A system and method for identifying abnormal perfusion patterns.

Abstract

The present disclosure addresses the above-mentioned needs and challenges by providing a system and method for identifying abnormal perfusion patterns in a subject. This is accomplished because the inventors have realized that normal and abnormal tissues have different effects on an oscillating input signal, for example in the form of a series of boluses of a fluorescent imaging agent, meaning that the pattern of the fluorescent output signal will change based on the condition of the tissue.

Description

[Technical field]

[0001] The present disclosure relates to systems and methods for continuously detecting and optionally classifying abnormal perfusion patterns in tissue by fluorescence imaging. [Background technology]

[0002] 2. Background of the Invention Surgeons are increasingly using fluorescence imaging during video-assisted surgery to aid in their clinical decision making. Today, fluorescence-guided surgery often involves administering a relatively large bolus of an optical contrast agent, such as indocyanine green (ICG), into a peripheral vein, waiting for distribution in the tissue, and then visually inspecting the tissue to identify areas with either low or high signal intensity. Low or no signal intensity corresponds to reduced tissue blood perfusion (i.e., ischemia), whereas high signal intensity corresponds to normal tissue blood perfusion. In general, it is important to know whether a patient has normal blood perfusion in the tissue of interest. In particular, this is important because abnormal blood perfusion can be a sign of, for example, cancerous or inflamed tissue.

[0003] Currently, there is a need for systems and methods that can objectively and reliably detect and / or identify areas having abnormal perfusion patterns so that, for example, cancerous or inflamed tissue that is not normally distinguishable from normal tissue by the human eye may be directly detected, for example, during a surgical procedure.

[0004] Another existing method is to assess tissue perfusion by visually assessing selected regions of interest while contrast agents are administered. By observing the influx and outflow of fluorescence, information about tissue perfusion can be inferred. Since the influx-related increase in signal intensity takes only a few seconds, it is nearly impossible to discern between the influx rate and timing in various regions across the image, relying on the surgeon's sole visual assessment. Thus, a limitation of this approach is that perfusion is only characterized as "sufficient" or "insufficient", i.e., it is not a quantitative assessment. It is therefore important to develop systems and methods that can objectively quantify tissue perfusion, for example, based on the dynamics of influx and outflow of fluorescence.

[0005] A major limitation common to all existing dynamic applications is that only one defined area can be visualized per ICG assessment, and the large dose makes it impractical to perform multiple such measurements without significant washout periods between measurements. In practice, this means that existing ICG fluorescence uses are limited to gathering information from specific defined anatomical structures of interest.

[0006] The challenge is that abnormal tissues, e.g., cancerous and inflamed tissues, are rarely distinguishable from normal tissues to the human eye. As a consequence, surgeons often miss the full extent of the cancer / inflammation, leading to poorer patient outcomes and possibly disease progression and reoperations. Another consequence of this is increased costs associated with medical staff and hospital services. Another risk with current methods is that surgeons remove too much tissue, because surgeons currently do not have real-time tools to identify the border between healthy and non-healthy tissues. Thus, surgeons need tools that can detect and identify cancerous and inflamed tissues in patients, preferably continuously and in real-time. This has the potential to allow safer and faster tissue dissection, substantially reducing surgery times.

[0007] Thus, there is a need for systems and methods that can not only identify hypoperfusion and vascular anatomical structures, but also detect and identify abnormal perfusion patterns in patient tissues, such as cancerous and inflamed tissues. Summary of the Invention [Means for solving the problem]

[0008] Summary of the Invention The present disclosure addresses the above-mentioned needs and challenges by providing a system and method for identifying abnormal perfusion patterns in a subject. This is accomplished because the inventors have realized that normal and abnormal tissues have different effects on an oscillating input signal, for example in the form of a series of boluses of a fluorescent imaging agent, meaning that the pattern of the fluorescent output signal changes based on the state of the tissue. In other words, tissue can be seen as a filter that distorts the input signal, and the methods of the present disclosure can derive information about this filter based on how the input signal is distorted.

[0009] The present inventors have previously developed a novel dosing regimen for fluorescence imaging that enables continuous perfusion monitoring, based on automated administration of time-spaced microboluses of at least one fluorescent agent. This approach creates a fluorescent signal from perfused tissue that oscillates with a predetermined pattern on a time basis. This facilitates the ability to track and register tissue perfusion in the background while the surgeon operates in white light, alerting the surgeon to any deviations in tissue perfusion in the anatomical structures in the camera's focus at any given time. This is further described in pending PCT application entitled "System and method for automatic perfusion measurement" as PCT / EP2019 / 065648 (published as WO 2019 / 238912 A1), which is incorporated by reference herein in its entirety.

[0010] Additionally, the present inventors have previously developed a novel method for identifying blood vessels in tissues of a subject. This method is described in pending PCT application entitled "Fluorescent anatomical mapping" as PCT / EP2020 / 087507, which is incorporated herein by reference in its entirety. This method utilizes detection of the phase shift of oscillating fluorescent signals in arteries relative to veins to generate a vascular map of superficial tissues in organs or connective tissues (including fat). The vascular map generated from the fluorescent signals can be run in the background and, at the surgeon's request, can be overlaid on a white light image, for example, as an augmented reality.

[0011] Thus, the identified vessels can be visualized and mapped to medical personnel, for example, before and during tissue dissection. A major advantage is that automated and continuous vessel identification can be performed in the background while the surgeon works while viewing the normal white-light camera image. The microbolus procedure can be initiated at the beginning of the operation and can be performed in the background, monitored by a computer system that measures and receives the fluorescent signal and controls the interval and administration of the microbolus regimen. The surgeon can then shift to a computer-generated "vessel view" showing the vessels in the area of ​​interest at any time during the operation. Advantageously, the continuous identified vessels can be superimposed in real time onto the white-light image, so that otherwise hidden vessels appear in the white-light image in real time as augmented reality objects.

[0012] The inventors have now realized that a novel dosing regimen involving a variety of fluorescent signals from a patient, generated from a controlled input signal using many small boluses, may be used to detect tissue regions with abnormal perfusion patterns (e.g., cancerous and inflamed tissues). This may be due, among other things, to the use of a variety of input signals, including many small doses ("microboluses") of at least one fluorescent imaging agent (e.g., ICG). It is also possible due to a novel method of analyzing the fluorescent output signal obtained from the patient and / or by analyzing the fluorescent images obtained from the patient's tissues. The disclosed method allows for the generation of a priori knowledge of each patient's particular "normal" tissues, and further allows for this knowledge to be used to identify the characteristic inflow and outflow patterns of abnormally perfused tissues (e.g., cancerous and inflamed tissues). The disclosed system and method may then flag this tissue as "abnormally perfused," thereby making the tissue known to the surgeon, who may ultimately make an informed decision on how to proceed.

[0013] One embodiment of the present disclosure relates to a computer-implemented method for detecting (and / or identifying) one or more regions having an abnormal perfusion pattern in a tissue of a subject, for example during a medical procedure, comprising the step of acquiring, preferably continuously, a fluorescent image of the tissue. The fluorescent image is associated with a fluorescent output signal that correlates with an input signal, preferably defined by a series of boluses of at least one fluorescent imaging agent. The series of boluses is preferably administered with a predefined and / or controlled duration between successive boluses. The method may further comprise the step of analyzing the fluorescent image. At least one tissue region having a normal perfusion pattern may be identified or selected, for example, manually, semi-automatically, and / or automatically. The normal perfusion pattern may be identified, and the identification may be provided in the intensity domain and / or in the time domain. Once the normal perfusion pattern is determined, possible tissue regions having an abnormal (non-normal) perfusion pattern may be detected in the fluorescent image, whereby abnormally perfused tissue may be detected in the patient.

[0014] One method of detecting, identifying and / or classifying tissue regions with normal and / or abnormal perfusion is to utilize the concept of body kernel as described herein.Preferably, at least a first body kernel is obtained, where the at least first body kernel is a filter imposed by the subject's body on the at least one fluorescent imaging agent in tissue regions with normal perfusion patterns, and / or at least a second body kernel is obtained, where the at least second body kernel is a filter imposed by the subject's body on the at least one fluorescent imaging agent in tissue regions with abnormal perfusion patterns.In that respect, the at least first body kernel and / or the at least second body kernel can be understood as at least one transfer function between the input signal and the fluorescent output signal.

[0015] Another embodiment relates to a computer-implemented method for determining a perfusion-related body kernel of a tissue of a subject, the body kernel being defined as a filter imposed by the body of the subject on a (certain) fluorescent imaging agent, the method comprising the steps of successively acquiring fluorescent images of the tissue, the fluorescent images being associated with a fluorescent output signal correlated with an input signal defined by a series of boluses of the fluorescent imaging agent, the series of boluses being administered with a predefined and / or controlled duration between successive boluses. At least one region of interest (ROI) may be selected in the fluorescent image corresponding to a tissue region. The body kernel of the ROI may then be determined by deconvolving the ROI with respect to the input signal, i.e., the convolution of the input signal with the body kernel corresponds to the fluorescent output signal from the tissue region. If this tissue region has normal perfusion dynamics, predicted output signals for other tissue regions, e.g., adjacent / proximitized tissue regions, can then be continuously calculated by successively convolving controlled and known input signals with the body kernel. Tissue regions with abnormal perfusion dynamics can then be continuously detected and / or monitored, preferably in real time, since a comparison between the fluorescence output signal and the predicted output signal can be provided for virtually any tissue region in the analyzed fluorescence image, as long as a suitable body kernel is provided. Such a comparison can then almost immediately reveal differences between the fluorescence output signal and the predicted output signal - and any differences provide an indication of an abnormal perfusion pattern. The main advantage of the disclosed approach is that it provides a systematic methodology that can be performed in real time in the background during a medical procedure, continuously analyzing acquired fluorescence images and automatically detecting abnormal perfusion patterns, thereby similarly detecting abnormally perfused tissue regions, which can be shown to appropriate medical personnel on a display, e.g., by augmented reality.All that is required to detect abnormal perfusion patterns in acquired fluoroscopic images is a known (oscillating) input signal and at least one relevant body kernel that is associated with a normal perfusion pattern, and such relevant body kernel can also be determined automatically during the procedure, e.g., as an initial step.

[0016] Yet another embodiment relates to a computer-implemented method for establishing a time-domain perfusion reference of a subject, comprising the steps of continuously measuring a fluorescence output signal correlated with an input signal defined by a series of boluses of the fluorescence imaging agent, where the series of boluses are administered with a predefined and / or controlled duration between successive boluses, and defining the subject-specific time-domain perfusion reference as fluorescence output signal versus time. The subject-specific time-domain perfusion reference may also be defined as the time difference between a bolus injection and the corresponding peak fluorescence output signal. The subject-specific time-domain perfusion reference may be determined with a more simple mechanism, e.g., a photodiode-based finger clip, i.e., without image acquisition, and may serve as, for example, a clock reference for defining and / or identifying normal perfusion patterns.

[0017] Any of the disclosed methods may further include identifying blood vessels in the fluorescence image. In the context of this disclosure, perfusion in blood vessels is not an example of a normal perfusion pattern because blood vessels are not "tissues". However, there may still be an advantage to identifying where blood vessels are located so that blood vessels are not selected as an example of a normal perfusion pattern. For example, systems and methods for identifying blood vessels in tissues of a subject during a medical procedure are further described in commonly assigned PCT / EP2020 / 087507.

[0018] A further aspect of the present disclosure is a system for identifying abnormal perfusion patterns in a tissue of a subject, e.g., during a medical procedure, the system comprising: - successively acquiring fluorescent images of said tissue, wherein said fluorescent images are associated with a fluorescent output signal correlated with an input signal defined by a series of boluses of at least one fluorescent imaging agent, wherein said series of boluses are administered with a predefined and / or controlled duration between successive boluses, - analyzing said fluorescent image; - identifying in said fluorescence image at least one tissue region having normal perfusion based on said analysis; - defining normal perfusion patterns in the intensity domain and in the time domain; and - detecting in said fluorescence image tissue regions that may have abnormal (non-normal) perfusion patterns; The present invention further relates to a system configured for:

[0019] The system of the present disclosure is preferably configured to perform the methods disclosed herein, which may be provided by a system having at least one processor and a memory having instructions stored thereon, which when executed by one or more processors, cause the system to perform any of the methods of the present disclosure.

[0020] The systems and methods of the present disclosure may be of great use to surgeons during surgery, in any organ and for any indication, particularly where vascular structure and anatomy may be important. Commonly owned pending application PCT / EP2020 / 087507 described how vascular maps may be generated in real time and superimposed on white light images to allow surgeons to otherwise see vessels hidden in the image.

[0021] The present disclosure builds on this approach by providing new functionality that allows for the identification of regions with abnormal perfusion patterns, such that, for example, inflamed tissue that is otherwise indistinguishable from normal tissue can be directly detected during, for example, a surgical procedure.

[0022] The disclosed approach used within a surgical procedure can be understood as a decision support in surgery. However, the use of the disclosed approach is not limited to use during a surgical procedure. The disclosed approach can also be advantageously applied to provide information about vascular structure, perfusion maps and anatomy, pre- or post-operatively, or even for monitoring wound healing, vascular anatomy and perfusion patterns in patients who have not undergone any surgery. In such cases, the disclosed approach can be understood as a physical examination tool, in the same way as a CT scan.

[0023] Various examples are provided herein in which the optical contrast agent is ICG. However, the approach of the present disclosure is not limited to ICG. Because many types, even several different contrast agents may be used simultaneously when contrast agents are used, and the body kernel principle of the present disclosure generally applies to fluorescence imaging using contrast agents. Various examples of applicable ICG administration are listed herein. For other contrast agents, the same dose may be applied, and if not, it is not a challenge to evaluate the relevant dose for other contrast agents, especially those listed herein. It is a matter of selecting a dose small enough so that one or more body kernels can be determined so that a predefined oscillating input signal that forms the basis of a fluorescence output signal is generated, so that several boluses can be injected at time intervals of several seconds or minutes so that a fluorescence signal can be detected.

