Oct image processing program and oct image processing device

Abstract

To provide an OCT image processing program and an OCT image processing device capable of reducing an influence of artifacts generated in a deep layer image more appropriately.SOLUTION: A control part acquires a shallow layer image and a deep layer image generated on the basis of motion contract data acquired for the same biological tissue. The shallow layer image is an image of a shallow layer region (region shallower than the deep layer region) of a biological tissue. The deep layer image is an image of a deep layer region deeper than the shallow layer region of the biological tissue. The control part calculates a weight for correction for correcting a signal image so that the correlation between the shallow layer image and the deep layer image becomes small. The control part corrects the deep layer image according to the calculated weight for correction. The control part calculates the weight for correction by a robust estimation method.SELECTED DRAWING: Figure 2

Description

[Technical Field]

[0001] The present disclosure relates to an optical coherence tomography (OCT) image processing program and an OCT image processing device used to process OCT images of biological tissue acquired based on the principles of OCT. [Background technology]

[0002] Conventionally, a technique for acquiring motion contrast data of biological tissue (e.g., the fundus of a test eye) based on the principles of OCT has been proposed. Motion contrast data is data obtained by processing multiple OCT signals acquired at different times from the same position on the biological tissue. The motion contrast data expresses information on the movement of the biological tissue (e.g., the movement of blood flow within blood vessels in the biological tissue). Note that data indicating the position of blood vessels in the biological tissue (angiography data) is an example of motion contrast data.

[0003] By processing the motion contrast data, it is also possible to acquire images of multiple regions of biological tissue at different depths. Here, we consider a case where an image (shallow image) of a first depth region (shallow region) and an image (deep image) of a second depth region (deep region) deeper than the first depth region are acquired based on the motion contrast data acquired for the same biological tissue. In this case, signals derived from tissue movement in the shallow region (e.g., blood flow) will also appear in the deep image as artifacts (hereinafter referred to as "projection artifacts").

[0004] The ophthalmologic image processing device described in Patent Document 1 sets a weight that reduces the correlation between a superficial image and a deep image, and corrects the deep image according to the set weight. Because blood vessels in superficial regions differ from blood vessels in deep regions, ideally, the correlation between the superficial image and the deep image would be small. However, the greater the influence of projection artifacts, the greater the correlation between the superficial image and the deep image. Therefore, Patent Document 1 aims to reduce the influence of projection artifacts by correcting the superficial image to reduce the correlation between the superficial image and the deep image. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Publication No. 2019-150405 Summary of the Invention [Problem to be solved by the invention]

[0006] In the technology described in Patent Document 1, a weight that reduces the correlation between a shallow image and a deep image is set as a weight for correcting a deep image. In this case, an appropriate weight may not be set due to the influence of a large number of outliers present in the image information of at least one of the shallow image and the deep image. As a result, the influence of projection artifacts may not be reduced appropriately.

[0007] A typical object of the present disclosure is to provide an OCT image processing program and an OCT image processing device that can more appropriately reduce the influence of artifacts that occur in deep layer images. [Means for solving the problem]

[0008] An OCT image processing program provided by a typical embodiment of the present disclosure is an OCT image processing program executed by an OCT image processing device that processes images acquired by an OCT device, wherein the OCT device is capable of generating motion contrast data by processing multiple OCT signals acquired at different times from the same position on biological tissue, and the OCT image processing program is executed by a control unit of the OCT image processing device, whereby the OCT image processing device executes the following steps: an image acquisition step of acquiring a shallow image, which is an image of a superficial region of biological tissue, and a deep image, which is an image of a deep region deeper than the superficial region, which are generated based on the motion contrast data; a correction weight calculation step of calculating a correction weight for correcting the deep image so as to reduce the correlation between the shallow image and the deep image; and an image correction step of correcting the deep image in accordance with the correction weight; and in the correction weight calculation step, the correction weight is calculated using a robust estimation method.

[0009] An OCT image processing device provided by a typical embodiment of the present disclosure is an OCT image processing device that processes images acquired by an OCT device, wherein the OCT device is capable of generating motion contrast data by processing multiple OCT signals acquired from the same position on biological tissue at different times, and a control unit of the OCT image processing device executes an image acquisition step of acquiring a shallow image, which is an image of a superficial region of the biological tissue, and a deep image, which is an image of a deep region deeper than the superficial region, generated based on the motion contrast data; a correction weight calculation step of calculating a correction weight for correcting the deep image so as to reduce the correlation between the shallow image and the deep image; and an image correction step of correcting the deep image in accordance with the correction weight, wherein the correction weight is calculated using a robust estimation method in the correction weight calculation step.

