple averaging) to generate the vascular signal. Using a 3×3 pixel window on an Intel® Core™ i7-3820 processor with MATLAB R2016a (The MathWorks, Inc.), it took about 420 ms to compute the decorrelation of a pair of B-scans (1024×1024 pixels per B-scan). If a larger window is used for processing, the computation time increases significantly (e.g., 990 ms for a 5×5 pixel window). On the other hand, data processing for the short time series method and speckle dispersion is much faster, taking about 64 ms and about 27 ms, respectively, to process each set of five co-located B-scans. This feature indicates the future implementation of short time series methods on fast scanning OCT systems to enable intraprocedural or real-time visualization of the microvasculature.
[0062] Consideration The method proposed in the above described embodiment takes as input a short time series of OCT B-scans, i.e. a series of at least three images taken at the same location, and performs a discrete Fourier transformation to determine the frequency content in order to image blood vessels. The higher magnitude observed at non-zero (high-pass) frequencies (here up to 28 Hz) in blood vessels creates a clear contrast to distinguish blood vessels from the surrounding stationary tissue. The method can also be easily applied to OCT scans taken using conventional scanning parameters for imaging microvascular networks. In a case study on human skin, short time series OCTA shows a moderate but consistently improved vascular contrast compared to speckle decorrelation and speckle dispersion, especially for smaller blood vessels. Although the in vivo comparison was performed on skin tissue, the application of the short time series OCTA method to other biological tissues, such as the retina, is also envisaged.
[0063] The number of B-scans taken from the same position is an important parameter for practical implementation of the short time series method according to the present invention. In this study, we chose five to maximize the vascular / static tissue contrast of the skin while minimizing the amount of collected data and the corresponding total acquisition time.
[0064] Thus, the short time series OCTA method according to a specific embodiment of the present invention has demonstrated its performance in imaging tissue microvasculature in vivo, where the flow signature in the frequency domain was analyzed via Fourier transformation of the time series of OCT signals in five B-scans from the same lateral position. The angiographic signal is calculated as the average of non-zero (high-pass) frequency components and clearly distinguishes blood vessels from stationary tissue, as demonstrated in flow phantoms and in human skin in vivo. Weighting the angiographic signal by the inverse of the average value of the OCT signal showed improved vessel detectability. The imaging performance of short time series OCTA was evaluated by comparison with commonly used speckle decorrelation and speckle dispersion methods, showing consistent and significantly improved results, especially improved visualization of small blood vessels and evidence of increased vascular density in human dermal microvascular network.
[0065] It should also be understood that the reference herein to a prior art document is not an admission that the document forms part of the general knowledge in the art in Australia or anywhere else.
Claims
1. 1. A method for detecting flow from a series of images of a material, comprising: providing a series of at least three images of the region of the material, each image comprising a plurality of voxels or regions of interest, such that the at least three images of the region of the material provide an intensity I(t) of each voxel or region of interest as a function of time t for at least three time points; - Fourier transforming I(t) for each voxel or region of interest, where I(t) comprises the intensities at the at least three time points, to obtain a distribution I(ω) of frequencies ω; A frequency ω in a frequency band greater than the first frequency H , a larger amplitude I L (ω H ) as a first visual characteristic, and further associating voxels or target regions having a smaller amplitude I S (ω H analysing I(ω) for each voxel or region of interest, including associating voxels or regions of interest having a second visual property with I(ω) and generating a processed image of the region of the material containing said voxels or regions of interest; The larger amplitude I L (ω H ) is related to the flow, and the smaller amplitude I S (ω H ) is associated with a stationary region, and said first frequency is 0.5, 1.2 or 3 Hz.
2. 2. A method for detecting flow from a series of images of a material as claimed in claim 1, comprising: A method of detecting flow from a series of images of a material, wherein the flow is blood flow within a blood vessel.
3. 3. A method for detecting flow from a series of images of a material according to claim 1 or 2, comprising: A method of detecting flow in a series of images of a material, wherein the first visual characteristic and the second visual characteristic differ in grey, color or intensity.
4. A method for detecting flow from a series of images of a material according to any one of claims 1 to 3, comprising: The step of analyzing I(ω) may include determining whether the contrast of the processed image is I L (ω H ) and voxels related to I S (ω H 2. A method for detecting flow in a series of images of a material, comprising:
5. A method for detecting a flow from a series of images of a substance according to claim 4, which is dependent on claim 2, further comprising the steps of:
13. A method for detecting flow in a series of images of a material, the method comprising generating the processed image having contrast-enhanced blood vessels.
6. A method for detecting flow from a series of images of a material according to any one of claims 1 to 5, comprising: The step of analyzing I(ω) L (ω H ) and I S (ω H ) at the frequency ωL of the lower frequency band, L 16. A method for detecting flow from a series of images of a material comprising: division by a function of the magnitude of the shift.
7. A method for detecting flow from a series of images of a material according to any one of claims 1 to 6, comprising: I L (ω H ) and I S (ω H ) is a respective average value of amplitudes in a predetermined frequency band greater than said first frequency.
8. A method for detecting flow from a series of images of a material according to any one of claims 1 to 7, comprising: I L (ω H ) is an amplitude about a frequency of 0 Hz.
9. A method for detecting flow from a series of images of a material according to any one of claims 1 to 8, comprising:
13. A method for detecting flow from a series of images of a material, wherein the step of providing at least three series of images comprises providing at least three series of depth images.
10. 10. A method for detecting flow from a series of images of a material according to claim 9, comprising:
11. A method for detecting flow from a series of images of a material, wherein the depth images are OCT images of a series of OCT B-scans comprised of OCT A-scans.
11. 11. A method for detecting flow from a series of images of a material according to claim 10, comprising: A method of detecting flow from a series of images of a material, wherein the OCT images comprise a series of OCT B-scans at different locations within the material to obtain a volumetric image.
12. 12. A method for detecting flow from a series of images of a material according to claim 11, comprising:
11. A method for detecting flow from a series of images of a material, wherein the step of providing a series of at least three images comprises acquiring OCT optical spectra and then converting the spectral intensity distributions associated with the OCT A-scans into spatial intensity distributions to form an image by inverse Fourier transforming each acquired OCT optical spectrum.
13. A method for detecting flow from a series of images of a material according to any one of claims 1 to 12, comprising: A method of detecting flow from a series of images of a material, characterized in that the material is a biological tissue of the human eye and skin, and tissue within the eye and skin.
Citation Information
Patent Citations
Image quality improving method for dynamic image
JP1995023926A
Magnetic resonance imaging apparatus
JP2011147560A
Filtering of image sequences
US20100034441A1