[0024] The present disclosure further relates to a computer program having instructions that, when executed by a computing device or computing system, cause said computing device or computing system to perform any of the methods disclosed herein. [Brief description of the drawings]

[0025] [Figure 1] FIG. 1A shows an example of an intensity curve after a bolus of ICG was provided to a subject, and FIG. 1B shows the corresponding intensity curve in which the hemodynamic parameters perfusion slope, slope start, slope end, maximum intensity, washout slope, washout start and washout slope end were calculated and plotted.

[0026] [Figure 2-1] FIG. 2A shows an oscillatory time intensity fluorescence curve in which oscillations are disrupted due to the onset of ischemia in a human subject.

[0027] FIG. 2B shows an expansion of the time interval around t=3800 seconds of the graph before the onset of ischemia occurs.

[0028] [Figure 2-2] FIG. 2C shows idealized data with and without ischemic conditions.

[0029] FIG. 2D shows idealized data in which only a portion of the oscillating time-intensity fluorescence curve can be detected.

[0030] [Diagram 3] FIG. 3A shows sequential measurements of a human subject injected with a microbolus.

[0031] FIG. 3B shows an expanded view of the interval shown in FIG. 3A.

[0032] [Figure 4] FIG. 4 shows measurements in a human subjected to venous occlusion, where blood flow is only partially restricted.

[0033] [Diagram 5] Figure 5A shows a snapshot of an ICG analysis running on a humanoid subject (right forearm) The image is taken at a very early stage when a microbolus of ICG has just been administered and is beginning to enter the artery.

[0034] FIG. 5B shows a snapshot taken a few seconds later than FIG. 5A, in which multiple arteries can be identified.

[0035] [Figure 6] FIG. 6A shows a snapshot several seconds later than FIG. 5B, when the fluorescence intensity from a particular microbolus of ICG is reaching its peak value.

[0036] FIG. 6B shows a snapshot approximately one minute later than FIG. 6A, illustrating the washout phase in which veins can be identified.

[0037] [Figure 7] FIG. 7 shows an edge-filtered version of FIG. 6B.

[0038] [Figure 8] FIG. 8 shows an image in which an artery is identified in the sequence of images shown in FIGS. 5A-6B.

[0039] [Figure 9] FIG. 9 shows images in which veins have been identified in the sequence of images shown in FIGS. 5A-6B.

[0040] [Figure 10] FIG. 10 shows an image in which the arteries and veins shown in FIGS. 8-9 have been visually enhanced in red (arteries) and blue (veins) and the images overlaid so that they are easily distinguishable.

[0041] [Figure 11]Figure 11 shows four plots that help illustrate how the body kernel can be estimated from a measured ICG signal and how the body kernel can be used to estimate the measured signal. In this example, synthetic data is used.

[0042] [Figure 12] Figure 12 shows essentially the same visual representation as Figure 11 via four plots, but now the input signal contains two pulses that change the measured signal compared to that of Figure 11.

[0043] [Figure 13] FIG. 13 shows an example of how an estimated body kernel is used to infer / estimate what a measured fluorescence output signal ("ICG signal") should look like based on knowledge of a previously estimated body kernel.

[0044] [Figure 14] Figure 14 shows the same input signal, body kernel, and estimated signal as previously shown in Figure 13. However, in this scenario, the measured signal is different than we would normally expect from the associated region of interest.

[0045] [Figure 15-1] 15 shows an example of how a body kernel can be estimated for a region of interest and then used to estimate a predicted ICG signal. The measured signal can be compared against the predicted signal, so that abnormal perfusion patterns can be detected. [Figure 15-2] Same as above. [Figure 15-3] Same as above.

[0046] [Figure 16]16 shows an example of the method of the present disclosure using real-life data, i.e., human data. In this example, the measured data was obtained from the subject's forearm.

[0047] [Figure 17-1] 17 shows one graph of measured fluorescence output signal versus time at three cycles, PR, P1, and P2, where the body kernel is determined from PR and used to calculate predicted output signals at P1 and P2. At P1, the predicted output signal follows the measured fluorescence output signal, whereas at P2 there is a clear deviation due to ischemia. [Figure 17-2] Same as above. [Figure 17-3] Same as above. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0048] Detailed Description of the Invention An input signal as used herein may be characterized by a predefined frequency with a period between 30 seconds and 15 minutes, e.g., between 1 and 10 minutes, or between 1 and 2 minutes, 2 and 3 minutes, 3 and 4 minutes, 4 and 5 minutes, 5 and 6 minutes, 6 and 8 minutes, or 8 and 10 minutes, over a period of at least 10 minutes, or at least 15 minutes, or at least 30 minutes, or at least 1 hour, or at least 2 hours, corresponding to a bolus injection that may be seen to create a predetermined pattern. The input signal does not necessarily have a fixed and / or constant frequency, so long as the series of boluses is administered with a predefined and / or controlled duration between successive boluses. The resulting fluorescent signal may be understood as oscillating in intensity following this predetermined pattern, at least for blood vessels and / or normally perfused tissues. As also described herein, this predetermined pattern may result from a controlled injection of a series of small boluses of at least one fluorescent agent, such as indocyanine green (ICG).

[0049] White light images of the tissue may also be continuously received and acquired, so that at least one white light image of the tissue may be generated (wherein, for example, identified blood vessels, normally perfused tissue regions, and / or abnormally perfused tissue regions may be visually enhanced, for example, by superimposing their corresponding elements onto the white light image, preferably also visually enhancing their corresponding elements, for example, by high contrast colors), and the identified blood vessels, normally perfused tissue regions, and / or detected abnormally perfused tissue regions may be displayed on a screen such that they appear as augmented reality objects.

[0050] Intraoperative Fluorescence Imaging Perfusion (e.g., blood flow) may be imaged intraoperatively and assessed in real time using near-infrared light from a surgical microscope or camera to acquire a video of fluorescence in the near-infrared region excited from a fluorescent vascular contrast agent administered intravenously as a tracer. The state of intraoperative perfusion may thereby be ascertained in real time. In the present disclosure, perfusion in tissues and / or blood vessels is used to identify blood vessels, normally perfused tissues and / or abnormally perfused tissues utilizing fluorescence imaging, but is not necessarily limited to the intraoperative use of a surgical camera.

[0051] The systems and methods of the present disclosure may provide enhanced information of tissue characteristics including the location of superficial and deeper blood vessels, normally perfused tissue and / or abnormally perfused tissue, particularly when different fluorescent imaging agents are used, because careful selection of different fluorescent imaging agents provides the option of obtaining perfusion information from different depths in the tissue.

[0052] During medical procedures involving fluorescence imaging, such as diagnostic, screening, examination and / or surgical procedures, a solution containing a fluorescent imaging agent (e.g., ICG) is injected intravenously and the molecules are excited by an infrared light source (e.g., a laser with a wavelength in the infrared wavelength range (e.g., approximately 780 nm). Fluorescence with a wavelength of approximately 830 nm is then emitted from the excited imaging agent molecules and can be recorded, for example, with an imaging device in the form of a camera. A filter can be provided to block the excitation light, since the excitation intensity is typically much greater than the fluorescence intensity. The excitation intensity can be approximately 1 W per emission angle, whereas the fluorescent power per pixel can be approximately 0.15 pW. Despite the difference being several orders of magnitude, a good signal-to-noise ratio (SNR) can be achieved. The recorded fluorescence provides an image of the perfusion in the imaged tissue, making it possible to see deeper due to the penetration depth of 5-10 mm for ICG. Because the ICG molecule is bound to proteins in the blood, the video images contain information about the level of perfusion - but that information can be difficult for the surgeon to quantitate intraoperatively when looking only at the acquired video images.

[0053] In the systems and methods of the present disclosure, the fluorescent imaging agent is selected from the group of indocyanine green (ICG) and fluorescein isothiocyanate, rhodamine, phycoerythrin, phycocyanin, allophycocyanin, orthophthalaldehyde, fluorescamine, rose bengal, trypan blue, fluorogold, green fluorescent protein, flavin, methylene blue, porphysomes, cyanine dyes, IRDDye800CW, CLR 1502 in combination with a targeting ligand, OTL38 in combination with a targeting ligand, or combinations thereof.

[0054] Indocyanine green (ICG) is a cyanine dye used in medical diagnostics and is by far the most common dye for perfusion assessment. It has a peak spectral absorption at about 800 nm. These infrared frequencies penetrate the retinal layer, allowing ICG angiography to image deeper circulatory patterns than fluorescein angiography. ICG binds tightly to plasma proteins and is confined to the vasculature. It is administered intravenously and is excreted from the body by the liver into the bile with a half-life of about 3-4 minutes, depending on liver function. ICG sodium salt is usually available in powder form and can be dissolved in a variety of solvents; 5% (<5% depending on the batch) Sodium iodide is usually added to ensure better solubility. Sterile lyophilisates of water-ICG solutions are approved in many European countries and in the United States as diagnostic agents for intravenous use under the names ICG-Pulsion, IC-Green and VERDYE.

[0055] The absorption and fluorescence spectrum of ICG is in the near infrared region. Typically, a laser with a wavelength of approximately 780 nm is used for excitation. At this wavelength, it is possible to detect the fluorescence of ICG by filtering out the light scattered from the excitation light.

[0056] Although the toxicity of ICG is classified as low, administration is not without risk (e.g., during pregnancy). It is known that ICG decomposes under the influence of UV light into toxic waste products, producing many as yet unknown substances. Thus, it is in the patient's interest that the dose of ICG used during fluorescence imaging be minimized as shown herein.

[0057] Fluorescein is another dye that is widely used as a fluorescent tracer for many applications. Fluorescein has an absorption maximum of 494 nm and an emission maximum of 512 nm (in water). It is therefore suitable for use in combination with ICG, because the absorption and emission wavelengths of the two dyes are several hundred nanometers apart.

[0058] According to one embodiment, the fluorescent imaging agent is bound to a molecule that targets abnormal tissue (e.g., a tumor targeting molecule) and given to a subject as a pre-surgery drug. The tumor targeting molecule then binds to tumor tissue inside the subject. The fluorescent imaging agent immobilized on the tumor tissue thereby indicates areas of tumor tissue (because such areas shine brighter than other areas). As a result, these areas can be more easily identified in images / videos. The tissue does not have to be tumor tissue, but can be other types of abnormal tissue (e.g., inflamed tissue). The important aspect is that the fluorescent imaging agent is bound to a molecule that targets abnormal tissue. Thus, this approach can be used in combination with any of the methods disclosed herein to enhance detection of abnormal tissue.

[0059] Combination of two types of molecules The present disclosure further relates to a computer-implemented method for detecting (and / or identifying) one or more regions having abnormal perfusion patterns, wherein at least two fluorescent imaging agents are used to simultaneously generate two different fluorescent signals. The at least two fluorescent imaging agents may be selected from the list of imaging agents provided elsewhere herein. Preferably, the at least two fluorescent imaging agents have different emittance wavelengths, which allows fluorescent images to be simultaneously obtained from at least two different depths of tissue. Thus, the advantage of using two fluorescent imaging agents (with different emittance wavelengths) is that a hierarchical analysis is provided, where images can be simultaneously obtained from different depths of tissue. Preferably, the at least two different depths are separated by at least 0.5 cm, preferably at least 1 cm, even more preferably at least 1.5 cm. As an example, the two depths may be 0.5 cm and 2 cm, as measured from the skin level of the subject. The use of two different fluorescent imaging agents may be applied to any of the computer-implemented methods disclosed herein.

[0060] Dosage regimen The present disclosure further relates to a method for automated perfusion assessment of a subject's anatomical structure, comprising administering intravenously a bolus of about 1 / 10 of the usual dose used for perfusion assessment. For indocyanine green (ICG), the usual bolus is 0.1-0.3 mg / kg body weight. According to the present disclosure, a bolus of less than 0.01 mg / kg body weight, preferably less than 0.005 mg / kg body weight, more preferably less than 0.0049 mg / kg body weight of a first fluorescent imaging agent such as ICG may be used, even more preferably less than 0.0048 mg / kg body weight, even more preferably less than 0.0047 mg / kg body weight, most preferably less than 0.0046 mg / kg body weight, and even more preferably less than 0.004 mg / kg body weight of a first fluorescent imaging agent may be used. For other fluorescent imaging agents described herein, the bolus is similarly reduced according to the present disclosure. As mentioned above, the agent can be injected by a controllable injection pump, for example, as a series of boluses with a defined time between subsequent boluses. After injection of each bolus, the fluorescence emission from the anatomical structure can be measured. This method can be combined with the method of the present disclosure for automatic perfusion assessment of a subject's anatomical structure, particularly for bolus administration regimens, to sequentially identify blood vessels in tissue.

[0061] The smallest bolus that provides a quantifiable fluorescence emission representative of perfusion of an anatomical structure and / or identifiable blood vessels can be determined after administering a series of increasing boluses, which may contain gradually increasing or decreasing amounts of agent, for example, the amount may increase or decrease in 10% increments from one bolus to a subsequent bolus.

[0062] The boluses are preferably provided as a regular series of injections with each bolus at a defined and regular time interval. The interval between the boluses may be between 5 and 600 seconds, for example between 5 and 300 seconds, for example between 10 and 180 seconds, for example between 10 and 140 seconds, for example between 10 and 90 seconds, for example between 15 and 80 seconds, for example between 20 and 70 seconds, for example between 30 and 60 seconds). In another embodiment, the interval between the boluses may be between 5 and 600 seconds, for example between 10 and 600 seconds, for example between 15 and 600 seconds, for example between 15 and 300 seconds, for example between 30 and 240 seconds, for example between 45 and 240 seconds, for example between 90 and 240 seconds, for example between 90 and 120 seconds. Preferably, it may be between 60 and 600 seconds, or between 120 and 600 seconds. The interval between boluses is preferably long enough to allow measurement / calculation of one or more perfusion-related parameters for each bolus in the anatomical structure (e.g., perfusion slope, slope onset, and washout slope). For boluses injected at very short time intervals, it is for example only possible (and sufficient) to assess slope onset and / or washout slope.