[0010] According to the OCT image processing program and OCT image processing device according to the present disclosure, the influence of artifacts occurring in deep layer images can be more appropriately reduced. [Brief explanation of the drawings]

[0011] [Figure 1] 1 is a block diagram showing a schematic configuration of an OCT device (OCT image processing device) 1. FIG. [Figure 2] 1 is a flowchart of OCT image processing executed by an OCT device (OCT image processing device) 1. [Figure 3] 10 is an explanatory diagram for explaining an example of a method for acquiring an interference signal for a two-dimensional measurement region 55 on a living tissue. FIG. [Figure 4] 10A and 10B are explanatory diagrams illustrating an example of a method for acquiring interference signals of multiple frames at different times from the same scanning line. [Figure 5] FIG. 10 is a diagram showing an example of the results of detecting layers and boundaries from a B-scan image. [Figure 6] FIG. 1 is a diagram showing an example of a shallow image 60 and a deep image 70 generated based on the same motion contrast data. [Figure 7] FIG. 1 is a diagram showing an example of a state in which each of a shallow layer image 60 and a deep layer image 70 is divided into a plurality of local regions LA. [Figure 8] FIG. 10 is a diagram showing a co-occurrence histogram of a shallow image 60 and a deep image 70. [Figure 9] 9 is an explanatory diagram showing the relationship between the co-occurrence histogram shown in FIG. 8 and the slopes of the lines showing the two correction weights w and w'. FIG. [Figure 10] FIG. 10 is a diagram showing an example of corrected images 80, 80' acquired by correcting a deep layer image 70 according to two correction weights w, w', respectively. DETAILED DESCRIPTION OF THE INVENTION

[0012] <Summary> The OCT image processing device exemplified in the present disclosure processes images acquired by the OCT device. The OCT image processing program is executed by a control unit of the OCT image processing device. The OCT device can generate motion contrast data by processing multiple OCT signals acquired at different times from the same position on biological tissue. The control unit of the OCT image processing device executes an image acquisition step, a correction weight calculation step, and an image correction step. In the image acquisition step, the control unit acquires a superficial image and a deep image generated based on the motion contrast data acquired for the same biological tissue. The superficial image is an image of a superficial region of the biological tissue (a region shallower than the deep region). The deep image is an image of a deep region of the biological tissue that is deeper than the superficial region. Here, projection artifacts are considered to be blood vessels in the superficial region that are captured in the deep image, so in areas where projection artifacts appear, it is considered that the same object (e.g., blood vessels) is captured in both the shallow image and the deep image. When removing projection artifacts, a process is performed to remove objects that are common to both the shallow image and the deep image. This processing can be achieved by reducing the correlation between the shallow image and the deep image. In the correction weight calculation step, the control unit calculates correction weights for correcting the deep image so that the correlation between the shallow image and the deep image is reduced (e.g., so that the correlation is minimized as much as possible). In the image correction step, the control unit corrects the deep image according to the calculated correction weights. In the correction weight calculation step, the control unit calculates the correction weights using a robust estimation method.

[0013] One possible method for decorrelating shallow and deep images is principal component analysis. As mentioned above, projection artifacts appear in both shallow and deep images, and are therefore considered to be the first principal component when the shallow and deep images are subjected to principal component analysis. Therefore, by visualizing the second principal component score, it is thought that a deep image from which the projection artifacts have been removed can be obtained.

[0014] However, while the method using principal component analysis produces an image that appears to have reduced projection artifacts, the resulting image contains both deep image information and shallow image information, making it difficult to say that this is an appropriate process for allowing a user to observe a deep image.

[0015] Therefore, we consider correcting the deep image using the gradient of the first principal component, rather than the principal component score whose coordinates are transformed by principal component analysis. The gradient of the first principal component is used as the correction weight w, and the shallow image E shallow and deep image E deep By subtracting from E, the projection artifact is removed. deep This process is shown by the following equation: E´ deep =E deep -w×E shallow

[0016] It is believed that the processing shown in the above equation will result in a deep image with projection artifacts removed. However, the gradient obtained by principal component analysis is generally different from the gradient of the line that minimizes the residual from the shallow image to the deep image. To minimize the residual from the shallow image to the deep image, one possible method is to calculate a line that converts the pixels of the shallow image to the pixels of the deep image using the least squares method, and then apply the above equation using the gradient of the calculated line as the correction weight w.

[0017] However, the calculation process of the correction weight w using the least squares method described above is based on the assumption that only projection artifacts that appear in both the shallow and deep images are present. In reality, there are many outliers, such as blood vessels that appear only in the deep image, blood vessels that appear only in the shallow image, and noise. In the above-described process, the influence of many outliers reduces the accuracy of the weight calculation using the least squares method, and there have been cases where the projection artifacts were not properly removed.

[0018] In contrast, the OCT image processing device of the present disclosure uses a robust estimation method to reduce the influence of outliers and calculate the correction weight w, which makes it easier to more appropriately reduce the influence of artifacts that occur in deep layer images.

[0019] In the correction weight calculation step, the correction weight may be calculated by a robust estimation method using shallow image weights calculated from the shallow image. As described above, the projection artifacts appearing in the deep image are considered to be the shallow image being projected onto the deep image. Therefore, by calculating the correction weights by a robust estimation method using the weights of the shallow image itself (shallow image weights), the correction weights are calculated in a state where the influence of outliers is more appropriately excluded. As a result, the influence of artifacts occurring in the deep image is more appropriately reduced.

