Clinical decision support system and computer implementation method

Abstract

To provide a computer based clinical decision support system (CDSS) for enhancing determination of a classification of a lymphedema induced fluorescence pattern, and a computer implemented method of enhancing determination of a classification of a lymphedema induced fluorescence pattern.SOLUTION: Described herein are a computer based clinical decision support system (CDSS) 62 and a method for determining a classification of a lymphedema induced fluorescence pattern. A fluorescence image 7 is determined from a measurement of a fluorescence signal in a tissue of a body part 4, to which a fluorescent agent 8 has been added. The CDSS 62 includes: an input interface 64 through which the fluorescence image 7, which is specific to a patient 6, is provided as an input feature to an artificial intelligence (Al) model 68; a processor 66 that performs an inference operation in which the fluorescence image 7 is applied to the Al model 68 to generate a classification of a lymphedema induced fluorescence pattern; and a user interface (UI) 14 through which the classification of the lymphedema induced fluorescence pattern is communicated to a user.SELECTED DRAWING: Figure 8

Description

Technical Field

[0005] ,

[0001] The present invention relates to a computer-based clinical decision support system (CDSS) configured to output a classification of lymphoedema-induced fluorescence patterns based on fluorescence images determined from measurements of fluorescence signals in tissues of body parts to which a fluorescent agent has been added. Further, the present invention relates to a computer-implemented method of determining a classification of lymphoedema-induced fluorescence patterns using the CDSS, the classification being based on fluorescence images determined by measuring fluorescence signals within tissues of body parts to which a fluorescent agent has been added.

[0002] Further, the present invention relates to a method of diagnosing lymphoedema and a method of long-term treatment of lymphoedema.

Background Art

[0003] Lymphoedema is the accumulation of lymph fluid in the tissues of the body. Oxygenated blood is pumped from the heart to the tissues via the arteries, while deoxygenated blood returns to the heart via the veins. Due to the much higher pressure level on the arterial side than on the venous side, the colourless fluid portion of the blood is pushed into the spaces between the cells. Typically, more fluid is pushed out than is reabsorbed on the venous side. The excess fluid is transported by the lymphatic vessels. Further, the fluid carries away local foreign bodies such as larger proteins and cell debris. Once in the lymphatic system, this fluid containing the transported substances is referred to as lymph or lymph fluid.

[0004] The lymphatic system includes lymphatic vessels having one-way valves similar to venous valves for transporting lymph fluid to the next lymph node. The lymph nodes remove certain substances and purify the fluid before it returns to the bloodstream. <00000​​When lymphatic flow is obstructed or not performed at the desired level, the lymphatic system becomes blocked, causing lymphatic fluid to accumulate in the interstitial space between tissue cells. This accumulation resulting from impaired lymphatic transport is called lymphedema. The accumulation of lymphatic fluid can trigger an inflammatory response that damages the cells surrounding the affected area. It can further lead to fibrosis, which can progress to hardening of the affected tissue.

[0006] Lymphedema is a lifelong condition for which there is no cure or drug therapy; therefore, early diagnosis and appropriate early intervention to improve drainage and reduce fluid load are crucial for the patient's health and recovery. Possible treatments, such as lymphatic massage and compression bandages, prior to surgery, depend on the severity, which is defined by the World Health Organization (WHO) as a four-stage system. Stage 1: Normal lymphatic flow. No signs or symptoms. Stage 2: Fluid accumulation accompanied by swelling. Stage 3: Persistent swelling of the affected limb or body part that does not subside when the limb is lifted. Stage 4: Elephantiasis (severe deformities of the limbs), thickening of the skin with wart-like growths, and extensive scarring.

[0007] The technique commonly used to diagnose lymphatic system dysfunction is a manual examination of the affected limb or body part by a physician. A known imaging technique is lymphoscintigraphy. In this technique, a radioactive tracer is injected into the tissue of the affected body part, followed by MRI (magnetic resonance imaging), CT (computed tomography), PET-CT (positron emission tomography) scan, or ultrasound imaging.

[0008] A relatively new imaging technique is infrared fluorescence imaging using fluorescent dyes, such as ICG (indocyanine green). ICG is a green medical dye that has been used for over 40 years. The dye fluoresces when excited by near-infrared light with wavelengths of 600 nm to 800 nm. This excitation causes ICG to emit fluorescence in the 750 nm to 950 nm range. The fluorescence of the ICG dye can be detected using a CCD or CMOS sensor or camera. The fluorescent dye is administered to the tissue of an affected limb or body part, and the concentration and flow of lymphatic fluid can be tracked based on the detected fluorescence. [Overview of the project] [Problems that the invention aims to solve]

[0009] The object of the present invention is to provide a computer-based clinical decision support system (CDSS) for enhancing the determination of classification of lymphedema-induced fluorescence patterns, and a computer implementation method for enhancing the determination of classification of lymphedema-induced fluorescence patterns.

[0010] Furthermore, an objective of the present invention is to provide an enhanced method for diagnosing lymphedema and an enhanced method for long-term treatment of lymphedema. [Means for solving the problem]

[0011] This objective is addressed by a computer-based clinical decision support system (hereinafter referred to as "CDSS") configured to output a classification of lymphedema-induced fluorescence patterns based on fluorescence images determined from the measurement of fluorescence signals in tissues of body parts to which a fluorescent agent has been added. An input interface that provides patient-specific fluorescence images as input features to an artificial intelligence (AI) model, A processor that performs inference operations where fluorescence images are applied to an AI model to generate classification of lymphedema-induced fluorescence patterns, A user interface (UI) that communicates the classification of lymphedema-induced fluorescence patterns to the user, It is equipped with.

[0012] Early diagnosis and appropriate treatment are crucial for the successful treatment and prevention of lymphedema. Fluorescence imaging is an appropriate approach for this early diagnosis. Fluorescence imaging analyzes the distribution and progression of a fluorescent agent as it propagates with the lymph fluid. Before initiating fluorescence measurements, the fluorescent agent is administered to the tissue of the patient's body undergoing medical examination. In a healthy lymphatic system, the fluorescent agent is rapidly taken up by the lymphatic system (along with the lymph fluid) and rapidly transported through the lymphatic vessels. The normal rhythmic contraction of the lymphatic vessels can be observed in time-resolved fluorescence images. It is possible to see how the fluorescent agent pushes up the limbs while one-way valves within the lymphatic vessels prevent fluid backflow.

[0013] In lymphedema, obstruction of lymphatic fluid leads to increased pressure within the lymphatic vessels. This causes lymphatic fluid to leak into the subcutaneous tissue. As lymphedema progresses, a characteristic reflux pattern can be seen on fluorescence imaging.

[0014] Fluorescence images are typically captured at two time points. The first image is captured during the initial transient phase, and at least the second image is captured during the later plateau phase. In this procedure, the patient remains still for several minutes initially while a certain amount of fluorescent agent is subcutaneously injected. Immediately after injection, during the so-called "transitional" phase, observable fluorescent images show lymphatic flow, allowing for the measurement of, for example, lymphatic pump function. These images are captured while the patient is still. Subsequently, the patient is allowed to move freely, and lymphatic circulation is assessed based on the observable patterns. These patterns allow for the classification of lymphedema severity stages, such as the severity staging defined by the WHO. Depending on the degree of patient movement, the fluorescent agent can reach a plateau phase approximately two hours after injection. Essentially, lymphatic circulation can be assessed to some extent between 2 and 72 hours after injection.

[0015] The fluorescence images analyzed by CDSS according to the embodiment are always fluorescence images taken during the late plateau stage in which a typical lymphatic reflux pattern can be observed.

[0016] Depending on the stage of lymphedema, characteristic reflux patterns can be observed in fluorescence imaging, which are clinically classified into several types. When the lymphatic system is healthy, a linear pattern can be observed. This linear pattern indicates the reflux of lymph fluid in the lymphatic vessels or veins. In the early stages of lymphedema, a so-called "splashback" pattern may be observed. As lymphedema progresses further, a so-called "starburst" pattern develops. In the severe and advanced stages of lymphedema, lymph fluid and the fluorescent agent along with it flow throughout the skin and subcutaneous tissue, resulting in a so-called "diffuse" pattern.

[0017] Advantageously, CDSS performs automated classification of lymphedema-induced fluorescence patterns and communicates this to the user. This greatly helps medical professionals in finding the correct diagnosis for lymphedema.

[0018] According to a favorable embodiment, the classification of lymphedema-induced fluorescence patterns is the stage of lymphedema severity and / or the clinical type of fluorescence pattern. The stage of severity may be, for example, a stage according to the WHO definition. The clinical type is as described above. Possible clinical types that can be identified are, for example, linear patterns, splash patterns, stardust patterns, and diffuse patterns.

[0019] In clinical practice, distinguishing between different types of fluorescence patterns is often difficult. Furthermore, assigning a specific stage of lymphedema to a specific fluorescence image can be challenging. Advantageously, CDSS assists clinicians during this decision-making process.

[0020] In a further advantageous embodiment, the fluorescence image and the corresponding visible light image are provided to the artificial intelligence model as input features through an input interface. In other words, the fluorescence image and / or a combination of the fluorescence image and the visible light image can be used as input to the artificial intelligence model. Since the clinical findings of lymphedema involve specific symptoms that are more or less visible on the surface of body parts, such as the skin, information that can be derived from the visible light image can assist and thereby improve the process of finding the classification of fluorescence patterns.

[0021] Within the context of this specification, the term “image” should not be interpreted as being limited to 2D images. Images may be reconstructed 3D images. Images may be captured, for example, by a stereoscopic imaging unit, or calculated from 2D images plus additional depth information, for example, by a scanner or any other suitable device.

[0022] In other words, the terms "fluorescent image" and "visible light image" are not limited to 2D images. These images can also be 3D or 2D images that include additional information, such as depth information. 2D images can be captured using a suitable camera. 3D images can be captured using, for example, a stereo camera. 2D or 3D images may also be line scans that can be captured using an image line scanner. Images can further be LIDAR scans (and additional color information taken from, for example, a 2D image) that include depth information, similar to 3D images. Data from a LIDAR scanner can be added to a 2D image as additional information. Combining 2D image data with LIDAR scan data yields information similar to that of a 3D image, because the 2D image data can be combined with the corresponding depth information.