[0063] For ICG, the amount of the fluorescent imaging agent is preferably between 0.0001 and 0.001 mg / kg body weight / bolus, for example between 0.001 and 0.01 mg / kg body weight / bolus, preferably between 0.0005 and 0.005 mg / kg body weight / bolus, more preferably between 0.001 and 0.004 mg / kg body weight / bolus. The initial amount of the fluorescent imaging agent is advantageously at least 0.001 mg / kg body weight, preferably less than 0.005 mg / kg body weight. Subsequent boluses may then increase from one bolus to the subsequent bolus by at least as much as 0.001 mg / kg body weight, preferably less than 0.005 mg / kg body weight / bolus. For other types of fluorescent imaging agents, the dose is preferably selected based on its fluorescence relative to ICG. Thus, fluorescent imaging agents with higher emission rates are preferably administered at correspondingly lower doses. The dose can be, for example, substantially inversely linear to the quantum yield of the fluorescent imaging agent. The dose can further be based on the absorption and emission spectra for ICG.

[0064] The bolus is preferably a liquid volume between 0.5 μL and 10 mL (e.g., 0.5 to 5 mL). That is, the amount of the first fluorescent imaging agent is preferably dissolved in a liquid (typically, water). In one embodiment, a volume of an isotonic solution (e.g., saline) is injected immediately after the injection of the bolus of the fluorescent imaging agent, for example, in which case the volume of the isotonic solution is 1 to 20 mL (e.g., 2.5 to 15 mL, e.g., 5 to 10 mL). Subsequent injection of a volume of an isotonic solution may typically be applied when the bolus is injected, for example, into a peripheral vein.

[0065] In a further embodiment of the present disclosure, a second fluorescent imaging agent is administered, said second fluorescent imaging agent having an emission maximum that differs from the emission maximum of the first fluorescent imaging agent by at least 50 nm, or by at least 100 nm. The first and second fluorescent imaging agents are preferably administered alternately. Advantageously, the interval between administrations of different fluorescent imaging agents is half the interval between subsequent administrations of the same fluorescent imaging agent.

[0066] In a further embodiment of the disclosed method, a series of fluorescent images of an anatomical structure and / or tissue are generated for assessment of perfusion and / or identification of blood vessels. The fluorescence may be automatically detected by illuminating the anatomical structure / tissue with a light source capable of exciting a fluorescent imaging agent, and the emission is quantified and / or analyzed through the series of fluorescent images of the anatomical structure / tissue.

[0067] The period between boluses may be determined by a computer configured to detect the perfusion gradient caused by each bolus. Furthermore, the amount of fluorescent imaging agent in the bolus may be controlled by a computer configured to determine a minimum bolus that corresponds to a minimum fluorescent emission representative of perfusion of the anatomical structure. This computer may be part of the system of the present disclosure.

[0068] In further embodiments, the perfusion assessment includes locating perfusion complications in an anatomical structure. Thus, the perfusion assessment may be used in conjunction with diagnostic or surgical procedures. For example, the procedures include laparoscopic diagnostic examinations, laparoscopic exploration, laparoscopic surgery involving traditional laparoscopy, robotic surgery, and open surgery. The procedures may alternatively include the creation of an anastomosis (e.g., intestinal anastomosis, wound, plastic surgery, cardia surgery, or cancer).

[0069] Further embodiments of the present disclosure relate to a fluorescent imaging agent for use in the methods disclosed herein. Still further embodiments relate to the use of a fluorescent imaging agent in the preparation of a medicament for use in the methods of automated vascular perfusion assessment and / or sequential identification as disclosed herein.

[0070] In a further embodiment of the present disclosure, the fluorescent imaging agent is repeatedly injected. In certain cases, longer periods may be necessary, such as at least 2 minutes, preferably at least 3 minutes, even more preferably at least 4 minutes, even more preferably at least 5 minutes, most preferably at least 8 minutes, and most preferably at least 10 minutes. Here, the fluorescent imaging agent is not injected to allow the fluorescent imaging agent to be washed out, thereby reducing its background level. Once the background level is reduced to an acceptable level (e.g., below a certain percentage of maximum fluorescent intensity) or until substantial fluorescence cannot be measured, the injection of the fluorescent imaging agent can be continued.

[0071] Vibration Dynamics The inventors have previously realized that the measurement and analysis of repeatable bolus injections can be further extended from the interpretation and quantification of single inflow and / or single outflow phases to the analysis of oscillating fluorescence dynamics. This controlled variation of the fluorescence dynamics due to the control of the input signal can reveal physical perfusion characteristics that were not previously available without invasive measurements. The input signal is like a Morse code sent to the body, and the measured fluorescence dynamics, and especially the analysis of the fluorescence dynamics in combination with the known input signal, provide new information about the perfusion patterns in the examined tissue.

[0072] The systems and methods of the present disclosure may be configured for the controlled injection of small boluses (e.g., minimal boluses) at regular intervals, thereby creating the input signal described above. These boluses may result in periodic variations in the input signal that, when measured, take the approximate form of a sinusoid, depending, for example, on the injection time interval. In such a curve, the measured intensity signal is predicted to increase with the influx of fluorescent imaging agent from a given bolus, then decrease during the washout phase of that bolus before increasing again with the subsequent bolus, etc., resulting in a periodic (sinusoidal) pattern.

[0073] In one embodiment, the system of the present disclosure can be configured to recognize parameters of the vibration intensity curve (e.g., frequency and / or amplitude). The trained system can then, in turn, anticipate both the directionality and regularity of the upcoming signal dynamics. The system preferably uses the measured values ​​to recognize the vibration pattern, so that the system can then detect discrepancies between the measured and predicted values. The measured values ​​can further be used successively to improve the pattern recognition, i.e., predicted values. Alternatively or additionally, injection parameters (e.g., bolus frequency, dose and flow rate) can be used to determine the predicted values ​​(i.e., vibration patterns).

[0074] If the system anticipates the predicted value, it may detect and alert the onset of an ischemic condition at an early point - ideally immediately. Detection of an ischemic condition may be a function of the predicted value and its detected value (e.g., as a threshold). However, with the disclosed approach of sequentially identifying vessels, an ischemic condition, or any type of disruption or breach in a vessel or vascular network, may be visually observed by appropriate medical personnel almost immediately.

[0075] The deviation from the expected sinusoidal pattern may be caused, for example, by the onset of an ischemic condition in at least a portion of the anatomical structure / tissue visible in the video image, or by a local change in perfusion to the indicated region. An illustration showing this change in dynamics due to the onset of ischemia in a human subject is shown in FIG. 2A, and a closer zoom is shown in FIG. 2B. As can be seen, it is possible to detect a transition from a regular oscillatory fluorescent signal to an ischemic plateau. However, it should be noted that changes to the perfusion of the anatomical structure of interest may result in other measured patterns in addition to the ischemic plateau. One example is a venous occlusion, where the outflow of blood from the anatomical region is blocked or reduced, resulting in a change in oscillatory dynamics due to crowding or pooling of the fluorescent agent in the indicated region. As can be seen in FIG. 4, the result is not a plateau while the periodic oscillations cease.

[0076] However, the systems and methods of the present disclosure may also be configured for controlled injection of small boluses (e.g., minimal boluses) at irregular intervals, so long as the injection of the bolus is controlled such that the input signal generated is controlled and known. The body kernel approach of the present disclosure allows for the calculation of a predictive signal so long as the input signal is known. Thus, in that respect, the approach of the present disclosure does not rely on a regular input signal.

[0077] The system described herein can observe and detect changes in the perfusion level of a given region in a video image within seconds. This can be detected in regions that have been observed for long periods of time (e.g., minutes), where the dynamics are continuously visualized and the predicted signal can be continuously calculated. An illustration highlighting the difference between signals that one might expect to observe for ischemic / healthy tissue regions is shown in FIG. 2C. However, it can equally be determined in anatomical regions that have only been visualized for short time intervals (e.g., 10-20 seconds), because the described system is trained to predict and detect certain phases of that regular dynamic signal at a given time in tissue, consisting of regular rises and falls in the time-intensity signal. See FIG. 2D, which illustrates what an anatomical region of interest might look like if it drifts in and out of focus of the recorded image.

[0078] Preferably, the approach of the present disclosure may utilize tracking technology that may run independently in the background, while the surgeon is only exposed to a visible white light signal and is therefore only interrupted / informed by a warning signal, e.g., during detection of an abnormal perfusion pattern.

[0079] Oscillating input signal The disclosed method utilizes an input signal that varies in bolus volume versus time. Thus, the input signal is defined by a series of boluses of at least one fluorescent imaging agent, the series of boluses being administered with a predefined and / or controlled duration between successive boluses. One way to provide an oscillating input signal is to inject a microbolus of a fluorescent imaging agent (e.g., indocyanine green (ICG)) at predefined regular or irregular intervals over a period of time, such that the input signal oscillates according to the bolus injections. The bolus volume may be constant, but may be controllably varied for one or more of the boluses. Injection at regular intervals provides an input signal with a predetermined frequency.

[0080] Thus, the input signal may be defined in terms of volume of fluorescent imaging agent versus time. The fluorescent imaging agent may be injected intravenously, and the imaging agent may be excited by a light source having a wavelength or range of wavelengths suitable for exciting the imaging agent, as previously described. The microboluses allow for successive injections of the fluorescent imaging agent at controllable, predetermined and / or regular or irregular intervals over a period of one hour or even several hours, for example at regular intervals between 1-10 minutes, for example at 1-2 minutes, 2-3 minutes, 3-4 minutes, 4-5 minutes, 5-6 minutes, 6-8 minutes or 8-10 minutes intervals. In one embodiment, the duration between the successive boluses is between 5 seconds and 10 minutes over the duration of a medical procedure, which may range from a few minutes to several hours. Thus, the input signal is preferably defined by a series of boluses of at least one fluorescent imaging agent, where the series of boluses are administered with a predefined and / or controlled duration between the successive boluses. A series of boluses of the at least one fluorescent imaging agent can be provided into the subject's vein during image acquisition, thereby generating the input signal.

[0081] Fluorescence image acquisition The first step of the disclosed computer-implemented method is preferably to obtain a fluorescent image of the anatomical structure / tissue of a subject. Additionally or alternatively, the first step can be to obtain a fluorescent output signal from the tissue of the subject. Fluorescence can be automatically detected by illuminating the anatomical structure / tissue with a light source that can excite the fluorescent imaging agent, and the emission is quantified and / or analyzed through a series of fluorescent images of the anatomical structure / tissue. Any of the disclosed methods can include analyzing the fluorescent image and / or analyzing the fluorescent output signal, for example, detecting at least one peak fluorescent signal for each bolus.

[0082] Calculating the Body Kernel In general, all of the methods disclosed herein rely on the presence of an input signal that is introduced (e.g., intravenously) into the subject's body. The input signal oscillates in time and includes a series of microboluses (aka microdoses), each of which includes a fluorescent agent as described in more detail previously. In general, the input signal becomes smeared over time as it passes through the body, and the measured fluorescent output signal is a distorted or altered version of the input signal due to the effects of the body. This is illustrated in Figures 11-14. The degree of smearing / distortion depends on the body and which part of the body is being investigated using the disclosed methods. As an example, fingers and toes are located away from the heart and consequently affect the input signal differently (i.e., give a different response) than body parts closer to the heart. In this respect, the body / body part can be thought of as a filter that affects the input signal.

[0083] The body kernel should be understood here as a function, which represents or describes how the body affects the input signal, i.e. the body kernel can be understood as a filter as described above. In general, the body kernel is not known a priori. However, the inventors have realized that the body kernel can be estimated based on the fluorescence signal measured from the body - since the input signal is known.

[0084] A good idea comes from the world of image and signal processing. In image processing, a kernel (also known as a convolution matrix, or mask) is a small matrix that is used for blurring, sharpening, embossing, edge detection, etc. This is achieved by performing a convolution between the kernel and the image. Convolution is a type of matrix operation, the process of adding each element of an image to its local neighborhood weighted by the kernel.

[0085] In normal image processing, a kernel is applied to an image to extract more information from the image, i.e., the kernel is known a priori. In the context of the present disclosure, a body kernel is convolved with an input signal to estimate an output signal - and thereby detect abnormally perfused tissue regions. However, the actual body kernel is typically not known, because it is context dependent. Thus, one aspect of the present disclosure relates to the determination of the body kernel, which is where deconvolution comes in.

[0086] Deconvolution is the inverse operation of convolution. As mentioned above, convolution can be used to apply a filter in image processing, and if the filter is known, deconvolution can be applied to recover the original signal. In microscopic imaging, deconvolution is an image processing technique utilized to improve the contrast and resolution of digital images captured by a microscope.

[0087] Within image processing, the goal of deconvolution is typically to find a solution f to a convolution equation of the form: f*g=h, where h is some recorded signal and f is some signal we are trying to recover, but which was convolved with a filter or distortion function g before it was recorded. The function g may represent an instrument transfer function or a driving force applied to a physical system.

[0088] In this case, f represents the known input signal, h represents the measured output signal, and the function g represents the body kernel. With knowledge of the body kernel g, the output signal h can be calculated as GuessIt is possible to do this because the input signal f is controlled and therefore known. And if the output signal can be estimated, e.g. if the estimated output signal represents a normal perfusion pattern, it is also possible to detect normal and / or abnormal perfusion patterns. And it is possible to identify and / or detect normally and / or abnormally perfused tissue regions. And one way to get there is to determine the body kernel, for example by applying deconvolution, which is a well-known technique.

[0089] Thus, the body kernel, defined as the filter imposed on the fluorescent imaging agent by the subject's body, can be determined by deconvolving the measured fluorescent signal from a tissue (e.g., a region of interest in the tissue), thereby obtaining the body kernel of the ROI. The deconvolution (in time) can be performed in a variety of ways known in the art (e.g., through optimization formulation or in Fourier space). In that regard, the body kernel can be understood as a transfer function between an input signal (i.e., a bolus of fluorescent agent) and the fluorescent output signal (e.g., the fluorescent image and its analysis). Estimation of the body kernel is illustrated in Figures 11-15 (using synthetic data) and Figure 16 (using real-life data). For example, by continuously acquiring a fluorescent image of the tissue, the fluorescent image is associated with a fluorescent output signal that is correlated with the input signal as defined herein. At least one region of interest (ROI) may be selected in the fluorescence image either manually, semi-automatically and / or automatically; a body kernel of the ROI may be determined by deconvolving the ROI with an input signal, such that the convolution of the input signal with the body kernel represents a fluorescence output signal from the ROI. The duration between the successive boluses may be selected according to the tissue type of the ROI.

