Method and apparatus for examining fluorescence images of mammalian tissue, and medical treatment device.

Abstract

This specification discloses an improved inspection method for inspecting a fluorescence image captured from mammalian tissue that has been made photosensitive with a fluorescent agent and irradiated with excitation light. The fluorescence image (FI) includes an array of pixels each having a fluorescence signal value, and the fluorescence signal values include a first number (N1) of fluorescence signal values representing mammalian tissue in the fluorescence image. The improved inspection method determine a reference value (IR) such that a second number (N2) of the fluorescence signal values included in the first number of fluorescence signal values is less than or equal to the reference value, and the remainder of the first number of fluorescence signal values exceeds the reference value, and the second number is a predetermined fraction of the first number (S6); the average fluorescence signal value ( R ) and standard deviation ( R )を決定する; In the fluorescence image (FI), perform image segmentation to distinguish a target region (TR) showing a part of the mammalian tissue (MT) identified as tumor tissue from a reference region (RR) showing the remaining part of the mammalian tissue (MT) (S8). This segmentation includes, for each pixel (p), determining that the pixel is part of the target region (TR) when the contrast-to-noise ratio CNR(p) of the pixel exceeds a predetermined threshold (T CNR ), and determining that the pixel is part of the reference region (RR) otherwise, where the contrast-to-noise ratio CNR(p) of the pixel is CNR(p) = (FI(p) - μ_R) / (c?σ_R), where FI(p) is the fluorescence signal value of pixel (p) and c is a predetermined constant. Identify at least one contour (B) of the target region (TR) (S9). See Figure 1

Description

[Background technology]

[0001] The present invention relates to an image processing method. This invention relates to an image processing apparatus. The present invention further relates to a medical treatment device including an inspection device.

[0002] Treatment for most solid tumors consists of radical surgical removal of all tumor tissue. However, distinguishing between normal and tumor tissue during surgery remains difficult. Therefore, pathological evaluation 2–5 days post-surgery is often performed, and tumor-positive margins are frequently found. According to the literature, the percentage of tumor-positive margins (TPMs) ranges from 10% to 35%, depending on the type of tumor. For example, Orosco, RK et al. Positive surgical margins in the 10 most common solid cancers. Sci. Rep. 8, 56-86 (2018). https: / / doi.org / 10.1038 / s41598-018-23403-5 Please refer to the following. If tumor tissue is present at or near the edge of the resected tissue, the risk of local recurrence and distant metastasis increases, and survival rates decrease. As a result, TPM requires additional treatment such as reoperation, radiation therapy, and systemic therapy. However, this comes with an increase in complications and a greater psychological burden on the patient. Therefore, it is extremely important to be able to accurately observe tumor tissue during surgery, but current optical techniques and the surgeon's visual and tactile information alone are insufficient to properly determine the tumor boundary. Therefore, new technologies that allow real-time observation of tumors are being researched with the aim of reducing the number of TPMs and lowering the need for additional treatment and complications.

[0003] One imaging technique that is attracting attention is fluorescence molecular imaging (FMI). This is because it allows for real-time visualization of tumors both inside the patient's body (i.e., in vivo) and immediately after resection (outside the body). The tissue to be examined is prepared with a fluorescent agent (FA) by administering the FA to the patient or by impregnating the tissue with the fluorescent agent. FAs are non-targeted fluorescent dyes such as indocyanine green (ICG), or targeted fluorescent dyes used to image tumor tissue or infection and track drug therapy. While FMI studies in the near-infrared (NIR) spectral range (700-900 nm) have shown promising results, complex problems remain. For example, light scattering and absorption by biological components such as water and blood contribute to the attenuation of excitation light, thereby reducing the sensitivity and contrast of the fluorescence image. These factors can contribute to the appearance of false-positive TPMs in fluorescence images, even when TPMs are not present in the patient. Therefore, the resulting fluorescence images (FI) are not always directly suitable for guidance to surgeons and other healthcare professionals, or for use in medical devices.

[0004] In "A Guideline for Clinicians Performing Clinical Studies with Fluorescence Imaging" (J. Nucl. Med., 63 (2022) 640) by W. Heeman et al., a method for determining the CNR ratio of tissue samples after examination by a pathologist is described. Here, CNR refers to the contrast-to-noise ratio calculated by the pathologist for the entire target region. [Overview of the Initiative]

[0005] According to the first objective, an improved examination method for examining fluorescent images of mammalian tissue is provided to facilitate the performance of medical procedures by specialists or surgical equipment. According to the second objective, an improved examination device is provided for examining fluorescent images of mammalian tissue to facilitate the performance of medical procedures by specialists or surgical equipment. According to the third objective, an improved medical treatment device, including an improved testing device, is provided.

[0006] The fluorescence image to be examined consists of an array of pixels, each having a fluorescence signal value, and is obtained through the preliminary steps defined below. The fluorescence signal values ​​include a first number of fluorescence signal values ​​representing mammalian tissue in the fluorescence image. The improved examination method includes subsequent steps for processing the fluorescence image.

[0007] In the preliminary step, mammalian tissue (e.g., human tissue) is made photosensitive with a fluorescent agent, then irradiated with excitation light, and a fluorescence image is obtained from the irradiated photosensitive tissue. Fluorescent agents are used to visualize different types of tissue, such as tumor tissue and non-tumor tissue. In one example, the fluorescent agent is a targeted fluorescence tracer such as cetuximab-IRDye800CW or Hexbix, used to image tumor tissue and / or infection and track drug therapy. In another example, the fluorescent agent is a non-targeted fluorescent dye such as indocyanine green (ICG) for imaging tissue perfusion. The excitation light irradiated onto the tissue in vivo or in vitro is typically in the infrared region. Fluorescent agents can be administered to the patient or used to penetrate tissue.

[0008] Subsequent steps in the examination method include, among other things, performing image segmentation to distinguish, in fluorescence imaging, a target region showing a portion of mammalian tissue identified as tumor tissue from a reference region showing a portion of mammalian tissue identified as healthy tissue.

[0009] Before performing this image segmentation, a reference value is determined such that a second fluorescence signal value included in the first number of fluorescence signal values is less than or equal to the reference value, and the remainder of the first number of fluorescence signal values exceeds the reference value. The second number is a predetermined ratio of the first number. In one example, a histogram of the fluorescence signal values included in the first number of fluorescence signal values is obtained, and the reference value is the value of the predetermined k-th q-quantile of the histogram. The predetermined k-th q-quantile is the expected tumor tissue area ratio r of the tumor tissue area to the total tissue area in the fluorescence image (FI) eに従って選択されるべきである . The ratio k / q must not exceed that tumor tissue area ratio, and the ratio k / q must not be too small. For example, ?0.1*r?_e?k / q?0.95*r_e

[0010] This improved method is used, for example, in the in vitro examination of mammalian tissues excised by a specialist to remove tumor tissues. Considering that the mammalian tissues excised by a specialist often contain a considerable amount of normal tissues, as tissues that are easily identified as tumors are surrounded by tissues that may appear normal at first glance but may develop into tumor tissues later, and it is sometimes practically impossible to accurately track the boundary of the tumor tissue during excision.

[0011] Therefore, in such cases, the ratio of k / q is often selected as follows. 0.5?k / q?0.9

[0012] The average fluorescence signal value F B and the standard deviation S are calculated. By selecting the k / q ratio not to be too small, that is, at least 0.1*r e 、好ましくは0.2に選択されることにより、正常組織の平均蛍光信号値および蛍光信号の標準偏差を信頼性高く推定するのに十分な量のデータが得られる。k / q比が大きすぎないように、すなわち最大でも0.95*r e 、好ましくは0.9 , the estimation of this statistical data being affected by the fluorescence signal data of the tumor tissue is avoided.