[0023] According to another advantageous embodiment, the CDSS is a direct link to an image capture processing device configured such that the input interface measures a fluorescence signal within the tissue of a body part and is configured to image the surface of the body part, the tissue to which the fluorescent agent is added forms part of the body part, the image capture processing device comprises an image capture device, the image capture device comprises an illumination unit configured to illuminate the tissue with excitation light having a wavelength suitable for generating emission light by excitation emission of the fluorescent agent, a fluorescence imaging unit configured to capture a fluorescence image by spatially resolving the measurement of the emission light to provide a fluorescence image, and a visible light imaging unit configured to capture a corresponding visible light image of a part of the surface of the body part And, equipped with, The fluorescence imaging unit and the visible light imaging unit are configured such that the viewing directions and / or viewpoints of the fluorescence image and the corresponding visible light image are linked via a known relationship. ,and This is further enhanced in that regard.

[0024] Advantageously, the image capture processing device can be part of an endoscope or a laparoscope.

[0025] The fluorescence imaging unit and the visible light imaging unit configured such that the viewing directions and / or viewpoints of the fluorescence image and the corresponding visible light image are linked via a known fixed relationship are advantageous in that corresponding images can be captured. The images are linked via a known fixed spatial relationship of the capture device. This significantly improves the success rate when applying a stitching algorithm to the images, and the stitching results in larger fluorescence images and larger visible light images. The larger images are larger in that they show parts of the body part that are larger than the parts of the body that can be captured by a single fluorescence image and a single visible light image respectively.

[0026] According to one embodiment, the CDSS is such that a large fluorescence image and the corresponding large visible light image are provided to an artificial intelligence model through an input interface as input features, The fluorescence imaging unit and the visible light imaging unit are further configured to repeatedly capture fluorescence images and visible light images to provide a series of fluorescence images and a series of visible light images. The image capture processing device further comprises a processing device, and the processing device is To generate a large visible light image of a body part, The set of stitching parameters is determined and applied. stitching algorithm For a series of visible light images Apply Rus Equipped with a ticking unit, The stitching unit is further configured to apply a stitching algorithm to a series of fluorescence images in order to generate a large fluorescence image. Apply to a series of fluorescence images. Stitching algorithm So , large Visible light image generate The set of stitching parameters determined at that time but Applicable So This further strengthens the system.

[0027] In the context of this specification, the term “stitching” is not limited to the combination of two or more 2D images. Stitching can also be performed based on other image data, for example, 3D image data. For example, further information such as depth information captured by a LiDAR scanner can be taken into consideration when running a stitching algorithm.

[0028] Stitching can be performed based on two or more 2D images, resulting in a larger 2D image. When a stitching algorithm is performed based on multiple 3D images, the result is a larger 3D image. It is also possible to perform the stitching process in which multiple 2D images and additional information (e.g., data from a LiDAR scan or similar device) are processed, and the result is a larger 3D image. In this case, the stitching process involves reconstructing a 3D image from the 2D image data and additional information. Image stitching involves identifying unique special features that are visible in two images that correspond to each other and need to be stitched together. This requires that the fields of view of the two images to be joined in the stitching process partially overlap.

[0029] Taking into consideration this requirement for the stitching process, according to one embodiment, the step of repeatedly capturing a fluorescence image and a visible light image to provide a series of fluorescence images and a series of visible light images is performed such that each subsequent image of the sequence is captured and the fields of view of the subsequent images overlap at least slightly.

[0030] However, the aforementioned additional information that may be used for stitching is not limited to, for example, depth information assigned to each individual pixel in a 2D image. During the acquisition of a series of fluorescence and visible light images, the image acquisition device may also acquire further data, i.e., additional information, such as data relating to the spatial orientation of the image acquisition device. Spatial orientation refers to, for example, the image in the coordinate system of the laboratory. captureThis can be the orientation of the device. This orientation can be characterized by three Cartesian coordinates x, y, and z, along with the viewing direction (e.g., a vector in the laboratory coordinate system) and the tilt angle of the image-capturing device around the viewing direction (e.g., the rotation angle of the image-capturing device around the viewing direction). This information can be captured for each single image (or pair of images including a fluorescent image and a visible light image) in a series of visible light and fluorescence images. When performing the stitching process, this additional information indicating the orientation of the image-capturing device in space can be used. In particular, the orientation of the image-capturing device, defined, for example, in the laboratory coordinate system, can be recalculated for the orientation of the image-capturing device relative to a body part. This information can be used when performing 3D reconstruction of a body part during the stitching process.

[0031] Advantageously, the stitching of a series of fluorescence images is performed based on a set of stitching parameters previously determined when performing visible light image stitching. Fluorescence images typically offer rare special features that make them suitable for the stitching process. Since visible light images and fluorescence images are linked by a known, certain relationship regarding viewing direction and / or viewpoint, the parameters of the stitching algorithm used for visible light images can also be applied to the stitching of fluorescence images. This significantly improves the stitching process, resulting in larger fluorescence images of better quality.

[0032] Examples of fluorescent agents include ICG (indocyanine green) or methylene blue. In the context of this specification, the terms “fluorescent dye” or “dye” (also referred to as “fluorescent dye” or “fluorophore”) refer to the molecular component that makes a molecule fluorescent. The component is a functional group in a molecule that absorbs energy at a specific wavelength and re-emits energy at a different specific wavelength. In various embodiments, fluorescent agents include fluorescent dyes, their analogues, their derivatives, or combinations thereof. Suitable fluorescent dyes include, but are not limited to, indocyanine green (ICG), fluorescein, methylene blue, isosulfan blue, patent blue, cyanine 5 (Cy5), cyanine 5.5 (Cy5.5), cyanine 7 (Cy7), cyanine 7.5 (Cy7.5), cypate, silicon rhodamine, 5-ALA, IRDye 700, IRDye 800 CW, IRDye 800 RS, IRDye 800 BK, porphyrin derivatives, Illuminare-1, ALM-488, GCP-002, GCP-003, LUM-015, EMI-137, SGM-101, ASP-1929, AVB-620, OTL-38, VGT-309, BLZ-100, ONM-100, and BEVA 800.

[0033] The stitching algorithm, for example, is a panoramic stitching algorithm, in which images are analyzed and special, characteristic features are extracted. These features are then linked together in multiple images, and image transformations (e.g., shift, rotation, stretching along one or more axes, or keystone correction) are performed. The positions of the linked features are used to determine image transformation parameters, also called stitching parameters. Following image orientation, the images are merged, thereby "stitching" them together. This transformation can be performed on both visible light images and fluorescence images. Finally, two images are output. This output may include a step of displaying the images on a screen, for example. For example, the two images may be shown side by side or as an overlay image.

[0034] Often, the function of the lymphatic system is affected in the extremities. However, the CDSS is not limited to the examination of the extremities. This system can be applied to the examination of general body parts that may be the torso, head, neck, back, or any other part of the body, not just the extremities of the patient. It is also possible for the body part to be an organ. The CDSS can be applied during incision surgery. The same applies when the surgery is a minimally invasive surgery performed using an endoscope or laparoscope.

[0035] In the context of this specification, “visible light image” refers to an image of a real-world situation. It reproduces an image impression similar to what the human eye can see. Unlike the human eye, a visible light image can be a color image, a grayscale image, or even a pseudo-color scale plot. A visible light image shows the surface of a body part containing tissue to which a fluorescent agent has been administered. If the tissue is placed on the surface of a body part, imaging of the surface of the body part includes imaging of the surface of the tissue.

[0036] According to another advantageous embodiment, the viewing direction and field of view of the fluorescent and visible light images are identical, and in particular, the fluorescent and visible light images are captured through the same objective lens, and more specifically through a single objective lens. The objective lens is similar to a camera lens and comprises one or more lenses.

[0037] Advantageously, fluorescence and visible light images can be captured by an imaging device comprising a prism assembly and a plurality of image sensors assigned to it. Fluorescence and visible light enter the prism assembly as a common beam, passing through the same incident plane of the prism assembly. The prism structure includes filters to separate the visible wavelength range from the infrared wavelength range, where excitation emission of the fluorescent agent typically occurs. Different wavelength bands, namely visible light (also abbreviated as Vis) and infrared light (also abbreviated as IR), are directed to different sensors. Capture of visible light and fluorescence images through a single objective lens allows for perfect alignment of the viewing direction and perspective of the two images. In particular, the viewing direction and viewpoint of the visible light and fluorescence images are identical.

[0038] In an advantageous embodiment, the acquisition of the fluorescence image and the acquisition of the visible light image are performed simultaneously without any time switching between the fluorescence image signal and the visible light image signal.

[0039] To have an advantage, Computer implementation The method eliminates the need for signal time switching. This is advantageous because the fluorescence image (infrared image) and the visible light image are captured precisely and simultaneously using separate image sensors. Therefore, it is possible to capture images at high frame rates of 60fps or more. When time switching is applied, it is usually not possible to achieve such high frame rates. Also, when capturing fluorescence and visible light images with separate sensors, the sensors can be positioned to focus on each other. This improves image sharpness. Furthermore, the sensor settings can be adjusted to the individual requirements for image acquisition of the visible light and fluorescence images. This relates to adjustments such as sensor gain, noise reduction, and exposure time.

[0040] In yet another advantageous embodiment, the steps of capturing a fluorescence image, illuminating the tissue with excitation light, and simultaneously capturing a visible light image are performed by a single image acquisition device. When illumination and image acquisition are integrated into a single device, the overall process of measuring fluorescence signals and simultaneously acquiring visible images can be improved.