[0090] A subject, i.e., a patient, may have multiple body kernels, where each body kernel represents how a particular body part affects the input signal. Thus, body kernels may be understood to be spatially dependent, i.e., local phenomena. Thus, different body parts / regions of interest of the same subject may be associated with different body kernels. However, in a series of fluorescent images showing the tissue of a subject, most of the tissue is usually normally perfused, in that it is possible to determine a single body kernel that represents the normal perfusion pattern of normally perfused tissue. Such a single body kernel may be generated from multiple regions of interest, multiple tissue regions, or simply the average or median from multiple pixels, etc. In that regard, it is noted that blood vessels may be identified and "removed" from the determination of the body kernel, since blood vessels do not represent normally perfused tissue. The body kernel may be represented mathematically as a matrix in one or two dimensions (or higher).

[0091] The use of a body kernel is a great advantage since it provides a systematic knowledge of what the predicted fluorescence output signal from the tissue should look like. In conclusion, this improves the speed of the method since instead of not knowing what to expect, the actual measured signal can be compared with the calculated predicted output signal, possibly in real time. The predicted output signal is calculated by convolving the input signal with the relevant body kernel. Once a body kernel has been estimated for a particular region of interest, e.g., a normally perfused region, the body kernel can be used in combination with a known input signal to calculate a predicted output signal and compare it with its corresponding fluorescence output signal, thereby detecting abnormal perfusion patterns. The assessment of the similarity between the actual measured fluorescence output signal and the predicted output signal can be performed using various mathematical methods, e.g., to compare two graphs / curves or two images. Similarity in this context can be understood as, for example, the distance between said graphs / curves, e.g. the root-mean-square difference between functions, the weighted least squares (WLS) between curves, the Hausdorff distance, the difference area between two curves, χ 2 as a function of the input signal / curve, or generally in phase or frequency, particularly in frequency composition, and / or amplitude of the signals / curves being compared. However, as also described herein, detection, identification and / or classification of normal and / or abnormal perfusion patterns and / or regional comparison may also be provided by the body kernel level comparison / match, which may often be the preferred option since the body kernel is essentially independent of the input signal.

[0092] Defining normal perfusion and identifying abnormal perfusion In order to detect abnormal perfusion patterns, it is usually necessary to define normal perfusion patterns. Thus, the methods of the present disclosure preferably include a step of defining normal perfusion patterns, for example, by identifying tissue regions that have normal perfusion. The inventors have found a number of ways to identify and define normal perfusion. The different approaches are described below.

[0093] One method of estimating normal perfusion is a manual definition of normal perfusion. According to one embodiment of the disclosed method, at least one tissue region having normal perfusion is identified manually, for example, by a surgeon selecting the tissue region as having normal perfusion. In this approach, the surgeon or physician simply selects the tissue region that is deemed to have normal perfusion (i.e., visual assessment). This has the drawback that the assessment is based on the surgeon's / physician's experience and is not completely objective. Once the region is designated, a spatially localized body kernel can be estimated for the region. Subsequently, the body kernel of the region deemed to have normal perfusion can be convolved with the input signal to calculate its predicted output signal (output signal predicted from a tissue region having normal perfusion). The fluorescence output signal from a region with abnormal perfusion (e.g., due to inflammation) deviates from the predicted output signal. Thus, by comparing the fluorescence output signal with the predicted signal, the region of abnormal perfusion can be detected / identified. Thus, one embodiment of the disclosed method further comprises determining a body kernel of at least one region of interest (ROI) in the fluorescence image and convolving the input signal with the body kernel, thereby defining a normal perfusion pattern of said ROI.

[0094] Another method of estimating normal perfusion is to analyze the fluorescence output signals from multiple areas / regions of interest distributed across different regions of the image. If the fluorescence output signals of multiple or these regions are substantially similar, these regions may be assumed to represent normal perfusion patterns and one or more associated body kernels may be determined. This approach differs from the first approach in that one or more associated body kernels may be determined automatically. A spatially local body kernel may be determined for each region assigned as normal perfusion, and such local body kernels may be assigned a spatial confidence region that is deemed valid for signal inference. Conversely, a region providing an output signal that correlates with the calculated predicted output signal within a predefined tolerance may be considered to have a normal perfusion pattern.

[0095] Yet another way to identify or aid in identifying normal perfusion is to have a reference detection elsewhere on the patient's body. As an example, if a surgeon is operating on a patient's arm, the reference can be obtained on the patient's finger, for example, using a light-emitting diode finger clip. The light-emitting diode finger clip can be similar to those used for pulse and blood oxygen level measurement (i.e., pulse oximeters). The light-emitting diode finger clip should have a wavelength corresponding to the peak absorption frequency of the fluorescent imaging agent (e.g., ICG) used in the tissue, and it should preferably further include an excitation light source and a photosensitive element with peak sensitivity at the wavelength emitted by the fluorescent imaging agent. This allows the intensity of the fluorescent signal to be measured transcutaneously. Thus, one embodiment of the disclosed method includes measuring the intensity of the fluorescent output signal in a transcutaneous manner by means other than image acquisition, for example, using a photodiode and / or a light-emitting diode finger clip. The intensity is preferably measured continuously, i.e., as intensity versus time. In this approach, a reference signal obtained by the light emitting diode finger clip is considered to be representative of a normal perfusion pattern. This can be compared to the fluorescent output signal from one or more regions of interest in the acquired fluorescent image. Regions of interest that compare sufficiently, e.g., in phase, frequency composition, and / or amplitude, with the transcutaneously measured signal can be considered to represent tissue regions having normal perfusion patterns, and one or more associated body kernels can then be determined. This approach can also thereby automatically identify normal perfusion.

[0096] Another or complementary method of defining normal perfusion and / or normal perfusion patterns is by comparing previous fluorescent output signals (e.g., a database of fluorescent output signals that have been labeled as having normal perfusion patterns, possibly also labeled with respect to the type of tissue and / or anatomical region being examined). Again, the kernel approach of the present disclosure is applied, where labeled body kernels have been generated with respect to normally perfused tissue regions such that there is a database of labeled body kernels. One or more of these labeled body kernels can then be directly applied as a method of defining normal perfusion patterns in the approach of the present disclosure.

[0097] Alternatively, or supplementally, if normal perfusion patterns are to be identified, multiple normally labeled body kernels may be convolved with the current input signal to generate multiple predicted output signals for normally perfused tissue. The predicted output signals may be compared with the current fluorescence output signal such that one or more normal perfusion patterns may be identified in one or more regions of interest in the fluorescence output image, thereby identifying one or more normally perfused tissue regions. One or more body kernels representing normally perfused tissue may then be generated based on the input signal and the fluorescence output signals from the normally perfused tissue regions.

[0098] Thus, there are many methods to automate the identification of normal perfusion patterns.

[0099] Classification of perfusion patterns The present disclosure further relates to methods for classifying abnormal perfusion patterns, i.e., where relevant tissue regions identified as abnormally perfused may be classified, for example, with respect to the cause of the abnormal perfusion pattern, e.g., the relevant tissue regions may be classified as cancerous tissue, glandular tissue, thyroid tissue, tumor tissue, inflamed tissue, ischemic tissue, etc. Tumor tissue may be cancerous tissue, but the tumor may also be benign, whereby it is not cancerous but still has an abnormal perfusion pattern.

[0100] Thus, one or more of the tissue regions detected as having abnormal perfusion patterns based on the at least first body kernel and / or the at least second body kernel may be classified as cancerous tissue, glandular tissue, thyroid tissue, tumor tissue, inflamed tissue, or ischemic tissue, wherein the classification is preferably based on the at least one second body kernel being labeled.

[0101] However, just as abnormal perfusion patterns may be classified, normal perfusion patterns may encompass several types of tissue, such as muscle tissue, body fat, ligament tissue, organ tissue, as well as arteries and veins. Thus, one or more of the tissue regions detected as having normal perfusion patterns based on the at least first body kernel and / or the at least second body kernel may be classified as muscle tissue, body fat, ligament tissue, organ tissue, vein, or artery. In that regard, the classification is preferably based on the at least one first body kernel being labeled.

[0102] The body kernel approach of the present disclosure can again be applied, but this time the body kernel becomes a filter imposed by the subject's body on the fluorescent imaging agent in normally or abnormally perfused tissue regions, and labeled body kernels have been generated for classified abnormally perfused or normally perfused tissue regions such that a database of labeled body kernels exists.

[0103] Thus, when an abnormal or normal perfusion pattern is detected, for example, by the approach of the present disclosure, a number of labeled body kernels can be convolved with known input signals to generate a number of predicted output signals. The predicted output signals can be compared with the fluorescence output signals from the detected normally or abnormally perfused tissue regions. This can be provided in real time and / or with previously acquired images. Alternatively and / or supplementally, a body kernel can be calculated from the detected normal or abnormal perfusion pattern by deconvolution as described herein. The calculated body kernel can then be compared with the labeled body kernel and, if possible, classified thereby. That is, classification can be provided by comparing the body kernels and / or by comparing the actual signal with the predicted signal calculated from the labeled body kernel.

[0104] The abnormally perfused tissue regions, which have been detected, may also be classified by comparing the fluorescence output signal from the abnormally perfused tissue region with previously acquired fluorescence output signals (i.e., labeled fluorescence output data) from the classified abnormally perfused tissue regions. However, the drawback is that the fluorescence output signal is dependent on the input signal. The body kernel approach of the present disclosure is independent of the input signal. That is, even if the labeled body kernels are based on different types of input signals, they may still be applied and convolved with the current input signal, thereby providing a much more accurate labeled predicted output signal for comparison with the measured fluorescence output signal from the abnormally perfused tissue region.

[0105] Once one or more tissue regions have been detected and similarly classified as normal or abnormal, they may be marked accordingly, e.g., visually in the image, e.g., by augmented reality, preferably in real time, such that these regions are marked for the surgeon.

[0106] For example, abnormal perfusion patterns in tumors can be caused by uncoordinated and increased proliferation of blood vessels. Due to the uncoordinated nature, fluorescent molecules can remain longer in tumors and therefore have a slower washout than is typical for normal tissue. This change in fluorescence dynamics can be detected by the approach of the present disclosure.

[0107] For example, abnormal perfusion patterns in inflamed tissues may be caused by dilated blood vessels in inflammation, which results in increased perfusion of the inflamed tissue that may result in increased inflow of blood. This change in fluorescence dynamics can be detected by the approaches of the present disclosure.

[0108] Application of Artificial Intelligence (AI) The disclosed approach may be improved through the use of artificial intelligence, e.g., machine learning, neural networks, etc., trained to identify normal and / or abnormal perfusion patterns and, in the case of abnormal perfusion patterns, classify the abnormal perfusion pattern, e.g., via the disclosed body kernel approach (where a database of labeled body kernels exists), thereby also training to identify the cause of the abnormal perfusion pattern, if possible.

[0109] For example, in a form of supervised learning of neural networks, the network is trained using a large amount of labeled data, e.g., in the form of images or videos and / or labeled body kernels as described above. The data may be labeled by experts such as surgeons and physicians who may identify abnormal tissues such as glandular tissues in images and / or videos. The labeled data may then be stored in a database ("knowledge bank") and later retrieved and compared to new measurements. The labeled data in the database may then be refined as more data from more surgeries is collected (from different patients). Once the "knowledge bank" is of sufficient size (e.g., data from 500-1000 surgeries), the neural network is configured to compare the locally estimated body kernels from new surgeries with the "true" body kernels stored in the database. If the estimated body kernels correlate / match with the body kernels in the database, the tissue region is preferably automatically flagged / auto-segmented so that the surgeon may check the region in more detail.

[0110] Identifying Arteries and Veins One embodiment of the disclosed method includes identifying arteries and veins in the identified blood vessels. This identification between arteries and veins is advantageously based on the oscillating fluorescent signal, i.e., a predetermined pattern of the input signal. The bolus hemodynamics is different for arteries and veins, e.g., when a bolus of ICG spreads in a patient, the fluorescent signal appears first in the arteries, then in the microcirculation of the surrounding tissue, and after a while in the veins. Thus, the time difference between the bolus injection and the arterial signal, tissue signal, and vein can be determined, e.g., from the first bolus or the first few boluses by analyzing a series of corresponding images, looking for signals in that order. That is, the novel microbolus dosage regimen currently used imposes a number of time and / or wavelength dynamics that can be exploited. Since only one or a few microboluses of a fluorescent imaging agent are required to determine patient-specific / situation-specific time differences between the bolus injection and the arterial signal, and between the arterial and venous signals, these time differences can be utilized in subsequent controlled microbolus administrations to successively distinguish arteries and veins. The time differences between the arterial and tissue signals, and between the tissue and venous signals, can also be utilized. Furthermore, the washout period of each microbolus involves different hemodynamics of arteries, veins, and surrounding tissues, which can be utilized to identify blood vessels and distinguish between arteries and veins, even during the washout period between bolus injections.

[0111] Identifying blood vessels One example of identifying blood vessels in a fluorescent image is by image filtering, for example, filtering the acquired fluorescent image based on a time-varying image gradient. Due to the oscillating fluorescent signal and hemodynamics in the tissue, there are periodically regions in the acquired image where the image brightness changes significantly. These regions are most often either arteries or veins, so by constantly applying appropriate image filtering, blood vessels will appear substantially constantly during successive bolus administrations.

