Computer system operation method, program and system

Abstract

A method and system for characterizing tissue of a subject is provided. The method includes acquiring and receiving data for multiple time series of fluorescence images, identifying one or more attributes of the data related to clinical characterization of tissue, classifying the data into clusters characterizing the tissue based on the one or more attributes such that data within the same cluster are more similar to each other than data within different clusters, receiving data for a target time series of fluorescence images, associating each of a plurality of subregions in the target time series of fluorescence images with a corresponding cluster, and generating a target spatial map based on the clusters corresponding to the plurality of subregions in the target time series of fluorescence images. The generated spatial map may then be used as input for tissue diagnosis using supervised machine learning.

Description

[Technical Field]

[0001] References to Related Applications This application claims the benefit of priority to U.S. Provisional Patent Application Nos. 62 / 368,960 and 62 / 368,971, both filed July 29, 2016, entitled "METHODS AND SYSTEMS FOR CHARACTERIZING TISSUE OF A SUBJECT USTILIZING MACHINE LEARING," which applications are incorporated herein by reference in their entireties.

[0002] The present invention relates generally to the field of imaging, and more particularly to the acquisition and / or processing of medical images to characterize tissue of interest and / or to predict and display clinical data about the tissue using machine learning. [Background technology]

[0003] Blood flow is a general term used to define the movement of blood through blood vessels and can be quantified, for example, by volumetric flow (i.e., volume / time) or transit velocity (i.e., distance / time). Tissue perfusion is distinct from vascular blood flow in that tissue perfusion defines the movement of blood through blood vessels within a tissue volume. More specifically, tissue perfusion relates to the microcirculatory blood flow per unit tissue volume, which provides oxygen and nutrients to, and removes waste products from, the capillary beds of the perfused tissue. Perfusion is associated with nutrient vasculature (i.e., tiny blood vessels known as capillaries). These nutrient vasculature includes blood vessels associated with the exchange of metabolites between blood and tissues, as opposed to larger diameter non-nutrient blood vessels.

[0004] There are many situations in which healthcare professionals desire to accurately assess blood flow and / or tissue perfusion within tissue. For example, when treating wounded tissue in a patient, healthcare professionals must accurately assess blood flow and / or tissue perfusion within and around the wound site, as poor tissue perfusion adversely affects the healing process. Accurate assessment of blood flow and / or tissue perfusion increases the likelihood of successful healing of both acute (e.g., surgical) and chronic wounds. Assessment of perfusion dynamics is also important in other medical applications. Such applications may include, for example, preoperative evaluation of patients undergoing plastic reconstructive surgery (e.g., flap transfer) and assessment of cardiac tissue viability and function during cardiac surgery (e.g., coronary artery bypass surgery, partial left ventriculectomy, left ventriculoplasty via Batista procedure, etc.).

[0005] Some advances have been made in the use of imaging techniques, such as fluorescence imaging techniques, to assess blood flow and / or tissue perfusion. Typically, fluorescence imaging techniques involve the administration of a bolus of an imaging agent (e.g., indocyanine green (ICG)). After administration, the imaging agent circulates through the target tissue, e.g., the vascular or lymphatic system, and emits a fluorescent signal when illuminated with appropriate excitation light. A fluorescence imaging system acquires images of the imaging agent fluorescence emitted as the imaging agent bolus passes through the target tissue within the imaging field of view. For example, images may be acquired as the bolus enters the tissue through arterial blood vessels, passes through the tissue's capillary system, and exits the tissue through venous blood vessels. When the images are displayed as a moving image on a monitor, clinicians may observe the passage of the imaging agent through the vasculature, represented as a time-varying change in fluorescence intensity. Based on visual perception of the fluorescence intensity, clinicians may make relative, qualitative decisions about the tissue's blood flow and / or perfusion status and subsequent healing potential. However, qualitative visual assessment of such images is not always sufficient for a number of reasons, especially when visual information is ambiguous. For example, such visual assessments are limited because many parameters, such as image brightness, image contrast, and image noise, can be affected by factors other than the blood flow and / or perfusion characteristics of the tissue. Furthermore, mere visual assessments are subjective (e.g., visual assessments may vary from clinician to clinician, and a clinician's visual assessment protocol may also vary somewhat from patient to patient and / or from imaging session to imaging session) and do not support a standardized protocol for assessing blood flow and / or tissue perfusion. Finally, due to clinician memory deficiencies and inaccurate recollection of past visual assessments, it can be difficult to reliably and consistently compare and track a patient's blood flow and / or perfusion status over time and across multiple imaging sessions.

[0006] Some attempts have been made to use machine learning algorithms for tissue diagnosis. Such approaches appear to rely on visible light images of wounds and therefore classify wounds based on their superficial appearance, without considering other important factors (e.g., blood flow patterns) that may be more indicative of tissue characteristics and / or condition (e.g., tissue health). The methods and systems described herein utilize the advantages of machine learning algorithms for superior pattern recognition in the context of tissue medical imaging, including blood flow dynamics observed in various types of tissue, including wound tissue and / or lymphatic tissue. As a result, visual representations of flow and / or perfusion patterns will be more accurate and intuitive than those previously demonstrated. Summary of the Invention

[0007]

[0006] Aspects of systems and methods for characterizing tissue of a subject are described herein. In general, in one aspect, a method for characterizing tissue of a subject includes receiving a plurality of time series of fluorescence images of the subject, identifying one or more attributes of the data related to clinical characterization of the tissue, and classifying the data into a plurality of clusters based on the one or more attributes of the data such that data within the same cluster are more similar to each other than to data within different clusters, the clusters characterizing the tissue.

[0008] In certain aspects, the method may further include receiving data for a target time series of fluorescence images of the target, associating each of a plurality of subregions in the target time series of fluorescence images with a corresponding cluster, and generating a target space (cluster) map based on the clusters associated with the plurality of subregions in the target time series of fluorescence images.

[0009] The method may further include receiving a plurality of object spatial maps, receiving metadata associated with each object spatial map, storing each object spatial map and its associated clinical data in a database record, and using the database record as input to a supervised machine learning algorithm to generate a predictive model, which may be used to predict clinical data associated with a target time series of fluorescence images of the subject.

[0010] In a further aspect, a system for characterizing tissue of a target includes one or more processors and a memory having instructions retained thereon that, when executed by the one or more processors, cause the system to perform a method.

[0011] According to one aspect, a method for characterizing tissue of a subject is provided. The method includes receiving data of a plurality of time series of fluorescence images of the subject, the time series of fluorescence images being or having been acquired by an image acquisition system. The method includes identifying one or more attributes of the data related to clinical characterization of the tissue. The method also includes classifying the data into a plurality of clusters based on the one or more attributes of the data such that data within the same cluster are more similar to each other than data within different clusters, the clusters characterizing the tissue. The method may also include generating a characterization output representative of the tissue based on the classified clusters.

[0012] Optionally, the data for the multiple time series of fluorescence images of the subject comprises raw data, pre-processed data, or a combination thereof. Optionally, the pre-processed data is pre-processed by applying data compression, principal component analysis, auto-encoding, or a combination thereof. Optionally, the attributes of the data related to the clinical characterization of the tissue are identified for multiple sub-regions within the time series of fluorescence images of the subject.

[0013] Optionally, at least one of said sub-regions is a pixel or voxel in said time series of fluorescence images.Optionally, at least one of said sub-regions is a group of pixels or a group of voxels in said time series of fluorescence images of said object.

[0014] The one or more attributes of the data of the plurality of time series of fluorescence images of the object include a time-intensity curve, a coefficient, a spatial location, an onset time, a time to flush, a maximum fluorescence intensity, blood inflow, blood outflow, or a combination thereof.

[0015] Optionally, the clusters characterize the tissue based on the spatial distribution of the clusters, properties of the clusters, cluster data, or a combination thereof. Optionally, the properties of the clusters include the shape of the clusters. Optionally, each cluster is represented by a centroid. A cluster centroid may indicate which of the one or more attributes of the data in the multiple time series of fluorescence images of the subject contribute to data classification.

[0016] Classifying the data of the plurality of time series of fluorescence images of the object into the plurality of clusters comprises classifying the data into 10 or fewer clusters. Optionally, classifying the data of the plurality of time series of fluorescence images of the object into the plurality of clusters comprises classifying the data into 7 clusters.

[0017] Optionally, classifying the data of the plurality of time series of fluorescence images of the subject into the plurality of clusters comprises applying an unsupervised clustering algorithm, wherein the clustering algorithm may be a k-means algorithm.

[0018] Optionally, the method includes generating a spatial map based on the plurality of clusters, wherein the spatial map may represent differences in blood flow, perfusion patterns, or a combination thereof, among a plurality of subregions in the time series of fluorescence images.

[0019] Optionally, the method comprises training a machine learning model based on said classified data, wherein said machine learning model may be trained with a supervised machine learning algorithm.

[0020] Optionally, the method includes, after receiving data for a target time series of fluorescence images of the target, associating each of a plurality of subregions in the target time series of fluorescence images with a corresponding cluster, generating an object spatial map based on the clusters associated with the plurality of subregions in the target time series of fluorescence images, and optionally displaying the spatial map, wherein generating the object spatial map may include assigning at least one of an intensity value and a color to each subregion in the target time series of fluorescence images based on the associated cluster.

[0021] According to one aspect, a method for predicting clinical data for tissue of a subject is provided. The method includes receiving a plurality of object spatial maps generated as described above and receiving metadata associated with each object spatial map. The method includes storing each object spatial map and its associated clinical data in a record in a database. The method includes using the records in the database as inputs for a supervised machine learning algorithm to generate a predictive model that characterizes the tissue.

[0022] Optionally, the metadata includes clinical data, non-clinical data, or a combination thereof. The clinical data may include a diagnosis of a tissue abnormality, a predicted healing time for a wound, a proposed treatment plan, or a combination thereof.

[0023] According to one aspect, a method for predicting clinical data to characterize tissue of a subject is provided. The method includes receiving data of a subject's time series of fluorescence images of the subject, the subject's time series of fluorescence images being or being acquired by an image acquisition device. The method includes characterizing the subject's tissue by using the generated predictive model to predict clinical data associated with the subject's time series of fluorescence images of the subject. The method may include generating a characterization output representative of the tissue.