[0041] Furthermore, according to another advantageous embodiment, the CDSS comprises an image capturing device comprising a dichroic prism assembly configured to receive fluorescent and visible light through an incident plane, the prism assembly comprising a first prism, a second prism, a first compensator prism located between the first and second prisms, a further dichroic prism assembly for splitting visible light into three optical components, and a second compensator prism located between the second prism and the further dichroic prism assembly, The first prism and the second prism each have a cross-section with at least five corners, and each corner has an interior angle of at least 90 degrees. ,So Each has an incident surface and a corresponding exit surface, and in a direction parallel to the normal of the incident surface The first prism and the second prism each at The incident beam that enters the incident surface, The first prism and the second prism Each prism reflects twice, and the resulting beam is parallel to the normal of the emission surface. to Through the exit surface The first prism and the second prism That Reka Each is designed to be fired from, Each of the first prism and the second prism The normals of the incident plane and the exit plane are perpendicular to each other. When light enters the first prism through the incident plane, the light is partially reflected toward the exit plane of the first prism. , the Light travels a first path length from the incident plane of prism 1 to the exit plane of prism 1, partially enters prism 2 through prism 1 compensator prism, and is partially reflected toward exit plane of prism 2. , the The second path length is traveled from the incident plane of prism 1 to the exit plane of prism 2. The first prism is further enhanced by being larger than the second prism so that the path lengths of the first and second prisms are equal.

[0042] Advantageously, the aforementioned five-prism assembly allows for the capture of two fluorescence imaging wavelengths and three colors for visible light imaging, such as red, blue, and green. The five-prism assembly is advantageous in that the optical paths of the light traveling from the incident plane to each of the sensors are of the same length. Therefore, all sensors are in focus, and furthermore, there is no timing gap between the sensor signals. Advantageously, the device does not require time switching of the received signals.

[0043] In yet another embodiment, the image capturing device defines first, second, and third optical paths for directing fluorescent and visible light to first, second, and third sensors, respectively, and the image capturing device further comprises a dichroic prism assembly configured to receive fluorescent and visible light through an incident surface, the dichroic prism assembly comprising a first prism, a second prism, and a third prism, each prism having first, second, and third exit surfaces, the first exit surface being provided with a first sensor, and the second exit surface being provided with a second sensor. A third sensor is provided on the third emission surface, and in particular, a first filter is provided in the first optical path, a second filter is provided in the second optical path, and a third filter is provided in the third optical path. The first, second, and third filters are red / blue patterned filters that, in any order, consist of a green filter, an infrared filter, and a red / blue patterned filter, with half of the light received by the green filter passing through the blue filter and half of the light received by the red / blue patterned filter passing through the red filter in an alternating pattern.

[0044] Furthermore, according to another embodiment, the first, second, and third filters are, in any order, a red / green / blue patterned filter (RGB filter), a first infrared filter, and a second infrared filter, and in particular, the first and second infrared filters have different transmission wavelengths.

[0045] In other words, the first and second infrared filters are for filtering IR light in different IR wavelength intervals, for example, a first IR band in which a typical fluorescent dye emits a first fluorescence peak, and a second IR band in which a typical fluorescent dye emits a second fluorescence peak. Typically, the second IR band is located at a higher wavelength compared to the first IR band. The first and second infrared filters can also be tuned to the emission bands of different fluorescent agents. Thus, for example, the emission of a first fluorescent agent can pass through the first filter (and be blocked in particular by the second filter) and be detected on the corresponding first sensor, and the emission of a second fluorescent agent can pass through the second filter (and be blocked in particular by the first filter) and be detected on the corresponding second sensor. For example, the first filter can be configured to measure the fluorescence emission of methylene blue, and the second filter can be configured to measure the fluorescence emission of ICG.

[0046] According to an advantageous embodiment of the present invention, the CDSS is further enhanced in that an illumination unit, a fluorescence imaging unit, and a visible light imaging unit are arranged in a single image acquisition device, and this single image acquisition device further includes a measuring unit configured to measure the distance between the surfaces of body parts to be captured in a visible light image. Furthermore, the CDSS may be further configured so that the image acquisition device outputs a signal indicating the measured distance. Measurements at different distances may be performed to optimize illumination and image acquisition in order to find the best image acquisition conditions. This best-fit distance may then be stored in the imaging system as a target distance for subsequent measurements.

[0047] For example, visible and / or audio signals may be output by the image acquisition device. These signals can guide the operator when operating the image acquisition device so that image acquisition is performed at at least a nearly constant distance from the surface of the body part. The integration of distance sensors and user-assisted outputs (visual or optical signals) in the image acquisition device allows the operator to acquire images with more uniform illumination. This improves the quality of fluorescence signal measurements.

[0048] To determine the best matching distance, the image acquisition device can repeatedly capture fluorescence and visible light images of the same portion of the body surface while measuring the distance. Multiple sets of fluorescence and visible light images can be captured at different distances. The set of images can then be analyzed, taking into account the imaging quality, to determine the best matching distance that yields the highest quality image. With this in mind, the output signal, which may be an audio or optical signal, can also indicate the deviation of the measured distance from the best matching distance. This directly informs the operator whether optimal acquisition conditions, particularly with respect to illumination, are being applied during the measurement.

[0049] According to yet another advantageous embodiment of the present invention, the CDSS further has an input interface that is a direct link to an electronic patient record, and patient-related data, in particular the patient's age, sex, height, weight, BMI (body mass index), fat mass, muscle mass, daily exercise level, occupation, skin color, medication status, presence or absence of vascular disease, in particular varicose veins or venous edema, disease, in particular dialysis or diabetes, blood albumin level, renal function, hepatic function, cardiac function, blood hemoglobin (Hb) concentration, blood estimate, lipid metabolism rate, blood glucose concentration, urea nitrogen (blood urea BUN, amount of UN), ABI (ankle-brachial index) value, and the same level before the onset of lymphedema. PlaceThe system is further enhanced in that data on lymphatic function measurements, endocrine information, and hormone levels are provided through an input interface as additional input features to the artificial intelligence model. By further considering this additional input information, the output, i.e., the classification of lymphedema induced by fluorescence patterns, can be enhanced. For example, the input interface may include hardware that allows the user to manually input data such as the patient's age, sex, height, weight, BMI (body mass index), fat mass, muscle mass, daily exercise level, occupation, skin color, medication status, presence or absence of vascular disease, especially varicose veins or venous edema, presence or absence of disease, especially dialysis or diabetes, blood albumin levels, renal function, hepatic function, cardiac function, blood hemoglobin (Hb) concentration, blood estimates, lipid metabolism rate, blood glucose concentration, urea nitrogen (blood urea BUN, amount of UN), ABI (ankle-brachial index) value, data on lymphatic function measurements at the same location (40A) before the onset of lymphedema, endocrine information, and hormone levels. Downloading this information from electronic patient records, if available, will speed up this process.

[0050] As mentioned above, the input features of the AI ​​model can be either 2D or 3D images, and the fluorescence image may be a combination of a visible light image and a fluorescence image. Furthermore, it is possible to use two or more fluorescent agents. For example, fluorescent agents such as ICG (indocyanine green) and methylene blue can be used. The image to be analyzed may be a single-dye image, a combined fluorescence image, i.e., two or more dyes that can be further combined with a visible light image. It may be a stitched image or a reconstructed image. For example, it is also possible to process an overlay image containing a visible light image and a fluorescence image.

[0051] Based on this information, the AI ​​model can perform image segmentation using algorithms such as watersheds, thresholding, clustering, and histograms. It is also possible to define user-defined regions of interest and / or segment lymphatic details within the image.

[0052] A software module for performing interference operations based on an AI model may be a pre-trained AI model for classifying fluorescence patterns. The AI ​​may also include multiple pre-trained AI models. The user can then select from different AI models or systems to best suit their needs.

[0053] In yet another advantageous embodiment, the input interface may include a user interface for correcting the automatically generated classification of the fluorescence pattern. This is particularly applicable to the automatic generation of the severity and clinical type stage of the fluorescence pattern. The corrected classification may be stored or assigned in relation to the fluorescence image and patient-related data, in addition to the automatic result for the classification which is the output of the AI ​​model. It is also possible to override the AI ​​prediction data.

[0054] In yet another advantageous embodiment, the AI ​​model can be customized through user-specific training. The user can select a training mode and input their own data, meaning their own fluorescence images as ground truth. The user-trained AI model is then stored in addition to the pre-trained models. The user can choose which AI model to use. The user-trained AI model may be particularly advantageous when a proprietary staging system that deviates from, for example, a common staging system for lymphedema, such as the WHO definition, is used.

[0055] This objective is further addressed by a computer-implemented method for determining the classification of lymphedema-induced fluorescence patterns using a computer-based clinical decision support system (CDSS), where the classification is based on fluorescence images determined by measuring the fluorescence signal in the tissue of a body part to which a fluorescent agent has been added. Computer implementation The method is, Receiving patient-specific fluorescence images as input features for an artificial intelligence (AI) model through an input interface, The processor performs the inference operation. death , fluorescence image but Applied to an AI model, it generates a classification of lymphedema-induced fluorescence patterns. Ruko Toto, The classification of lymphedema-induced fluorescence patterns is communicated to the user through a user interface (UI), Includes.

[0056] The same or similar advantages mentioned with respect to CDSS also apply to computer implementations in the same or similar manner and are therefore not repeated.

[0057] Furthermore, book Computer implementation The method can be advantageously enhanced in that the classification of lymphedema-induced fluorescence patterns corresponds to the stage of lymphedema severity and / or the clinical type of fluorescence pattern.

[0058] Furthermore, book Computer implementation The method is advantageously enhanced in that fluorescence images and corresponding visible light images are provided to the artificial intelligence (AI) model as input features through an input interface.

[0059] According to yet another advantageous embodiment, Computer implementationThe method involves an input interface configured to measure fluorescence signals within the tissue of a body part and receiving fluorescence images and corresponding visible light images via a direct link to an image capture processing device configured to image the surface of the body part, wherein the tissue to which the fluorescent agent is added forms part of the body part, and the image capture processing device comprises an illumination unit, a fluorescence imaging unit and a visible light imaging unit. Computer implementation The method is A process in which an illumination unit illuminates the tissue with excitation light having a wavelength suitable for generating emitted light through the excitation and emission of a fluorescent agent, To provide a fluorescence image, the process involves capturing the fluorescence image using a fluorescence imaging unit by spatially resolved measurement of emitted light, A process of capturing a visible light image by a visible light imaging unit by capturing a corresponding visible light image of a portion of the surface of a body part. And, further including, The fluorescence imaging unit and the visible light imaging unit are configured such that the viewing direction and / or viewpoint of the fluorescence image and the corresponding visible light image are linked through a known relationship. ru, It is strengthened in that respect.