[0013] The average fluorescence signal value μ R and the standard deviation? Rを用いてImage segmentation is performed, and in the fluorescence image (FI), the target region (TR) and the reference region (RR) are distinguished. The target region (TR) is the region in the fluorescence image (FI) corresponding to a part of the mammalian tissue (MT) identified as tumor tissue in this segmentation. The reference region (RR) is the region in the fluorescence image (FI) corresponding to the remaining part of the mammalian tissue. In this segmentation operation, the contrast-to-noise ratio CNR(p) is determined for each pixel (p). Pixels for which the contrast-to-noise ratio CNR(p) exceeds a predetermined threshold are classified as part of the target region (TR), and otherwise are classified as part of the reference region (RR). The contrast-to-noise ratio CNR(p) of a pixel is defined as follows. CNR(p)=(FI(p)-μ_R) / (c?σ_R ), where FI(p) is the fluorescence signal value of the pixel (p), and c is a predetermined constant. Thus, in contrast to the method known from Heeman, the CNR ratio is calculated on a pixel-by-pixel basis and there is no prior knowledge provided by a pathologist. In an embodiment where the predetermined threshold is 1 and the predetermined constant is 2, the best results are obtained. However, it can be changed according to further requirements. For example, if it is necessary to consider a safety margin when determining the target region, a predetermined threshold less than 1 and / or a predetermined constant less than 2 can be selected. In that case, the estimated target region (TR) will become larger and will show not only the part identified as tumor tissue within the mammalian tissue (MT), but also the part of the mammalian tissue at risk of becoming tumor tissue surrounding the part identified as tumor tissue.

[0014] This method further includes identifying the contour of a target region. The contour can be displayed on a screen, for example, superimposed on a fluorescence image, assisting medical professionals in intervening in selected tissue areas corresponding to the target region in the fluorescence image. Interventions may include, for example, treating the selected tissue area with therapeutic radiation, supplying therapeutic drugs to the selected tissue area displayed in the target region, selectively activating therapeutic drugs in the selected tissue area, or excising the selected tissue area. Instead of superimposing the contour on a fluorescence image, it is also possible to superimpose the contour on a natural image of the tissue, i.e., an image that appears to have been taken under ambient light conditions. This allows medical professionals to monitor the tissue while performing medical interventions, as if they were directly viewing the tissue under ambient light conditions, rather than monitoring the fluorescence response of the image. In yet another example, the contour is projected onto the tissue.

[0015] In some embodiments of this method, the background is further acquired from the fluorescence image, and the method provides preliminary image segmentation for distinguishing between a foreground region representing mammalian tissue and a background region representing the background in the fluorescence image. In one example, the background is formed by a carrier surface on which mammalian tissue (e.g., completely excised tissue or a section thereof) is placed for in vitro examination. In another example, the fluorescence image is acquired in vivo with the background placed in the camera's field of view as a reference. In preliminary image segmentation, each pixel is determined to be significantly above the average background fluorescence value. If a pixel's fluorescence value (FSV) is determined to be significantly above the average background fluorescence value, that pixel is classified as part of the foreground region; otherwise, it is classified as part of the background region. For example, if the difference between a pixel's fluorescence value (FSV) and the background fluorescence value exceeds a value obtained by multiplying the standard deviation of the background fluorescence values ​​by a predetermined coefficient, that pixel's fluorescence value (FSV) is determined to be significantly above the average background fluorescence value. The predetermined coefficient is selected, for example, in the range of 1 to 5, for example, approximately 2.

[0016] In one example, the mean fluorescence signal value and standard deviation are estimated in the calibration step, and a calibration fluorescence image of only the background is captured before capturing the fluorescence image, and the mean fluorescence signal value and standard deviation of the fluorescence signal values ​​in the calibration fluorescence image are determined. A rough but useful estimate of the average fluorescence signal value can be obtained as follows: μ_est=min??+ max? / 2;σ_est=(max - min) / (2√3)

[0017] T min and max are the minimum and maximum fluorescence values, respectively, among all fluorescence values ​​in the fluorescence image.

[0018] In another example, the mean fluorescence signal value and the standard deviation of the fluorescence signal values ​​are calculated from a designated region (BP) representing the background of a fluorescence image (FI) (S2A). For example, the operator can specify a rectangular region in the fluorescence image (FI) that represents part of the background. Based on the mean fluorescence signal value and the standard deviation of the fluorescence signal values ​​in this region, complete preliminary image segmentation can be performed.

[0019] After preliminary image segmentation, the risk of edge effects can be reduced by applying an expansion process that extends the foreground region by one pixel or more. Corrections can also be applied, such as removing regions smaller than a threshold area value from the areas identified during preliminary image segmentation. Examples include small regions initially identified as the foreground or small regions initially identified as the background. Typically, the largest initially identified foreground region is selected for subsequent processing, while smaller regions are considered part of the background.

[0020] If preliminary image segmentation is applied, subsequent image segmentation to determine one or more target regions is applied to the portion of the image determined to be the foreground region. [Brief explanation of the drawing]

[0021] These and other aspects of the present invention are disclosed in more detail in the accompanying drawings. Figure 1 schematically illustrates the steps of an improved examination method for examining fluorescence images of mammalian tissue. Figure 2 shows the optional steps of the improved inspection method. Figures 3A and 3B show further optional steps of the improved inspection method. Figures 4A and 4B show the segmentation steps of the improved inspection method. Figures 5A, 5B, and 5C illustrate the application of this method to fluorescence images obtained from sample tissue. Figures 6A, 6B, and 6C show the application of this method to fluorescence images obtained from further sample tissues. Figure 7 schematically illustrates the steps of another improved examination method for examining fluorescence images of mammalian tissue. Figures 8-10 show the embodiment of Figure 7 applied to fluorescence images of sample tissue. Figures 11-14 illustrate the application of the embodiment of Figure 7 using heuristic information from the method of claim 1. Figure 15 schematically shows an inspection device according to one embodiment of the present invention. Figure 16 schematically shows a medical treatment device according to one embodiment of the present invention. Figures 17A and 17B show the tray containing the sample and its fluorescence image, respectively, as captured in visible light. Figures 18A–18F show the image of Figure 17B divided according to different quartiles. Figures 19A, 19B, and 19C illustrate further approaches for detecting the boundary between normal and diseased tissue. [Modes for carrying out the invention]

[0022] Figure 1 schematically shows steps S5-S9 of an improved examination method for examining fluorescence images obtained from mammalian tissue. The fluorescence image is obtained through the following preparation steps S1-S4.

[0023] Next, mammalian tissues are made photosensitive with fluorescent agents. Fluorescent agents are used to visualize various types of tissues, including tumor tissue and non-tumor tissue. For example, fluorescent agents include targeted fluorescent tracers such as cetuximab-IRDye800CW, or hexylaminolevulinic acid (abbreviated as Hexilvix), which is used for imaging tumor tissue and infections, and for tracking drug therapy. Another example is non-targeted fluorescent dyes such as indocyanine green (ICG), which is used for imaging tissue perfusion. The excitation light irradiated onto tissues in vivo or in vitro is typically in the infrared region. Fluorescent agents can be administered to patients or used to penetrate tissues.

[0024] In preparation step S3, excitation light is shone on mammalian tissue that has been made photosensitive with a fluorescent agent, and in preparation step S4, a fluorescence image of the mammalian tissue is acquired. The fluorescence image includes an array of pixels, each having a fluorescence signal value. The fluorescence signal value includes the number of first fluorescence signal values ​​representing the mammalian tissue in the fluorescence image. The first number is the number of pixels in the fluorescence image (FI) when the image shows only mammalian tissue, but the first number may be less if, for example, the image also has a background.

[0025] In step S6 of the improved testing method, a reference value is determined such that the second fluorescence signal value included in the first fluorescence signal value is less than or equal to a reference value, and the remaining number of first fluorescence signal values ​​exceeds a reference value. The second number is a predetermined proportion of the first number.