[0020] In the correction weight calculation step, the correction weights may be calculated by weighted regression analysis (e.g., weighted least squares method, etc.) as a robust estimation method. With weighted regression analysis, by setting a weight for each piece of data, it is possible to suppress the influence of low reliability data (data that is likely to be an outlier) on the regression results. Therefore, by using weighted regression analysis, the correction weights are calculated in a state where the influence of outliers is more appropriately excluded.

[0021] However, the specific method for calculating the correction weights may be changed. For example, the correction weights may be calculated by at least one of the following methods: Random Sample Consensus (RANSAC), M-estimation, and Least Median of Squares (LMedS).

[0022] In the correction weight calculation step, when m is the shallow image weight, x is the brightness of the shallow image, and y is the brightness of the deep image, the correction weight w may be calculated by the following (Equation 1). In this case, the correction weight is calculated in a state where the influence of outliers is more appropriately excluded.

number

[0023] As an example, the shallow layer image may be normalized so that the minimum brightness is 0 and the maximum brightness is 1, and then squared and used as the shallow layer image weight m. In this case, the brightness of the background becomes almost 0, so that strong signal information tends to remain.

[0024] In the correction weight calculation step, the control unit may perform a calculation process of a correction weight for each local region. In the image correction step, the control unit may correct the deep image for each local region according to the correction weight calculated for each local region. In this case, if the influence of the projection artifact in a certain local region is large, the correlation between the shallow image and the deep image will be large, and therefore the correction weight will be large. On the other hand, if the influence of the projection artifact in a certain local region is small, the correlation between the shallow image and the deep image will be large, and therefore the correction weight will be small. Therefore, by performing processing for each local region, the influence of artifacts occurring in the deep image can be more appropriately reduced.

[0025] In addition, when calculating the correction weights using the shallow image weights, the control unit may also calculate the shallow image weights from the shallow images for each local region, and may also calculate the correction weights using the shallow image weights for each local region. In this case, since the shallow image weights are also calculated for each local region, the influence of outliers is also appropriately reduced according to the local region.

[0026] <Embodiment> A typical embodiment according to the present disclosure will be described below. As an example, the OCT device 1 of this embodiment can process OCT signals acquired using biological tissue of the fundus of a subject's eye E as a specimen. However, at least some of the techniques exemplified in this disclosure can be applied even when processing OCT signals of biological tissue other than the fundus of the subject's eye E or biological tissue other than the subject's eye E (for example, skin, digestive organs, brain, or blood vessels (including cardiovascular vessels)). OCT data is data acquired based on the principles of optical coherence tomography (OCT).

[0027] In this embodiment, the OCT device 1 itself performs various processes described below, thereby functioning as an OCT image processing device that processes OCT images. However, devices that can function as an OCT image processing device are not limited to the OCT device 1. For example, a PC or the like that can acquire the OCT signal or OCT image generated by the OCT device 1 may also function as an OCT image processing device.

[0028] The schematic configuration of an OCT device (OCT image processing device) 1 of this embodiment will be described with reference to Fig. 1. The OCT device 1 includes an OCT section 10 and a control unit 30. The OCT section 10 includes an OCT light source 11, a coupler (light splitter) 12, a measurement optical system 13, a reference optical system 20, a light receiving element 22, and a front observation optical system 23.

[0029] The OCT light source 11 emits light (OCT light) for acquiring an OCT signal. The coupler 12 splits the OCT light emitted from the OCT light source 11 into measurement light and reference light. The coupler 12 of this embodiment also combines and causes interference between the measurement light reflected by the biological tissue of the subject (in this embodiment, the fundus tissue of the subject's eye E) and the reference light generated by the reference optical system 20. That is, the coupler 12 of this embodiment serves both as a branching optical element that branches the OCT light into measurement light and reference light, and as a combining optical element that combines the reflected light of the measurement light with the reference light. Note that the configuration of at least one of the branching optical element and the combining optical element may be changed. For example, an element other than a coupler (e.g., a circulator, a beam splitter, etc.) may be used.

[0030] The measurement optical system 13 guides the measurement light split by the coupler 12 to the subject and returns the measurement light reflected by the subject to the coupler 12. The measurement optical system 13 includes a scanning unit 14, an irradiation optical system 16, and a focus adjustment unit 17. The scanning unit 14 is driven by a driving unit 15 to scan (deflect) the measurement light in a two-dimensional direction intersecting the optical axis of the measurement light. In this embodiment, two galvanometer mirrors capable of deflecting the measurement light in different directions are used as the scanning unit 14. However, another device for deflecting light (e.g., at least one of a polygon mirror, a resonant scanner, an acousto-optical element, etc.) may also be used as the scanning unit 14. The irradiation optical system 16 is located downstream of the scanning unit 14 in the optical path (i.e., on the subject side) and irradiates the measurement light onto the tissue of the subject. The focus adjustment unit 17 adjusts the focus of the measurement light by moving an optical member (e.g., a lens) included in the irradiation optical system 16 in a direction along the optical axis of the measurement light.