[0060] Furthermore, book Computer implementation The method involves providing a large fluorescence image and a corresponding large visible light image as input features to an artificial intelligence (AI) model through an input interface. The fluorescence imaging unit and the visible light imaging unit repeatedly capture fluorescence images and visible light images to provide a series of fluorescence images and a series of visible light images. The image capture processing device further comprises a processing device that includes a stitching unit, Computer implementation The method is To generate large visible light images of body parts, a stitching unit is used The set of stitching parameters is determined and applied. stitching algorithm In a series of visible light images Applicable work To the extent , The process involves a stitching unit further applying a stitching algorithm to a series of fluorescence images in order to generate a larger fluorescence image. And, further including, Applied to a series of fluorescence images Stitching algorithm So , large Visible light image generate The set of stitching parameters determined at that time but Applicable So ru ,and In that respect, it can be strengthened to an advantage.

[0061] It is even more advantageous if the viewing direction and field of view of the fluorescence image and the visible light image are the same. In particular, the fluorescence image and the visible light image are captured through one of the same objective lenses.

[0062] Furthermore, book Computer implementation The method can be enhanced in that the acquisition of fluorescence images and visible light images is performed simultaneously, particularly without time switching between the fluorescence image signal and the visible light image signal.

[0063] Furthermore, according to an advantageous embodiment, the steps of capturing a fluorescence image, illuminating the tissue with excitation light, and simultaneously capturing a visible light image are performed by a single image capture device.

[0064] Furthermore, according to another embodiment, Computer implementation The method may further include a step of measuring the distance between the surface of the body part to be captured in the visible light image and the capturing device. Furthermore, the imaging device may output a signal indicating the measured distance. The method may further include a step of repeatedly capturing fluorescence and visible light images of the same portion of the body part's surface while measuring the distance. Multiple sets of fluorescence and visible light images may be captured at different distances. The sets of images can be analyzed considering the imaging quality to determine the best matching distance and obtain the highest quality image. Thus, the output signal can indicate the deviation of the measured distance from the best matching distance.

[0065] In yet another advantageous embodiment, a method for determining the classification of lymphedema-induced fluorescence patterns is a method in which the measurement of fluorescence signals is performed on tissue to which at least first and second fluorescent agents are added, and a step of capturing a fluorescence image is performed. To capture a first fluorescence image in a first wavelength range, which is generated by illuminating the tissue with first excitation light having a first wavelength suitable for generating emitted light by the first excitation emission of a first fluorescent agent, To capture a second fluorescence image within a second wavelength range, which is generated by illuminating the tissue with a second excitation light having a second wavelength suitable for generating emitted light by the second excitation emission of a second fluorescent agent, Includes, The first and second fluorescence images are provided to the artificial intelligence (AI) model as input features through the input interface. The input interface receives the first and second fluorescence images as input features for the artificial intelligence (AI) model. The processor can be further enhanced by performing inference operations by applying the first and second fluorescence images to an AI model to generate classifications of lymphedema-induced fluorescence patterns.

[0066] Furthermore, book Computer implementation The method can be enhanced in that patient-related data, particularly data concerning the patient's gender, age, or BMI, is provided through an input interface as further input features to an artificial intelligence (AI) model, via a direct link to electronic patient records.

[0067] The purpose of this is a method for diagnosing lymphedema, The process involves administering a fluorescent agent to a part of the patient's body, A step of determining the classification of lymphedema-induced fluorescence patterns using a computer-based clinical decision support system (CDSS), wherein the classification is based on fluorescence images determined by measuring the fluorescence signal in the tissue of a body part to which a fluorescent agent has been added. The method includes, Receiving patient-specific fluorescence images as input features for an artificial intelligence (AI) model through an input interface, This involves the processor performing inference operations, where fluorescence images are applied to an AI model to generate classifications of lymphedema-induced fluorescence patterns, and To derive diagnostic results, particularly regarding the stage of lymphedema, from the classification of lymphedema-induced fluorescence patterns, The classification and diagnostic results of lymphedema-induced fluorescence patterns are communicated to the user through the user interface (UI), This is further resolved by methods that include the following.

[0068] The method for diagnosing lymphedema can be performed with a higher level of accuracy and reliability. This method is advantageous because it does not rely on the judgment of the medical professional interpreting the fluorescence pattern. Such judgments are inevitably of lower quality due to the personal experience of the medical professional. This method is more objective and therefore more accurate.

[0069] The purpose of this is a method for the long-term treatment of lymphedema, A step of diagnosing the severity of lymphedema by performing a method for diagnosing lymphedema on a patient according to one or more of the embodiments described above, A process for providing treatment to a patient, wherein the treatment is customized according to the diagnostic results regarding the severity of lymphedema, A process comprising repeatedly performing the steps of diagnosing the severity of lymphedema and administering treatment to the patient, wherein in each iteration, the treatment is adjusted to the stage in which lymphedema was detected, particularly to the stage in which lymphedema was detected. This is further resolved by methods including the following.

[0070] The long-term treatment method according to aspects of the present invention is particularly useful because, in contrast to conventional methods, the diagnosis of lymphedema provides objective results regarding the severity of the disease. The success of long-term treatment can be analyzed from an objective standpoint.

[0071] CDSS applies artificial intelligence models. Machine learning (ML) is a branch of artificial intelligence. Machine learning algorithms build models based on sample data known as training data in order to make predictions or decisions without being explicitly programmed to do so.

[0072] Machine learning (ML) has two common modes: supervised ML and unsupervised ML. Supervised ML uses prior knowledge (e.g., correlating inputs to outputs or results) to learn the relationship between inputs and outputs. The goal of supervised ML is to learn a function that best approximates the relationship between training inputs and outputs given a portion of training data, so that the ML model can implement the same relationship when given inputs to produce the corresponding outputs. Unsupervised ML is the training of ML algorithms using unclassified and unlabeled information, allowing the algorithm to act on that information without guidance. Unsupervised ML is useful in exploratory analysis because it can automatically identify structures within the data.

[0073] Common tasks in supervised machine learning are classification and regression problems. Classification problems, also called categorization problems, aim to classify an item into one of several categorical values ​​(e.g., Is this object an apple or an orange?). Regression algorithms aim to quantify some items (e.g., by assigning a score to a given input value). Some examples of commonly used supervised ML algorithms include logistic regression (LR), naive Bayes, random forest (RF), neural networks (NN), deep neural networks (DNN), matrix factorization, and support vector machines (SVM).

[0074] Some common tasks for unsupervised machine learning include clustering, representation learning, and density estimation. Some examples of commonly used unsupervised machine learning algorithms are K-means clustering, principal component analysis, and autoencoders.

[0075] Another type of machine learning is federated learning (also known as collaborative learning), which trains algorithms across multiple distributed devices holding local data without exchanging data. This approach contrasts with traditional centralized machine learning techniques where all local datasets are uploaded to a single server, as well as more classical distributed methods that often assume local data samples are uniformly distributed. Federated learning allows multiple actors to build a common, robust machine learning model without sharing data, thus enabling them to address critical issues such as data privacy, data security, data access rights, and access to heterogeneous data.

[0076] In some cases, the AI ​​model may be trained continuously or periodically before the processor performs an inference operation. Then, during the inference operation, patient-specific input features provided to the AI ​​model may propagate from the input layer, through one or more hidden layers, to an output layer corresponding to the classification of lymphedema-induced fluorescence patterns. For example, the stage of the severity or level of lymphedema, and / or the clinical type of fluorescence pattern, e.g., one of linear, splash, stardust, or diffuse patterns.

[0077] During and / or following the inference process, the classification of lymphedema-induced fluorescence patterns may be communicated to the user via the user interface (UI). For example, fluorescence images or overlay images may be displayed on the screen along with information about the lymphedema-induced fluorescence patterns. This could be, for example, a report showing the level of lymphedema or the clinical type of fluorescence pattern corresponding to the confidence level generated by the AI. The report may also include suggested diagnostic and / or treatment options.

[0078] Further features of the present invention will become apparent from the description of embodiments according to the present invention, along with the claims and accompanying drawings. Embodiments according to the present invention may satisfy individual features or combinations of features.

[0079] The present invention is described below based on exemplary embodiments without limiting the general intent of the invention, and all disclosures of details of the present invention not described in further detail herein are expressly referenced to the drawings. [Brief explanation of the drawing]

[0080] [Figure 1] This diagram shows a schematic representation of an image capture processing device that forms part of the CDSS (Chronic Distress Scoreboard). [Figure 2] A schematic diagram of the image acquisition device and the processing device of the image acquisition processing device is shown. [Figure 3] Examples of visible light images and their corresponding fluorescence images are shown. [Figure 4] This figure shows a large overlay image partially generated from the visible light image and fluorescence image shown in Figure 3(a) and Figure 3(b). [Figure 5] This shows the internal prism assembly of the image acquisition device. [Figure 6] A schematic diagram of an endoscope or laparoscope including an image acquisition device is shown. [Figure 7] A flowchart of the stitching algorithm is shown. [Figure 8] A schematic diagram of a computer-based clinical decision support system (CDSS) is provided. [Figure 9A] An example of a fluorescence pattern is shown. [Figure 9B] An example of a fluorescence pattern is shown. [Figure 9C] An example of a fluorescence pattern is shown. [Figure 9D] An example of a fluorescence pattern is shown. [Figure 10] A schematic diagram of another internal prism assembly for an image acquisition device is shown. [Modes for carrying out the invention]

[0081] In drawings, elements of the same or similar type, or their corresponding parts, are given the same reference number to avoid the need to reintroduce items.

[0082] Figure 1 shows an image capture and processing device 2 that can function as an input interface to a computer-based clinical decision support system (CDSS). The CDSS will be explained with reference to Figure 8. First, let's focus on data acquisition by the image capture and processing device 2.