[0112] Another more accurate approach is to identify blood vessels based on the inherent phase difference between the fluorescent signal associated with each blood vessel and surrounding tissue. This is due to the hemodynamics in the subject / patient's body. The global oscillating fluorescent signal as disclosed herein is imposed by controlled and repeated injection of small boluses of fluorescent agent such as ICG. Each bolus of the fluorescent agent also produces a fluctuating fluorescent signal at a more local level due to the hemodynamics in the subject. Each small bolus of fluorescent agent reaches arteries, tissues and veins at different times, and the disclosed approach utilizes these time differences to identify blood vessels, distinguish blood vessels and tissues, and distinguish arteries, veins and tissues. The time difference between the fluorescent signals originating from arteries, tissues and veins can be seen as a phase difference in the successive evolution of the fluorescent signal. That is, at any time during the oscillating fluorescent signal, there is a phase difference at different positions in the corresponding acquired fluorescent image that arises from the fluctuating hemodynamics in arteries, veins and surrounding tissues, i.e., due to the inherent time difference originating from the hemodynamics of the subject. The subject-specific time difference may be known or the first of several boluses may be used to determine the subject-specific time difference, i.e., to determine at least one time difference selected from the following group: The time difference between the bolus injection and the arterial or venous fluorescent signal, the time difference between the arterial and venous fluorescent signals, and the time difference between the arterial or venous fluorescent signal and the tissue fluorescent signal. Thus, any of the methods disclosed herein may analyze at least a portion of the fluorescent images / signals to identify the following groups: - the time difference between the bolus injection and the arterial or venous or tissue fluorescent signal, - the time difference between the arterial and venous fluorescent signals, and - the time difference between the arterial or venous fluorescence signal and the tissue fluorescence signal, determining at least one time difference selected from: sequentially identifying tissues and / or blood vessels in said fluorescence image based on said time differences and said fluorescence output signals. It may further include:

[0113] This may be provided once for some of the bolus injections or for each of the bolus injections.

[0114] One or more of these time differences can be "translated" into a corresponding phase / phase difference or one or more of blood vessels, arteries, veins, surrounding tissue, and / or something else. Thus, knowing the expected phases of arteries, veins, and surrounding tissue, it is possible to thereby associate each pixel, or ROI, in the fluorescence image with a classification of either artery, vein, surrounding tissue, or something else.

[0115] The predicted phase of the arteries, veins and surrounding tissues at a given time point can be directly related to a predefined oscillation pattern that determines the period and oscillation frequency of the oscillating fluorescent signal. The predicted phase can be either estimated, calculated and / or determined during the medical procedure, for example, as an initial learning period in which the oscillating fluorescent signal is observed for one or several microboluses to measure its associated hemodynamics in a specific situation. Thereby, the predicted phase of the arteries, veins and surrounding tissues can be directly related to the time point at which the bolus is injected. Thus, in the subsequent medical procedure, the time point of each bolus injection provides information of the predicted phase of the arteries, veins and surrounding tissues.

[0116] Thus, the disclosed method may include determining the phase difference between a fluorescent signal originating from an artery in the tissue and a fluorescent signal originating from a vein in the tissue, and optionally a fluorescent signal originating from tissue surrounding the blood vessel, and relating this phase difference to a predetermined vibration pattern.

[0117] Thus, analyzing the phase and / or phase difference in a sequence of fluorescence images is one way to identify blood vessels in a fluorescence image. That is, blood vessels can be identified / detected in a fluorescence image sequence at a pixel-by-pixel level by knowing the phase of the fluorescence signal - and this can be detected at any time during the microbolus dose regimen. This phase information, coupled with the knowledge of a given bolus dose, i.e., that said fluorescence signal is oscillating in a given pattern, provides the necessary information whether a pixel in an image, or a region of interest (ROI), is a blood vessel, and further whether a pixel in an image, or a region of interest, is an artery, a vein, tissue or something else. Thereby, the actual mapping of blood vessels (including arteries and veins) can be provided, for example, pixel-by-pixel or group of pixels, or ROI-by-ROI, mapped in the form of each pixel - or ROI - classified as a blood vessel (preferably including an artery or a vein), surrounding tissue or something else as required, in substantially each fluorescence image.

[0118] The disclosed approach may provide medical personnel with a constantly updated map of blood vessels (including arteries and veins) within the surgeon's field of view / anatomical region of interest. In particular, the identified arteries and veins may be superimposed onto a white light image and visually enhanced so that the arteries and veins are visually distinguishable (e.g., each has a distinct high contrast color) and displayed on a screen.

[0119] Once identified and possibly mapped, in a fluorescence image, or a sequence thereof, said vessels (preferably including the identified arteries and veins) can be tracked in subsequent fluorescence images, even if motion (e.g. peristalsis) is involved. Tracking of the identified vessels and / or tissue regions with normal perfusion patterns and / or tissue regions with abnormal perfusion patterns can be provided, for example, by using tracking methods available in the prior art and known to the skilled person. An example of tracking in fluorescence images is disclosed in WO 2018 / 104552.

[0120] system As noted above, the present disclosure further relates to systems configured to perform any of the methods disclosed herein.

[0121] In one embodiment, the system further includes a controllable injection pump for holding at least one first fluorescent imaging agent, the injection pump configured to inject a series of predefined boluses of the first fluorescent imaging agent into a vein of the subject, thereby generating the input signal.

[0122] The fluorescent image may be received by a suitable imaging unit (e.g., a camera, e.g., a surgical, laparoscopic or microscopic camera (e.g., a video camera)) that may be part of the system of the present disclosure. The analysis may be provided by a processing device, either locally or as part of a cloud service. The system of the present disclosure may be configured to perform all of the steps of the method of the present disclosure.

[0123] The predefined bolus preferably corresponds to the first fluorescent imaging agent of less than 0.01 mg ICG per kg body weight, more preferably less than 0.005 mg ICG per kg body weight. The predefined bolus may also correspond to the first fluorescent imaging agent of less than 0.5 mg ICG. Preferably, the fluorescent agent is ICG, and the amount of ICG in the predefined bolus is preferably less than 0.01 mg ICG per kg body weight of the subject, most preferably less than 0.005 mg ICG per kg body weight of the subject. Thus, preferably, the amount of ICG in the predefined bolus is less than 1 mg ICG or less than 0.5 mg ICG, most preferably less than 0.25 mg ICG.

[0124] The system may be configured to inject a bolus at intervals of between 5 and 600 seconds (eg, between 15 and 300 seconds, eg, between 45 and 210 seconds, eg, between 90 and 120 seconds).

[0125] A further advantage of the disclosed approach is the opportunity to identify local networks of blood vessels: by briefly clamping a freely visible vessel, its associated perfused region is delimited, and by subsequently observing its fluctuating ICG signal, its associated network becomes clearly visible, since perfusion in that region changes rapidly.

[0126] digestive tract Complications related to the digestive tract are often related to local hemodynamics. Thus, changes in normal hemodynamic status may be an indicator of increased risk of complications. Perfusion assessment of the digestive tract, especially in and near the surface of the digestive tract, such as the tissue of the digestive tract wall, may therefore be an important diagnostic tool when examining the digestive tract, for example, in laparoscopic diagnostic examinations, laparoscopic surgery with laparoscopy or traditional laparoscopy, or robotic surgery, and open surgery, for example, to diagnose or localize complications. Perfusion assessment is also important during the surgical procedure of creating an anastomosis, which may be provided to establish communication between two previously separate parts of the digestive tract. As an example, an intestinal anastomosis establishes communication between two previously separate parts of the intestine, typically restoring intestinal continuity after removal of a pathological condition affecting the intestine. Intestinal anastomoses may be provided, for example, for 1) restoration of intestinal (e.g., bowel) continuity following resection of pathological intestine, and 2) bypass of unresectable pathological intestine (e.g., bowel). Certain pediatric conditions may also require intestinal anastomoses. Resection of pathological bowel may be performed in the following situations: Intestinal gangrene due to vascular compromise caused by mesenteric vascular disease, prolonged intestinal obstruction, intussusception, or intestinal volvulus Malignant tumors Benign conditions (e.g., intestinal polyps, intussusception, roundworm infections associated with intestinal obstruction) Infections (e.g. tuberculosis complicated by stricture or perforation) ·Traumatic perforation Large perforations (traumatic) that are not amenable to primary closure Radiation enteritis complicated by bleeding, stricture, or perforation, Inflammatory bowel disease, ulcerative colitis, or Crohn's disease that is refractory to medical treatment or associated with complications (e.g., bleeding, perforation, toxic megacolon, dysplasia / carcinoma) Chronic constipation, idiopathic slow transit constipation, or Hirschsprung's disease: If the above conditions are refractory to medical treatment, a subtotal colectomy may be performed.

[0127] Bypass of unresectable pathological bowel may be performed in the following situations: Locally advanced tumors causing luminal obstruction Metastatic disease causing intestinal obstruction Poor general condition or other conditions that prevent major resection

[0128] Pediatric conditions that may require an intestinal anastomosis include: Congenital anomalies (e.g. Meckel's diverticulum, intestinal atresia, malrotation with intestinal volvulus leading to gangrene, meconium ileus, gastric duplication cyst, Hirschsprung's disease) Inflammatory conditions (e.g. necrotizing enterocolitis, enteritis, tuberculosis, intestinal perforation) Other conditions (e.g. intussusception, angiodysplasia, polypoid disease, ascariasis) As part of other surgical procedures (e.g., Kasai portoenterostomy, choledochal cyst, urinary diversion, pancreatic tumors)

[0129] Postoperative complications associated with anastomoses in the digestive tract are unfortunately frequent and often result from insufficient perfusion (capillary blood supply) at the anastomosis (i.e., the joining of two parts of the tube). Examples of insufficient or abnormal perfusion can lead to anastomotic leakage, which is a serious and frequent complication associated with, for example, colon surgery (more than 10% of the procedures result in complications). Within colon cancer surgery, more than 30% of patients with anastomotic failure die due to postoperative complications, and approximately 25% of the remaining patients suffer from a stoma for the rest of their lives. Risk factors associated with anastomotic failure include anastomotic tension, tissue damage, and especially reduced blood perfusion.

[0130] The present disclosure further relates to performing image analysis of one or more video sequences (wherein fluorescent images of said tissues are continuously acquired as described herein) representing at least a portion of the digestive tract, e.g., acquired pre-, intra- and / or post-operatively during surgery, particularly surgery involving the digestive tract, which may be the case for digestive tract surgery - thus, said video sequences may preferably include at least some outer portions of the digestive tract, such that perfusion in at least some of the digestive tract walls may be measured and assessed.

[0131] The digestive tract is an organ system in humans and other animals that ingests food, digests it, extracts and absorbs energy and nutrients, and excretes the remaining waste products as feces and urine. The digestive tract can be understood as the tube that carries food to the digestive tract. Thus, the term digestive tract, as used herein, includes the buccal cavity; the pharynx; the small intestine, including the duodenum, jejunum, and ileum; the stomach, including the esophagus, cardia, and pylorus; and the large intestine, including the cecum, colon, rectum, and anal canal.

[0132] clinical application Visualization of vascular anatomy and detection and classification of abnormal perfusion patterns as disclosed herein are of great importance during almost any type of surgery, since continuous detection of blood vessels reduces the risk of unintentional severing of blood vessels. Meanwhile, detection and optional classification of abnormally perfused tissue can help in identifying areas of malignant, cancerous, and / or inflamed tissue. The systems and methods of the present disclosure may also be used within the following clinical applications, among others:

[0133] Abdominal / General surgery Resection surgery, to rapidly identify the exact vessel location to be cut / ligated and to detect tissue areas with reduced perfusion. Ischemic bowel surgery, to rapidly determine which and where blood vessels are blocked and anatomical structures are poorly perfused. Acute abdomen, to detect underlying pathology, e.g. to help rule out ischemia. Repeat surgery, in previously operated patients with numerous surgical adhesions. To detect blood vessels and abnormal perfusion patterns, such as in cancer surgery, resection procedures performed by general surgeons. General surgery, involving virtually any organ in the abdominal tract, e.g. performing anastomosis or surgery on the stomach. General surgery, involving superficial or deep infections, to detect and map blood vessels and / or to detect, classify and / or map abnormally perfused tissue areas.

[0134] thyroid surgery Thyroid surgery, including the removal of thyroid tissue. Continuous detection of blood vessels and abnormal perfusion patterns is a major advantage since thyroid surgery carries the risk of excessive bleeding. Thyroid surgery also carries the risk of removal or removal of one or more portions of the parathyroid glands, and the disclosed approach may be applied to identify and map blood vessels in the relevant regions so that the parathyroid glands are more easily visible to the surgeon. In one embodiment, the images are acquired during thyroid surgery, where blood vessels and / or abnormally perfused tissue regions within one or more of the parathyroid glands are identified and visualized to medical personnel involved in the surgery.

[0135] Pelvic surgery Gynecological / Uro-Gynecologic Surgery, to quickly locate the exact vessels to be cut / ligated and to identify and identify vessels intraoperatively. Cancer surgery, to detect blood vessels and / or to detect and advantageously classify abnormally perfused tissue areas.

[0136] Kidney surgery Cancer surgery, to detect blood vessels and / or to detect and advantageously classify abnormally perfused tissue areas.

[0137] plastic surgery Skin grafting, to rapidly identify abnormally perfused tissue areas and / or exact vessel location in the donor to be cut / ligated and to monitor perfusion of the same vessels on the recipient; to monitor healing and vascularization in the acute intraoperative situation and also in the days that follow. · In all skin close to the surgical procedure, to detect and map abnormally perfused tissue areas in an anatomical region before the surgical procedure begins, and / or to detect and map vascular structures, as well as during the surgical procedure if the surgeon deems this necessary for further information. That is, the disclosed approach may be used as a clinical tool to identify and map a patient's blood vessels and their associated vascular anatomy days or weeks before a surgical procedure is to be performed, giving medical personnel time to carefully plan the procedure. This mapping may be done, for example, according to the disclosed approach, which measures the fluorescent signal through the skin of the area of ​​interest over a period of 30-45 minutes and creates a 2D or 3D map for medical personnel to use in planning. The identified vascular map may be integrated with other examinations (e.g., CT scan, MR scan, or ultrasound scan). The disclosed approach has the advantage of allowing identification and mapping of blood vessels that are too small to be accurately mapped in, for example, a CT scan.

[0138] Ear, Throat, and Neck Surgery A variety of procedures including: facial cosmetic surgery, tracheotomy, cancer, etc.