[0031] The reference optical system 20 generates reference light and returns it to the coupler 12. In this embodiment, the reference optical system 20 generates the reference light by reflecting the reference light split by the coupler 12 using a reflective optical system (e.g., a reference mirror). However, the configuration of the reference optical system 20 can also be changed. For example, the reference optical system 20 may transmit the light incident from the coupler 12 without reflecting it and return it to the coupler 12. The reference optical system 20 includes an optical path length difference adjustment unit 21 that changes the optical path length difference between the measurement light and the reference light. In this embodiment, the optical path length difference is changed by moving the reference mirror in the optical axis direction. Note that the configuration for changing the optical path length difference may be provided in the optical path of the measurement optical system 13.

[0032] The light receiving element 22 detects an interference signal by receiving interference light between the measurement light and the reference light generated by the coupler 12. In this embodiment, the principle of Fourier domain OCT is adopted. In Fourier domain OCT, the spectral intensity of the interference light (spectral interference signal) is detected by the light receiving element 22, and a complex OCT signal is acquired by Fourier transforming the spectral intensity data. Examples of Fourier domain OCT that can be adopted include spectral-domain OCT (SD-OCT) and swept-source OCT (SS-OCT). It is also possible to adopt, for example, time-domain OCT (TD-OCT).

[0033] In this embodiment, SD-OCT is employed. In the case of SD-OCT, for example, a low-coherence light source (broadband light source) is used as the OCT light source 11, and a spectroscopic optical system (spectrometer) that separates the interference light into individual frequency components (individual wavelength components) is provided near the light-receiving element 22 in the optical path of the interference light. In the case of SS-OCT, for example, a wavelength-scanning light source (tunable light source) that changes the emission wavelength at high speed over time is used as the OCT light source 11. In this case, the OCT light source 11 may include a light source, a fiber ring resonator, and a wavelength-selective filter. Examples of wavelength-selective filters include a filter that combines a diffraction grating and a polygon mirror, and a filter that uses a Fabry-Perot etalon.

[0034] In this embodiment, the scanning unit 14 scans a spot of measurement light within a two-dimensional measurement region to acquire three-dimensional OCT data (e.g., a three-dimensional tomographic image). However, the principle for acquiring three-dimensional OCT data can be changed. For example, three-dimensional OCT data may be acquired using the principle of line-field OCT (hereinafter referred to as "LF-OCT"). In LF-OCT, measurement light is simultaneously irradiated onto an irradiation line extending in a one-dimensional direction on tissue, and the reflected light of the measurement light and the interference light of the reference light are received by a one-dimensional light-receiving element (e.g., a line sensor) or a two-dimensional light-receiving element. The measurement light is scanned within the two-dimensional measurement region in a direction intersecting the irradiation line to acquire three-dimensional OCT data. Alternatively, three-dimensional OCT data may be acquired using the principle of full-field OCT (hereinafter referred to as "FF-OCT"). In FF-OCT, measurement light is irradiated onto a two-dimensional measurement region on tissue, and the reflected light of the measurement light and the interference light of the reference light are received by a two-dimensional light-receiving element. In this case, the OCT device 1 does not need to include the scanning unit 14.

[0035] The front observation optical system 23 is provided to capture a front observation image of the biological tissue of the subject (in this embodiment, the fundus of the subject's eye E) in real time. The front observation image in this embodiment is a two-dimensional image of the tissue when viewed from a direction along the optical axis of the OCT measurement light (front direction). In this embodiment, a scanning laser ophthalmoscope (SLO) is used as the front observation optical system 23. However, the front observation optical system 23 may be configured with a configuration other than an SLO (for example, an infrared camera that captures a front image by irradiating a two-dimensional imaging range with infrared light all at once).

[0036] Furthermore, the OCT device 1 can acquire (generate) an Enface image, which is a two-dimensional front image of tissue viewed from the front direction along the optical axis of the measurement light, based on the acquired three-dimensional OCT data. When an Enface image is acquired in real time, the acquired Enface image can also be used as the aforementioned front observation image. In this case, the front observation optical system 23 can be omitted. Enface image data may be, for example, integrated image data in which brightness values ​​are integrated in the depth direction (Z direction) at each position in the X and Y directions, integrated values ​​of spectral data at each position in the X and Y directions, brightness data at each position in the X and Y directions at a certain depth, or brightness data at each position in the X and Y directions in any layer of the retina (e.g., the retinal surface). The OCT device 1 of this embodiment can also generate an Enface image from motion contrast data. Motion contrast data is data obtained by processing multiple OCT signals acquired at different times from the same position on biological tissue. The motion contrast data represents information on the movement of biological tissue (e.g., the movement of blood flow within blood vessels in biological tissue). In this embodiment, an enface image of a specific layer is generated based on the motion contrast data, thereby generating an angiography image (blood vessel image) that shows the positions of blood vessels contained in the specific layer. Note that the OCT device 1 of this embodiment can also generate multiple OCT images (e.g., two-dimensional enface images) with different depth regions from motion contrast data acquired for the same biological tissue.