[0083] The image acquisition and processing device 2 is configured to measure fluorescence signals within the tissue of a body part 4 of patient 6. As a mere example, the body part 4 of patient 6 being examined is the arm. Fluorescence signal measurements can also be performed on other body parts 4 of patient 6, such as the legs, part of the head, the neck, the back, or any other part of the body. Measurements can also be performed during incision surgery. In this application scenario, the body part 4 could be, for example, an internal organ of patient 6. Fluorescence signal measurements can also be performed during minimally invasive surgery. In this application scenario, the image acquisition and processing device 2 is at least partially integrated with, for example, an endoscope or laparoscope. For example, the endoscope or laparoscope includes the image acquisition device 10.

[0084] Before the measurement begins, a fluorescent agent 8 is administered, or injected, into the tissue of the patient's body part 4. The capture of the fluorescent signal by the image acquisition processing device 2 within the tissue of body part 4 excludes the step of administering the fluorescent agent 8.

[0085] Fluorescent agent 8 is, for example, ICG. ICG (indocyanine green) is a green medical dye that has been used for over 40 years. ICG fluoresces when excited by near-infrared light with wavelengths of 600 nm to 800 nm. The emitted fluorescence is in the 750 nm to 950 nm range. Fluorescent agent 8 may also contain two different medical dyes. For example, fluorescent agent 8 may be a mixture of methylene blue and ICG.

[0086] Following the administration of the fluorescent agent 8, the patient's body portion 4 can be examined using an image capture device 10, which forms part of the image capture processing device 2, as indicated by the arrows in Figure 1.

[0087] During fluorescence imaging of lymphedema, images are typically captured at two time points. The first image is captured during the initial transient phase when the fluorescent agent is rapidly taken up by the lymphatic system, and the second image is captured during the late plateau phase. All fluorescence images processed by CDSS are captured during this late plateau phase.

[0088] The image acquisition device 10 is configured to image the surface 11 of a body part 4 and to detect the fluorescence signal generated from the illumination of a fluorescent agent 8 with excitation light. When the image acquisition device 10 is applied to surgery, the surface 11 of the body part 4 is, for example, the surface of an internal organ. In this case, the surface 11 of the body part 4 is identical to the surface of the tissue to which the fluorescent agent 8 has been administered. For the emission of light with an appropriate excitation wavelength, the image acquisition device 10 includes an illumination unit 16 (not shown in Figure 1).

[0089] The captured image is communicated to a processing device 12 that can implement CDSS. The CDSS is configured to output a classification of lymphedema-induced fluorescence patterns based on the fluorescence image captured by the image capture device 10. According to this embodiment, the image capture processing device 2 Image capture device 10 It functions as an input interface for artificial intelligence models.

[0090] Note that the input interface does not necessarily have to be provided by the image capture processing device 2. This will be explained later with reference to Figure 8. In the embodiment shown in Figure 1, the input interface is provided by the image capture device 10, as described above. It provides a fluorescence image specific to patient 6 to an artificial intelligence (AI) model implemented in the processing device 12. The processor of the processing device 12 performs interference operations in which the fluorescence image is applied to the AI ​​model to generate a classification of lymphedema-induced fluorescence patterns. A typical lymphedema-induced fluorescence pattern is shown in Figure 9, which will be explained in more detail below.

[0091] The classification of lymphedema-induced fluorescence patterns is communicated to the user, or in a given embodiment, to physician 3. The output is performed via a user interface, which may be a display 14. The output of the AI ​​model, i.e., the classification of lymphedema-induced fluorescence patterns, along with fluorescence images and visible light images, can be output on the display 14. The image capture device 10 can be operated by physician 3.

[0092] Figure 2 shows the image capture device 10 and processing device 12 of the image capture and processing device 2 in more detail. The image capture device 10 includes an illumination unit 16 configured to illuminate tissue with excitation light having a wavelength suitable for generating fluorescence by exciting the emission of a fluorescent agent 8. The illumination unit 16 is provided with, for example, multiple LEDs.

[0093] The image acquisition device 10 further comprises an objective lens 18 for capturing visible light and fluorescence. Light is directed through the objective lens 18 to a prism assembly 20. The prism assembly 20 is configured to separate fluorescence, particularly in the wavelength range of 750 nm to 950 nm, from the visible light that yields a visible light image. The fluorescence is directed to a fluorescence imaging unit 22, which is, for example, a CCD or CMOS sensor with additional wavelength filters and electronics as needed. The fluorescence imaging unit 22 is configured to capture a fluorescence image by spatially resolved measurement of emitted light, i.e., the excitation emission of the fluorescent agent 8, in order to provide a fluorescence image. Furthermore, there is a visible light imaging unit 24, which may be another CCD or CMOS sensor with additional different wavelength filters and electronics as needed. The prism assembly 20 is configured to direct visible light to the visible light imaging unit 24 so that the unit can capture a visible light image of a portion of the surface 11 of a patient's body part 4. Similarly, the prism assembly 20 is configured to direct fluorescence to the fluorescence imaging unit 22. Details of the prism assembly 20, fluorescence imaging unit 22, and visible light imaging unit 24 will be described later.

[0094] The image acquisition device 10 may be a scanning unit, such as an image line scan unit or a LiDAR scan unit. The image acquisition device 10 may also be a 3D camera suitable for capturing a pair of stereoscopic images that can compute a 3D image containing depth information. Of course, the image acquisition device 10 may be a combination of these devices.

[0095] Image data is communicated from the image acquisition device 10 to the processing device 12 via a suitable data link 26, which can be a wireless data link or a wired data link, such as a data cable.

[0096] The image acquisition device 10 is configured to operate the fluorescence imaging unit 22 and the visible light imaging unit 24 to simultaneously acquire visible light and fluorescence images. In particular, the image acquisition device 10 does not perform time switching between the fluorescence image signal and the visible light image signal. In other words, the sensors of the fluorescence imaging unit 22 and the visible light imaging unit 24 are used exclusively to acquire images in their respective wavelength ranges, meaning that the sensors of imaging units 22 and 24 are used to acquire fluorescence images in the IR spectrum or visible light images in the visible spectrum. Imaging units 22 and 24 are not used to acquire images in both wavelength ranges. This offers a significant advantage. For example, the sensors can be positioned to be precisely focused, which is not possible when an image sensor is used for both purposes, i.e., to acquire visible light and infrared light, because the focal positions for these different wavelengths are usually different. Furthermore, sensor parameters can be individually adjusted, for example, with respect to the required exposure time or sensor gain. Individual settings are advantageous because the IR signal is typically lower than the visible light signal.

[0097] The fluorescence imaging unit 22 and the visible light imaging unit 24 have a fixed spatial relationship with respect to each other. This is because the units are arranged on a single mounting structure or frame of the image acquisition device 10. Furthermore, the fluorescence imaging unit 22 and the visible light imaging unit 24 use the same objective lens 18 and prism assembly 20 for imaging the fluorescence image and the visible light image, respectively. These means that the fluorescence imaging unit 22 and the visible light imaging unit 24 are configured such that the viewing direction and viewpoint of the fluorescence image and the visible light image are linked through a known and fixed relationship. In a given embodiment, the viewing direction of the two images is the same, and both units 22, 24 image through the same objective lens 18.

[0098] The image acquisition device 10 can be configured to operate the fluorescence imaging unit 22 and the visible light imaging unit 24 to repeatedly acquire fluorescence and visible light images, thereby providing a series of fluorescence and visible light images. This operation can be performed by the processing device 12 operating the image sensors of the fluorescence imaging unit 22 and the visible light imaging unit 24. The series of images are typically acquired while an operator or physician 3 (see Figure 1) moves the image acquisition device 10 along the longitudinal direction L of a body part 4 of the patient 6. This movement may be performed so as to include overlapping portions of subsequent images in the series. In other words, details shown in the first image of a series of images are also shown in the second image of a subsequent image in the series. This is important for subsequent stitching processes as desired. To protect that corresponding features may be found in subsequent images, the frequency of image acquisition may be set to a sufficiently high value. Image acquisition can be initiated manually, for example by physician 3, or image acquisition can be controlled by the image acquisition device 10 so as to satisfy the described preconditions.

[0099] The image acquisition device 10 may further be configured to acquire its position and orientation while in motion. For example, the position and orientation of the image acquisition device 10 in a reference system of the examination room or of the patient 6 can be determined for each image or pair of images being captured. This information can be stored and communicated along with the image or pair of images, including visible and fluorescent images. This information may be useful for subsequent image reconstruction, such as generating a 3D image from a series of 2D images.

[0100] Two sets of images (i.e., a first set of visible light images and a second set of fluorescence images) or a set of image pairs (each image pair including a fluorescence image and a visible light image) imageOnce captured by the capture device 10 and received into the processing device 12, a series of visible light images can be processed by the stitching unit 28 (see Figure 2). The stitching unit 28 is configured to apply a stitching algorithm to the series of visible light images to generate a larger visible light image of the body part 4. The larger image is "larger" in that it shows a larger portion of the patient's body part 4 compared to a single image.

[0101] Within the context of this specification, the term “stitching” should not be understood as limiting the stitching process to the combination of two or more 2D images. Stitching can also be performed based on 3D images, and the result of this process is a larger 3D image. The stitching process can also be performed based on 2D images and additional information regarding the orientation of the view in which the 2D images were captured. Further information regarding the position of the image capture device 10 can also be taken into consideration. As described above, a larger 3D image can be generated based on these datasets, that is, a larger 3D image can be stitched together from a series of 2D images plus information regarding the position and orientation of the image capture device 10. For example, 3D scan data from a LIDAR sensor can be combined with 2D image information. In this case as well, the result of the stitching process is a larger 3D image. The term stitching, in particular, encompasses the process of reconstructing a 3D image from a dataset.

[0102] Regardless of the specific type of data being processed during stitching, the stitching algorithm begins with stitching visible light images. The stitching algorithm generates and applies a set of stitching parameters when performing the stitching operation. The detailed operation of the stitching unit 28 will be described later. The stitching unit 28 is configured to apply the stitching algorithm to a series of fluorescence images as well as a series of visible light images in order to generate a large fluorescence image. In this case as well, the stitching process is not limited to a combination of two or more 2D images. It is also possible to generate 3D fluorescence images in a similar manner to that described above for visible light images.