[0026] In step S7, what is the average fluorescence signal value of the fluorescence signal values ​​included in the second number of fluorescence signal values? R And the standard deviation? R This will be decided.

[0027] In step S8, image segmentation is performed in the fluorescence image FI to distinguish between the target region TR and the reference region RR, which represents the rest of the mammalian tissue MT. This step is performed pixel by pixel. For each pixel (p), the contrast-to-noise ratio (CNR) (p) of that pixel is set to a predetermined threshold (T CNR If the value exceeds ), the pixel is determined to be part of the target region TR; otherwise, it is determined to be part of the reference region RR. Here, the contrast-to-noise ratio CNR(p) of a pixel is defined as follows: CNR(p)=(FI(p)-μ_R) / (c?σ_R ),

[0028] Here, FI(p) is the fluorescence signal value of pixel(p), and c is a predetermined constant. The optimal values ​​are a threshold of 1 and a constant c of 2.

[0029] In step S9, at least one contour B of the target region TR is identified. In one example, at least one contour includes a primary contour B that indicates the boundary between the target region and the reference region. In another example, at least one contour includes a secondary contour that extends outward by a certain distance from the boundary between the target region and the reference region, extending the target region in a safety region. This reduces the risk that the normal-looking tissue near the target region will later grow into tumor tissue. In a specific embodiment, the secondary contour B' extends outward by a certain distance from the boundary so as to avoid crossing with a designated anatomical structure. In one example, both the primary and secondary contours are identified.

[0030] An embodiment of the improved method is described with reference to Figure 2. In this example, it is applied to in vitro examination of mammalian tissue, where the mammalian tissue is placed on a background. In this example, the mean fluorescence signal value and the standard deviation of the fluorescence signal value are determined from a portion BP designated as representing the background in the fluorescence image FI. The operator can easily specify a rectangular or square portion BP in the image that does not represent mammalian tissue using the user interface. Next, the statistical characteristics of the image data of this portion are determined and used to perform preliminary segmentation as shown in step S5 of Figure 1. Preliminary segmentation divides the fluorescence image (FI) into foreground and background regions based on the statistical characteristics of the background region estimated from the specified portion BP. Typically, the estimated statistical characteristics include the mean fluorescence signal value and the standard deviation of the fluorescence signal value of the background. These statistical characteristics can be efficiently estimated from the minimum and maximum fluorescence values ​​identified in the specified portion BP as follows: μ_est=min??+ max? / 2;σ_est=(max - min) / (2√3)

[0031] In preliminary image segmentation S5, it is determined for each pixel whether its fluorescence signal value (FSV) significantly exceeds the average background fluorescence signal value. If a pixel's fluorescence signal value (FSV) is determined to significantly exceed the average background fluorescence signal value, that pixel is classified as part of the foreground region; otherwise, it is classified as part of the background region. For example, if the difference between a pixel's fluorescence signal value (FSV) and the background fluorescence signal value exceeds a predetermined coefficient multiple of the standard deviation of the background fluorescence signal value, that pixel's fluorescence signal value (FSV) is determined to significantly exceed the average background fluorescence signal value. The predetermined coefficient is selected, for example, in the range of 1 to 5, for example, approximately 2. In one example, the mean fluorescence signal value and standard deviation are estimated in the calibration step, and a calibration fluorescence image of only the background is captured before capturing the fluorescence image, and the mean fluorescence signal value and standard deviation of the fluorescence signal values ​​in the calibration fluorescence image are determined. A rough but useful estimate of the average fluorescence signal value can be obtained as follows: μ_est = (min + max) / 2; σ_est = (max - min) / (2√3) Here, min and max are the minimum fluorescence value and the maximum fluorescence value among all the fluorescence values in the fluorescence image, respectively. In the example shown in Figure 2, the following statistical characteristics of the background were estimated. μ est ≈ 5.18; σ 推定値 ≈ 1.59

[0032] Based on this estimation, pixels with a fluorescence value of at least 8.36 were identified as the foreground FG (i.e., representing mammalian tissue), and pixels with lower fluorescence values were identified as the background BG.

[0033] Figure 3A schematically shows how the fluorescence image (FI) is divided into the background region BG and the foreground region FG by the boundary B. Figure 3B shows the state after further correcting the boundary to the corrected boundary B'. This correction is achieved by expanding the foreground FG by 1 pixel. The fluorescence signal values in the fluorescence image (FI) are composed of the first number N1 of fluorescence signal values in the foreground. These represent mammalian tissue in the fluorescence image. Figure 4A shows how the reference value IR is determined such that the second number N2 of fluorescence signal values included in the first N1 fluorescence signal values is below the reference value I R and the remaining fluorescence signal values among the first N2 fluorescence signal values exceed the reference value IR. The second fluorescence signal value N2は is a predetermined ratio of the first N1 fluorescence signal values. In this example, the predetermined ratio is 0.5. This means that the reference value IR is the median of the fluorescence signal values included in the first N1 fluorescence signal values. Further analysis reveals that the fluorescence signal values included in the second N2 fluorescence signal values have the following statistical characteristics. μR ? ≈ 37.54; σ R ≈ 12.91 Here, μ RR and σ RR r represents the estimated mean and estimated standard deviation of the fluorescence signal values ​​included in the fluorescence signal values ​​of the second number N2, respectively. Based on these statistical characteristics, image segmentation is performed to distinguish between the target region TR and the reference region RR in the fluorescence image. Here, the reference region refers to the remaining portion of the mammalian tissue other than the target region. In Figure 4B, the contrast-to-noise ratio CNR (p) of the pixel reaches a predetermined threshold (T CNR Pixels exceeding ) are identified as part of the target region TR, shown in black in Figure 4B. The remaining pixels, shown in white, are identified as part of the reference region RR. The contrast-to-noise ratio CNR (p) of a pixel is defined as follows: CNR(p)=(FI(p)-μ_R) / (c?σ_R ),

[0034] TFI(p) is the fluorescence signal value of the pixel (p), and c is a predetermined constant. In this example, the value of c is 2, and the predetermined threshold (T CNR The value of ) is 1.

[0035] Figures 5A, 5B, 5C, and 6A, 6B, and 6C show various examples of the application of the above method. The examples shown in Figures 5A, 5B, and 5C are sections of penile squamous cell carcinoma tissue excised from the first patient. The examples shown in Figures 6A, 6B, and 6C are sections of penile squamous cell carcinoma tissue excised from the second patient. Fluorescence images were acquired using a PEARL imaging device after the tissue was fluorescently treated with cetuximab-IRDye800CW. The captured wavelength range is 800 nm. In the figures, reference B' indicates the corrected boundary of the mammalian tissue relative to the background. References C, C1, and C2 show the contours of the target region within the tissue estimated by this method. Ground truth, i.e., the contour of the tumor tissue determined by the pathologist, is indicated by reference GT.

[0036] The above method can be used in combination with the scanline-based method. An example of the scan trajectory-based method is schematically shown in Figure 7.

[0037] The inspection method described therein, in step S10, obtains at least one fluorescence signal value vector of fluorescence signal values ​​along the scanning trajectory in the fluorescence image. In practice, the scanning trajectory is typically a straight line that coincides with the principal axis of the fluorescence image.

[0038] In step S11, one or more candidate scan trajectory segments (typically scan line segments) are determined for a given threshold ts, where the fluorescence signal value exceeds the threshold ts. This process is repeated for multiple thresholds. In step S12, it is verified whether this procedure has been performed for all of the multiple thresholds.

[0039] The remaining sections of the scan path are considered reference sections, and the statistical characteristics of the fluorescence signal values ​​of the pixels that make up that section are derived. Typically, the mean value μ of these fluorescence signal values ​​is used. rs And standard deviation?? rsが These are determined. The values ​​of these statistical properties depend on the choice of threshold ts, so μ rs (ts) and ?? rs (ts) It is written as follows.