[0037] The control unit 30 is responsible for various controls of the OCT device 1. The control unit 30 includes a CPU 31, a RAM 32, a ROM 33, and a non-volatile memory (NVM) 34. The CPU 31 is a controller that performs various controls. The RAM 32 temporarily stores various information. The ROM 33 stores programs executed by the CPU 31, various initial values, etc. The NVM 34 is a non-transitory storage medium that can retain its contents even when the power supply is cut off. An OCT image processing program for executing OCT image processing (see FIG. 2), which will be described later, may be stored in the NVM 34.

[0038] A microphone 36, a monitor 37, and an operation unit 38 are connected to the control unit 30. The microphone 36 inputs sound. The monitor 37 is an example of a display unit that displays various images. The operation unit 38 is operated by a user to input various operation instructions to the OCT device 1. The operation unit 38 may be various devices such as a mouse, a keyboard, a touch panel, or a foot switch. Note that various operation instructions may be input to the OCT device 1 by inputting sound into the microphone 36. In this case, the CPU 31 may determine the type of operation instruction by performing voice recognition processing on the input sound.

[0039] In this embodiment, an integrated OCT device 1 in which the OCT section 10 and the control unit 30 are built into a single housing is exemplified. However, it goes without saying that the OCT device 1 may include multiple devices in different housings. For example, the OCT device 1 may include an optical device that includes the OCT section 10 and a PC that is connected to the optical device via a wired or wireless connection. In this case, the control unit of the optical device and the control unit of the PC may both function as the control unit 30 of the OCT device 1.

[0040] 2 to 10, OCT image processing performed by an OCT image processing device (OCT device 1 in this embodiment) will be described. In the OCT image processing, a shallow layer image 60 and a deep layer image 70 (see FIGS. 6 and 7) are generated based on the same motion contrast data and are acquired. The deep layer image 70 is corrected so as to reduce the influence of projection artifacts that appear in the deep layer image 70 due to the signal of the shallow layer image 60. The CPU 31 of the OCT device 1 performs the OCT image processing shown in FIG. 2 in accordance with an OCT image processing program stored in the NVM 34.

[0041] The CPU 31 generates (acquires) motion contrast data of the biological tissue of the subject (in this embodiment, the fundus of the subject's eye E) (S1). First, the CPU 31 controls the front observation optical system 23 to start capturing a two-dimensional front image of the biological tissue from which an interference signal is to be acquired. In the example shown in FIG. 3, fundus blood vessels 53 and the like are captured in the two-dimensional front image 50. The two-dimensional front image 50 is repeatedly and intermittently captured and displayed on the monitor 37 as a moving image.

[0042] The CPU 31 acquires an interference signal for a measurement region 55 on biological tissue when a trigger signal for starting acquisition of the interference signal is generated. In this embodiment, the CPU 31 controls the driving of the scanning unit 14 by the driving unit 15, and causes the measurement light spot to scan within the two-dimensional measurement region 55, thereby acquiring an interference signal for the measurement region 55. As an example, in this embodiment, as shown in FIG. 3 , a plurality of linear scanning lines (scan lines) 58 for scanning the spot are set at equal intervals within the measurement region 55, and the measurement light spot is scanned on each scanning line 58, thereby acquiring an interference signal for the two-dimensional measurement region 55.

[0043] Specifically, the CPU 31 acquires at least two frames of interference signals from the same position on the biological tissue (the same scanning line 58 in the example shown in FIG. 3) at different times. In the example shown in FIG. 3, the CPU 31 first scans the first scanning line 58 of the multiple scanning lines 58 with the measurement light, thereby acquiring an interference signal detected by the light receiving element 22. Hereinafter, the direction in which the scanning lines 58 extend is referred to as the X direction. Scanning each scanning line 58 with the measurement light in the X direction once is referred to as a "B scan." A two-dimensional image generated by a B scan is referred to as a "B scan image." In the B scan image, each of multiple pixel rows extending in the direction along the optical axis of the measurement light is referred to as an "A scan image." Hereinafter, one frame of interference signal will be described as an interference signal acquired by one B scan. The Z direction is referred to as the direction along the optical axis of the measurement light. The Y direction is a direction intersecting both the X direction and the Z direction (in this embodiment, a direction perpendicular to the X direction).

[0044] When the first B scan for the first scanning line 58 is completed, the CPU 31 executes a second B scan on the first scanning line 58 to acquire a second frame of interference signals. As a result, as shown in FIG. 4, two frames of interference signals are acquired from the first scanning line 58 at different times. The CPU 31 can also acquire three or more frames of interference signals from the same position (for example, on the same scanning line 58). When it is possible to acquire multiple frames of interference signals from the same scanning line 58 at different times by scanning the scanning line 58 once with the measurement light (for example, when two measurement light beams with optical axes shifted by a predetermined interval are scanned at once), there is no need to scan the same scanning line 58 with the measurement light multiple times.

[0045] Upon completing acquisition of interference signals for multiple frames from the first scanning line 58, the CPU 31 moves the position where the B scan is performed parallel to the Y direction and executes processing for acquiring interference signals for multiple frames from the second scanning line 58. By executing the above processing for each of the multiple scanning lines 58, interference signals are acquired for the two-dimensional measurement region 55. Note that the directions of the first B scan and the second B scan on the same scanning line 58 may be reversed, or multiple B scans may be repeated in the same direction.