[0103] The stitching algorithm applied to the stitching of fluorescence images is the same algorithm used for stitching visible light images. Furthermore, the stitching of fluorescence images is performed using the same set of stitching parameters determined when performing the stitching of visible light images. This is possible because there is a fixed relationship between the viewing direction and field of view of the visible light image and the fluorescence image. Naturally, if the viewing direction and viewpoint of the visible light image and the fluorescence image are not the same, it is necessary to apply a fixed offset or a shift in the stitching parameters. This takes into account the known fixed spatial relationship between the IR and Vis image sensors and the corresponding optical system.

[0104] Following stitching, large visible light and large fluorescence images become available and can function as input to an AI model. Large visible light and fluorescence images can also be output. For example, the images are displayed side-by-side on display 14 along with the AI ​​model results, i.e., the classification of lymphedema-induced fluorescence patterns.

[0105] Unlike conventional examination systems, the display 14 can show corresponding visible light and fluorescence images along with the results of the AI ​​model. Details visible in the fluorescence image, such as high fluorescence intensity indicating characteristic patterns of lymphatic fluid accumulation or reflux, can be precisely located in the patient's body part 4 at the corresponding location shown in the visible light image. This allows the physician 3 to precisely locate areas where, for example, lymphatic fluid accumulation is found. This is extremely useful information for, for example, tailored specific treatment for the patient 6.

[0106] Visible light images and fluorescence images, particularly large visible light images and large fluorescence images, can also be superimposed to provide overlay images, particularly large overlay images, of body parts 4. This processing can be performed by the superposition unit 30 of the processing device 12. The overlay images can also be output via the display 14.

[0107] Figure 3 (a) This shows an example of a visible light image 5 in which a portion of the surface 11 of body part 4 of patient 6 is visible. As an example, patient 6 A portion of its legs is shown. Figure 3 (b) is, patient 6 The image shows the corresponding fluorescence image 7, determined by measuring the fluorescence signal of the fluorescent agent 8 applied to the leg tissue. High-intensity spots or regions of fluorescence are visible. This type of fluorescence image 7 These typically fall into the category of diffuse patterns indicating a late or severe stage of lymphedema. Different types of lymphedema-induced fluorescence patterns are described in more detail below. The observed fluorescence patterns are in patients 6 The image strongly indicates slow lymphatic transport and lymphatic accumulation resulting from lymphedema in the leg. Advantageously, physician 3 can identify the location of the area where slow lymphatic transport is occurring by comparing the fluorescence image 7 with the visible light image 5. Furthermore, physician 3 This provides, for example, the output of an AI model that is the stage of lymphedema severity and / or the clinical type of fluorescence pattern.

[0108] Figure 4 shows overlay image 9, and Figure 3 (a) and Figure 3 (b) In addition to the images shown, visible light image 5 and fluorescence image 7 have been stitched together. Exemplary single visible light image 5 and fluorescence image 7 can also be seen in Figure 4, projected between the dashed lines shown in the large overlay image 9. By stitching together the visible light image 5 and fluorescence image 7, a large overlay image 9 showing almost the entire body part 4 of patient 6 can be provided. The fluorescence signal can be shown in false color to clearly distinguish it from the features of the visible light image 5. Also, the overlay image 9 This can function as input to an AI model of CDSS.

[0109] Figure 5 shows a prism assembly 20 that can be implemented in the image acquisition device 10. The first prism P1 is a pentaprism. The incident light beam A, which is visible and fluorescent, enters the first prism P1 via the incident surface S1 and is partially reflected by surface S2, which is one of two surfaces not adjacent to the incident surface S1. Next, the reflected beam B is reflected off the first surface of the surface adjacent to the incident surface S1. The reflection angle can be less than the critical angle so that the reflection is not internal (the adjacent surface can be coated to avoid light leakage and reflect the desired target wavelength). Then, the reflected beam C intersects with the incident light beam A, exits the first prism P1 through the second surface of the surface adjacent to the incident surface S1, and heads towards the sensor D1. A portion of beam A passes through surface S2 and enters the compensating prism P2. Using the two non-internal reflections, the incident beam A can be directed towards the sensor D1 via beams B and C. Furthermore, there may be no gap between prisms P1 and P2, no gap between prisms P3 and P4, and no gap between prisms P2 and P3. Prism P2 is a compensator prism for adjusting the individual lengths of the optical path from the incident surface S1 to sensors D1 to D5.

[0110] prismFrom P2, beam D is incident on the second pentaprism P3. Similar to prism P1, inward reflection is used to intersect the beam itself. For brevity, the beam description will not be repeated except to state that in prism P3, beam portions E, F, and G correspond to beam portions A, B, and C in prism P1, respectively. Prism P3 also cannot use internal reflection to reflect the incident beam toward sensor D2. Two non-internal reflections can be used to direct the incident beam E toward sensor D2 via beams F and G.

[0111] Following prism P3 is another compensating prism P4. Finally, beam H is incident on a dichroic prism assembly containing prisms P5, P6, and P7, each having sensors D3, D4, and D5, respectively. The dichroic prism assembly is for splitting visible light into red, green, and blue components directed towards sensors D3, D4, and D5, respectively. Light enters the prism assembly through beam I. Optical coating C1 is positioned between prisms P5 and P6, and another optical coating C2 is positioned between prisms P6 and P7. Each optical coating C1 and C2 has different reflectivity and wavelength sensitivity. Optical coating At C1, the incident beam I is partially reflected off the same prism plane from which the light was incident (beam J). On that same plane, the beam, now labeled K, is reflected again towards sensor D3. beam From J beam The reflection to K is an internal reflection. Therefore, sensor D3 receives light reflected by coating C1, sensor D4 receives light from beam L (beams M and N) reflected by coating S2 in an analog manner, and sensor D5 receives light from beam O that crosses the prism without obstruction.

[0112] There is an air gap between prism P4 and prism P5. In the prism assembly 20, the following total path length may be defined for each endpoint channel (defined with respect to the sensor at the end of the channel): Sensor D1 (e.g., first near-infrared) path: A + B + C Sensor D2 (e.g., second near-infrared) path: A+D+E+F+G Sensor D3 (e.g., red) path: A+D+E+H+I+J+K Sensor D4 (e.g., blue) path: A+D+E+H+I+ L+M+N Sensor D5 (e.g., green) path: A+D+E+H+I+ L+O

[0113] The path length is A+B+C=A+D+E+F+G=A+D+E+H+l+J+K=A+D+E+H+l+ L+M+N =A+D+E+H+I+ L+O It will be aligned in such a way.

[0114] Matching the path lengths may involve adjusting the focal plane focus position difference for wavelengths detected by sensors D1-D5. That is, for example, the path length towards the sensor for blue (B) light may not be exactly the same as the path length towards the sensor for red (R) light. This is because the ideal distance for producing a sharp, focused image depends to some extent on the wavelength of light. Prisms can be configured to accommodate these dependencies. The length of D+H can be adjusted by the lateral displacement of compensator prisms P2 and P4, which can act as focus compensators due to the wavelength shift.

[0115] Larger gaps in path I can be used for additional filtering or filled with glass compensators for focus shift and compensation. A gap is required on the specific bottom surface of the red prism due to internal reflections in the path from beam J to beam K. Space can be provided between the prism output surface and each of sensors D1-D5 to provide additional filtering, or should be filled accordingly with glass compensators.

[0116] Sensors D1 and D2 are IR sensors configured to capture the fluorescence image 7. For example, sensors D1 and D2, along with appropriate electronics, are part of the fluorescence imaging unit 22. Sensors D3, D4, and D5 are for capturing the three components of the visible light image 5. For example, sensors D3, D4, and D5, along with appropriate electronics, are part of the visible light imaging unit 24. It is also possible to consider corresponding prisms that direct the light beam to the sensors, which are part of each unit, namely the fluorescence imaging unit 22 and the visible light imaging unit 24.

[0117] Figure 6 schematically shows an endoscope 50 or a laparoscope. Considering aspects of the present invention, the differences between a laparoscope and an endoscope are relatively small. Therefore, when the description refers to an endoscope, a laparoscopic configuration is also usually possible. For the sake of argument, the following refers to an endoscope 50.

[0118] The endoscope 50 can function as an input interface for the CDSS. The endoscope 50 includes an image capture device 10, which captures fluorescence images that may be input features for the AI ​​model, as described in more detail above. The image capture device 10 includes an objective lens 18 that captures fluorescence images 7 and visible light images 5. The objective lens 18 focuses light incident from the incident surface S1 of the prism assembly 20 onto sensors D1 to D5. The objective lens 18 can also be integrated into the last part of the endoscope to match the rear focal length of the prism.

[0119] The endoscope 50 includes an optical fiber 52 connected to a light source 54 that couples light to the endoscope 50. The light source 54 can provide white light for illuminating the surface 11 of the body part 4 and capturing a visible light image 5. The light source 54 can also be configured to emit excitation light suitable for exciting a fluorescent dye applied as a fluorescent agent to emit fluorescence. In other words, the light source 54 can be configured to emit both visible light and light in the IR spectrum.

[0120] Inside the shaft 56 of the endoscope 50, the optical fiber 52 splits into multiple fibers 51. The endoscope 50 may have a flexible shaft 56 or a rigid shaft 56. In the rigid shaft 56, a lens system consisting of lens elements and / or relay rod lenses can be used to guide light through the shaft 56. If the endoscope 50 has a flexible shaft 56, the fiber bundle 51 can be used to guide light from the light source 54 to the tip of the endoscope shaft 56. A fiber bundle 58 is located inside the shaft 56 of the endoscope 50 to guide light coming from the examination area from the distal tip of the endoscope shaft 56 (not shown in Figure 6) to the image acquisition device 10 located at the proximal end of the shaft 56. In another embodiment not shown, the entire image acquisition device 10 may be miniaturized and located at the distal tip or distal end of the endoscope shaft 56.