[0040] After determining these statistical characteristics, we determine which pixels (p) on the scan trajectory satisfy the requirements of CNR(p)?t. CNR(p)=(FI(p)-μ_rs (ts)) / (c?σ_rs (ts) ),

[0041] Here, c is a predetermined constant selected from the range of 1.5 to 3, for example, about 2, and tr is a threshold, for example, a value of 1. In step S13, it is determined which of the multiple thresholds ts best matches each set of candidate scan trajectory intervals with one or more sets of scan trajectory intervals obtained using the CNR requirements described above. In one example, the degree of match of the identified scan trajectory intervals is determined by the F-value described above.

[0042] Alternatively, the optimal value t optはThe following is determined: t_opt=■(argmin@ts)|ts-(μ_rs (ts)+2*σ_rs (ts))|

[0043] In the embodiment shown in Figure 7, steps S10 to S13 are repeated for multiple scan trajectories that are different from each other in the scan trajectory set. Therefore, in step S14, it is verified whether steps S10 to S13 have been performed for all scan trajectories in the scan trajectory set. Thus, for each scan trajectory in the scan trajectory set, the value t optが決定され、値t opt Based on this, the scan trajectory is divided into scan trajectory sections. As mentioned above, the scan trajectory is preferably a line that coincides with the principal axis of the fluorescence image. For example, it could be the set of all horizontal lines, or a subset of horizontal lines.

[0044] In the illustrated embodiment, steps S10 to S14 are repeated for multiple scan trajectory sets. Therefore, in step S15, it is verified whether steps S10 to S14 have been performed for multiple scan trajectory sets, for example, a set of horizontal scan lines and a set of vertical scan lines.

[0045] The method in Figure 7 can be used in various combinations with the method in Figure 1. For example, the method in Figure 1 can be used to find the optimal threshold t. optが予想される This indicates a range. For example, the range can be selected as follows: μ_r? t_opt?μ_r+4*σ_r or μ_r+σ_r? t_opt?μ_r+3*σ_r

[0046] As another example, the method in Figure 1 provides an extension range for limiting the extension range of the scan trajectory. This extension range can, for example, represent a line segment that crosses the contour line shown by the method in Figure 1 and extends a predetermined distance on both sides of the intersection. Alternatively, the method in Figure 1 can represent an extension range of the scan trajectory that passes through the target area and extends a predetermined distance on both sides of the target area.

[0047] As yet another example, the method in Figure 1 provides information indicating the direction of the contour of the target area to be identified. This information is, for example, the principal axis of the contour identified in the method in Figure 1. Using this information, the method in Figure 1 is optimally executed by performing analysis based on the scan trajectory using scan lines transverse to the principal direction.

[0048] A proper definition of a principal axis is the line segment that minimizes the average distance between that segment and a point on the contour line. Euclidean distance is an example of a distance measure, but other methods are also available. Principal axes can be found using the OpenCV tool "fitLine".

[0049] In practice, the fluorescence image (FI) is rotated before performing the method shown in Figure 7. In this case, the rotation aligns the principal axis of the contour with the coordinate axes of the fluorescence image, and the scan lines are reoriented along another coordinate axis of the fluorescence image (FI).

[0050] In the method shown in Figure 7 above, by changing the threshold ts, each set of candidate scan trajectory sections of the scan trajectory corresponds to the CNR described above. 要件 A threshold t opt ​​is determined that best matches the set of one or more scan trajectory sections obtained using the method. In this case, the statistical characteristic μ rs and ?? rsは , estimated from sections of the scan trajectory that are not identified as candidate scan trajectory sections. In the alternative approach, the scan trajectory is estimated to have intersections at the estimated locations, and the statistical property μ rs and ?? rsはThe estimated intersection is estimated from the scan trajectory section with the lowest average fluorescence value on the side of the estimated intersection. Based on the CNR index with these estimated statistical characteristics, the scan trajectory is divided into a reference scan trajectory section and a target scan trajectory section, and it is determined whether the estimated intersection coincides with the transition from the reference scan trajectory section to the target scan trajectory section, or vice versa. If they coincide, the estimated intersection becomes a candidate contour point. This approach is described in more detail in European Patent Application No. 23154545.0 filed by the same applicant on February 1, 2023. The method in Figure 1 can provide contour locations as heuristic information to limit the search range for the optimal intersection location in an alternative approach to the method in Figure 7.

[0051] The method shown in Figure 7 will be explained with reference to Figure 8. The upper part of Figure 8 shows a fluorescence image (FI) obtained from a mammalian tissue sample. The lower part shows the intensity values ​​along the scan trajectory L in the fluorescence image (FI). In the illustrated example, each intensity value on the scan line is obtained as a weighted average of the Gaussian distributions of pixels with the same x-coordinate in the strip W.

[0052] Figure 9 shows how, in step S11, for each of the multiple thresholds, the set of candidate scan trajectory segments in which the fluorescence signal value exceeds the threshold is determined. The top of Figure 9 shows a first example, where threshold 50 divides the scan into two candidate target segments T50a and T50b and two reference segments R50a and R50b. The bottom of Figure 9 shows how threshold 58 divides the scan into one candidate target segment T58 and one reference segment R58.

[0053] Statistical characteristics μ from fluorescence signal values ​​in one or more reference sections of the scan trajectory rs and ?? rsが推定され、これらの統計特性に基づいて、 Alternative partitions of the scan trajectory are determined according to the CNR criteria. Using the statistical properties of one or more reference sections, the threshold t is determined to best match the CNR-based partition. opt It is identified.

[0054] As shown in Figure 10, in step S13 of the method in Figure 7, the optimal threshold for this scanline is found to be 62. This divides the scanline into a reference section R62 and a target section T62, as shown at the top of Figure 10. As shown at the bottom of Figure 10, this identifies contour points Ts and Te, which indicate the start and end points of the target region T62, respectively, as the scanline traverses the positive x-direction.

[0055] As another example, Figure 11 shows how to apply the method of Figure 1 to identify the first contour C1 and the second contour C2 shown at the bottom of Figure 11 in the fluorescence image shown at the top of Figure 11.

[0056] The upper part of Figure 12 shows how the principal axis AX1 of the target region with contour C1 is determined. The lower part of Figure 12 shows how the fluorescence image (FI) is rotated so that the principal axis AX1 coincides with one of the principal axes of the fluorescence image (FI) (in this case, the y-axis). Note that for contours with more complex shapes, fitting can be done using multiple straight lines. The resulting rotated fluorescence image (FI) can be optimally scanned using scan lines aligned with the X-axis. Furthermore, the position of contour line C1 can provide heuristic information that reduces the scan line-based approach shown in Figure 7. For example, the heuristic information may indicate a spatial search range or an intensity search range.

[0057] In the example shown at the top of Figure 13, the start (+) and end ( ) of the target section, identified using the threshold-based version of the method in Figure 7 with heuristic information from the method in Figure 1, and aligned to the first contour C1 shown in Figure 11. In the example shown at the bottom of Figure 13, the start (+) and end ( ) of the target section, identified using the threshold-based version of the method in Figure 7 with heuristic information from the method in Figure 1, and aligned to the second contour C2 shown in Figure 11.

[0058] At the top of Figure 14, the identified points for each contour mapped to the original image are displayed.

[0059] The lower part of Figure 14 shows how isolated points are removed in the subsequent steps. A point is considered isolated if its local point density is below a threshold; that is, if the ratio of the number of points to the size of a given area centered on the point is less than a given value. Typically, the size of the area is selected in the range of 100 to 1000 pixels, and the minimum number of points that must be present within the area is selected in the range of 5 to 50. The best results are obtained when the number of points is in the range of 1 / 30 to 1 / 20 of the size of the area. In this example, the radius of the area is 40 pixels. That is, the area of ​​the area is approximately 500 pixels, and the minimum number of points is 20.