[0046] The CPU 31 performs a Fourier transform on the acquired interference signal to acquire a complex OCT signal. The CPU 31 performs image registration on the multiple complex OCT signals acquired at different times from the same position on the biological tissue, and corrects the phase difference between the multiple complex OCT signals. Next, the CPU 31 generates (acquires) motion contrast data based on the amount of change in the multiple complex OCT signals.

[0047] Next, the CPU 31 performs segmentation processing to detect at least one of layers and boundaries of biological tissue (layers and boundaries) from at least some of the B-scan images generated by the complex OCT signals (S2). FIG. 5 shows an example of the results of detecting multiple layers and boundaries from B-scan images. As an example, in this embodiment, a mathematical model trained by a machine learning algorithm is used. The mathematical model is pre-trained with multiple training data sets including B-scan images so that, when a B-scan image is input, it outputs a detection result of layers and boundaries shown in the input B-scan image. The CPU 31 inputs the B-scan image into the mathematical model to obtain a detection result of layers and boundaries in the B-scan image. However, other methods (e.g., methods using known image processing) can also be used for the segmentation processing. The processing in S2 is performed for each of the multiple B-scan images acquired for each of the multiple scan lines 58.

[0048] The CPU 31 generates (acquires) a shallow image 60 and a deep image 70 of the same biological tissue based on the motion contrast data acquired in S1 and the result of the segmentation process executed in S2 (S3). The shallow image 60 is an image of a first depth region (shallow region). The deep image 70 is an image of a second depth region (deep region) deeper than the first depth region.

[0049] As shown in FIG. 6 , in this embodiment, an Enface image, which is a two-dimensional front image of tissue viewed from the front direction along the optical axis of the OCT measurement light, is acquired as a shallow image 60 and a deep image 70. The Enface image data may be, for example, integrated image data in which brightness values ​​are integrated in the depth direction (Z direction) at each position in the X and Y directions, integrated values ​​of spectral data at each position in the X and Y directions, brightness data at each position in the X and Y directions at a certain depth, or brightness data at each position in the X and Y directions in any layer of the retina (e.g., the retinal surface). As described above, the motion contrast data generated in S1 contains information on the movement of biological tissue (e.g., the movement of blood flow in blood vessels in biological tissue). In this embodiment, a shallow image 60 and a deep image 70 (both Enface images) of a specific layer are acquired based on the motion contrast data. Therefore, the shallow image 60 and the deep image 70 in this embodiment are angiography images (vascular images) that show the positions of blood vessels contained in a specific layer.

[0050] The CPU 31 divides each image region of the shallow image 60 and the deep image 70 acquired in S3 into a plurality of local regions LA (S4). In the example shown in FIG. 7, each image region of the shallow image 60 and the deep image 70 is divided into 6 × 6 (36) grid-shaped local regions LA. Each local region LA has a uniform shape and size. However, the specific method for dividing the image region into a plurality of local regions can be changed. Note that each of the shallow image 60 and the deep image 70 is an image acquired from the same motion contrast data. Therefore, since the positions in the X and Y directions of the shallow image 60 and the deep image 70 are the same, the correspondence relationship between the regions in the X and Y directions of both images can be uniquely identified.

[0051] Next, the CPU 31 executes a process of correcting the deep layer image 70 so as to reduce the influence of projection artifacts appearing in the deep layer image 70 due to signals from the superficial layer image 60 (S6 to S10). Here, the algorithm of the correction process for the deep layer image 70 in this embodiment will be explained while comparing it with conventional correction processes. In reality, the blood vessels in the superficial layer region where the superficial layer image 60 is acquired are different from the blood vessels in the deep layer region where the deep layer image 70 is acquired. Therefore, ideally, the correlation between the superficial layer image 60 and the deep layer image 70 is considered to be small. However, as described above, if the superficial layer image 60 and the deep layer image 70 are acquired based on the same motion contrast data, signals derived from the movement of the tissue in the superficial layer region where the superficial layer image 60 is acquired (in this embodiment, blood flow in the blood vessels) will also appear in the deep layer image 70 as projection artifacts. In other words, the deep layer image 70 acquired in S3 is deep ”, and the shallow layer image 60 acquired in S3 is called “E shallow ”, the projection artifact is denoted by “p”, and the true depth image with the projection artifact removed is denoted by “E´ deep ", then the deep image 70 "E deep " is expressed by the following (Equation 2). E deep =E´ deep +p...(Formula 2)

[0052] The projection artifact "p" can be considered to be the shallow layer image 60 multiplied by a weight (hereinafter referred to as "correction weight") "w." Therefore, the projection artifact "p" can be expressed by the following (Equation 3). p=w×E shallow ...(Formula 3)

[0053] Therefore, if an appropriate correction weight "w" is calculated, the true deep image "E'" from which the projection artifacts have been removed can be obtained. deep " can be obtained by the following (Equation 4). E´ deep =E deep -w×E shallow ...(Formula 4)