[0121] Figure 7 shows a flowchart of a stitching algorithm that can be used for stitching visible light and fluorescence images. The flowchart is more or less self-evident and is explained very simply. First, a series of acquired images (S1) are transferred to the stitching unit 28 of the processing device 12. Next, the algorithm performs frame pre-selection (step S2). In this pre-selection step, frames suitable for stitching are selected. S3 represents the selected image to be stitched, which is then subjected to pre-processing (step S4). Feature extraction is performed on the pre-processed image (S5) (step S6). Once image features are extracted (S7), image matching is performed using the known image from S3 and the features extracted from step S7 (step S8). Based on the selected image (S9), the image transformation is estimated (step S10). This estimated image transformation (S11), also called the stitching parameter, is applied (step S12). The application of the transformation results in a transformed image (S13). Further image correction, such as exposure correction, can be performed (step S14). The transformed and corrected image (S15) is stitched together by identifying the locations of seams (i.e., lines along which the images are joined) (step S16). Data indicating the locations of the seams (S17) is used together with the transformed and corrected image (S12) to create a composite image (step S18). In a given embodiment, this results in a large visible light image or a large fluorescence image as a result of stitching (S19).

[0122] Furthermore, the image acquisition device 10 applied to capture visible light images 5 and fluorescence images 7 may further include a measuring unit 32 configured to measure the distance d (see Figure 1) between the surface 11 of the patient's body part 4 to be captured in the visible light image 5 and the image acquisition device 10. A distance sensor 33 communicating with the measuring unit 32 is, for example, an ultrasonic sensor, a laser distance sensor, or any other suitable distance measuring device. Furthermore, the image acquisition device 10 is configured to output a signal indicating the measured distance d. For example, the image acquisition device 10 outputs an optical or acoustic signal that gives the operator of the device 10 information about the best distance d for performing the measurement. Performing the measurement at a constant distance d significantly improves the measurement results, especially due to uniform illumination.

[0123] In addition to the distance sensor 33, the image acquisition device 10 may include an internal measuring unit (IMU) 35, which may be used to collect rotational data in pitch, yaw, and roll, as well as acceleration data in three spatial axes (x, y, and z). This information from the IMU may be used as additional data to improve the performance of the stitching algorithm, or to provide the operator with feedback for positioning and rotating the camera, thereby providing better images for the stitching algorithm.

[0124] The processing device 12 may further comprise an AI unit 60 that implements the CDSS. Figure 8 schematically shows a computer-based clinical decision support system (CDSS) 62. The CDSS comprises an input interface 64, a processor 66 that implements an AI model 68, and an output interface 70. The input interface 64 is coupled to an image capture processing device 2, which functions as a direct data link providing input for the AI ​​model 68. Furthermore, the input interface 64 is coupled to a user input interface 72, which can be a keyboard, a touchpad, or any other device suitable for the purpose. The input interface 64 is further coupled to a database 74 that holds patient-related data, particularly electronic patient records. The output interface 70 is coupled to a display 14, which may be identical to the display 14 of the image capture processing device 2 shown in Figure 1.

[0125] An exemplary CDSS62 is configured to provide a classification of lymphedema-induced fluorescence patterns based on fluorescence images 7. In various embodiments outlined above, the CDSS62 includes an image capture processing device 2, particularly an image capture device 10, as an input interface 64 in which fluorescence images 7 specific to the patient 6 are provided as input features to an artificial intelligence (AI) model 68. A processor 66 performs inference operations in which fluorescence images 7, large fluorescence images, or overlay images 9 are applied to the AI ​​model 68 to generate a classification of lymphedema-induced fluorescence patterns. A user interface (UI) in which the classification of lymphedema-induced fluorescence patterns, e.g., the stage of lymphedema severity and the clinical type of fluorescence pattern, is communicated to a user, e.g., a clinician, may be a display 14.

[0126] In some embodiments, the input interface 64 may be a direct data link between the CDSS 62 and one or more medical devices that generate at least some of the input features. For example, the input interface 64 may transmit fluorescence images directly to the CDSS 62 during therapeutic and / or diagnostic medical procedures. This can be done by an image capture processing device 2. Additionally or alternatively, the input interface 64 may be a classic user interface that facilitates interaction between the user and the CDSS 62. For example, the input interface 64 may facilitate a user input interface 72 in which the user can manually input further patient-related data, such as gender, age, or body mass index. Additionally or alternatively, the input interface 64 may provide the CDSS 62 with access to an electronic patient record from which one or more input features can be extracted. This can be done by a direct link to a database 74 that holds the respective information. In any of these cases, the input interface 64 is configured to collect one or more of the following input features related to a particular patient 6, namely fluorescence images 7, large fluorescence images, overlay images 9, and optionally additional information from the patient records as described above, at or before the time when the CDSS 62 is used to assess the level or stage of lymphedema.

[0127] Based on one or more of the above input features, the processor 66 uses the AI ​​model 68 to perform inference operations to generate classifications of lymphedema-induced fluorescence patterns. For example, the input interface 64 may deliver fluorescence images 7 into the input layer of the AI ​​model 68, which propagates these input features through the AI ​​model 68 to the output layer. The AI ​​model 68 can provide the computer system with the ability to perform tasks without being explicitly programmed by performing inferences based on patterns found in the analysis of the data. The AI ​​model 68 explores the study and construction of algorithms (e.g., machine learning algorithms) that can learn from existing data and make predictions about new data. Such algorithms operate by building an AI model from exemplary training data to make data-driven predictions or decisions, which are expressed as outputs or evaluations. The enhancement and training of the AI ​​model can also be performed by manual user input, for example, correcting for automatically generated stages of lymphedema severity or clinical types of fluorescence patterns.

[0128] AI models can be trained in the cloud using data from various locations, and the trained network can then be downloaded and used for various tasks. Using the cloud for training is particularly advantageous because it offers higher performance compared to local systems.

[0129] Figure 9 shows examples of clinical types of fluorescence patterns. Figure 9A shows a typical linear pattern. This pattern can be observed when the lymphatic system is in good condition and the fluorescent agent 8 is transported upward through the lymphatic vessels along with the lymph fluid. This finding is typically stage 1, meaning normal lymphatic flow and no signs or symptoms. Figure 9B shows a splash pattern, which can be observed when lymphatic flow is partially obstructed. Typically, stage 2 is assigned to this finding, meaning there is an accumulation of lymph fluid with the possibility of swelling. Figure 9C shows a stardust pattern, which can be observed when lymphedema progresses further. This pattern is typically assigned to stage 2 or 3, where permanent swelling of body part 4 may also be observed. In stage 3 lymphedema, this swelling typically does not dissipate with the elevation of the affected body part 4. Finally, Figure 9D shows a diffuse pattern indicating severe lymphedema. The fluorescent agent 8 is injected along with the lymph fluid into the skin and subcutaneous tissue of patient 6. This clinical finding is typically identified as stage 3 or 4, which means there is thickening of the skin with growth and deformation of the affected limb 4. Illustrated patterns A–D can be found anywhere on the body or limb 4 of patient 6, as shown by the dashed rectangles on the right side of Figure 9.

[0130] Figure 10 shows another embodiment of the prism assembly 20 of the image acquisition device 10. The prism assembly 20 comprises prisms P5, P6, and P7, which are configured to split light into red, green, and blue components toward, for example, sensors D3, D4, and D5. According to a further embodiment, the prism assembly 20 is configured to split incident light into a green component, a red / blue component, and an infrared component, and direct these toward sensors D3, D4, and D5. According to yet another embodiment, the prism assembly 20 is configured to split incident light into a visible light component directed toward a red / green / blue sensor (RGB sensor), a first infrared component of a first wavelength or wavelength interval, and a second infrared component of a second wavelength or wavelength interval, and direct these toward sensors D3, D4, and D5.

[0131] Light enters the prism assembly 20 through the indicated arrow. prism Optical coating C1 is placed between P5 and P6, and optical coating C2 is placed between prisms P6 and P7. Each optical coating C1 and C2 has different reflectivity and wavelength sensitivity. coating At C1, the incident beam I is partially reflected (beam J) from the same plane of prism P5 into which the light was incident. On that same plane, the beam, now labeled K, is reflected again toward filter F3 and sensor D3. beam From J beam The reflection to K is an internal reflection. Therefore, filter F3 and sensor D3 receive the light reflected by coating C1, and similarly, filter F4 and sensor D4 receive the light reflected by coating C The light from beam L (beams M and N) reflected by 2 is received. Filter F5 and sensor D5 receive the light from beam O that has passed through the prism without obstruction.

[0132] When referring to embodiments in which the incident light is split into red, green, and blue components, the coatings and filters are selected accordingly.

[0133] In embodiments where incident light is separated into green, red / blue, and infrared components, filter F3 can be a patterned filter (red / blue). In particular, arrays of red and blue filters exist in an alternating pattern. The pattern can consist of groups of 2x2 pixels filtered for one specific color. Filter F4 can be a green filter, meaning that the filter contains only green filters. There is a single pixel grid, and the light received at each pixel is filtered with the green filter. Filter F5 can be an IR filter. Each pixel is filtered with the IR filter.

[0134] In general, coatings C1 and C2 should be matched to filters F3, F4, and F5. For example, the first coating C1 can transmit visible light while reflecting IR light so that IR light is directed toward the IR filter F3. The second coating C2 may reflect red and blue light while being transparent to green light, so that filter F4 should be a red / blue patterned filter and F5 should be a green filter 23.

[0135] In a further embodiment in which the incident light is split into a visible light component (RGB), a first infrared component, and a second infrared component, the coatings C1, C2 and filters F3, F4, F5 are configured such that, for example, sensor D4 is a color sensor (RGB sensor) for detecting a visible light image of all three colors. Furthermore, sensor D3 may be configured to detect fluorescence of a first wavelength, and sensor D5 may be configured to detect fluorescence of a second wavelength.

[0136] Similarly, referring to the prism assembly 20 in Figure 5, the coatings S1, S2, S3, S4, C1, and C2, and the filters F1, F2, F3, F4, and F5, positioned in front of each of the sensors D1, D2, D3, D4, and D5, can be configured to detect up to four fluorescence wavelengths. For example, sensor D4 is a color sensor that detects all three visible light images. Sensor D3 is for detecting fluorescence at a first wavelength, sensor D5 is for detecting fluorescence at a second wavelength, sensor D1 is for detecting fluorescence at a third wavelength, and sensor D2 is for detecting fluorescence at a fourth wavelength.