[0060] In additional or alternative processing steps, point clouds with a maximum pixel intensity below a threshold value are excluded. This threshold value is, for example, the overall intensity mean plus a coefficient of the standard deviation, and is a configurable parameter. Furthermore, in additional or alternative processing steps, point clouds whose area (the area used for calculation) is less than a certain percentage of the total image area are also excluded.

[0061] Figure 15 schematically shows a mammalian tissue inspection device 1 configured to acquire a fluorescence image FI of mammalian tissue MT that has been photosensitive with a fluorescent agent and irradiated with excitation light. The fluorescence image FI includes an array of pixels, each having a fluorescence signal value. The fluorescence signal values ​​include a number N1 of first fluorescence signal values ​​representing the mammalian tissue in the fluorescence image. If a background is present in the image, the first number is less than the total number of pixels in the fluorescence image (FI). Otherwise, the first number N1 may be equal to the total number of pixels. In the illustrated example, the inspection device 1 acquires the fluorescence image (FI) from an external input 1i. Alternatively, the inspection device may be equipped with a camera for capturing images. The inspection device may also be equipped with a suitable excitation light source.

[0062] The inspection device 1 includes a reference value determination module 11 configured to determine the reference value IR such that the second number of fluorescence signal values ​​N2 included in the first number of fluorescence signal values ​​is less than or equal to the reference value IR, and the remaining fluorescence signal values ​​of the first number of fluorescence signal values ​​exceed the reference value IR. In one example, the second number is a predetermined fraction of the first number, for example, N2 / N1 = 0.5, in which case the reference value is the median. The statistical characterization module 12 determines the average fluorescence signal value of the fluorescence signal values ​​included in the second number of fluorescence pixel values ​​N2 (? R ) and standard deviation (? R ) will be decided.

[0063] The segmentation module 13 performs image segmentation in a fluorescence image (FI) to distinguish between a target region (TR) that represents the portion of mammalian tissue (MT) identified as tumor tissue and a reference region (RR) that represents the rest of the mammalian tissue (MT). For each pixel (p), the segmentation module determines that a pixel is part of the target region (TR) if its contrast-to-noise ratio CNR(p) exceeds a predetermined threshold (T CNR). 、そうでない場合、当該 It operates on a pixel-by-pixel basis, determining that a pixel is part of the reference region (RR). Here, the contrast-to-noise ratio (CNR(p)) of a pixel is defined as follows: CNR(p)=(FI(p)-?_R) / (c??_R ),

[0064] Here, FI(p) is the fluorescence signal value of the pixel (p), and c is a predetermined constant. The optimal value is T CNR = 1, c = 2.

[0065] The segmentation module 13 is further configured to identify contour C of the target region. Contour C is a primary contour indicating the boundary between the target region and the reference region. The segmentation module 13 is further configured to identify a secondary contour C' that extends outward by a certain distance from the boundary between the target region and the reference region. In the illustrated example, the segmentation module 13 is configured to generate a secondary contour C' that extends outward by a certain distance from the boundary in order to avoid intersection with a specified anatomical structure.

[0066] Figure 16 shows a medical treatment device 100 that, in addition to the elements of the inspection device, further comprises an excitation light source 7 for irradiating mammalian tissue, a camera 6 for acquiring a fluorescence image FI of the mammalian tissue, and a treatment device 5 for performing a medical procedure to excise or irradiate a tumor according to a constructed contour, or to activate a therapeutic substance within a range specified by the constructed contour.

[0067] This invention enables more accurate identification of the contours of affected tissues, such as tumor-infected or infected tissues. Accurately determining the location of these contours is crucial in treatment. For example, when surgically removing affected tissue, it is important not only to ensure that no affected tissue remains after surgery, but also to ensure that healthy tissue is not removed unnecessarily. Similarly, accurately determining the location of these contours is also crucial in the application of photodynamic therapy. This allows for localized activation of the therapeutic substance within the designated tissue area. Outside the designated area, the therapeutic substance remains unactivated, reducing damage to healthy tissue.

[0068] In one example, a therapeutic substance may be activated to act as a chemotherapeutic agent within a region indicated to contain a tumor. In another example, a therapeutic substance may be activated to have antibacterial activity within a region indicated to contain infected tissue.

[0069] Therefore, photodynamic therapy involves at least the following steps: A fluorescence image is acquired in living mammalian tissue by making it photosensitive with a fluorescent agent and irradiating it with excitation light. The resulting fluorescence image consists of an array of pixels, each with its own fluorescence signal value.

[0070] Fluorescent dyes are used to visualize affected tissues, such as tumor tissue or infected tissue. For example, fluorescent dyes include targeted fluorescent tracers like cetuximab-IRDye800CW and Hexbix, used for imaging tumor tissue and infections, and for tracking drug therapy. In another example, fluorescent dyes include non-targeted fluorescent dyes like indocyanine green (ICG), used for imaging tissue perfusion.

[0071] Examples of drugs used for imaging infected tissue include vancomycin-IRDye800CW and 1D9-IRDye800CW. 1D9 is a monoclonal antibody against Staphylococcus aureus (including MRSA). Other examples are listed in the attached references.

[0072] Examples of the latter application are shown in Figures 17A, 17B, and 18A-18F.

[0073] Figure 17A shows an image of a tray containing five sample forms I–V prepared as specified in the table below.

[0074] Figure 17B shows fluorescence images obtained from the same tray containing the sample, using a Pearl imaging system at a wavelength of 800 nm. sample Tracer (Yes / No) Are bacteria present? (Yes / No) I no no II no yes III yes yes IV yes no V yes yes

[0075] Sample 1 (I) is a sterile foam and does not contain a tracer.

[0076] The second sample, Form II, is immersed in a solution containing a Staphylococcus aureus culture but without a tracer.

[0077] The Staphylococcus culture is immersed in a solution containing 10 L of tracer tIRDye800CW solution.

[0078] Fourth sample. Foam IV is immersed in a sterile solution containing tracer tIRDye800CW.

[0079] The fifth sample, Form V, is immersed in a solution containing a Staphylococcus aureus culture and 20 μL of tracer tIRDye800CW solution.

[0080] Figures 18A–18F show the fluorescence image of Figure 17B, divided according to different quartile settings Q ranging from Q = 0.50 in Figure 18A to Q = 0.99 in Figure 18F. Therefore, Figures 18A–18F show the proportion of pixels (1-Q) with fluorescence intensity higher than the proportion of unselected pixels Q.

[0081] What is the average fluorescence signal value in step S7? R And the standard deviation? Rを決定するために使用される基準値は The settings are configured such that the proportion of Q in the pixels has a fluorescence intensity below the reference value, and the proportion of 1-Q has a fluorescence intensity above the reference value. It was found that the best results were obtained when Q = 0.70.

[0082] In this study, the inventors hypothesized that the fluorescence emission observed at the boundary of diseased tissue, such as tumor tissue or infected tissue, is stray light emission, i.e., fluorescence emitted from the diseased tissue scattered by healthy tissue near the boundary. Furthermore, the inventors predicted that, as a result, the intensity of the fluorescence emission decreases exponentially as one moves away from the boundary. That is, I(d) = I_0 e^(-α?d)

[0083] Here, I 0 is the intensity measured at the boundary of the affected tissue, I(d) is the intensity at a location in healthy tissue at a distance d from the boundary, and is a constant. Based on these observations, the following method will be explained with reference to Figures 19A-C.