[0054] Ideally, it is considered that the correlation between the shallow layer image 60 and the deep layer image 70 is small. Therefore, in the above-mentioned Patent Document 1, a weight that reduces the correlation between the shallow layer image 60 and the deep layer image 70 is calculated as a correction weight for correcting the deep layer image 70 by the following (Equation 5). w=Cov(E shallow ,E deep ) / Var(E shallow )...(Formula 5)

[0055] Here, a method for calculating the correction weight "w" in this embodiment will be described. Figures 8 and 9 are examples of co-occurrence histograms (sometimes called "inter-image co-occurrence histograms") in which the brightness values ​​of pixels at the same position in the shallow image 60 and the deep image 70 are plotted. In other words, the co-occurrence histograms shown in Figures 8 and 9 are obtained by calculating the occurrence frequency of pairs of brightness values ​​of pixels at the same coordinates in the shallow image 60 and the deep image 70 over the entire image area. In the examples shown in Figures 8 and 9, the brightness of each pixel on the co-occurrence histogram indicates the occurrence frequency.

[0056] In the example shown in FIG. 8, regions S and D in the co-occurrence histogram contain many pixels in parts that are common to both the superficial image 60 and the deep image 70 (mainly thick blood vessels in the superficial image 60 and parts where signals derived from blood vessels in the superficial image 60 appear in the deep image 70 as projection artifacts). Region S also contains pixels in parts that appear only in the superficial image 60 (mainly capillaries, etc.). Region D contains pixels in parts that appear only in the deep image 70 (mainly neovascularization, etc.). The information on pixels contained in regions S and D is likely to be information on pixels in parts that are unlikely to be affected by projection artifacts. In other words, the information on pixels contained in regions S and D is likely to be outliers that are unnecessary for calculating the correction weight "w" for reducing the influence of projection artifacts.

[0057] Therefore, as shown in Fig. 9, if a correction weight "w'" that simply reduces the correlation between the shallow image 60 and the deep image 70 is calculated, there is a high possibility that the calculated correction weight "w'" will be affected by information on pixels contained in only one of the region S and the region D as outliers. In fact, in the example shown in Fig. 9, the correction weight "w'" differs from the ideal correction weight "w" due to the influence of a large amount of pixel information in the portion that is only captured in the shallow image 60.

[0058] In contrast, the OCT device 1 of this embodiment calculates the correction weight "w" (see FIG. 9) by a robust estimation method using the shallow image weight "m" calculated from the shallow image 60. Projection artifacts appearing in the deep image 70 are considered to be the reflection of the shallow image 60 onto the deep image 70. Therefore, by calculating the correction weight "w" by a robust estimation method using the weight of the shallow image 60 itself (shallow image weight "m"), the correction weight "m" is calculated with the influence of outliers more appropriately excluded. As a result, the influence of artifacts occurring in the deep image 70 is more appropriately reduced.

[0059] As an example, the OCT device 1 of this embodiment calculates the correction weight by weighted regression analysis (e.g., weighted least squares method) using the shallow layer image weight "m." According to weighted regression analysis, by setting a weight for each piece of data, it is possible to suppress the influence of low reliability data (data that is likely to be an outlier) on the regression results. Therefore, by using weighted regression analysis, the correction weight "w" is calculated in a state where the influence of outliers is more appropriately excluded.

[0060] More specifically, in the OCT device 1 of this embodiment, when "m" is the shallow layer image weight, "x" is the brightness of the shallow layer image 60, and "y" is the brightness of the deep layer image 70, the correction weight "w" is calculated by the following (Equation 1). In this embodiment, the shallow layer image 60 is normalized so that the minimum brightness is 0 and the maximum brightness is 1, and then the result is squared and used as the shallow layer image weight m, thereby making the brightness of the background almost 0 and making it easier to retain information of strong signals.

number

[0061] Returning to the explanation of Fig. 2, the correction process (S6 to S10) of the deep layer image 70 in this embodiment will be described in detail. First, the CPU 31 identifies one of the local regions LA that has not yet been processed among the multiple local regions LA divided in S4 (S6). The CPU 31 calculates a shallow layer image weight "m" from the image information of the local region LA identified in S6 among the image regions of the shallow layer image 60 acquired in S3 (S7).

[0062] Next, the CPU 31 calculates a correction weight "w" for correcting the deep image 70 of the local region LA identified in S6 by a robust estimation method (in this embodiment, a weighted least squares method, which is one type of weighted regression analysis) using the shallow image weight "m" calculated in S7 (S8). As described above, in this embodiment, the correction weight "w" is calculated by (Equation 1). The CPU 31 corrects the image of the local region LA identified in S6, among the deep images 70 acquired in S3, according to the correction weight "w" calculated in S8 (S9). If correction has not been completed for all of the multiple local regions LA (S10: NO), the process returns to S6, and correction processing is performed for the next local region LA (S6 to S10). If correction has been completed for all of the multiple local regions LA (S10: YES), the process ends.