[0024] According to one aspect, there is provided use of a database to predict clinical data associated with a subject's time series of fluorescence images of the subject.

[0025] According to one aspect, a method for characterizing tissue of a subject is provided. The method includes receiving data of a subject time series of fluorescence images of the subject, the subject time series of fluorescence images of the subject being or having been acquired by an image acquisition device. The method also includes associating each of a plurality of subregions in the subject time series of fluorescence images with a corresponding category, the categories characterizing the tissue and defined based on one or more attributes related to a clinical characterization of the tissue such that data within the same category are more similar to each other than data within different categories. The method also includes generating a spatial map representing the tissue based on the categories associated with the plurality of subregions in the subject time series of fluorescence images. The method may also include displaying the spatial map.

[0026] According to one aspect, a method for characterizing tissue of a subject is provided. The method includes receiving data of a plurality of time series of fluorescent images, the plurality of time series of fluorescent images being or having been acquired by an image acquisition system. The method includes selecting feature vectors from the data, each feature vector characterizing one or more features of the data. The method includes generating a dataset including the feature vectors. The method includes classifying the dataset to generate a labeled dataset. The method includes generating a plurality of centroids representing the tissue characterization. The method may include displaying the tissue characterization output based on the plurality of centroids.

[0027] According to one aspect, a method for characterizing tissue of a subject is provided, the method including receiving a training data set including a plurality of feature vectors characterizing one or more features of a plurality of data entries, each data entry being at least a portion of a time-intensity curve of a training subregion in a training time series of fluorescence images, the time series of fluorescence images being or having been acquired by an image acquisition system.

[0028] According to one aspect, there is provided a system including one or more processors configured to cause the system to perform one or more of the methods. The system may include an image acquisition device configured to acquire a time series of fluorescence images.

[0029] Optionally, the system includes a display for displaying the space map image, the object space map image, or both.

[0030] Optionally, the one or more processors are further configured to overlay the space map image, the object map image, or both, onto the anatomical image of the tissue.

[0031] Optionally, the system includes a light source that provides excitation light to induce fluorescent emission from the fluorescent imaging agent in the tissue.

[0032] The system includes an image acquisition assembly that generates the time series of fluorescent images, the object time series of fluorescent images, or both based on the fluorescent emission.

[0033] According to one aspect, a system for processing a time series of images of tissue of a target is provided. The system includes a user interface. The system includes a processor configured to communicate with the user interface. The system includes a non-transitory computer-readable storage medium having instructions stored thereon that, when executed by the processor, cause the processor to perform any one of the methods. The processor may be in communication with an imaging system. The system may include the imaging system. The processor may be a component of the imaging system. The processor may be configured to control operation of the imaging system.

[0034] Optionally, the imaging system is a fluorescence imaging system and the time series of images may be a time series of fluorescence images. The fluorescence imaging system may include an illumination module configured to illuminate the tissue of the subject to stimulate fluorescence emission from a fluorescent imaging agent in the tissue of the subject. The fluorescence imaging system may include a camera assembly configured to acquire the time series of fluorescence images.

[0035] According to an aspect, a non-transitory tangible computer readable medium having computer executable program code means embodied therein for performing any one of the methods is provided.

[0036] According to one aspect, a kit is provided for processing a time series of fluorescent images of tissue of a subject, the kit including a system and a fluorescent imaging agent.

[0037] According to an aspect, there is provided a fluorescent imaging agent for use in a method or system. The fluorescent imaging agent may be used in a method or system for wound management. The wound management may include chronic wound management.

[0038] Optionally, the fluorescent imaging agent comprises indocyanine green, ICG. The fluorescent imaging agent may be ICG.

[0039] According to one aspect, a method for visualizing angiographic data is provided, comprising the steps of: a) receiving at least one time image sequence, wherein the time series of fluorescence images is or was acquired by an image acquisition system; b) dividing the at least one time image sequence into a plurality of time sequences of a spatial domain of the images of the time image sequence; c) automatically dividing the plurality of time sequences of the spatial domain into a plurality of clusters such that the sequences within a cluster are more similar to each other than sequences from different clusters; d) receiving an angiographic image sequence to be visualized; e) determining, for each pixel of the angiographic image sequence, to which cluster the time sequence of that pixel corresponds; and f) generating an image, wherein each pixel of the image is assigned a pixel value according to the cluster, the location of the pixel in the angiographic image sequence being determined to correspond to the cluster.

[0040] Optionally, step b) comprises determining, for each time sequence in the spatial domain, a feature vector representing the temporal image changes in that spatial domain.

[0041] The feature vector may be determined using a dimensionality-reducing machine learning algorithm, which may be based on principal component analysis, an autoencoding neural network, or a combination thereof.

[0042] Optionally, in step b), said time sequence in the spatial domain is a time sequence of individual pixels of said images of said time image sequence.

[0043] Optionally, step c) is performed using an unsupervised clustering algorithm, which may include a k-means algorithm.

[0044] Optionally, step c) includes automatically partitioning the plurality of time sequences in the spatial domain into a plurality of clusters using an unsupervised clustering algorithm, partitioning the plurality of time sequences in the spatial domain into a training data set and a test data set, using the training data set as input for a supervised machine learning algorithm to generate a predictive model, and testing the predictive model against the test data set, and step e) includes using the predictive model to determine which cluster the time sequence of pixels corresponds to.

[0045] Optionally, step c) comprises automatically dividing the plurality of time sequences in spatial domain into a plurality of clusters based on time dependence of intensities in the spatial domain.

[0046] Optionally, step c) comprises determining the plurality of clusters based on a cumulative classification error.

[0047] According to one aspect, a method for visualizing angiographic data is provided, comprising the steps of: a) obtaining a plurality of masks representing different time dependencies of intensity of a spatial region of an image; b) receiving an angiographic image sequence to be visualized; c) determining, for each pixel of the angiographic image sequence, with which mask the time sequence of that pixel best corresponds; and d) generating an image, wherein each pixel of the image is assigned a pixel value according to the mask, the pixel's position in the angiographic image sequence being determined to correspond to the mask.

[0048] Optionally, the plurality of masks are obtained by: e) receiving at least one time image sequence, f) dividing the at least one time image sequence into a plurality of time sequences of spatial domains of the images in the time image sequence, g) automatically dividing the plurality of time sequences of spatial domains into a plurality of clusters such that the sequences within the same cluster are more similar to each other than to sequences from different clusters, and h) generating, for each cluster, a mask representing the time dependence of the intensity of the spatial domain of the cluster, wherein each mask may represent the time dependence of the intensity of the centroid of the corresponding cluster.

[0049] According to one aspect, a method for predicting clinical data is provided, comprising the steps of: a) receiving a plurality of generated angiographic image visualizations; b) for each angiographic image visualization, storing data representing the angiographic image visualization in a record in a database; c) for each angiographic image visualization, storing associated clinical data in the corresponding record in the database; d) using the record in the database as input for a supervised machine learning algorithm to generate a predictive model; e) receiving an angiographic image sequence to be analyzed; f) visualizing the angiographic image sequence; and g) using the predictive model to predict clinical data associated with the angiographic image sequence.

[0050] According to one aspect, a method for predicting clinical data is provided, comprising the steps of: a) receiving an angiographic image sequence to be analyzed, wherein the time series of fluorescence images is or was acquired by an image acquisition system; b) visualizing the angiographic image sequence; c) using a predictive model to predict clinical data associated with the angiographic image sequence, the predictive model being obtained by: d) receiving a plurality of generated angiographic image visualizations; e) storing, for each angiographic image visualization, data representing the angiographic image visualization in a record in a database; f) storing, for each angiographic image visualization, clinical data associated with the angiographic image visualization in the corresponding record in the database; and g) using the records in the database as input for a supervised machine learning algorithm to generate the predictive model.

[0051] According to an aspect, use of a database for predicting clinical data associated with an angiographic image sequence is provided.

[0052] According to an aspect, there is provided using a predictive model to predict clinical data associated with an angiographic image sequence.

[0053] According to one aspect, a method for generating a plurality of masks is provided, the method comprising the steps of: a) receiving at least one time image sequence being or having been acquired by an image acquisition system; b) dividing the at least one time image sequence into a plurality of time sequences of spatial domains of the images in the time image sequence; c) automatically dividing the plurality of time sequences of spatial domains into a plurality of clusters such that the sequences within a cluster are more similar to each other than to sequences from different clusters; and d) generating, for each cluster, a mask representing the time dependence of the intensity of the spatial domain of the cluster. Each mask may represent the time dependence of the intensity of the centroid of the corresponding cluster.

[0054] According to one aspect, there is provided a use of a plurality of masks obtained by the method for visualizing angiographic data.

[0055] According to one aspect, a system for visualizing angiographic data is provided, comprising: a) a first receiver that receives at least one temporal image sequence being or having been acquired by an image acquisition system; b) a divider that divides the at least one temporal image sequence into a plurality of temporal sequences of spatial domains of the images of the temporal image sequence; c) a clusterer that automatically divides the plurality of temporal sequences of spatial domains into a plurality of clusters such that the sequences within a cluster are more similar to each other than sequences from different clusters; d) a second receiver that receives an angiographic image sequence to be visualized; e) a determiner that determines, for each pixel of the angiographic image sequence, with which cluster the temporal sequence of that pixel corresponds; and f) an image generator that generates an image, where each pixel of the image is assigned a pixel value according to the cluster, and where the location of the pixel in the angiographic image sequence is determined to correspond to the cluster.

[0056] According to one aspect, a system for visualizing angiographic data is provided, comprising: a) an acquisition unit that acquires a plurality of masks representing different time dependencies of intensity in a spatial region of an image; b) a receiver that receives an angiographic image sequence to be visualized; c) a determination unit that determines, for each pixel of the angiographic image sequence, with which mask the time sequence of that pixel best corresponds; and d) an image generator that generates an image, where each pixel of the image is assigned a pixel value according to the mask, and where the position of the pixel in the angiographic image sequence has been determined to correspond to the mask.