[0139] orthopedic surgery Amputation, to rapidly identify and map abnormally perfused tissue areas and / or identify and map blood vessels before a limb or anatomical area is amputated, to select the exact and best site for amputation, to reclose the skin and ensure optimal healing. To detect and map abnormally perfused tissue areas, such as infections, and / or to detect and map blood vessels where debridement or similar procedures may be required.

[0140] Cardiac surgery To rapidly detect and map abnormally perfused tissue areas and / or identify and locate peripheral blood vessels to the heart during CABG, bypass surgery.

[0141] vascular surgery Amputation surgery, to rapidly detect and map abnormally perfused tissue areas and / or to identify and locate the exact vessels to be cut / ligated as described above. ·Vascular harvesting for bypass surgery, to detect and map abnormally perfused tissue areas and / or to rapidly identify and locate the exact vessels to be cut / ligated. EXAMPLES

[0142] Working Example The intensity curves shown in Figures 1A-B are the result of injection of a bolus with a regular amount of fluorescent agent (ICG, in these cases). The amount of ICG in each bolus was selected such that the fluorescent emission is visible to the human eye. Examples are provided to illustrate various perfusion parameters that can be calculated after fluorescent imaging. These same parameters can also be determined, to a large extent, after injection of much smaller doses (i.e., the micro-dose approach disclosed herein, possibly with repeated and sequential measurements of perfusion and associated evaluation and identification of blood vessels).

[0143] FIG. 1A is an example of an intensity curve acquired from a tissue, e.g., from a region of interest in a video sequence, after a bolus of ICG has been provided to a subject. The same type of data could be obtained using a different contrast agent. The intensity is essentially zero until a sudden increase in intensity indicates the passage of ICG molecules in the imaged tissue, which are excited to fluoresce. The intensity peak is followed by a gradual washout of the ICG molecules. Intensity is shown in arbitrary units.

[0144] 1B is the corresponding intensity curve, where the hemodynamic parameters perfusion slope, slope start, slope end, maximum intensity, washout slope, washout start and washout slope end are calculated and plotted. Assessment of perfusion parameters is further disclosed in pending application WO 2018 / 104552, which is incorporated herein by reference in its entirety.

[0145] Figure 2A shows actual measurement data from a human subject, who was repeatedly injected with microboluses of ICG at regular intervals (in this example, the intervals were approximately 2 minutes). The first microbolus of ICG contained an amount of 0.00456 mg ICG / kg human subject body weight, and each subsequent microbolus of ICG contained an amount of 0.00456 mg ICG / kg human subject body weight. The same amount of ICG / kg human subject body weight was included. The time intensity curves show a substantially sinusoidal pattern that increases linearly over time. The increase in intensity over time is related to the ratio between the dose of fluorescent agent and the washout time (during which the fluorescent intensity decreases). At a certain time point (approximately t=3800 s, FIG. 2B), perfusion is restricted, causing the onset of ischemia. This is because a lack of oscillation can be seen after this time point, forming what can be described as the ischemic plateau.

[0146] FIG. 2C shows idealized data showing a sinusoidal time-intensity curve. The measured ROI intensity increases upon injection of the fluorescent imaging agent and decreases during the washout phase. At approximately t=3750 seconds, the measured data shows a constant measured ROI intensity value due to the onset of ischemia. Alternatively, if there was no ischemia, the measured value would instead be expected to follow a dashed line such that the measured ROI value continuously follows a sinusoidal pattern.

[0147] FIG. 2D shows idealized data showing sinusoidal time-intensity curves without ischemic conditions when the anatomical region of interest drifts in and out of focus. The dashed lines show the expected measurements if the ROI is continuously observable. If this is not possible, for example due to the anatomical region of interest drifting in and out of focus of the recorded image, the measured data may not be complete, but instead there may be gaps - time intervals where no measured data was acquired for the anatomical region of interest. Thus, the system can preferably recognize the sinusoidal pattern even when the recorded data is not complete, because the phase of the dynamics is known, for example because the predicted fluorescence output can be determined when the associated body kernel is obtained. If the system can accurately recognize the sinusoidal pattern, it is provided along with the predicted intensity value of the ROI at each time point, which can then be used for comparison with the measured value. If the measured value differs from the expected value, the system can be configured to provide a warning to the surgeon. Thus, the system may be configured to recognize the phase of the oscillating / sinusoidal pattern at a measured time point or interval, which is then compared to a predicted phase for that time point or interval, where the predicted phase is preferably based on the recognized oscillation pattern or / known frequency of repeated bolus injections. As a result, the system does not necessarily require consecutive measurements, but may instead be based on a predicted phase of the oscillation pattern in combination with the time information of the measured time point or interval, such that a specific phase of the oscillation pattern is predicted to be present at the measured interval.

[0148] FIG. 3A shows fluorescence intensity measurements of a human subject taken over a longer time interval (approximately 40 min), where the human subject was repeatedly injected with microboluses of ICG. The first microbolus of ICG contained an amount of 0.0046 mg ICG per kg of human subject body weight, and each subsequent microbolus of ICG contained the same amount of 0.0046 mg ICG per kg of human subject body weight. The intensity of seven separate ROIs was measured and assigned to separate colors in the graph. The measured fluorescence intensity exhibits a periodic sinusoidal pattern, where the frequency coincides with the injection frequency (approximately 120 s). The pattern increases substantially linearly due to accumulation of the fluorescent imaging agent due to the relatively short period of injection compared to the dose size. At approximately t=2000 s, the repeated injections of the fluorescent imaging agent were stopped, causing an approximately exponential decay of the fluorescence intensity.

[0149] Figure 3B shows an enlargement of the marked area of ​​Figure 3A, where smaller variations can be seen within the same ROI and between different ROIs, and at the same time, the periodic intensity pattern is distinct from the pattern of each ROI having the same period.

[0150] FIG. 4 shows a time-intensity plot of measurements made in a human subject by repeated injection of microboluses of a fluorescent imaging agent. The first microbolus of ICG contained an amount of 0.00456 mg ICG / kg human subject body weight, and each of the subsequent microboluses of ICG contained the same amount of 0.00456 mg ICG / kg human subject body weight. The graph shows the result of venous occlusion, where, between approximately t=62-78 min, perfusion is limited, but not completely prevented. In this case, the oscillatory dynamics of the measured fluorescence intensity ceases, and the measurements show an irregular increase during venous occlusion. It should therefore be noted that reduced perfusion does not necessarily result in a plateau, as typically occurs in other cases, during ischemic conditions.

[0151] FIG. 5A shows a snapshot of the ICG analysis tool running on a humanoid subject. The acquired images in FIGS. 5-10 show a site on the right forearm, i.e., the fluorescent signal is visible across the skin of the arm. The image in FIG. 5A was taken at a very early stage of the bolus injection, where the microbolus of ICG has just been injected and is beginning to enter the artery; i.e., some arteries can be identified. The four boxes in the image indicate the measurement regions (aka regions of interest (ROIs)), and four ICG intensity curves are shown on the right (one for each ROI). One of the ROIs is located in an artery, and its corresponding intensity curve is the highest. One ROI is located in a tissue region, and the associated intensity curve indicates that some ICG has already diffused into the tissue. Two ROIs are located in veins and are almost flat, indicating that ICG has not yet left the tissue region and been transported back via the veins. The intensity curves of the two venous ROIs coincide and cannot be distinguished from each other. The phase differences of the four ICG intensity curves are clearly visible, so if the expected phases of arteries, veins and tissues are known, the four ROIs in Figure 5A can be classified as arteries, veins and surrounding tissues, respectively.

[0152] FIG. 5B shows a snapshot several seconds later than FIG. 5A, where several arteries can be visually identified. There are still some dark areas where ICG has not yet spread, i.e., ICG is still entering the subject. All four ICG intensity curves of the four ROIs are steadily increasing. However, as can be seen from the curves, there is a clear phase difference for three groups of curves: 1) arteries, 2) tissue, and 3) veins. The arteries ROI is in front of the two other groups. The tissue ROI is "behind" the arteries curves, and the tissue ROI is "in front" of the veins ROI. Note that the phase difference in both directions is more or less equal, i.e., the tissue is approximately halfway through the movement of the ICG molecules from arteries to tissue to veins. Again, an ROI can be classified as either an artery, a vein, surrounding tissue, or other, if its expected phase is known. This can be provided for all pixels in the image, or for groups of pixels. That is, every pixel can be classified as an artery, vein, surrounding tissue, or other if its expected phase is known. That is, image filtering can provide a visual identification of blood vessels, but if the phases of the various signals are known for a given vibration pattern, all or most of the pixels in the image can be classified; much more detailed information is obtained. This can be utilized, for example, when overlaying blood vessels onto white light images, and also when it is needed to track objects in a sequence of images where motion, e.g., peristalsis, is occurring.

[0153] FIG. 6A shows a snapshot several seconds later than FIG. 5B, where the fluorescence intensity from a particular microbolus ICG is reaching its peak value. After this point, more ICG starts leaving the tissue region than the amount of ICG entering the tissue region. Comparing this image with the snapshots shown earlier in FIGS. 5A-B, it can be seen that almost all the regions are visible here. The darkest regions are now veins, which have not yet started to transport ICG away from the tissue. This is also reflected in the corresponding ICG intensity curves on the right. The ICG curves are all still growing. That is, we are still in the phase dominated by the arteries, but the concave shape of the curves indicated that the peak intensity is approaching. However, a phase difference is still discernible between the ICG intensity curves. In this snapshot, the blood vessels are not clearly identifiable because too much ICG has entered the tissue region from the arteries and not enough ICG has yet entered the veins. However, the veins may actually be identifiable as dark regions.

[0154] FIG. 6B shows a snapshot approximately one minute later than FIG. 6A, illustrating the washout phase where veins can be clearly identified. The corresponding perfusion analysis from the associated ICG curves shows that the venous ROI clearly has the highest intensity, but the time scale is also different in the washout phase. That is, it may require a longer time scale to evaluate the phase difference between the artery and the tissue, the vein, in the washout phase. However, the phase difference is visible in the ICG curve.

[0155] FIG. 7 shows an edge-filtered version of FIG. 6B illustrating one example of visual enhancement that may be provided to medical personnel during a medical procedure by utilizing the approach of the present disclosure. Once the phase difference is known for a particular situation, it is known when the arteries and veins are each optimally identified. In FIG. 7, the veins are visually enhanced, which allows the surgeon to avoid accidentally cutting any veins. A pixel-by-pixel approach is used in FIG. 7 in combination with edge filtering on FIG. 6B followed by smoothing filtering. The result is a visibly darkened entire image except for the main veins. This information may be overlaid / superimposed on any screen visible to the surgeon during surgery (even white light image screens) to improve the basis on which the surgeon draws his decisions.

[0156] Figure 8 shows an image in which arteries have been identified in the sequence of images shown in Figures 5A-6B. The arteries are visually enhanced in black, as can be seen in the greyscale image.

[0157] Figure 9 shows an image in which veins have been identified in the sequence of images shown in Figures 5A-6B. The veins are visually enhanced in black, as can be seen in the greyscale image.

[0158] FIG. 10 shows an image in which the arteries and veins shown in FIGS. 8-9 are visually enhanced in red (arteries) and blue (veins) and superimposed in an image such that the arteries and veins are clearly visible and easily identifiable. This is an example of an augmented reality (AR) view that may be provided to a surgeon during a medical procedure using the approach of the present disclosure. Red pixels have been mapped to arterial groups and blue pixels have been mapped to venous groups. These mappings may be continuously updated since a microbolus regimen may be performed in the background. In practice, a given group (e.g., arterial group) has a known phase that may be identified in an initial learning phase of the microbolus regimen (where the arterial phase may be associated with the time of each bolus injection). Once the arterial phase is known, a segment of the fluorescent signal (e.g., for a few seconds) is sufficient to assess the associated perfusion of one or more pixels or ROIs in the sequence of images and calculate a phase matching score. If the phase at a pixel / ROI matches the arterial phase, the phase matching score is high and the pixel may be identified as an artery, which may be colored red. A similar cycle may be performed for veins. Such phase matching assessment may be provided both during the inflow and washout periods of ICG. The disclosed approach of continuously identifying blood vessels in tissue may therefore be performed continuously in the background.

[0159] FIG. 11 shows four plots that serve to illustrate how the body kernel can be estimated from the measured ICG signal (aka fluorescence output signal) and how the body kernel can be used to calculate a predicted output signal (i.e., the estimated signal). The top left plot shows a graph of the input signal (ICG volume vs. time) of a single bolus injection of ICG. The top right plot shows the fluorescence output signal ("measured ICG signal"). The bottom left plot shows the estimated body kernel and the true body kernel, and the bottom right plot shows the predicted output signal ("estimated signal") along with the fluorescence output signal ("measured ICG signal"). In this example, synthetic data (i.e., computer generated) is used, since the figure is for illustrative purposes only. That is, when the calculated body kernel shown in the bottom left figure is convolved with the input signal shown in the top left graph, the result is the predicted signal shown in the bottom right. For this idealized situation, the predicted and measured signals shown in the bottom right graph are in perfect agreement. The same applies to the "true" and "estimated" body kernels at the bottom left. The above principles also apply to real-life data, as will be explained in connection with other figures.

[0160] FIG. 12 shows essentially the same visual representation as FIG. 11 through four plots. However, here the input signal includes two pulses, i.e., two boluses, which alter the measured signal compared to that of FIG. 11. The body kernel remains the same as in FIG. 11 because it is a measure of how the properties of the human body, i.e., the tissues investigated, filter or alter the input signal, thereby distorting / changing the measured output signal somewhat from the input signal. Similar to the data used in connection with FIG. 11, the data in FIG. 12 is synthetically created data to illustrate the basic principle of estimating the body kernel to infer a predicted measured signal. However, the important thing is that while the body kernel is the same as in FIG. 11 and FIG. 12, without knowledge of the body kernel it is virtually impossible to infer what kind of output signal the dual bolus input signal shown in FIG. 12 would produce. However, with knowledge of the body kernel, it can be seen that the dual bolus input signal shown in the upper left of FIG. 12 can be convolved with the body kernel, thereby calculating a predicted output signal that matches the measured output signal in the lower right of FIG. 12.