[0084] Figure 19A shows the measured fluorescence value I(x) as a function of position along the scanning path in the fluorescence image. In this case, the scanning path is the scan line in the x-direction of the image, but the scan line may be in other directions. The scanning path may also be a curve. As shown in Figure 19A, the scanning range extends from 0 pixels to approximately 210 pixels. The relatively high value of function I(x) at position B indicates that the pixel at coordinate x = 130 represents part of the affected tissue, illustrating how the boundary of the affected tissue at the left side of B is determined.

[0085] Figure 19B shows the logarithm of the measured fluorescence value I(x), log(I(x)). Considering the above observations, the value of log(I(x)) is expected to decrease in the direction away from the boundary according to the function. log(I(d))=-α?d?log(I(x0))

[0086] Therefore, the function log(I(x)) is expected to have a linear portion in the region near the lesion within the healthy tissue. To identify these regions, the linearity of the function log(I(x)) as a function of x is determined by matching the curve in the sliding window with a linear function. For example, the length of the sliding window is 30 pixels, and the linear function is matched using the least squares method. Different lengths are also applicable, but the length should not be too short in order to obtain a suitable signal-to-noise ratio. Preferably, the length is at least 10 pixels. The length should not be too long in order to obtain a sufficiently high resolution. However, this is in relation to the image resolution (mm.pixels). -1 This depends on the linearity function. In this case, where the resolution is 85 μm, the sliding window length is preferably 70 pixels or less. However, if the resolution increases by a certain coefficient, the maximum length of the window can also be increased by that coefficient. The sliding window is symmetric to avoid bias in the linearity function. However, it may also be considered to use an asymmetric window to correct for bias.

[0087] In Figure 19C, the correlation of the function log(I(x)) within the sliding window is shown as Linearity(x). Here, a value of 1 indicates the extreme case where the function log(I(x)) within the sliding window is perfectly linear, and a value of 0 indicates that the function log(I(x)) within the sliding window deviates to the maximum extent from a linear function. Note that in the region where the intensity I(x) is constant, the function Linearity(x) approaches 1. In practice, as shown by the dashed line in Figure 19C, we assume that the function log(I(x)) is linear when the function Linear(x) is 0.95 or greater. In the x-axis direction, the function falls below this threshold at position A. This is thought to be the position where the scan line intersects with the boundary of the affected tissue.

[0088] At this point, a transition occurs from a region where the function log(I(x)) increases linearly to a region where the function log(I(x)) takes on a more constant value. Point A' indicates the position on the scan line where the value becomes more constant. Similarly, the position to the right of B allows us to identify the location where the diseased tissue is in contact with healthy tissue.

[0089] This method can be repeated using different scan lines. For example, in this case, the scan lines extend in the x-direction, and by determining the boundary position for each scan line with a different y-coordinate, the contour of the lesioned tissue within the healthy tissue can be estimated. This method can also be applied to scan lines in different directions, either alternatively or additionally.

[0090] The method described with reference to Figures 19A-19C can be used independently, but it can also be used as a preprocessing step for further analysis. For example, based on the contours identified here, it is possible to estimate how many pixels in the image represent affected tissue and how many pixels represent healthy tissue. For example, if the tissue image consists of regions of N0 pixels representing affected tissue and regions of N2 pixels representing healthy tissue, the predetermined ratio used to determine the reference value IR is N2 / (N0+N2). This method is also applicable as a preprocessing step to the method described in the international patent application PCT / NL2024 / 050047 filed by the same applicant. In this application, this preprocessing step can be used to determine a tentatively assigned location.

[0091] Alternatively, the method described with reference to Figures 19A-19C can also be used as a post-processing step to apply corrections to segmentation results obtained by other methods.

[0092] In summary, the inspection method described herein with reference to Figures 19A-19C includes the following steps:

[0093] It is presumed that a fluorescence image of mammalian tissue, made photosensitive with a fluorescent agent and irradiated with excitation light, was captured, and a fluorescence image (FI) containing an array of pixels with each fluorescence signal value was obtained.

[0094] Fluorescence images may be acquired outside the body during surgery to confirm that the affected tissue has been completely removed. Alternatively, they may be acquired inside the body to assist the surgeon during the procedure.

[0095] Fluorescent dyes are used to visualize infected tissues, such as tumor tissue and other infected tissues. Representative dyes used for imaging infected tissue include vancomycin-IRDye800CW and 1D9-IRDye800CW. 1D9 is a monoclonal antibody against Staphylococcus aureus (including MRSA). Other examples are listed in the attached references.

[0096] A sequence of at least one fluorescence value I(x) is obtained from pixels along position x on the scan path. The logarithm of each value in the sequence of fluorescence values ​​I(x) is calculated. Alternatively, the logarithms of all fluorescence values ​​I(x) in the image can be calculated, and then a sequence of logarithms log(I(x)) can be obtained from pixels along position x on the scan path. However, it is generally preferable to first obtain the fluorescence value sequence and then apply the logarithmic function. This eliminates the need to apply the logarithmic function to pixels that are not involved in the calculation.

[0097] Next, we determine how closely the logarithmic sequence log(I(x)) approximates a linear function locally.

[0098] The transition from healthy tissue to affected tissue boundary is estimated at the point where linearity begins to decrease significantly, for example, below a threshold (such as 0.95).

[0099] The scan path can be a scan line in any direction or a curved path. However, a scan line or scan path with negligible curvature is recommended to ensure that the curvature of the path does not affect the linearity of the sequence of values ​​in the boundary region.

[0100] Appendix: Other Reference Materials Enhancement of antimicrobial photodynamic therapy for Staphylococcus aureus infections using potassium iodide. Bispo M, Suhani S, van Dijl JM.J Photochem Photobiol B. 2021 Dec;225:112334. doi:10.1016 / j.jphotobiol.2021.112334. Comparison of two fluorescent probes in preclinical non-invasive imaging and image-guided debridement surgery for staphylococcal biofilm implant infections. Park HY, Zoller SD, Hegde V, Sheppard W, Burke Z, Blumstein G, Hamad C, Sprague M, Hoang J, Smith R, Romero Pastrana F, Chuprina J, Miller LS, Lopez Alvarez M, Bispo M, Juan Austen M, Juan Dayl JM, Francis KP, Berntal NM.Sci Rep. 2021 Jan 15;11(1):1622. Doi: 10.1038 / s41598-020-78362-7. The fight against Staphylococcus aureus infections using light and photoimmune complexes. Bispo M, Anaya Sanchez A, Suhani S, Raineri EJM, Lopez Alvarez M, Hoiker M, Szymanoski W, Romero Pastrana F, Buist G, Horsewill AR, Francis KP, Van Dam GM, Van Austen M, Van Dayl JM. JCI Insight. 2020 Nov 19;5(22):e139512. Doi: 10.1172 / jci.insight.139512. Easy and reproducible synthesis of near-infrared fluorescent complexes containing small target molecules for microbial infection imaging. Riessing F, Bispo M, Lopez Alvarez M, Juan Austen M, Feringa BL, Van Dayl JM, Szymanski W.ACS Omega. 2020 8 26;5(35):22071-22080. Doi: 10.1021 / acsomega.0c02094. A new in vivo mouse model of shoulder implant infection. Shepard WL, Mossick GM, Smith RA, Hamad CD, Park HY, Zoller SD, Trica R, McCoy TK, Boswell R, Juan J, Truong N, Ceballos N, Clarkson S, Hori KR, Van Dale JM, Francis KP, Petrigliano FA, Berntal NM.J. Shoulder and Elbow Surgery 2020 July;29(7):1412-1424. Doi: 10.1016 / j.jse.2019.10.032. Multimodal imaging guides surgical management in preclinical spinal implant infection models. Zoller SD, Park HY, Olafsen T, Zamirpa C, Burke ZD, Blumstein G, Shepard WL, Hamad CD, Hori KR, Tseng JC, Chuprina J, McManus C, Lee JT, Bispo M, Romero Pastrana F, Raineli EJ, Miller JF, Miller LS, Van Daile JM, Francis KP, Berntal NM.JCI Insight. 7 February 2019;4(3):e124813. Doi: 10.1172 / jci.insight.124813. Non-invasive optical and nuclear imaging of Staphylococcus-specific infections using human monoclonal antibody-based probes. Romero Pastrana F, Thompson JM, Hoekker M, Hoekstra H, Dillen CA, Ortinez RV, Ashbaugh AG, Pickett JE, Linsen MD, Berntal NM, Francis KP, Buist G, Van Austen M, Van Dam GM, Solek DLJ, Miller LS, Van Dyle JM. Toxicology. January 1, 2018; 9(1):262-272. Doi: 10.1080 / 21505594.2017.1403004. Real-time in vivo imaging of invasive and biomaterial-associated bacterial infections using fluorescently labeled vancomycin. Van Austen M, Schäfer T, Gazendam JA, Olsen K, Tsompanidu E, De Goffor MC, Harmsen HJ, Crane LM, Lim E, Francis KP, Chang L, Olive M, Nzczakristos V, Van Dale JM, Van Dam GM.Nat Commun. 2013;4:2584. Doi:10.1038 / ncomms3584.