[0057] According to one aspect, there is provided a system for generating a plurality of masks, the system including: a) a receiver for receiving at least one time image sequence being or having been acquired by an image acquisition system; b) a divider for dividing the at least one time image sequence into a plurality of time sequences of spatial domains of the images in the time image sequence; c) a clusterer for automatically dividing the plurality of time sequences of spatial domains into a plurality of clusters such that the sequences within a cluster are more similar to each other than sequences from different clusters; and d) a generator for generating, for each cluster, a mask representing the time dependence of the intensity of the spatial domain of that cluster.

[0058] It will be appreciated that the method may be a computer-implemented method.

[0059] The method and system facilitate obtaining and generating visual representations of target tissues, which may be more accurate in terms of data representation and more intuitive for medical professionals to use for clinical decision-making. The method and system, and the generated visual representations of tissues, may be applicable to various types of tissues (e.g., various wounds, including chronic wounds, acute wounds, and pressure ulcers) and may provide a framework for automatically classifying tissues (e.g., wound tissues) and / or predicting clinical outcomes (e.g., healing timelines for wound tissues).

[0060] The methods, systems, and kits may be used for blood flow imaging, tissue perfusion imaging, lymphatic imaging, or a combination thereof, which may be performed during invasive, minimally invasive, or non-invasive surgery, or a combination thereof. Examples of invasive surgery that may involve blood flow and tissue perfusion include cardiac-related surgery (e.g., CABG on-pump or off-pump) and reconstructive surgery. Examples of non-invasive or minimally invasive surgery include the treatment and / or management of wounds (e.g., chronic wounds such as pressure ulcers). In this regard, for example, changes in wound size (e.g., diameter, area) over time, as well as changes in tissue perfusion within and / or around the wound, may be tracked over time by applying the methods and systems. Examples of lymphatic imaging include identification of lymph nodes, lymphatic drainage, lymphatic mapping, or a combination thereof. In some variations, such lymphatic imaging may relate to the female reproductive system (e.g., uterus, cervix, vulva).

[0061] It will be understood that any option mentioned in connection with any of the methods may be used in conjunction with other methods, systems, and kits, and vice versa. It will be understood that any of the options may be combined. It will be understood that any of the aspects may be combined. Below, embodiments and variations thereof are described. It will be understood that any of the embodiments and / or variations may be combined with the above-mentioned methods, systems, and kits. [Brief explanation of the drawings]

[0062] This patent or application document contains at least one color drawing. Copies of this patent or patent application publication with color drawing(s) will be provided by the Office upon request and payment of the necessary fee. The features will become apparent to those skilled in the art from the detailed description of illustrative embodiments, taken in conjunction with the accompanying drawings.

[0063] [Figure 1]FIG. 1 is a block diagram of an exemplary method for characterizing tissue of a subject, according to an embodiment.

[0064] [Figure 2A] Figure 2A is an illustration of a time series of images or a time series of interest, and Figure 2B is an illustration of a time-intensity curve generated for a sub-region in the time series of images or a time series of interest. [Figure 2B] Figure 2A is an illustration of a time series of images or a time series of interest, and Figure 2B is an illustration of a time-intensity curve generated for a sub-region in the time series of images or a time series of interest.

[0065] [Figure 3A] Figure 3A is an example time-intensity curve with several example parameters that approximate or otherwise characterize the time-intensity curve. Figure 3B shows a sample data set including several intensity versus time curves for individual pixels, where the intensity values ​​over time comprise a feature vector. Figure 3C shows the combination of pixel entries from various training sequences into a single matrix. Figures 3D and 3E show a schematic classification of pixel curves and the assignment of labels to each data sample. Figure 3F shows the determination of the optimal number of clusters for classification. [Figure 3B] Figure 3A is an example time-intensity curve with several example parameters that approximate or otherwise characterize the time-intensity curve. Figure 3B shows a sample data set including several intensity versus time curves for individual pixels, where the intensity values ​​over time comprise a feature vector. Figure 3C shows the combination of pixel entries from various training sequences into a single matrix. Figures 3D and 3E show a schematic classification of pixel curves and the assignment of labels to each data sample. Figure 3F shows the determination of the optimal number of clusters for classification. [Figure 3C]Figure 3A is an example time-intensity curve with several example parameters that approximate or otherwise characterize the time-intensity curve. Figure 3B shows a sample data set including several intensity versus time curves for individual pixels, where the intensity values ​​over time comprise a feature vector. Figure 3C shows the combination of pixel entries from various training sequences into a single matrix. Figures 3D and 3E show a schematic classification of pixel curves and the assignment of labels to each data sample. Figure 3F shows the determination of the optimal number of clusters for classification. [Figure 3D] Figure 3A is an example time-intensity curve with several example parameters that approximate or otherwise characterize the time-intensity curve. Figure 3B shows a sample data set including several intensity versus time curves for individual pixels, where the intensity values ​​over time comprise a feature vector. Figure 3C shows the combination of pixel entries from various training sequences into a single matrix. Figures 3D and 3E show a schematic classification of pixel curves and the assignment of labels to each data sample. Figure 3F shows the determination of the optimal number of clusters for classification. [Figure 3E] Figure 3A is an example time-intensity curve with several example parameters that approximate or otherwise characterize the time-intensity curve. Figure 3B shows a sample data set including several intensity versus time curves for individual pixels, where the intensity values ​​over time comprise a feature vector. Figure 3C shows the combination of pixel entries from various training sequences into a single matrix. Figures 3D and 3E show a schematic classification of pixel curves and the assignment of labels to each data sample. Figure 3F shows the determination of the optimal number of clusters for classification. [Figure 3F]Figure 3A is an example time-intensity curve with several example parameters that approximate or otherwise characterize the time-intensity curve. Figure 3B shows a sample data set including several intensity versus time curves for individual pixels, where the intensity values ​​over time comprise a feature vector. Figure 3C shows the combination of pixel entries from various training sequences into a single matrix. Figures 3D and 3E show a schematic classification of pixel curves and the assignment of labels to each data sample. Figure 3F shows the determination of the optimal number of clusters for classification.

[0066] [Figure 4] FIG. 1 is a block diagram of an exemplary method for characterizing tissue of a subject, according to an embodiment.

[0067] [Figure 5A] Figure 5A is a block diagram of an exemplary method for predicting clinical data. Figure 5B is an illustration of combining subject metadata / clinical data with a spatial map and using it as input to a database or registry, and further use in training a classification neural network. Figure 5C is an illustration of using the methods and systems described herein in de novo data classification for predicting clinical data and / or for diagnosis. [Figure 5B] Figure 5A is a block diagram of an exemplary method for predicting clinical data. Figure 5B is an illustration of combining subject metadata / clinical data with a spatial map and using it as input to a database or registry, and further use in training a classification neural network. Figure 5C is an illustration of using the methods and systems described herein in de novo data classification for predicting clinical data and / or for diagnosis. [Figure 5C]Figure 5A is a block diagram of an exemplary method for predicting clinical data. Figure 5B is an illustration of combining subject metadata / clinical data with a spatial map and using it as input to a database or registry, and further use in training a classification neural network. Figure 5C is an illustration of using the methods and systems described herein in de novo data classification for predicting clinical data and / or for diagnosis.

[0068] [Figure 6] FIG. 1 is a block diagram of an exemplary method for characterizing tissue of a subject and / or predicting clinical data.

[0069] [Figure 7] FIG. 1 is an illustration of an exemplary fluorescence imaging system configured to characterize tissue of a target.

[0070] [Figure 8] FIG. 1 is an illustration of an exemplary illumination module of a fluorescence imaging system configured to characterize tissue of a target.

[0071] [Figure 9] 1 is an exemplary camera module of a fluorescence imaging system configured to characterize tissue of interest.

[0072] [Figure 10] 1 shows the centroids generated for breast tissue.

[0073] [Figure 11A] 11A to 11F illustrate the application of the method and system to breast tissue in reconstructive surgery. [Figure 11B] 11A to 11F illustrate the application of the method and system to breast tissue in reconstructive surgery. [Figure 11C] 11A to 11F illustrate the application of the method and system to breast tissue in reconstructive surgery. [Figure 11D]11A to 11F illustrate the application of the method and system to breast tissue in reconstructive surgery. [Figure 11E] 11A to 11F illustrate the application of the method and system to breast tissue in reconstructive surgery. [Figure 11F] 11A to 11F illustrate the application of the method and system to breast tissue in reconstructive surgery.

[0074] [Figure 12A] Figure 12A shows the generated center of gravity for a subject's foot, and Figures 12B and 12C show the application of the methods and systems described herein to foot tissue. [Figure 12B] Figure 12A shows the generated center of gravity for a subject's foot, and Figures 12B and 12C show the application of the methods and systems described herein to foot tissue. [Figure 12C] Figure 12A shows the generated center of gravity for a subject's foot, and Figures 12B and 12C show the application of the methods and systems described herein to foot tissue.

[0075] [Figure 13] 13 and 14 illustrate an exemplary training method according to an embodiment. [Figure 14] 13 and 14 illustrate an exemplary training method according to an embodiment.

[0076] [Figure 15] 13 and 14. An exemplary use of neural networks to predict clinical data (healing duration) of a wound based on a model trained on fluorescent images of the wound / tissue as described in connection with FIGS.

[0077] [Figure 16] 1A-1C are schematic diagrams illustrating exemplary clinical applications including training and prediction of clinical data, according to various embodiments herein. DETAILED DESCRIPTION OF THE INVENTION

[0078] Implementations and embodiments of various aspects and variations of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings. Various fluorescence imaging and / or processing systems and methods are described herein. At least two variations of the imaging and / or processing systems and methods are described, although other variations of the fluorescence imaging and / or processing systems and methods may include any suitable combination of aspects of the systems and methods described herein, including combinations of all or some of the described aspects. Exemplary embodiments are described more fully below with reference to the accompanying drawings, but the exemplary embodiments may be embodied in different ways and should not be construed as limited to those described herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey exemplary implementations to those skilled in the art. Various devices, systems, methods, processors, kits, and imaging agents are described herein. At least two variations of the devices, systems, methods, processors, kits, and imaging agents are described, although other variations may include any suitable combination of aspects of the devices, systems, methods, processors, kits, and imaging agents described herein, including combinations of all or some of the described aspects.