[0161] FIG. 13 shows another example of how an estimated body kernel is used to infer what a measured fluorescence output signal ("ICG signal") should look like by calculating a predicted output signal (estimated signal). In this example, the input signal is known (top left plot) and the estimated body kernel is known from the estimation shown in FIG. 11 and FIG. 12. In the bottom right plot, the estimated body kernel is convolved with the input signal to obtain an estimated signal. The measured signal is further plotted to illustrate the usefulness of using a body kernel in the estimation. As can be seen, the measured signal looks very similar to the estimated signal. Thus, once the body kernel is estimated for a particular region of interest, this body kernel can be used to infer what the measured ICG signal is predicted to look like, even when using input signals different from the input signals used to estimate the body kernel. These plots were generated using synthetic data for illustrative purposes only.

[0162] Figure 14 shows the same input signal and the same body kernel as shown in Figure 13, and therefore also the estimated signal. However, in this example, the measured ICG signal (measured fluorescence output signal) is different from the predicted signal (which can be seen in the bottom right plot). Thus, in this scenario, the measured signal is different from what we would normally expect from the relevant region of interest, which is an indication that the perfusion pattern from the tissue in the region is abnormal. Conclusion: such tissue can be flagged as abnormal and further investigated by the surgeon / physician. These plots were generated using synthetic data for illustrative purposes only.

[0163] FIG. 15 shows another example of how a body kernel can be estimated for a region of interest and subsequently used to calculate a predicted output signal. The actual measured signal can then be compared to the predicted signal, thereby detecting abnormal perfusion patterns. The following references "top left" etc. refer to the figure as viewed in a landscape orientation. The top left plot shows an input signal (intensity vs. time) including a single pulse, i.e., a single bolus injection. The top center plot shows a measured fluorescent signal (intensity vs. time), and the top right plot shows an estimated body kernel calculated by deconvolving the measured signal and the input signal. The perfusion pattern is identified as having a normal perfusion pattern. Once the body kernel for the region is determined, it can be used in combination with other input signals (e.g., having a different pattern of pulses) to generate predicted output signals for the same or other tissue regions, thereby looking for tissue regions with abnormal perfusion patterns. The bottom two rows show the measured signal (middle column) from an input signal containing three pulses (left column). For the bottom two rows, the same body kernel (estimated in relation to the input signal containing a single pulse in the top row and labeled normal) is used to generate a predicted output signal. In the scenario shown in the middle row, it can be observed that the measured signal aligns and correlates with the predicted output signal, indicating a normal perfusion pattern in the tissue region being measured. However, in the scenario shown in the bottom row, the measured output signal deviates significantly from the predicted output signal, indicating an abnormal perfusion pattern in the tissue region being measured. Thus, the disclosed method can identify tissue as having either a normal or abnormal perfusion pattern. This can be used to flag / classify the tissue as either normal or abnormal. In this example, synthetic data (with noise on the measured signal) is used because this example serves for illustrative / illustrative purposes. The body kernel can be calculated continuously, and thereby also used for comparison purposes. This is illustrated in the right column of FIG. 15.In the middle row, the graphs in the right column show a body kernel that has been calculated and labeled normal compared to a body kernel calculated from the measured signal in the middle column of the middle row, and the two body kernels are observed to be similar. However, in the bottom row, it is observed that the body kernel calculated from a tissue region with an abnormal perfusion pattern ("Estimated") is very different from the body kernel labeled normal ("Predicted").

[0164] FIG. 16 shows an example of the disclosed method using real-life data, i.e., data from a human. The top left plot shows an input signal that includes a single pulse (i.e., equivalent to a single injection). Note that this input signal is estimated / approximated because the injection was performed manually using a syringe. The true input signal is not known in this case. The top right plot shows the measured fluorescence output signal (intensity vs. time) from a subject. In this example, the measured data was obtained from the subject's forearm. The bottom left plot shows the estimated body kernel for the region of interest on the forearm. The bottom right plot shows the measured signal with the estimated signal superimposed. It is observed that the measured signal correlates well with the estimated signal for a significant portion of the signal. This indicates a normal perfusion pattern in the region of interest. The tail of the distribution can be ignored, or a more accurate body kernel can be obtained, or the input signal can be varied. A more accurate body kernel can be obtained by measuring over a longer period of time.

[0165] 17 is a graph of measured fluorescence output signal versus time for three cycles, from left to right: PR, P1, and P2. A body kernel is calculated based on the fluorescence output signal for cycle PR and the known input signal (shown here). The calculated body kernel is shown in the upper right corner of FIG. 17.

[0166] In period P1, the measured fluorescence output includes many pulses resulting from the corresponding bolus injection. In the graphs on the middle row, the fluorescence output pulses from P1 are isolated, plotted together, and normalized to the same pulse length (known from the input signal). It can be seen that the pulses from P1 follow substantially the same pattern. In the graphs on the middle row, the fluorescence output pulses from P1 are isolated, plotted together, and normalized to the same pulse length, and also plotted together with the predicted output signal calculated from the known input signal convolved with the calculated body kernel. It can be seen that the predicted signal follows substantially the same pattern as the measured pulses. However, as can be seen in period P2, it can be seen that the predicted output signal (calculated from the known input signal) matches very poorly with the fluorescence output signal, and abnormal perfusion patterns (the reason being ischemia) can be easily detected.

[0167] Further details of the invention 1. A computer-implemented method for detecting (and / or identifying) one or more regions having an abnormal perfusion pattern in a tissue of a subject, for example during a medical procedure, the method comprising: - successively acquiring fluorescent images of said tissue, wherein said fluorescent images are associated with a fluorescent output signal correlated with an input signal defined by a series of boluses of at least one fluorescent imaging agent, wherein said series of boluses are administered with a predefined and / or controlled duration between successive boluses, - analyzing said fluorescence image; - obtaining, defining and / or determining a normal perfusion pattern at least in the intensity domain and / or in the time domain, and - detecting in said fluorescence image tissue regions likely to have abnormal (non-normal) perfusion patterns; The method includes:

[0168] 2. The method of claim 1, wherein the duration between subsequent boluses is between 5 seconds and 5 minutes or even 10 minutes, over a period of at least 2 or 3 minutes, or at least 5 minutes or at least 10 minutes, or at least 15 minutes, or at least 30 minutes, or at least 1 hour.

[0169] 3. A method according to any of the preceding items, comprising obtaining at least a first body kernel, which is a filter imposed by the subject's body onto the at least one fluorescent imaging agent, in a tissue region having a normal perfusion pattern.

[0170] 4. A method according to any of the preceding items, comprising obtaining at least a second body kernel, which is a filter imposed by the subject's body onto the at least one fluorescent imaging agent, in a tissue region having an abnormal perfusion pattern.

[0171] 5. The method according to any of the preceding items, wherein the at least a first body kernel and / or the at least a second body kernel are at least one transfer function between the input signal and the fluorescence output signal.

[0172] 6. The method according to any of the preceding items, comprising identifying at least one tissue region having normal perfusion.

[0173] 7. The method of any of the preceding items, wherein the input signal is defined in terms of volume of fluorescent imaging agent versus time.

[0174] 8. The method according to any of the preceding items, comprising identifying blood vessels in the fluorescent image.

[0175] 9. The method of any of the preceding items, wherein at least one tissue region having normal perfusion is identified manually, for example by a surgeon selecting the tissue region as having normal perfusion.

[0176] 10. The method according to any of the preceding items, wherein the normal perfusion pattern is determined automatically.

[0177] 11. The method of any of the preceding items comprising measuring, preferably continuously, the intensity of the fluorescence output signal, preferably the intensity versus time, in a transcutaneous manner by means other than image acquisition, for example using a photodiode and / or a light emitting diode finger clip.

[0178] 12. The method according to item 11, wherein the transcutaneously measured intensity (e.g., continuously measured intensity versus time) is used to define a normal perfusion pattern.

[0179] 13. The method of any of the preceding items, comprising obtaining at least one body kernel.

[0180] 14. The method according to any of the preceding items 13, wherein the body kernel is a filter imposed by the subject's body on the fluorescent imaging agent in tissue regions with normal perfusion patterns.

[0181] 15. The method according to any of the preceding items 13 to 14, wherein the body kernel is a filter imposed by the subject's body on the fluorescent imaging agent in tissue regions having abnormal perfusion patterns.

[0182] 16. A method according to any one of the preceding items 13 to 15, wherein the body kernel is obtained by selecting at least one region of interest (ROI) in the fluorescence image corresponding to at least one tissue region, and deconvolving the ROI with the input signal to determine a body kernel for the ROI, such that the convolution of the input signal with the body kernel corresponds to the fluorescence output signal from the tissue region.

[0183] 17. A method according to any of the preceding items, comprising the step of calculating a predicted output signal, preferably in at least one region of interest in the fluorescent image, based on the input signal.

[0184] 18. The method according to any of the preceding items 17, wherein a tissue region having an abnormal perfusion pattern is detected by comparing the fluorescence output signal of the tissue region with the predicted output signal of the tissue region.

[0185] 19. A method according to any one of items 17 to 18, wherein the predicted output signal is determined by convolving the input signal with an associated body kernel.

[0186] 20. A method according to any of the preceding items, comprising analyzing a plurality of regions of interest (ROIs) within the fluorescence image, the ROIs being distributed across different tissue regions, and one or more of these ROIs being defined as having a normal perfusion pattern if the fluorescence output signals of a plurality of these ROIs are substantially similar and / or the associated body kernels of a plurality of these ROIs are substantially similar, thereby automatically defining and / or determining a normal perfusion pattern.

[0187] 21. A method according to any of the preceding items, wherein the sequence of acquired fluorescent images is analyzed and pixels in the fluorescent images are classified as either 1) artery, 2) vein, 3) tissue, or 4) other based on the phase of the fluorescent signal at each pixel relative to the input signal.

[0188] 22. A method according to any of the preceding items, wherein one or more of the tissue regions detected as having a normal perfusion pattern based on the at least a first body kernel and / or the at least a second body kernel are classified as muscle tissue, body fat, ligament tissue, a vein, or an artery, the classification being based on the at least one first body kernel being labeled.

[0189] 23. The method of any of the preceding items, wherein one or more of the tissue regions detected as having an abnormal perfusion pattern based on the at least a first body kernel and / or the at least a second body kernel are classified as cancerous tissue, glandular tissue, tumor tissue, inflamed tissue, or ischemic tissue, wherein the classification is based on the at least one second body kernel being labeled.

[0190] 24. The method of any of the preceding items, wherein tissue regions having abnormal perfusion patterns are detected by comparing body kernels, for example, by comparing the first body kernel and the second body kernel.

[0191] 25. A method according to any of the preceding items, wherein tissue regions having abnormal perfusion patterns are detected and / or classified by comparing the at least one second body kernel with one or more labeled body kernels.

[0192] 26. The method of any of the preceding items, wherein the sequence of acquired fluorescent images is analyzed and pixels in the fluorescent images are classified as either 1) normally perfused tissue, 2) abnormally perfused tissue, or 3) other based on the phase of the fluorescent signal at each pixel relative to the input signal.

[0193] 27. A computer-implemented method for determining a perfusion-related body kernel of tissue in a subject, the body kernel being defined as a filter imposed by the body of the subject on a fluorescent imaging agent, the method comprising: - successively acquiring fluorescent images of said tissue, wherein said fluorescent images are associated with a fluorescent output signal correlated with an input signal defined by a series of boluses of said fluorescent imaging agent, wherein said series of boluses are administered with a predefined and / or controlled duration between successive boluses, - selecting at least one region of interest (ROI) in said fluorescence image corresponding to a tissue region, and - deconvolving the ROI against the input signal to determine a body kernel for the ROI, such that a convolution of the input signal with the body kernel corresponds to the fluorescence output signal from the tissue region; The method includes:

[0194] 28. The method according to any of the preceding items 27, wherein the duration between the successive boluses is selected according to the tissue type of the ROI.

[0195] 29. A method according to any of the preceding items 27 to 28, wherein the body kernel is a filter imposed by the subject's body on the fluorescent imaging agent in tissue regions having normal perfusion patterns.

[0196] 30. A method according to any of the preceding items 27 to 28, wherein the body kernel is a filter imposed by the subject's body on the fluorescent imaging agent in tissue regions having abnormal perfusion patterns.

[0197] 31. A method according to any one of items 27 to 29 above, comprising the step of calculating a predicted output signal by convolving the input signal with the body kernel.

[0198] 32. - determining the body kernel of at least one ROI according to the method according to any of items 27 to 31, - convolving said input signal with said body kernel, thereby defining a normal perfusion pattern of said ROI; 22. The method according to any one of the above items 1 to 21, further comprising:

[0199] 33. A computer-implemented method for establishing a time-domain perfusion reference for a subject, the method comprising: - continuously measuring a fluorescent output signal correlated with an input signal defined by a series of boluses of said fluorescent imaging agent, said series of boluses being administered with a predefined and / or controlled duration between successive boluses; and - defining a subject-specific time domain perfusion reference as the fluorescence output signal versus time; The method includes:

[0200] 34. The method according to any of the preceding items 33, wherein the fluorescence output signal is measured in a transcutaneous manner, for example by a light-emitting diode finger clip.

[0201] 35. The method according to any one of the preceding items 33 to 34, wherein the intensity of the fluorescent output signal is measured by a photodiode.

[0202] 36. The method according to any one of items 33 to 35 above, comprising a step of analyzing the fluorescent output signal, for example, detecting at least one peak fluorescent signal for each bolus.

[0203] 37. The method according to any of the preceding items 33 to 36, wherein the subject-specific time domain perfusion reference is defined as the time difference between a bolus injection and the corresponding peak fluorescence output signal.