[0063] FIG. 10 is a diagram comparing an example of the results of a conventional correction process (correction process according to a correction weight “w′” that simply reduces the correlation between the shallow image 60 and the deep image 70) with the correction process of this embodiment (correction process according to a correction weight “w” by a robust estimation method using a shallow image weight “m”). The two corrected images 80 and 80′ shown in FIG. 10 are both generated by processing the same deep image 70 according to the correction weights. However, in the corrected image 80′ generated by the conventional correction process (correction process according to the correction weight “w′”), signals due to blood vessels present in the superficial image 60 still remain as projection artifacts. This is thought to be due to the influence of outliers, which prevent the correction weight “w′” from being properly calculated. In contrast, in the corrected image 80 generated by the correction process of this embodiment (correction process according to the correction weight “w”), it can be seen that signals due to blood vessels present in the superficial image 60 are more appropriately removed than in the corrected image 80′. As described above, according to the processing of this embodiment, the influence of the projection artifacts occurring in the deep layer image 70 can be more appropriately reduced.

[0064] In this embodiment, the CPU 31 executes the following steps for each of the local regions LA: calculating a shallow image weight "m" (S7); calculating a correction weight "w" (S8); and correcting the deep image 70 according to the correction weight "w" (S9). Therefore, if the influence of the projection artifact in a certain local region LA is large, the correlation between the shallow image 60 and the deep image 70 increases, and the correction weight "w" increases. On the other hand, if the influence of the projection artifact in a certain local region LA is small, the correlation between the shallow image 60 and the deep image 70 increases, and the correction weight "w" decreases. Furthermore, since the shallow image weight "m" is calculated for each local region LA, the influence of outliers is also appropriately reduced according to the local region LA. Therefore, by performing the process for each local region LA, the influence of artifacts occurring in the deep image 70 can be more appropriately reduced.

[0065] The techniques disclosed in the above embodiments are merely examples. Therefore, the techniques exemplified in the above embodiments can be modified. It is also possible to execute only some of the techniques exemplified in the above embodiments. For example, the process of calculating the shallow image weight "m" (S7), the process of calculating the correction weight "w" (S8), and the process of correcting the deep image 70 according to the correction weight "w" (S9) may be performed for the entire image area rather than for each local area LA. Even in this case, the influence of outliers can be easily appropriately eliminated, and the influence of artifacts occurring in the deep image 70 can be easily appropriately reduced.

[0066] The process of acquiring the shallow image 60 and the deep image 70 in S3 of Fig. 2 is an example of an "image acquisition step." The process of calculating the correction weight "w" in S7 and S8 is an example of a "correction weight calculation step." The process of correcting the deep image 70 according to the correction weight "w" in S9 is an example of an "image correction step." [Explanation of symbols]

[0067] 1. OCT device (OCT image processing device) 31 CPU 34 NVM 60 Shallow layer image 70 Deep Images 80 Corrected image LA local area

Claims

1. An OCT image processing program executed by an OCT image processing device that processes an image acquired by the OCT device, the OCT device is capable of generating motion contrast data by processing multiple OCT signals acquired from the same position on biological tissue at different times; The OCT image processing program is executed by a control unit of the OCT image processing device, an image acquisition step of acquiring a shallow image, which is an image of a superficial region of a biological tissue, and a deep image, which is an image of a deep region deeper than the superficial region, generated based on the motion contrast data; a correction weight calculation step of calculating a correction weight for correcting the deep image so that the correlation between the shallow image and the deep image is reduced; an image correction step of correcting the deep image according to the correction weight; is performed by the OCT imaging device, The OCT image processing program is characterized in that, in the correction weight calculation step, the correction weight is calculated by a robust estimation method.

2. 2. The OCT image processing program according to claim 1, The OCT image processing program is characterized in that, in the correction weight calculation step, the correction weight is calculated by a robust estimation method using shallow layer image weights calculated from the shallow layer images.

3. 3. The OCT image processing program according to claim 1, The OCT image processing program is characterized in that, in the correction weight calculation step, the correction weight is calculated by weighted regression analysis as the robust estimation method.

4. 4. The OCT image processing program according to claim 3, In the correction weight calculation step, when m is the shallow layer image weight, x is the brightness of the shallow layer image, and y is the brightness of the deep layer image, the correction weight w is calculated by (Equation 1). An OCT image processing program characterized by the above. [Equation 3]

5. 5. The OCT image processing program according to claim 1, In the correction weight calculation step, the calculation process of the correction weight is performed for each local region, In the image correction step, the deep layer image is corrected for each local region according to the correction weight calculated for each local region.

6. An OCT image processing device that processes an image acquired by an OCT device, the OCT device is capable of generating motion contrast data by processing multiple OCT signals acquired from the same position on biological tissue at different times; The control unit of the OCT image processing device an image acquisition step of acquiring a shallow image, which is an image of a superficial region of a biological tissue, and a deep image, which is an image of a deep region deeper than the superficial region, generated based on the motion contrast data; a correction weight calculation step of calculating a correction weight for correcting the deep image so that the correlation between the shallow image and the deep image is reduced; an image correction step of correcting the deep image according to the correction weight; Run The OCT image processing apparatus is characterized in that, in the correction weight calculation step, the correction weight is calculated by a robust estimation method.

Images