[0079] Generally, corresponding or like reference numerals will be used throughout the drawings to refer to the same or corresponding parts wherever possible.

[0080] Spatially relative terms, such as "below," "below," "lower," "above," and the like, may be used herein to facilitate the description of the relationship of one element or feature to another element or feature as depicted in the figures. It will be understood that the spatially relative terms are intended to encompass different configurations of the device in use or operation in addition to the configuration depicted in the figures. For example, if the device in the figures were turned over, elements described as "below" or "below" other elements or features would now be positioned "above" those other elements or features. Thus, the exemplary term "below" can encompass both an above and below orientation. The device may be otherwise configured (e.g., rotated 90 degrees or in other orientations), and the spatially relative descriptors used herein may be interpreted accordingly.

[0081] The methods and systems described herein facilitate obtaining and generating visual representations of target tissues, which may be more accurate in terms of data representation and more intuitive for medical professionals to use for clinical decision-making. The methods and systems described herein, and the generated visual representations of tissues, are applicable to various types of tissues (e.g., various wounds, including chronic wounds, acute wounds, and pressure ulcers), and can provide a framework for automatically classifying tissues (e.g., wound tissue) and / or predicting clinical outcomes (e.g., healing timelines for wound tissues).

[0082] The methods and systems described herein use, in part, machine learning or deep learning. Machine learning-based methods and systems make it easier to solve problems for which there are no algorithmic solutions or for which solutions are very difficult to find. Medical diagnosis and tissue characterization based on tissue imaging are tasks particularly suited to machine learning algorithms due to the complexity of the physiological processes occurring in the human body. Machine learning can be used to discover medically relevant features and patterns in large datasets, which helps medical professionals make medical diagnoses more accurately, quickly, and consistently, regardless of the professional's experience.

[0083] The accuracy of a trained predictive model depends on the quantity and quality of its input data. As a result, most previously proposed automatic wound classification frameworks rely on large databases of wound images, where the sample inputs are images. Classical supervised machine learning methods rely on millions of labeled data samples to train data generation. This poses a problem for medical imaging data, such as fluorescence imaging data, because such data must be large and labeled to be usable in classical supervised machine learning.

[0084] Furthermore, the quality of data and the amount of useful information it contains are important factors that determine how well a classical machine learning model or algorithm can learn. Several challenges arise when using such models in the context of imaging data, such as fluorescence imaging data. These challenges include replication with missing values ​​in the dataset and feature selection relevant to model creation. Additional challenges can arise in the context of learning algorithms and optimization. For example, if a model performs poorly on a test dataset, one must be able to establish the cause of failure and adjust the model accordingly, which can be challenging for medical imaging data. The methods and systems described herein circumvent the "big data requirement" of current machine learning models by using the temporal dimension of pixel intensities, thereby enabling the construction of a training set from only a handful of patient sequences.

[0085] Additionally, by applying a clustering machine learning algorithm in various embodiments of the methods and systems of the present invention, the training data set is classified automatically, without the involvement of a clinical expert.

[0086] 1 , an example method 100 for characterizing tissue of a target may include receiving 112 data of multiple time series of fluorescence images of the target tissue, where the multiple time series of fluorescence images are acquired / imaged using an image acquisition / imaging device or system; identifying 114 one or more attributes of the data (e.g., various features of an angiographic curve including raw intensity values ​​over time, maximum intensity, inflow rate, outflow rate, perfusion onset period, duration of arterial / capillary / venous phases, as described herein) related to a clinical characterization of the tissue; classifying 116 the data into multiple clusters based on the one or more attributes of the data such that data within the same cluster are more similar to each other than data within different clusters, where the clusters characterize the tissue; and generating a tissue characterization output (based on the classified clusters). In some embodiments, the feature vector in connection with the identifying step may be for every individual pixel, or may include a combination of similar features from neighboring pixels. The identifying step may be manual (e.g., using intensity versus time values), automated (e.g., algorithmically assisted via principal component analysis as described herein), or a combination thereof. In a further aspect, the method may further include receiving data of a subject time series of fluorescence images of the subject (e.g., data acquired / derived from imaging of a patient for whom diagnosis and / or evaluation is sought), associating 120 each of a plurality of subregions in the subject time series of fluorescence images of the tissue with a corresponding cluster, and generating 122 a subject spatial map of the tissue based on the clusters associated with the plurality of subregions in the subject time series of fluorescence images. In certain aspects, the method may further include displaying 122 a the subject spatial map (e.g., an image). Throughout the specification, “spatial map” and / or “subject spatial map” are used interchangeably with “cluster map” and / or “subject cluster map.”Throughout the specification, "subject" includes human subjects and animal subjects (eg, mammals).

[0087] In certain embodiments, at least portions of the method may be performed by a computer system separate from the medical imaging system. For example, some or all of step 112 of receiving a time series of fluorescence images of tissue, step 114 of identifying one or more attributes of the data, step 116 of classifying the data into multiple clusters, and step 118 of further receiving data for the target time series of fluorescence images, step 120 of associating each of multiple subregions in the target time series of fluorescence images with a corresponding cluster, step 122 of generating the target space map, and step 122a of displaying the target space map may be performed by a computer system at an off-site location away from the medical site (e.g., where the fluorescence imaging system is located), or by a computer system located in the medical setting but not embodied in the imaging system. In these embodiments, the time series of fluorescence images and / or the target time series may be received as a result of transfer of image data from a data-carrying medium (e.g., a hard drive, cloud storage, etc.) or via network communication (e.g., a wired connection, the Internet, a wireless network based on appropriate wireless technology standards, etc.). For example, the method may include a client-server architecture, where the imaging system may include client hardware that sends image data to a computational server and loads processed data (e.g., ranking map images and intermediate outputs of various steps of the methods described herein) onto the imaging system. After the client hardware of the imaging system loads the processed data, the imaging system may further process the data and / or display the processed data according to the methods described herein.

[0088] In some embodiments, at least a portion of the method is performed by one or more processors in a computer system incorporated into a medical imaging system, such as at a clinical site. For example, some or all of steps of acquiring / receiving 112 a time series of fluorescence images of tissue and / or receiving 118 a data of a subject time series of fluorescence images, identifying 114 one or more attributes of the data related to a clinical characterization of the tissue, classifying 116 the data into multiple clusters, associating 120 each of multiple subregions in the subject time series of fluorescence images with a corresponding cluster, generating 122 a data of the subject space, and displaying 122 a data of the subject space may be performed by the computer system of the medical imaging system. In some of these embodiments, the method may further include generating 110 a data of the time series of fluorescence images prior to receiving 118 a data of the time series of fluorescence images.

[0089] As described above, conventional medical imaging techniques, such as fluorescence imaging techniques, offer limited opportunities for clinicians to accurately assess blood flow and / or tissue perfusion in a target tissue. For example, when visually assessing a fluorescence image capturing the passage of a dye bolus through tissue, the clinician's assessment of blood flow and / or tissue perfusion is confounded by parameters unrelated to the tissue's perfusion characteristics (e.g., brightness, image contrast, image noise). Additionally, a clinician's mere visual image assessment is subjective and can vary from clinician to clinician, patient to patient, and / or imaging session to imaging session.

[0090] The methods and systems described herein are useful for characterizing tissue, predicting clinical data and outcomes, and providing image data to users in a manner that enables more effective clinical decision-making to further facilitate predicting clinical outcomes. In particular, an object space map (e.g., an image) generated according to the methods described herein (e.g., 122 in FIG. 1 ) for a subject (e.g., a patient) undergoing or undergoing medical imaging succinctly illustrates relative differences between image elements, such as pixels (or voxels), or between different regions of the imaged target tissue, with respect to clinically relevant attributes. In some embodiments, the object space map (e.g., 122 in FIG. 1 ) may be a visualization of how different areas of the imaged target tissue differ in terms of their healing state, tissue properties, and / or other tissue conditions. For example, the object space map image may visualize inflammation, malignancy, disease, or other tissue abnormalities in a manner that is easily perceptible and identifiable by humans. As further described herein, by promoting standardized protocols for assessing blood flow and / or tissue perfusion and providing an avenue for comparing and tracking patient assessments over time and across multiple imaging sessions, these generated visualizations reduce ambiguity and the influence of clinician subjectivity, thus enabling clinicians to make more consistent clinical assessments and / or treatment decisions.

[0091] Although various exemplary embodiments are described herein in the context of fluorescence image time series and / or object time series, the methods are applicable to other image sources generated as time series of the dynamic behavior of imaging agents within tissue, and for other medical purposes. For example, images may be obtained from computed tomography (CT) angiography using radiopaque contrast dyes to assess blood flow and tissue perfusion. As another example, images may be obtained from positron emission tomography (PET) using fluorodeoxyglucose (FDG) or other radiotracers to assess metabolic activity and enable evaluation of pathology and / or provide information usable for evaluating pathology. As another example, images may be obtained from contrast-enhanced ultrasound imaging, which utilizes the use of gas-filled microbubble contrast media administered intravenously into the systemic circulation. Such ultrasound imaging using microbubble contrast agents enhances ultrasound backscatter or reflection, producing unique ultrasound images with increased contrast due to the large difference in echogenicity (i.e., the ability of the object to reflect ultrasound) between the gas within the microbubbles and soft tissue. For example, contrast-enhanced ultrasound can be used to obtain images of blood perfusion and blood flow within organs. Generation of tissue image time series and object time series and associated data

[0092] In one embodiment, as shown in FIG. 1 , method 100 includes generating 110 a time series of fluorescent images of tissue and / or generating 118a a target time series of fluorescent images of tissue of a target prior to receiving 112 a time series and / or receiving 118 a target time series of fluorescent images. The time series of fluorescent images and / or the target time series of fluorescent images may be generated by a fluorescent imaging technique using a fluorescent imaging agent, such as indocyanine green (ICG) dye, as the fluorescent imaging agent. When administered to a subject, ICG binds to blood proteins and circulates with the blood within the tissue. Although reference is made herein to fluorescent agents and fluorescent dyes, suitable imaging agents other than fluorescent agents and dyes may be used depending on the type of imaging technique used to generate the time series of images in embodiments where the time series of images and / or the target time series of images is not fluorescence-based.