[0204] 38. - Establishing a time-domain perfusion reference for the subject according to the method according to any of items 33 to 37, - correlating an analysis of said fluorescence images with said time domain perfusion reference, thereby identifying normal perfusion patterns; 22. The method according to any one of the above items 1 to 21, further comprising:

[0205] 39. Analyze at least a portion of the fluorescent images / signals to identify the following groups: - the time difference between the bolus injection and the arterial or venous or tissue fluorescent signal, - the time difference between the arterial and venous fluorescent signals, and - the time difference between the arterial or venous fluorescence signal and the tissue fluorescence signal, determining at least one time difference selected from: sequentially identifying tissues and / or vessels in the fluorescence image based on the time differences and the fluorescence output signals; The method according to any of the preceding items, comprising:

[0206] 40. The method of any of the preceding items, wherein a series of boluses of at least one fluorescent imaging agent are provided into the subject's vein during said image acquisition, thereby generating said input signal, and wherein said series of boluses are administered with a predefined duration between successive boluses.

[0207] 41. The method according to any of the preceding items, wherein the fluorescent imaging agent is indocyanine green (ICG) and each bolus of ICG corresponds to less than 0.01 mg of ICG per kg of body weight.

[0208] 42. The method according to any of the preceding items, wherein the fluorescent imaging agent is ICG and each bolus of ICG corresponds to less than 0.005 mg of ICG per kg of body weight.

[0209] 43. The method according to any of the preceding items, wherein the fluorescent imaging agent is ICG, and each bolus of ICG corresponds to less than 0.004 mg ICG per kg body weight, more preferably less than 0.003 mg ICG per kg body weight, even more preferably less than 0.002 mg ICG per kg body weight, and most preferably less than 0.001 mg ICG per kg body weight.

[0210] 44. The method of any of the preceding items, wherein the series of boluses are injected automatically by a controllable injection pump.

[0211] 45. The method of any of the preceding items, wherein the imaged tissue is part of an anatomical structure in the digestive tract, preferably selected from the oral cavity; the pharynx; the small intestine, including the duodenum, pharynx, and ileum; the stomach, including the esophagus, cardia, and pylorus; and the large intestine, including the cecum, colon, rectum, and anal canal.

[0212] 46. ​​The method of any of the preceding items, wherein the imaged tissue undergoes peristaltic movement during the medical procedure.

[0213] 47. The method according to any of the preceding items, wherein the imaged tissue is part of an internal organ of the subject, or part of the skin of the subject, or part of a wound of the subject.

[0214] 48. A method according to any of the preceding items, wherein the images are acquired during thyroid surgery and blood vessels in one or more of the parathyroid glands are identified and made visible to medical personnel involved in the surgery.

[0215] 49. The method according to any of the preceding items, wherein at least two fluorescent imaging agents are used to simultaneously generate two different fluorescent signals.

[0216] 50. The method of claim 49, wherein the at least two fluorescent imaging agents have different emittance wavelengths and the fluorescent images are acquired simultaneously from at least two different depths in the tissue.

[0217] 51. The method according to item 50, wherein the at least two different depths are separated by at least 1 cm, preferably at least 1.5 cm.

[0218] 52. The method according to any of the preceding items, wherein the fluorescent imaging agent is conjugated to a molecule that targets abnormal tissue (e.g., a tumor targeting molecule).

[0219] 53. The method of any of the preceding items, wherein the at least one fluorescent imaging agent is selected from the group of indocyanine green (ICG), infracyanine green (IfCG), brilliant blue green (BBG), and bromophenol blue (BPB), fluorescein isothiocyanate, rhodamine, phycoerythrin, phycocyanin, allophycocyanin, orthophthalaldehyde, fluorescamine, rose bengal, trypan blue, fluorogold, green fluorescent protein, flavin, methylene blue, porphysome, cyanine dyes, IRDDye800CW, CLR 1502 in combination with a targeting ligand, OTL38 in combination with a targeting ligand, or a combination thereof.

[0220] 54. A computer program having instructions that, when executed by a computing device or computing system, cause said computing device or computing system to perform the method according to any of the preceding items.

[0221] 55. A system for identifying abnormal perfusion patterns in a tissue of a subject, for example during a medical procedure, the system comprising: - successively acquiring fluorescent images of said tissue, wherein said fluorescent images are associated with a fluorescent output signal correlated with an input signal defined by a series of boluses of at least one fluorescent imaging agent, wherein said series of boluses are administered with a predefined and / or controlled duration between successive boluses, - analyzing said fluorescence image; - identifying at least one tissue region having normal perfusion in said fluorescence image based on said analysis; - defining normal perfusion patterns in the intensity domain and in the time domain, and - detecting in said fluorescence image tissue regions likely to have abnormal (non-normal) perfusion patterns; The system is configured for.

[0222] 56. The system of item 55, comprising: an injection pump controllable for holding at least one first fluorescent imaging agent, the injection pump configured to inject a series of predefined boluses of the first fluorescent imaging agent into a vein of the subject, thereby generating the input signal.

[0223] 57. A system described in any of items 55 to 56, wherein the fluorescent agent is ICG, the amount of ICG in the predefined bolus is less than 0.005 mg per kg of body weight, and the system is configured to inject the bolus at intervals between 5 seconds and 10 minutes.

[0224] 58. A system according to any one of items 55 to 57, wherein the system is configured to carry out the steps according to any one of items 1 to 53.

Claims

1. 1. A computer-implemented method for detecting one or more regions having an abnormal perfusion pattern in a tissue of a subject during a medical procedure, the method comprising: - successively acquiring fluorescent images of said tissue, wherein said fluorescent images are associated with a fluorescent output signal correlated with an input signal defined by a series of boluses of at least one fluorescent imaging agent, wherein said series of boluses are administered with a predefined and controlled duration between successive boluses, - analyzing said fluorescence image, - obtaining, defining and / or determining a normal perfusion pattern at least in the intensity domain and / or in the time domain, and - obtaining at least a first body kernel, which is a filter imposed by the body of said subject on said at least one fluorescent imaging agent, in a tissue area having a normal perfusion pattern, and / or - obtaining at least a second body kernel, which is a filter imposed by the body of said subject on said at least one fluorescent imaging agent in a tissue region having an abnormal perfusion pattern, As a result, said at least first body kernel and / or said at least second body kernel become at least one transfer function between said input signal and said fluorescence output signal; - detecting possible tissue regions having abnormal perfusion patterns in said fluorescence image based on said at least first body kernel and / or said at least second body kernel, The method includes:

2. 2. The method of claim 1, wherein the duration between successive boluses is between 5 seconds and 5 minutes, possibly up to 10 minutes, over a period of at least 2 minutes, or 3 minutes, preferably at least 5 minutes, and the input signal is defined in terms of volume of fluorescent imaging agent versus time.

3. The method of any of claims 1 to 2, comprising identifying at least one tissue region having normal perfusion and determining said normal perfusion pattern therefrom.

4. 10. The method of claim 1, comprising automatically detecting and defining normal perfusion patterns at least in the time domain.

5. 5. A method according to claim 1, further comprising identifying blood vessels, preferably continuously, in the fluorescence image, thereby excluding identified blood vessels from the assessment of normal and abnormal perfusion patterns in the tissue.

6. The method of claim 1 or 4, wherein the at least one tissue region having normal perfusion is identified manually, for example by a surgeon, by selecting the tissue region as having normal perfusion.

7. 3. The method of claim 1, further comprising continuously measuring the intensity versus time of the fluorescence output signal in a transcutaneous manner by means other than the image acquisition, wherein the transcutaneously measured intensity versus time is used to define normal perfusion patterns.

8. 3. The method of claim 1, wherein a body kernel is obtained by selecting at least one region of interest (ROI) in the fluorescence image corresponding to at least one tissue region and deconvolving the ROI with the input signal to determine a body kernel for the ROI, such that convolution of the input signal with the body kernel corresponds to the fluorescence output signal from the tissue region.

9. 9. The method of claim 1, further comprising calculating at least one predicted output signal based on said input signals and said body kernel.

10. 10. The method of claim 9, wherein tissue regions having abnormal perfusion patterns are detected by comparing the fluorescent output signal of the tissue region to the predicted output signal of the tissue region.

11. 3. The method of claim 1, further comprising classifying tissue regions detected as having abnormal perfusion patterns by utilizing a plurality of labeled body kernels.

12. 3. The method of claim 1, wherein tissue regions having abnormal perfusion patterns are detected by comparing body kernels, e.g., by comparing the first body kernel and the second body kernel.

13. 10. The method of claim 1, wherein tissue regions having abnormal perfusion patterns are detected and / or classified by comparing the at least one second body kernel with one or more labeled body kernels.

14. 3. The method of claim 1, comprising analyzing a plurality of regions of interest (ROIs) in the fluorescence image, the ROIs being distributed across different tissue regions, and one or more of the ROIs being defined as having a normal perfusion pattern if the fluorescence output signals of a plurality of the ROIs are substantially similar and / or the associated body kernels of a plurality of the ROIs are substantially similar.

15. 3. The method of claim 1, wherein the at least one fluorescent imaging agent is selected from the group of indocyanine green (ICG), infracyanine green (IfCG), brilliant blue green (BBG), and bromophenol blue (BPB), fluorescein isothiocyanate, rhodamine, phycoerythrin, phycocyanin, allophycocyanin, orthophthalaldehyde, fluorescamine, rose bengal, trypan blue, fluorogold, green fluorescent protein, flavin, methylene blue, porphysome, cyanine dyes, IRDDye800CW, CLR 1502 in combination with a targeting ligand, OTL38 in combination with a targeting ligand, or a combination thereof.

16. 3. The method of claim 1, wherein the fluorescent agent is ICG, and wherein the amount of ICG in the predefined bolus is less than 0.005 mg per kg of body weight, and the bolus is injected at an interval between 5 seconds and 5 minutes.

17. 5. The method of claim 1, wherein one or more of the tissue regions detected as having normal perfusion patterns based on the at least a first body kernel and / or the at least a second body kernel are classified as muscle tissue, body fat, ligament tissue, veins, or arteries, wherein the classification is based on the at least one first body kernel being labeled.

18. 3. The method of claim 1, wherein one or more of the tissue regions detected as having abnormal perfusion patterns based on the at least a first body kernel and / or the at least a second body kernel are classified as cancerous tissue, glandular tissue, tumor tissue, inflamed tissue, or ischemic tissue, wherein the classification is based on the at least one second body kernel being labeled.

19. 1. A computer-implemented method for determining a perfusion-related body kernel of tissue in a subject, the body kernel being defined as a filter imposed by the subject's body on a fluorescent imaging agent, the method comprising: - successively acquiring fluorescent images of said tissue, wherein said fluorescent images are associated with a fluorescent output signal correlated with an input signal defined by a series of boluses of said fluorescent imaging agent, wherein said series of boluses are administered with a predefined and / or controlled duration between successive boluses, - selecting at least one region of interest (ROI) in said fluorescence image corresponding to a tissue region, and - deconvolving the ROI against the input signal to determine a body kernel for the ROI, such that the convolution of the input signal with the body kernel corresponds to the fluorescence output signal from the tissue region, and the body kernel is a transfer function between the input signal and the fluorescence output signal.

20. 20. The method of claim 19, wherein the body kernel is a filter imposed by the subject's body on a fluorescent imaging agent in tissue regions having normal perfusion patterns, or the body kernel is a filter imposed by the subject's body on a fluorescent imaging agent in tissue regions having abnormal perfusion patterns.

21. A computer program comprising instructions which, when executed by a computing device or a computing system, cause said computing device or a computing system to carry out the method according to any of claims 1 or 2.

22. 1. A system for determining a perfusion-related body kernel of tissue in a subject, the body kernel being defined as a filter imposed on a fluorescent imaging agent by the body of the subject, the system comprising: - successively acquiring fluorescent images of said tissue, wherein said fluorescent images are associated with a fluorescent output signal correlated with an input signal defined by a series of boluses of said fluorescent imaging agent, wherein said series of boluses are administered with a predefined and / or controlled duration between successive boluses, - selecting at least one region of interest (ROI) in said fluorescence image corresponding to a tissue area, and - deconvolving the ROI with respect to the input signal to determine a body kernel of the ROI, such that the convolution of the input signal with the body kernel corresponds to the fluorescence output signal from the tissue region, and the body kernel is a transfer function between the input signal and the fluorescence output signal.

23. 23. The system of claim 22, wherein the body kernel is a filter imposed by the subject's body on a fluorescent imaging agent in tissue regions having normal perfusion patterns, or the body kernel is a filter imposed by the subject's body on a fluorescent imaging agent in tissue regions having abnormal perfusion patterns.

24. A system for identifying abnormal perfusion patterns in a tissue of a subject, for example during a medical procedure, the system comprising: a controllable injection pump for carrying at least one first fluorescent imaging agent, the injection pump configured to inject a series of predefined boluses of the first fluorescent imaging agent into a vein of the subject, thereby generating an input signal; The system comprises: - successively acquiring fluorescent images of said tissue, wherein said fluorescent images are associated with a fluorescent output signal correlated with said input signal defined by said series of boluses of said at least one fluorescent imaging agent, wherein said series of boluses are administered with a defined and controlled duration between successive boluses, - analyzing said fluorescence image, - automatically determining normal perfusion patterns, at least in the time domain; and - detecting in said fluorescence image tissue regions likely to have abnormal perfusion patterns; The system is configured for.

25. 25. The system of claim 24, wherein the at least one fluorescent imaging agent is selected from the group of indocyanine green (ICG), infracyanine green (IfCG), brilliant blue green (BBG), and bromophenol blue (BPB), fluorescein isothiocyanate, rhodamine, phycoerythrin, phycocyanin, allophycocyanin, orthophthalaldehyde, fluorescamine, rose bengal, trypan blue, fluorogold, green fluorescent protein, flavin, methylene blue, porphysome, cyanine dyes, IRDDye800CW, CLR 1502 in combination with a targeting ligand, OTL38 in combination with a targeting ligand, or a combination thereof.

26. 26. The system of claim 24, wherein the fluorescent agent is ICG, and wherein the amount of ICG in the predefined bolus is less than 0.005 mg per kg of body weight, and wherein the system is configured to inject the bolus at intervals between 5 seconds and 5 minutes.

27. The system of claim 24, configured to perform any of the steps of claims 1 to 2.

Images