Claims

1. A method for examining a fluorescence image taken from mammalian tissue that has been photosensitive with a fluorescent agent and irradiated with excitation light, wherein the fluorescence image (FI) comprises an array of pixels having respective fluorescence signal values, and the fluorescence signal values ​​comprise a first number (N1) of fluorescence signal values ​​representing the mammalian tissue in the fluorescence image, The reference value (IR) is determined such that the number of second fluorescence signal values ​​(N2) included in the number of first fluorescence signal values ​​is less than or equal to the reference value, the remaining number of first fluorescence signal values ​​exceeds the reference value, and the second number is a predetermined fraction of the first number (S6); The average fluorescence signal value of the fluorescence signal values ​​included in the second number of fluorescence signal values ​​(? Ｒ ) and standard deviation (? Ｒ ）を決定する； Image segmentation is performed to distinguish between a target region (TR) showing a portion of mammalian tissue (MT) identified as tumor tissue or infected tissue and a reference region (RR) showing the rest of the mammalian tissue (MT) (S8). This division is performed when the contrast-to-noise ratio (CNR) (p) of each pixel (p) reaches a predetermined threshold (T ＣＮＲ The process includes determining that if the pixel exceeds ) the threshold, it is part of the target region (TR), and otherwise, it is part of the reference region (RR), and the contrast-to-noise ratio CNR(p) of the pixel is CNR(p)=(FI(p)-μ_R) / (c?σ_R ), Here, FI(p) is the fluorescence signal value of pixel(p), and c is a predetermined constant. Identify at least one contour (B) of the target region (TR) (S9).

2. The inspection method according to claim 1 is characterized in that the predetermined threshold is 1 and the predetermined constant is 2.

3. An inspection method according to claim 1 or 2, characterized in that a fluorescence image (FI) is captured as yet another image of the background (S4), and further comprising the following steps: In the fluorescence image (FI), preliminary image segmentation is performed to distinguish between the foreground region (FR) representing mammalian tissue (MT) and the background region (BR) representing the background (S5). This division involves determining, for each pixel, that it is part of the foreground region (FR) if its fluorescence signal value (FSV) significantly exceeds the average fluorescence signal value determined for the background, taking into account the standard deviation of the fluorescence signal values ​​(FSV) in the background; otherwise, it is determined that the pixel is part of the background region (BR).

4. The inspection method according to claim 3, wherein the mean fluorescence signal value and the standard deviation of the fluorescence signal value are determined from a portion (BP) designated to represent the background of the fluorescence image (FI) (S2A).

5. A testing method according to any one of claims 1 to 4, wherein the fluorescence image of mammalian tissue is captured in vivo.

6. A testing method according to any one of claims 1 to 4, wherein the fluorescence image of the mammalian tissue is taken outside the body with the mammalian tissue placed in the background.

7. An inspection method according to any one of claims 2 to 6, characterized in that the step of performing preliminary image division (S5) further includes an expansion operation.

8. A testing method according to any one of claims 1 to 7, wherein the predetermined k-quartile is the k-th percentile in which k is in the range of 25 to 75.

9. The inspection method according to claim 8, wherein the percentile is the median.

10. An inspection method according to any one of claims 1 to 9, further comprising an evaluation based on a scan trajectory, which includes the following steps: S10: Obtain at least one fluorescence signal value vector of fluorescence signal values ​​in the fluorescence image (FI) along the scanning trajectory; For each of the multiple thresholds, a set of candidate scan trajectory sections whose fluorescence signal value exceeds the threshold is determined (S11), and the statistical properties of the scan trajectory sections that are not candidate scan trajectory sections are determined. Based on segmentation using the determined statistical characteristics and contrast-to-noise ratio, it is determined which of several thresholds best matches the image segmentation of the fluorescence image (FI) along the scan trajectory for each set of candidate scan trajectory sections (S13).

11. The inspection method according to claim 10, wherein the procedure described therein is repeated for a plurality of mutually different scan trajectories among a set of scan trajectories (S14).

12. The inspection method according to claim 11, wherein the procedure described therein is repeated for a plurality of scan trajectory sets (S15).

13. An inspection method according to any one of claims 10 to 12, further comprising the step of rejecting isolated points (S16).

14. An inspection method according to any one of claims 1 to 13, further, At least one contour determines at least one principal axis. To obtain at least one fluorescence signal value vector of fluorescence signal values ​​in a fluorescence image (FI) along a scan line orthogonal to the principal axis; Based on contour information at the intersection with the scan line, the system selectively identifies one or more sections of the scan line in which the contrast-to-noise ratio exceeds a predetermined level, each of which scan line sections includes a first endpoint indicating a transition (NT) from normal tissue to tumor tissue or infected tissue, and a second endpoint indicating a transition (TN) from tumor tissue or infected tissue to normal tissue.

15. The method according to claim 14, wherein the principal axis is defined as a line segment that minimizes the average distance between the line segment and a point on the contour.

16. An inspection method according to claim 14 or 15, wherein, prior to the step of acquiring at least one fluorescence signal value vector (S22), the fluorescence image (FI) is rotated (S21) so that the principal axis of the contour coincides with the coordinate axes of the fluorescence image (FI), and the scan lines are oriented along the other coordinate axes of the fluorescence image (FI).

17. An inspection method according to claim 14, 15, or 16, characterized in that the information underlying the selectively performed step (S23) includes the intensity of the fluorescence image (FI) at the intersection of contours.

18. An inspection method according to claim 14, 15, or 16, characterized in that the information underlying the selectively performed step (S23) includes the coordinates of the intersection.

19. An inspection method according to any one of claims 14 to 18, further comprising the step (S24) of combining and displaying the coordinates of the start and end points of a plurality of contour lines within a single image.

20. An inspection method according to claim 19, wherein the coupling step (S24) includes performing a reverse rotation according to the principal axis of the contour line.

21. An inspection method according to any one of claims 10 to 16, further, Determine the density of each endpoint in each region of each endpoint. We will remove endpoints whose density is below a predetermined threshold.

22. An inspection method according to any one of claims 1 to 21, wherein at least one contour includes a main contour (C) that indicates the boundary between the target area and the reference area.

23. An inspection method according to any one of claims 1 to 22, wherein at least one contour includes a secondary contour (C') that extends outward by a certain distance from the boundary between the target area and the reference area.

24. The inspection method according to claim 23, wherein the secondary contour (C') extends at a distance outside the boundary so as to avoid intersecting with a particular anatomical structure.