[0093] In some embodiments, a fluorescent imaging agent (e.g., ICG) may be administered to a subject (e.g., into a vein, artery, or other tissue) as a bolus at a concentration suitable for imaging. In some embodiments in which a method is performed to assess tissue perfusion, a fluorescent imaging agent may be administered to a subject by injection into a vein or artery of the subject, where the dye bolus circulates through the vascular system and passes through the capillary system. In some embodiments in which multiple fluorescent imaging agents are used, such agents may be administered simultaneously (e.g., in a single bolus) or sequentially (e.g., in separate boluses). In some embodiments, the fluorescent imaging agent may be administered via a catheter. In some embodiments, the fluorescent imaging agent may be administered to a subject within one hour prior to taking measurements to generate a time series of fluorescent images and / or a subject time series. For example, the fluorescent imaging agent may be administered to a subject within 30 minutes prior to taking measurements. In other embodiments, the fluorescent imaging agent may be administered at least 30 seconds before taking measurements. In some embodiments, the fluorescent imaging agent may be administered simultaneously with taking measurements.

[0094] In some embodiments, the fluorescent imaging agent may be administered at various concentrations to achieve a desired circulating blood concentration. For example, in one embodiment for tissue perfusion assessment in which the fluorescent imaging agent is ICG, the fluorescent imaging agent may be administered at a concentration of about 2.5 mg / mL to achieve a circulating blood concentration of about 5 μM to about 10 μM. In some embodiments, the upper concentration limit for administration of the fluorescent imaging agent is the concentration at which the fluorescent imaging agent becomes clinically toxic in the circulation, and the lower concentration limit is the limit of the device used to acquire the time series of fluorescent images and detect the fluorescent imaging agent circulating in the blood. In some embodiments, the upper concentration limit for administration of the fluorescent imaging agent is the concentration at which the fluorescent imaging agent becomes self-deactivating. For example, the circulating concentration of ICG may range from about 2 μM to about 10 mM.

[0095] Thus, in variations, the method may include administering a fluorescent imaging agent or other imaging agent to a subject and generating or acquiring a time series of fluorescent images and / or a subject time series of fluorescent images before processing the generated data. In other variations, the method may exclude the step of administering the fluorescent imaging agent or other imaging agent to the subject. For example, the time series of fluorescent images and / or the subject time series of fluorescent images may be based on measurements of a fluorescent imaging agent, such as indocyanine green (ICG) dye, already present in the subject, and / or may be based on an autofluorescence response (e.g., native tissue autofluorescence or induced tissue autofluorescence), or may be based on measurements of a combination of autofluorescence and exogenous fluorescence resulting from the fluorescent imaging agent.

[0096] In some embodiments, suitable fluorescent imaging agents include agents that are circulatable with blood (e.g., fluorescent dyes that are circulatable with blood components such as lipoproteins and blood plasma serum) and that fluoresce when exposed to appropriate excitation light energy. Fluorescent imaging agents may include fluorescent dyes, analogs thereof, derivatives thereof, or combinations thereof. Fluorescent dyes may include non-toxic fluorescent dyes. In some embodiments, the fluorescent imaging agent optimally fluoresces in the near-infrared spectrum. In some embodiments, the fluorescent imaging agent is or includes a tricarbocyanine dye, such as indocyanine green (ICG). In other embodiments, the fluorescent imaging agent may be or include fluorescein isothiocyanate, rhodamine, phycoerythrin, phycocyanin, allophycocyanin, orthophthalaldehyde, fluorescamine, rose bengal, trypan blue, fluorogold, green fluorescent protein, flavin (e.g., riboflavin), methylene blue, porphysomes, cyanine dyes (e.g., cathepsin-activated Cy5 or Cy5.5 combined with a targeting ligand), IRDye800CW, CLR 1502 combined with a targeting ligand, OTL38 combined with a targeting ligand, methylene blue, or a combination thereof, and such fluorescent imaging agents are excited using excitation light wavelengths appropriate for each imaging agent. In some embodiments, the fluorescent imaging agent is or includes methylene blue, ICG, or a combination thereof. In some embodiments, analogs or derivatives of fluorescent imaging agents may be used. For example, fluorochrome analogs or derivatives may include fluorochromes that have been chemically altered but that maintain the ability to fluoresce when exposed to light energy of an appropriate wavelength. In embodiments where some or all of the fluorescence is derived from autofluorescence, one or more of the fluorophores that produce the autofluorescence may be an endogenous tissue fluorophore (e.g., collagen, elastin, NADH, etc.), 5-aminolevulinic acid (5-ALA), or a combination thereof.

[0097] In some embodiments, the fluorescent imaging agent may be provided as a lyophilized powder, solid, or liquid. The fluorescent imaging agent may be provided in a vial (e.g., a sterile vial), which may allow for reconstitution at the appropriate concentration by preparing a sterile fluid with a sterile syringe. Reconstitution may be performed using an appropriate carrier or diluent. For example, the fluorescent imaging agent may be reconstituted with an aqueous diluent immediately prior to administration. Any diluent or carrier that will maintain the fluorescent imaging agent in solution may be used. As an example, ICG can be reconstituted with water. In some embodiments, once the fluorescent imaging agent is reconstituted, it may be mixed with additional diluents and carriers. In some embodiments, the fluorescent imaging agent may be conjugated to other molecules (e.g., proteins, peptides, amino acids, synthetic polymers, sugars, etc.) to enhance solubility, stability, imaging properties, or a combination thereof. Additional buffers, including trisaminomethane, HCl, NaOH, phosphate buffer, and HEPES, may be added.

[0098] Those skilled in the art will appreciate that although fluorescent imaging agents are described in detail above, other imaging agents may be used in connection with the systems, methods and techniques described herein, depending on the medical imaging modality.

[0099] In some variations, fluorescent imaging agents according to one or more of the various embodiments used in combination with the methods, systems, and kits described herein may be used for blood flow imaging, tissue perfusion imaging, lymphatic imaging, bile imaging, or a combination thereof, which may be performed during invasive, minimally invasive, non-invasive, or a combination thereof. Examples of invasive procedures that may involve blood flow and tissue perfusion include cardiac-related procedures (e.g., CABG on-pump or off-pump) and reconstructive surgery. Examples of non-invasive or minimally invasive procedures include the treatment and / or management of wounds (e.g., chronic wounds such as pressure ulcers). In this regard, changes in a wound over time, such as changes in wound size (e.g., diameter, area), and changes in tissue perfusion within and / or around the wound, may be tracked over time using the methods and systems. Examples of lymphatic imaging include identification of lymph nodes, lymphatic drainage, lymphatic mapping, or a combination thereof. In some variations, such lymphatic imaging may relate to the female reproductive system (eg, uterus, cervix, vulva).

[0100] In embodiments involving cardiac or any vascular application, the imaging agent (e.g., ICG alone or in combination with other imaging agents) may be injected intravenously or may already be injected intravenously. For example, the imaging agent may be infused intravenously through a central venous line, bypass pump, and / or cardioplegia line and / or other vascular system to flush and / or perfuse the coronary vasculature, capillaries, and / or grafts. ICG may be administered to vascular grafts or other vasculature as a dilute ICG / blood / saline solution, such that the final concentration of ICG in the coronary arteries or other vasculature (depending on the application) is approximately the same as or lower than that resulting from an injection of approximately 2.5 mg (i.e., 1 ml of 2.5 mg / ml) into the central line or bypass pump. ICG may be prepared, for example, by dissolving 25 mg of solid in 10 ml of sterile aqueous solvent, which may be provided together with the ICG by the manufacturer. One milliliter of ICG solution may be mixed with 500 ml of sterile saline (e.g., by injecting 1 ml of ICG into a 500 ml saline bag). Thirty milliliters of the dilute ICG / saline solution may be added to 10 ml of the subject's blood. The blood may be obtained aseptically from a central arterial line or a bypass pump. ICG in the blood binds to plasma proteins, helping to prevent leakage from the blood vessels. Mixing of the ICG and blood may be performed using standard sterile techniques within a sterile surgical field. For each graft, 10 ml of the ICG / saline / blood mixture may be administered. Instead of administering ICG by injection through the graft wall with a needle, ICG may be administered with a syringe attached to the (open) proximal end of the graft. When the graft is implanted, the surgeon routinely attaches an adapter to the proximal end of the graft. In this case, before performing the first anastomosis, a saline-filled syringe is attached, the distal end of the graft is closed, and saline is injected into the graft, applying pressure to the graft and assessing the integrity of the conduit (e.g., for leaks and side branches).In other aspects, the methods, doses, or combinations thereof described herein in connection with cardiac imaging may be used in any vasculature and / or tissue perfusion imaging application.

[0101] Lymphatic mapping is an important part of effective surgical staging for cancers that spread through the lymphatic system (e.g., breast, gastric, and gynecological cancers). Removal of multiple lymph nodes from a particular lymphatic basin can cause significant complications, including acute or chronic lymphedema, paresthesia, and / or seroma formation. Indeed, if a sentinel lymph node is negative for metastasis, the surrounding lymph nodes are also often negative. Identification of tumor-draining lymph nodes (LNs) has become an important step for staging cancers that spread through the lymphatic system, for example, in breast cancer surgery. LN mapping involves using dyes and / or radiotracers to identify LNs for either biopsy or resection and for subsequent pathological evaluation of metastases. The goal of lymphadenectomy during surgical staging is to identify and remove LNs at high risk for local spread of cancer. Sentinel lymph node (SLN) mapping has emerged as an effective surgical strategy in the treatment of breast cancer. It is generally based on the following concept: metastasis (spread of cancer to the axillary LNs), if present, must be located in the SLN, which is defined in the art as the first LN or group of lymph nodes to which cancer cells from the primary tumor have most likely spread. If an SLN is negative for metastasis, the surrounding second and third LNs should also be negative. The main advantage of SLN mapping is that it reduces the number of subjects who undergo traditional partial or complete lymphadenectomy and therefore suffer from associated conditions such as lymphedema and lymphocysts.