[0137] All specified features, including those obtained solely from the drawings, and individual features disclosed in combination with other features, are considered important to the present invention, both individually and in combination. Embodiments of the present invention can be realized by individual features or combinations of features. Features described in combination with the expressions "in particular" or "especially" should be treated as preferred embodiments. [Explanation of Symbols]

[0138] 2 Image acquisition and processing device 3. Doctor 4 body parts 5. Visible light image 6 patients 7. Fluorescence images 8. Fluorescent dyes 9 Overlay Images 10 Image acquisition devices 11 Surface 12 Processing devices 14 displays 16 Lighting Units 18 Objective lenses 20 Prism Assembly 22 Fluorescence imaging unit 24 Visible light imaging unit 26 Data Links 28 stitching units 30 Overlay Units 32 measuring units 33 Distance Sensor 35 Internal Measurement Unit 50 Endoscopes 52 Optical Fibers 51 Fiber 54 Light source 56 shaft 58 Fiber Bundles 60 AI units 62 CDSS 64 Input Interfaces 66 processors 68 AI Models 70 Output Interfaces 72 User Input Interface 74 Databases P1 First pentaprism P2, P4 Compensation Prism P3 Second Pentaprism P5, P6, P7 Dichroic Prism Assembly A Incident light beam B Light beam S1 entrance plane D1 ~ D5 sensor C1, C2 coating F1 ~ F5 Filter L Longitudinal Direction d distance

Claims

1. A computer-based clinical decision support system configured to output a classification of lymphedema-induced fluorescence patterns based on fluorescence images determined from the measurement of fluorescence signals in tissues of body parts to which a fluorescent agent has been added, An input interface that provides an artificial intelligence model with the patient-specific fluorescence image and the corresponding visible light image as input features, A processor that performs an inference operation in which the fluorescence image and the visible light image are applied to the artificial intelligence model to generate the classification of the lymphedema-induced fluorescence pattern, A user interface through which the classification of the lymphedema-inducing fluorescence patterns is communicated to the user, A clinical decision support system equipped with the following features.

2. The clinical decision support system according to claim 1, wherein the classification of the lymphedema-induced fluorescence pattern is the stage of lymphedema severity and / or the clinical type of the lymphedema-induced fluorescence pattern.

3. The input interface is configured to measure the fluorescence signal within the tissue of the body part and is a direct link to an image capture processing device configured to image the surface of the body part, wherein the tissue to which the fluorescent agent is added forms part of the body part, and the image capture processing device comprises an image capture device. The aforementioned image acquisition device, A lighting unit configured to illuminate the tissue with excitation light having a wavelength suitable for generating emitted light by the excitation emission of the fluorescent agent, To provide the aforementioned fluorescence image, a fluorescence imaging unit is configured to capture the fluorescence image by spatially resolved measurement of the emitted light, The system comprises a visible light imaging unit configured to capture the corresponding visible light image of a portion of the surface of the body part, The clinical decision support system according to claim 1, wherein the fluorescence imaging unit and the visible light imaging unit are configured such that the viewing direction and / or viewpoint of the fluorescence image and the corresponding visible light image are linked through a known relationship.

4. The large fluorescence image and the corresponding large visible light image are provided as input features to the artificial intelligence model through the input interface. The fluorescence imaging unit and the visible light imaging unit are further configured to repeatedly capture the fluorescence image and the visible light image to provide a series of fluorescence images and a series of visible light images. The aforementioned image capture processing device further comprises a processing device, The processing device is The system includes a stitching unit that applies a stitching algorithm to a series of visible light images, where a set of stitching parameters is determined and applied, in order to generate the large visible light image of the body part. The stitching unit is further configured to apply the stitching algorithm to the series of fluorescence images in order to generate the large fluorescence image, The clinical decision support system according to claim 3, wherein the stitching algorithm applied to the series of fluorescence images applies the set of stitching parameters determined when generating the large visible light image.

5. The image capture device comprises a dichroic prism assembly configured to receive fluorescent and visible light through the incident surface. The dichroic prism assembly comprises a first prism, a second prism, a first compensator prism located between the first and second prisms, a further dichroic prism assembly for splitting the visible light into three optical components, and a second compensator prism located between the second prism and the further dichroic prism assembly. The first prism and the second prism each have a cross-section with at least five corners, each corner having an interior angle of at least 90 degrees, and each has an incident surface and an exit surface, and are designed such that an incident beam entering the incident surface of the first prism and the second prism in a direction parallel to the normal to the incident surface is reflected twice inside the first prism and the second prism, and exits from the first prism and the second prism through the exit surface parallel to the normal to the exit surface. The normals to the incident surface and the normals to the exit surface of the first and second prisms are perpendicular to each other. When light enters the first prism through the incident surface, the light is partially reflected toward the exit surface of the first prism and travels a first path length from the incident surface of the first prism to the exit surface of the first prism, and the light is partially entered into the second prism via the first compensator prism, partially reflected toward the exit surface of the second prism and travels a second path length from the incident surface of the first prism to the exit surface of the second prism, The clinical decision support system according to claim 3, wherein the first prism is larger than the second prism such that the length of the first path and the length of the second path are the same.

6. The clinical decision support system according to claim 1, wherein the input interface is further a direct link to an electronic patient record, and the patient's age, sex, height, weight, BMI (body mass index), fat mass, muscle mass, daily exercise level, employment status, skin color, medication status, presence or absence of vascular disease, varicose veins or venous edema, disease, presence or absence of dialysis or diabetes, blood albumin level, renal function, hepatic function, cardiac function, blood hemoglobin (Hb) concentration, blood estimated values, lipid metabolism rate, blood glucose concentration, urea nitrogen, ABI (ankle-brachial index) value, lymphatic function measurement data at the same location before the onset of lymphedema, endocrine information, and hormone level data are provided through the input interface as further input features to the artificial intelligence model.

7. A computer implementation method for determining the classification of lymphedema-induced fluorescence patterns using a computer-based clinical decision support system, The aforementioned classification is based on fluorescence imaging determined by measuring the fluorescence signal in the tissue of a body part to which a fluorescent agent has been added. The aforementioned computer implementation method is The patient's unique fluorescence image and the corresponding visible light image are received as input features for an artificial intelligence model through an input interface. The processor performs inference operations, the fluorescence image and the visible light image are applied to the artificial intelligence model to generate the classification of the lymphedema-induced fluorescence pattern, The classification of the lymphedema-induced fluorescence patterns is communicated to the user through a user interface. Computer implementation methods, including those mentioned above.

8. The computer implementation method according to claim 7, wherein the classification of the lymphedema-induced fluorescence pattern is the stage of lymphedema severity and / or the clinical type of the lymphedema-induced fluorescence pattern.

9. The input interface is configured to measure the fluorescence signal within the tissue of the body part and receives the fluorescence image and the corresponding visible light image via a direct link to an image capture processing device configured to image the surface of the body part, wherein the tissue to which the fluorescent agent is added forms part of the body part, and the image capture processing device comprises an illumination unit, a fluorescence imaging unit and a visible light imaging unit. The aforementioned computer implementation method is A step of illuminating the tissue with the illumination unit using excitation light having a wavelength suitable for generating emitted light by the excitation emission of the fluorescent agent, To provide the aforementioned fluorescence image, the steps include capturing the fluorescence image by the fluorescence imaging unit through spatially resolved measurement of the emitted light, The process further includes capturing the visible light image by the visible light imaging unit by capturing the corresponding visible light image of a portion of the surface of the body part, The computer implementation method according to claim 7, wherein the fluorescence imaging unit and the visible light imaging unit are configured such that the viewing direction and / or viewpoint of the fluorescence image and the corresponding visible light image are linked through a known relationship.

10. The large fluorescence image and the corresponding large visible light image are provided as input features to the artificial intelligence model through the input interface. The fluorescence imaging unit and the visible light imaging unit repeatedly capture the fluorescence image and the visible light image to provide a series of fluorescence images and a series of visible light images. The image capture processing device further comprises a processing device equipped with a stitching unit, The aforementioned computer implementation method is The process involves applying a stitching algorithm, in which a set of stitching parameters is determined and applied by the stitching unit, to a series of visible light images in order to generate the large visible light image of the body part, The process further includes the step of applying the stitching algorithm to the series of fluorescence images by the stitching unit in order to generate the large fluorescence image, The computer implementation method according to claim 9, wherein the stitching algorithm applied to the series of fluorescence images is the set of stitching parameters determined when generating the large visible light image.

11. The measurement of the fluorescence signal is performed on tissue to which at least the first and second fluorescent agents have been added. The step of capturing the aforementioned fluorescence image is: To capture a first fluorescence image in a first wavelength range, which is generated by illuminating the tissue with first excitation light having a first wavelength suitable for generating emitted light by the first excitation emission of the first fluorescent agent, This includes capturing a second fluorescence image within a second wavelength range, which is generated by illuminating the tissue with a second excitation light having a second wavelength suitable for generating emission light by the second excitation emission of the second fluorescent agent, The first and second fluorescence images are provided to the artificial intelligence model as input features through the input interface. The input interface receives the first and second fluorescence images as input features of the artificial intelligence model, The computer implementation method according to claim 9, wherein the processor performs the inference operation by applying the first and second fluorescence images to the artificial intelligence model to generate the classification of the lymphedema-inducing fluorescence patterns.

12. The computer implementation method according to claim 7, wherein the patient's age, sex, height, weight, BMI (body mass index), fat mass, muscle mass, daily exercise level, employment status, skin color, medication status, presence or absence of vascular disease, varicose veins or venous edema, disease, presence or absence of dialysis or diabetes, blood albumin level, renal function, hepatic function, cardiac function, blood hemoglobin (Hb) concentration, blood estimated values, lipid metabolism rate, blood glucose concentration, urea nitrogen, ABI (ankle-brachial index) value, lymphatic function measurement data at the same location before the onset of lymphedema, endocrine information, and hormone level data are provided through the input interface as further input features to the artificial intelligence model via a direct link to the electronic patient record.

Images