25. The inspection method according to claim 1, further comprising the following steps: Obtaining at least one sequence of logarithmic fluorescence signal values ​​for each subsequent pixel arranged along a scan trajectory in a fluorescence image, wherein the scan trajectory begins at a position identified as part of a reference region in the fluorescence image, To calculate an index of local linearity of logarithmic fluorescence signal values ​​as a function of position in at least one array; The method involves identifying a location within a sequence, where the indicator shows that the logarithmic fluorescence signal value as a function of the location within the sequence is no longer linear, and the identified location is a candidate boundary location for the boundary between normal tissue and tumor or infected tissue.

26. A testing device (1) for examining mammalian tissue, A fluorescence image (FI) is obtained of mammalian tissue (MT) that has been made photosensitive with a fluorescent agent and irradiated with excitation light. The fluorescence image includes an array of pixels, each having a fluorescence signal value, and each fluorescence signal value includes a first number of fluorescence signal values ​​representing the mammalian tissue in the fluorescence image. The average fluorescence signal value of the fluorescence signal values ​​included in the second number of fluorescence pixel values ​​( ? Ｒ ) and standard deviation (? Ｒ ）を決定する； Image segmentation is performed to distinguish between a target region (TR) representing a portion of mammalian tissue (MT) identified as tumor tissue or infected tissue in a fluorescence image (FI), and a reference region (RR) representing the rest of the mammalian tissue (MT). Here, the device determines that for each pixel (p), the contrast-to-noise ratio CNR (p) of the pixel is at a predetermined threshold (T ＣＮＲ The system is configured to determine if the pixel exceeds the specified threshold, that it is part of the target region (TR), and otherwise determine that the pixel is part of the reference region (RR), and the contrast-to-noise ratio CNR(p) of the pixel is, CNR(p)=(FI(p)-μ_R) / (c?σ_R ), Here, FI(p) is the fluorescence signal value of pixel(p), and c is a predetermined constant. Identifies the contour of the target region (TR).

27. The inspection apparatus according to claim 26, characterized in that the predetermined threshold is 1 and the predetermined constant is 2.

28. An inspection apparatus according to claim 26 or 27, configured to acquire a fluorescence image (FI) representing the background in addition to mammalian tissue (MT), and having the following configuration: In a fluorescence image (FI), preliminary image segmentation is performed to distinguish between a foreground region (FR) representing mammalian tissue (MT) and a background region (BR) representing the background, wherein the device is configured to determine that each pixel is part of the foreground region (FR) if its fluorescence signal value (FSV) significantly exceeds the mean fluorescence signal value determined for the background, taking into account the standard deviation of the fluorescence signal values ​​(FSV) in the background; otherwise, it determines that the pixel is part of the background region (BR);

29. The inspection apparatus according to claim 28 is further configured to acquire a calibration fluorescence image of only the background before acquiring a fluorescence image, and to determine the average fluorescence signal value and standard deviation from the calibration fluorescence image.

30. An inspection apparatus according to claim 26, configured to determine the average fluorescence signal value and the standard deviation (S2A) of the fluorescence signal value from a portion (BP) of a fluorescence image (FI) designated as representing the background.

31. An inspection device according to any one of claims 26 to 30, configured to acquire a fluorescence image of mammalian tissue captured in vivo.

32. An inspection apparatus according to any one of claims 26 to 31, configured to acquire a fluorescence image taken in vitro from mammalian tissue placed on a background.

33. An inspection apparatus according to any one of claims 26 to 32, wherein the apparatus is configured to perform an expansion calculation (S5) after preliminary image division.

34. An inspection device according to any one of claims 26 to 33, wherein a predetermined k-th quartile is the k-th percentile in the range of 25 to 75.

35. The inspection apparatus according to claim 34, wherein the percentile is the median.

36. An inspection apparatus according to any one of claims 26 to 35, further configured to perform an evaluation based on a scan trajectory, which includes the following steps: S10: Obtain at least one fluorescence signal value vector of fluorescence signal values ​​in the fluorescence image (FI) along the scanning trajectory; For each of the multiple thresholds, a set of candidate scanning trajectory intervals in which the fluorescence signal value exceeds the threshold is determined (S11). It is determined which of the multiple thresholds best matches each set of candidate scan trajectory sections to the second image segmentation of the fluorescence image (FI) along the scan trajectory (S13).

37. An inspection device according to claim 36, configured to repeat the procedure described therein for a plurality of mutually different scanning trajectories among a set of scanning trajectories.

38. An inspection device according to claim 37, configured to repeat the procedure described therein for multiple sets of scanning trajectories.

39. An inspection apparatus according to any one of claims 26 to 38, further comprising the step of rejecting isolated points (S16).

40. An inspection apparatus according to any one of claims 26 to 39, further comprising the following configuration: Determine at least one principal axis of at least one contour (S20). At least one fluorescence signal value vector is obtained for fluorescence signal values ​​in the fluorescence image (FI) along a scan line orthogonal to the principal axis (S22). Based on contour information at the intersection with the scan lines, one or more sections of the scan lines in which the contrast-to-noise ratio exceeds a predetermined level are selectively identified (S23). Each of the scan line sections includes a first endpoint representing a transition (NT) from normal tissue to tumor tissue or infected tissue, and a second endpoint representing a transition (TN) from tumor tissue or infected tissue to normal tissue.

41. The inspection apparatus according to claim 40, wherein the main axis is defined as a line segment that minimizes the measured distance between the line segment and a point on the contour.

42. An inspection apparatus according to claim 40 or 41, configured to rotate a fluorescence image (FI) before acquiring at least one fluorescence signal value vector (S22), wherein the rotation aligns the principal axis of the contour with the coordinate axes of the fluorescence image (FI) and directs the scan lines along the other coordinate axes of the fluorescence image (FI).

43. An inspection apparatus according to claim 40, 41, or 42, wherein the information underlying the selectively performed step (S23) includes the intensity of a fluorescence image (FI) at the intersection of contours.

44. An inspection apparatus according to claim 40, 41, or 42, wherein the information underlying the selectively performed step (S23) includes the coordinates of the intersection.

45. An inspection apparatus according to any one of claims 40 to 44, configured to display a combination of coordinates of multiple contour points in a single image (S24).

46. An inspection device according to claim 45, wherein the inspection device is configured to perform reverse rotation according to the principal axis of the contour before performing the coupling (S24).

47. An inspection device according to any one of claims 26 to 46, further configured to selectively remove isolated candidate points.

48. An inspection apparatus according to any one of claims 26 to 43, further comprising a camera (6) for acquiring a fluorescence image (FI) of mammalian tissue (MT).

49. The inspection apparatus according to claim 48, further comprising an excitation light source (7) for irradiating mammalian tissue.

50. An inspection apparatus according to any one of claims 26 to 49, wherein at least one contour includes a main contour (C) that indicates the boundary between a target area and a reference area.

51. An inspection apparatus according to any one of claims 26 to 50, wherein at least one contour includes a secondary contour (C') that extends outward by a certain distance from the boundary between the target area and the reference area.

52. The inspection apparatus according to claim 51, wherein the secondary contour (C') extends at a distance outside the boundary so as to avoid intersecting with a particular anatomical structure.

53. An inspection apparatus according to claim 26, further comprising the following configuration: At least one sequence of logarithmic fluorescence signal values ​​for each subsequent pixel positioned along the scan trajectory in the fluorescence image is obtained, where the scan trajectory begins at a position identified as part of a reference region in the fluorescence image. The index of local linearity of the logarithmic fluorescence signal value is calculated as a function of position in at least one array. The indicator identifies the location within the sequence, showing that the logarithmic fluorescence signal value as a function of the location within the sequence is no longer linear, and the identified location is a candidate boundary location for the boundary between normal tissue and tumor or infected tissue.

54. A medical treatment device (100) comprising the inspection device according to claim 49, and a medical treatment instrument (5) for excising or irradiating a tumor according to a constructed contour, or for activating a therapeutic substance in a range specified by the constructed contour.

Images