[0102] The current standard of care for SLN mapping involves the injection of a tracer to identify lymphatic pathways from the primary tumor. The tracer used may be a radioisotope (e.g., technetium-99 or Tc-99m) for intraoperative localization with a gamma probe. Radioactive tracer techniques (known as scintigraphy) are limited to hospitals with access to radioisotopes, require the involvement of a nuclear physicist, and do not provide real-time visual guidance. A colored dye, isosulfan blue, has also been used, but this dye cannot be seen through skin or fatty tissue. In addition, blue stains can appear as chest tattoos that remain for several months, and subdermal injections can cause skin necrosis. Rarely, allergic reactions, including anaphylaxis, have been reported. Severe anaphylactic reactions have occurred (in approximately 2% of patients) after isosulfan blue injection. Symptoms include dyspnea, shock, angioedema, hives, and itching. Reactions are more likely in patients with a history of bronchial asthma and in patients with allergies or drug reactions to triphenylmethane dyes. Isosulfan blue is known to interfere with measurements of oxygen saturation by pulse oximetry and methemoglobin by gas analyzers. Use of isosulfan blue may cause transient or long-term blue staining (tattooing).

[0103] In contrast, fluorescence imaging according to various embodiments used in SLN visualization and mapping facilitates direct intraoperative real-time visual identification of LNs and / or afferent lymphatic channels, facilitates high-resolution optical real-time guidance through skin and adipose tissue, and facilitates visualization of blood flow, tissue perfusion, or a combination thereof.

[0104] In some embodiments, visualization and / or classification of lymph nodes during fluorescence imaging may be based on imaging with one or more imaging agents, which may further be based on visualization and / or classification with a gamma probe (e.g., technetium Tc-99m, a clear, colorless aqueous solution that is injected around the areola according to standard practice), other conventionally used colored imaging agents (isosulfan blue), and / or other evaluations, such as histology. For example, a subject's chest may be injected with two doses of approximately 1% isosulfan blue (for comparison) and two doses of an ICG solution having a concentration of approximately 2.5 mg / ml. The injection of isosulfan blue may precede the injection of ICG, or vice versa. For example, an anesthetized subject may be injected with 0.4 ml (0.2 ml at each site) of isosulfan blue in the periareolar area of ​​the chest using a TB syringe and a 30G needle. For the right breast, the subject may be injected at the 12 o'clock and 9 o'clock positions, and for the left breast, the subject may be injected at the 12 o'clock and 3 o'clock positions. The total dose of intradermal injection of isosulfan blue for each breast may be approximately 4.0 mg (0.4 ml of a 1% solution: 10 mg / ml). In another exemplary embodiment, a subject may first receive an ICG injection followed by isosulfan blue (for comparison). One 25 mg ICG vial may be reconstituted with 10 ml of sterile water for injection to produce a 2.5 mg / ml solution immediately prior to ICG administration. For example, using a TB syringe and a 30G needle, a subject may be injected with approximately 0.1 ml (0.05 ml at each site) of ICG in the periaorola area of ​​their breasts (injections may be made at the 12 and 9 o'clock positions for the right breast, and at the 12 and 3 o'clock positions for the left breast). The total dose of intradermal injection of ICG for each breast may be approximately 0.25 mg (0.1 ml of a 2.5 mg / ml solution) per breast. ICG may be injected, for example, at a rate of 5 to 10 seconds per injection. When ICG is injected intradermally, its protein-binding properties allow it to be rapidly absorbed by the lymph and transported through the ducts to the LNs. In one embodiment, ICG may be provided in the form of a sterile, lyophilized powder containing 25 mg ICG with less than 5% sodium iodide.ICG may be packaged with an aqueous solvent consisting of sterile water for injection, which is used to reconstitute the ICG. In some embodiments, the ICG dosage (mg) for sentinel lymphatic mapping of breast cancer ranges from about 0.5 mg to about 10 mg, depending on the route of administration. In some embodiments, the ICG dosage may be about 0.6 mg to about 0.75 mg, about 0.75 mg to about 5 mg, or about 5 mg to about 10 mg. The route of administration may be, for example, subdermal, intradermal (e.g., the area around the areola), under the areola, in the skin overlying the tumor, intradermal in the areola closest to the tumor, subdermal within the areola, intradermal above the tumor, around the areola over the entire breast, or a combination thereof. NIR fluorescence-positive LNs (e.g., using ICG) may be represented, for example, as black-and-white NIR fluorescence images and / or as fully or partially color (white light) images, fully or partially unsaturated white light images, enhanced color images, overlays (e.g., fluorescence with other images), composite images (e.g., fluorescence incorporated into other images) that may have various colors, levels of unsaturation, or ranges of color to highlight / visualize predetermined features of interest. Images may be further processed for further visualization and / or other analysis (e.g., quantification). Lymph nodes and lymphatic vessels may be visualized (e.g., intraoperatively, in real time) using fluorescence imaging systems and methods according to various embodiments of ICG and SLNs alone or ICG and SLNs combined with a gamma probe (Tc-99m) in accordance with the American Society of Breast Surgeons (ASBrS) practice standards for SLN biopsy of breast cancer patients. Fluorescent imaging of the LNs may begin by tracing the lymphatic channels from the injection site to the axillary LNs. Once a visual image of the LNs is identified, LN mapping and identification may be performed through the incised skin, and LN mapping may be performed until an ICG-visualized lymph node is identified. For comparison, mapping with isosulfan blue may be performed until a "blue" lymph node is identified.LNs identified with ICG alone or in combination with other imaging techniques (e.g., isosulfan blue and / or TC-99m) may be labeled for resection. Subjects may have various stages of breast cancer (e.g., IA, IB, IIA).

[0105] In some embodiments, for example, in gynecological cancers (e.g., uterine, endometrial, vulvar, and cervical malignancies), ICG may be administered intercellularly for visualization of lymph nodes, lymphatic channels, or a combination thereof. When injected intercellularly, ICG's protein-binding properties allow it to be rapidly absorbed by the lymph and transported through the ducts to the SLN. ICG may be provided for injection in the form of a sterile, lyophilized powder containing 25 mg ICG (e.g., 25 mg / vial) with less than 5% sodium iodide. The ICG may then be reconstituted with commercially available water for injection (sterile) before use. According to one embodiment, a vial containing 25 mg of ICG was reconstituted with 20 ml of water for injection to obtain a 1.25 mg / ml solution. To obtain a total dose of 5 mg of ICG per subject, subjects are injected with a total of 4 ml of this 1.25 mg / ml solution (4 x 1 ml injections). The cervix may also be injected four (4) times (for comparison purposes) with 1 ml of a 1% 10 mg / ml solution of isosulfan blue, for a total dose of 40 mg. Injections may be administered while the subject is under anesthesia in the operating room. In certain embodiments, the ICG dose (mg) for sentinel lymph node detection and / or mapping of gynecological cancers ranges from about 0.1 mg to about 5 mg, depending on the route of administration. In certain embodiments, the ICG dose may be about 0.1 mg to about 0.75 mg, about 0.75 mg to about 1.5 mg, about 1.5 mg to about 2.5 mg, or about 2.5 mg to about 5 mg. The route of administration may be, for example, cervical injection, vulvar peritumoral injection, hysteroscopic endometrial injection, or a combination thereof. To minimize spillage of isosulfan blue or ICG that may interfere with the mapping procedure when the LNs are excised, mapping may be performed on one hemipelvic site, with both isosulfan blue and ICG mapping performed before the LNs are excised. LN mapping for clinical stage I endometrial cancer may be performed according to the NCCN Guidelines for Uterine Neoplasia, SLN Algorithm for Surgical Staging of Endometrial Cancer, and SLN mapping for clinical stage I cervical cancer may be performed according to the NCCN Guidelines for Cervical Neoplasia, Surgery / SLN Mapping Algorithm for Early Stage Cervical Cancer.Therefore, identification of LNs may be based on ICG fluorescence imaging alone, or in combination with colorimetric dyes (isosulfan blue) and / or radiotracers, or by co-administration of colorimetric dyes (isosulfan blue) and / or radiotracers.

[0106] Lymph node visualization may be qualitative and / or quantitative. Such visualization may include, for example, lymph node detection, detection rate, and anatomical distribution of lymph nodes. Lymph node visualization according to various embodiments may be used alone or in combination with other variables (e.g., vital signs, height, weight, demographics, surgical predictors, relevant medical history and underlying conditions, tissue visualization and / or evaluation, Tc-99m visualization and / or evaluation, concomitant therapies). Follow-up visits may occur on the day of discharge and at a later date (e.g., one month).

[0107] Lymph contains high levels of proteins, and ICG can therefore bind to endogenous proteins upon entering the lymphatic system. Fluorescence imaging (e.g., ICG imaging) for lymphatic mapping, when used in accordance with the methods and systems described herein, offers the following exemplary advantages: a high signal-to-background ratio (or tumor-to-background ratio) due to the lack of significant autofluorescence in NIR; real-time visualization features of lymphatic mapping; tissue determination (i.e., structural visualization); rapid excretion and disappearance after entering the vasculature; and avoidance of non-ionizing radiation. Furthermore, NIR imaging has superior tissue penetration (approximately 5 to 10 mm of tissue) compared to visible light (1 to 3 mm of tissue). For example, the use of ICG also facilitates visualization through the peritoneum overlying para-aortic lymph nodes. While tissue fluorescence is observable with NIR light for extended periods, it is not visible with visible light, and therefore does not affect the pathological evaluation or treatment of LNs. Furthermore, fluorescence is easier to detect during surgery than blue lymph node staining (isosulfan blue). In other aspects, the methods, doses, or combinations thereof described herein in connection with lymphatic imaging may be used in any vasculature and / or tissue perfusion imaging application.

[0108] Tissue perfusion is related to the microcirculatory blood flow per unit tissue volume, which provides oxygen and nutrients to the capillary bed of the perfused tissue and removes waste products from it. Tissue perfusion is a phenomenon related to but distinct from blood flow within blood vessels. Quantifying blood flow through blood vessels may be expressed in terms that define flow (i.e., volume / time) or in terms that define speed (i.e., distance / time). Tissue blood perfusion defines the movement of blood through capillaries, such as arterioles, tubules, and venules, within a tissue volume. Quantifying tissue blood perfusion is expressed in terms of blood flow through tissue, i.e., volume / time / tissue volume (or tissue mass). Perfusion is related to nutrient vessels (e.g., capillaries), which include vessels associated with the exchange of metabolites between blood and tissue, rather than larger-diameter non-nutrient vessels. In some embodiments, quantifying the target tissue may involve calculating or determining a parameter or quantity related to the target tissue, such as rate, size, volume, time, distance / time, and / or volume / time, and / or change, which may relate to any one or more of the aforementioned parameters or quantities. However, compared to blood movement through larger diameter vessels, blood movement through individual capillaries can be highly irregular. This is primarily due to vasomotion, where spontaneous oscillations in vascular pulses manifest as pulsations in the movement of red blood cells.

[0109] In some embodiments, by interstitial administration, a fluorescent imaging agent, such as ICG, may be used for fluorescent imaging of lymph nodes and delineation of lymphatic vessels in the cervix and uterus during lymphatic mapping in patients with solid tumors. For solid tumors, this procedure is a component of intraoperative management. Intraoperative fluorescent imaging can be performed during lymphatic mapping, for example, by using a fluorescent agent, such as ICG, in a PINPOINT® fluorescent imaging system (available from Novadaq Technologies Inc.).

[0110] In some embodiments, fluorescent imaging agents, such as ICG, administered intradermally may be used for fluorescent imaging of lymph nodes and delineation of lymphatic vessels in the chest during lymphatic mapping in patients with solid tumors. For solid tumors, such procedures are a component of intraoperative management. For example, intraoperative fluorescent imaging can be performed during lymphatic mapping using a fluorescent agent, such as ICG, with the SPY-PHI portable handheld imaging system (available from Novadaq Technologies Inc.).

[0111] In some embodiments, fluorescent imaging agents, such as ICG, administered intradermally (including subcutaneously), may be used for fluorescent imaging of lymph nodes and delineation of lymphatic vessels in skin tissue during lymphatic mapping in patients with solid tumors, for which this procedure is a component of intraoperative management (e.g., melanoma). Intraoperative fluorescent imaging can be performed during lymphatic mapping, for example, by using a fluorescent imaging agent, such as ICG, with the SPY® Elite and SPY-PHI portable handheld imaging systems (available from Novadaq Technologies Inc.).

[0112] In some embodiments, by interstitial administration, a fluorescent imaging agent, such as ICG, may be used for fluorescent imaging of lymph nodes and delineation of lymphatic vessels during lymphangiography of primary and secondary lymphedema of the extremities. For example, intraoperative fluorescent imaging can be performed during lymphatic mapping by using a fluorescent imaging agent, such as ICG, with the SPY® Elite and SPY-PHI portable handheld imaging systems (available from Novadaq Technologies Inc.).

[0113] In some embodiments, via intravascular administration, fluorescent imaging agents, such as ICG, may be used for fluorescent imaging of blood flow and tissue perfusion during vascular and / or organ transplant surgery. Intraoperative fluorescent imaging (e.g., angiography) can be performed using fluorescent imaging agents, such as ICG, with SPY® Elite, LUNA, and SPY-PHI fluorescent imaging systems (available from Novadaq Technologies Inc.).

[0114] In some embodiments, via intravascular administration, fluorescent imaging agents, such as ICG, may be used for fluorescent imaging of blood flow and tissue perfusion during vascular, gastrointestinal, organ transplant, plastic, micro, and / or reconstructive surgery, including general minimally invasive surgery. Intraoperative fluorescent imaging (e.g., angiography) can be performed using fluorescent imaging agents, such as ICG, with the SPY® Elite, LUNA, SPY-PHI, and PINPOINT® fluorescent imaging systems (available from Novadaq Technologies Inc.).

[0115] In some embodiments, via intravascular administration, a fluorescent imaging agent, such as ICG, may be used during intraoperative cholangiography for fluorescent imaging of the bile duct. Such imaging can be performed using a fluorescent imaging agent, such as ICG, in a PINPOINT® fluorescent imaging system (available from Novadaq Technologies Inc.).

[0116] One or more embodiments are directed to fluorescent imaging agents for use in, for example, the imaging systems and methods described herein. In one or more embodiments, the use may include blood flow imaging, tissue perfusion imaging, lymphatic imaging, or a combination thereof, which may occur during invasive surgery, minimally invasive surgery, non-invasive surgery, or a combination thereof. The fluorescent agent may be included in a kit described herein.

[0117] In one or more embodiments, the invasive procedure may include cardiac or reconstructive surgery, which may include coronary artery bypass graft (CABG) surgery, which may be on-pump and / or off-pump.

[0118] In one or more embodiments, the minimally invasive or non-invasive surgery may include wound repair surgery.

[0119] In one or more embodiments, the lymphatic imaging may include identification of lymph nodes, lymphatic drainage, lymphatic mapping, or a combination thereof. The lymphatic imaging may relate to the female reproductive system.

[0120] The methods and processes described herein may be performed by a computer, processor, manager, or controller, or by code or instructions executed in hardware or other circuitry. Having described in detail the algorithms underlying the methods (or the operation of a computer, processor, or controller), the code or instructions for implementing the operations of the method embodiments may transform a computer, processor, or controller into a dedicated processor for performing the methods described herein.

[0121] Other embodiments may also include a computer-readable medium, such as a non-transitory computer-readable medium, for carrying the code or instructions described above. The computer-readable medium may be a volatile or non-volatile memory or other storage device, which may be removably or permanently coupled to a computer, processor, or controller that executes the code or instructions to perform the method embodiments described herein.

[0122] In some variations, the time series of fluorescence images and / or the subject time series of fluorescence images include multiple separate image frames (e.g., fluorescence image frames) or data representing the separate frames, which are ordered by the time of acquisition. For example, the time series of fluorescence images and / or the subject time series of fluorescence images can be acquired using a fluorescence imaging system in which the subject receives an intravenous injection of ICG immediately prior to the procedure, the tissue is illuminated with light at the excitation wavelength of the ICG, and the resulting fluorescence emission from the dye as it passes through the target tissue is imaged. The fluorescence images are then stored as a series of separate frames or data representing the separate frames (e.g., a compressed movie), which are ordered by the time of their acquisition.

[0123] In certain embodiments, separate image frames in a time series are spatially aligned or registered. For example, a typical time series of fluorescence images and / or a subject time series of fluorescence images may be recorded for two to three minutes, during which some subject movement is unavoidable. As a result, the same anatomical feature may appear in different positions in image frames acquired at different times during the image time series acquisition period. Such misalignment may introduce errors in subsequent analysis, in which the fluorescence level of each pixel or group of pixels is tracked over time. To help reduce errors, the generated image frames may be spatially aligned (registered) with each other. In certain embodiments, image registration or alignment refers to the process of determining a spatial transformation that maps a point from one image to a homologous point in a second image.

[0124] Image registration may be an iterative process. For example, according to an exemplary embodiment, image registration may use one or more of the following set of components: two input images, a transformation, a metric, an interpolator, and an optimizer. The transformation maps the fixed image space to the dynamic image space. The optimizer is required to search the parameter space. An Insight Segmentation and Registration Toolkit (ITK) (http: / / itk.org / )-based implementation of the transformation may be used in searching for the optimum value of the metric. The metric compares how well two images match each other. Finally, the interpolator evaluates the intensity of the dynamic image at non-grid locations. This procedure is performed for all frames included in the analysis to align the entire time series of fluorescence images. The components loop through the range of input sequence frames, subtract a background image for baseline correction, apply a noise reduction filter, and then register consecutive sets of images.

[0125] In some variations, the data for the multiple time series of fluorescence images and / or the target time series of fluorescence images comprises image data, and may comprise raw data, unprocessed data, or a combination thereof. In some variations, the time series of fluorescence images and / or the target time series of fluorescence images is pre-processed, for example, to extract selected data, to calculate baseline intensities, to perform an image enhancement process, or a combination thereof.

[0126] For example, extracting selected data includes cropping to locate and remove specific data from image time series data. For example, during a fluorescent imaging procedure on a subject, an operator may begin recording a fluorescent image time series and / or a target time series of fluorescent images long before the fluorescent imaging agent reaches the target tissue. As a result, the fluorescent image time series may have a significant number of "dark" frames early on, which may add unnecessary computation time for frames that do not contain meaningful data. To alleviate this problem, cropping can be used to remove these "dark" frames from the beginning of the fluorescent image time series. Additionally, when a fluorescent imaging agent (e.g., ICG) is administered to a subject, the fluorescent signal from the imaging agent as it passes through the target tissue typically undergoes a series of phases: a rapid increase in fluorescence intensity as the imaging agent enters the tissue through the arterial blood vessels, followed by a period of stable fluorescence as the imaging agent passes through the capillary system, followed by a slow decrease in fluorescence intensity due to venous outflow of the imaging agent, followed by a period of residual fluorescence as the imaging agent retained on the inner surface of the vasculature is released into the bloodstream. This final "residual" phase may last for several minutes. It is not a direct indication of blood flow and does not provide meaningful perfusion information, so cropping can be used to find residual phases and remove them from subsequent analysis steps.

[0127] In certain embodiments, preprocessing may include calculating a baseline intensity. For example, when a time series of fluorescent images and / or a target time series of fluorescent images is being generated by a fluorescent imaging system, various external factors, such as camera noise, thermal noise, and / or the presence of residual fluorescent dye from previous administration, may contribute to the fluorescence of the recorded series. To minimize the impact of such factors on the analysis, a baseline intensity may be calculated for each series, and the analysis of the data may be adjusted accordingly.

[0128] In some aspects, preprocessing may include an image quality verification process. For example, such a process may include an onset brightness test in embodiments where acquisition of a time series of fluorescence images begins too late, such that the imaging agent has already begun passing through the target tissue by the time the first frame is acquired. In this scenario, the time series of fluorescence images cannot be reliably analyzed or processed because information about the onset of perfusion is lost. As a result, such a series may be rejected.

[0129] In some embodiments, the image quality verification process may include an intensity variation test. For example, such a test may be used in situations where the fluorescence imaging system is suddenly moved during image acquisition, a foreign object appears in the field of view, or light from an external source illuminates the scene while the sequence is being acquired. All of these events can significantly distort the results of subsequent analysis. Therefore, a time series of fluorescence images that undergo such a test will fail the verification procedure (be identified as unsuitable for further processing). According to an exemplary embodiment, the intensity variation test involves calculating the difference between the average intensities of adjacent frames in the time series of fluorescence images and comparing it to a selected intensity difference threshold. To pass verification, the difference in intensity between all consecutive frames must be within the limits specified by the selected intensity difference threshold.

[0130] In certain aspects, the image quality verification process may include an intensity peak location test to check whether the acquisition of a time series of fluorescence images was stopped prematurely. For example, the intensity peak location test ensures that a sufficient number of frames were acquired to cover all phases of dye bolus passage through the tissue. According to an exemplary embodiment, the fluorescence intensity peak location test involves finding the frame with the maximum mean fluorescence intensity and verifying that it is not the last frame of the time series of fluorescence images. If this condition is not met, it is a strong indication that the fluorescence intensity value has not yet reached its maximum and therefore such a time series of fluorescence images is not suitable for further analysis.

[0131] In certain embodiments, the image quality verification process may further include a maximum fluorescence intensity test, the purpose of which is to filter time series of fluorescence images where the images are either too dark (most of the pixels below a predetermined threshold) or oversaturated (most of the pixels above a predetermined saturation threshold).

[0132] Curvature of the tissue surface, excessive movement during image acquisition, dark or oversaturated images, foreign objects in the imaging field, and extraneous light or shading can affect the quality of the fluorescence image time series and / or the subject time series of fluorescence images, and thus the subsequent processing of such image data. To mitigate these issues, well-structured imaging protocols and fluorescence imaging systems designed to minimize such issues may be used.

[0133] In some embodiments, the data may be preprocessed, for example, by applying data compression, principal component analysis, autoencoding, or a combination of these approaches or other known preprocessing methods. Preprocessing may vary depending on the type of data and / or imaging application. In some embodiments, preprocessing may include calculating coefficients, spatial location, onset time, time to flush, maximum fluorescence intensity, blood inflow, blood outflow, or a combination thereof. Attributes of data related to the clinical characterization of tissues

[0134] As shown in FIG. 1 , the described method includes identifying one or more attributes of data (e.g., fluorescence imaging-derived data) related to clinical characterization of tissue. In one embodiment, the one or more attributes of data (e.g., 114 in FIG. 1 ) of the multiple time series of fluorescence images include multiple time-intensity curves for multiple sub-regions or calculation regions in the time series of fluorescence images. Each time-intensity curve corresponds to a corresponding sub-region or calculation region in the fluorescence images. In some variations, at least one of the sub-regions or calculation regions may be an image element, such as a single pixel or group of pixels, a voxel or group of voxels, or other area or volume spatially defined in the time series of fluorescence images. The size of each sub-region or calculation region may be the same as the size of all other sub-regions or calculation regions, or may be different from the size of all or some of the other sub-regions or calculation regions. In some variations, the boundaries and / or distribution of one or more sub-regions or computational regions may be given (e.g., one computational region for each pixel or voxel, or one computational region for a group of 2x2 pixels or a block of 2x2x2 voxels), while in other embodiments, the boundaries and / or distribution of one or more sub-regions or computational regions may be determined by a user, such as a medical professional.

[0135] A separate time-intensity curve may be generated for each of some or all of the multiple subregions or computational regions. As shown schematically in FIGS. 2A and 2B, a given time-intensity curve 212 (FIG. 2B) corresponding to a particular subregion or computational region 210 (FIG. 2A) describes the intensity of the fluorescence signal observed across a time series (i.e., over time) of a fluorescence image of tissue in that subregion or computational region. In some embodiments, the time-intensity curve may describe all phases (e.g., arterial, capillary, venous, and residual in an angiography application), a subset of a phase or combination of phases, a subset of all phases, or a derivative thereof (e.g., including a determination based on first and second time derivatives of the fluorescence intensity change, pixel by pixel or voxel by voxel). A processor implemented in the fluorescence imaging system that generates the fluorescence image of the tissue, or a processor separate from the fluorescence imaging system that generates the fluorescence image, may generate all or some of the time-intensity curves.

[0136] In one embodiment, as shown in Figure 2B, the time-intensity curve 212 includes a region of increasing intensity, a region of peak intensity, a plateau region, a region of decreasing intensity, or a combination thereof. In the context of fluorescence imaging (e.g., fluorescence angiography), as shown in Figure 3, the time-intensity curve 312 may represent the passage of a bolus of a fluorescent imaging agent (e.g., a fluorescent dye) through tissue as a series of phases. The series of phases may be an arterial phase, a capillary phase, a venous phase, a residual phase, or a combination thereof.

[0137] The shape of the time-intensity curve (or a portion thereof), the area under the time-intensity curve, or a combination thereof may be indicative of the distribution of the fluorescent imaging agent within the tissue of interest, the blood flow within the tissue, or a combination thereof. In some applications, the distribution of the imaging agent within the tissue of interest is indicative of a tissue characteristic, a tissue state (e.g., inflammation, malignancy, abnormality, disease), or a combination thereof.

[0138] In certain embodiments, the one or more attributes of the data of the multiple time series of fluorescence images (e.g., 114 in FIG. 1 ) may include a time-intensity curve, coefficients, spatial location, onset time, time to flush, maximum fluorescence intensity, blood inflow, blood outflow, or combinations thereof, as described herein, for multiple sub-regions or calculation regions in the time series of fluorescence images. In further embodiments, the one or more attributes of the data of the multiple time series of fluorescence images may include the contributions (e.g., statistical properties) of neighboring pixels, intensity gradients in space and time, or combinations thereof.

[0139] In some variations, the multiple time series of fluorescence images (e.g., 112) may be derived from a healthy subject, a population of healthy subjects, a healthy tissue region within the target tissue of the subject, a healthy tissue region outside the target tissue of the subject, a combination of two or more of such surrogates, or in some variations, a further combination of such surrogates that takes into account the context in the time series of fluorescence images. Additionally, the time series of fluorescence images (e.g., 112) may be specific to a particular modality (e.g., a systemic condition such as diabetes), condition, clinical context, or combination of these factors in which the tissue (e.g., wound tissue) is being evaluated. Classifying Data into Clusters

[0140] 1 , the method includes classifying 116 the data into a plurality of clusters based on one or more attributes of the data such that data within the same cluster are more similar to each other than data within different clusters, where the clusters characterize an organization. The number of clusters into which the data is classified may be optimized and determined for a particular application. In one aspect, classifying the data into a plurality of clusters includes classifying the data into a selected number of clusters (e.g., 10 or fewer clusters).

[0141] In some embodiments, when multiple time series of fluorescent images are received, feature vectors may be selected from the data, each feature vector characterizing one or more features of the data, and a dataset including the feature vectors may be generated. For example, for a selected imaging modality (or multiple modalities) (e.g., chronic wound, acute wound, pressure ulcer), for a selected anatomical feature (e.g., foot, heel, shin, chest, etc.), or for a combination thereof, a user may select multiple representative field sequences (e.g., approximately 3-5) covering a wide range of tissue conditions (e.g., wounds) and their different stages. For example, in a time series of fluorescent images of tissue, all field sequences can be treated as three-dimensional data (two spatial dimensions and one temporal dimension), so the time dimension can be utilized, and the intensity versus time curves of individual pixels (time-intensity curves) can be used as feature vectors to generate the dataset. This approach facilitates overcoming the "big data" requirements imposed by conventional techniques using machine learning algorithms. Fluorescence imaging systems, such as the SPY® fluorescence imaging system, SPY-PHI® fluorescence imaging system, PINPOINT® fluorescence imaging system, and LUNA® fluorescence imaging system, all available from Novadaq Technologies Inc., record a sequence of frames, each of which can generate a large number of pixels. As a result, every individual pixel (or computational domain, as described herein) represents a single sample of the dataset, while its intensity values ​​over time constitute a feature vector. Thus, the dataset includes a collection of intensity versus time curves, as shown in FIG. 3B. In some embodiments, the dataset may be generated by combining pixel entries from different training sequences into a single matrix, as shown in FIG. 3C.

[0142] In interpreting and processing data derived from fluorescence imaging time series, for example, time-intensity curves are selected as attributes related to the clinical characterization of tissue, and one of the challenges in such interpretation and processing is finding an accurate and uniform way to classify time-intensity curves. It is known in the art that the dynamics of blood flow and / or perfusion through tissue are directly related to its viability and healability. As a result, it is desirable to establish what represents meaningful differences in the many observed intensity versus time curves and what can be ignored as noise. The methods and systems described herein remove the "human factor," thus making it easier to identify blood flow and / or perfusion patterns that appear to be highly correlated with the health of the imaged tissue.

[0143] In some embodiments, clusters are classified using an algorithm that facilitates finding natural groupings in the data, such that items within the same cluster are more similar to each other than items from different clusters. Classification involves, for example, dividing a dataset into several distinct categories of pixel curves (e.g., FIG. 3D ), followed by assigning each data sample its appropriate label. To achieve this, known unsupervised clustering (partitioning) algorithms, such as k-means++, may be used. In further embodiments, other clustering algorithms, such as Density-Based Spatial Clustering of Applications with Noise (DBSCAN) or hierarchical clustering (agglomerative or divisive), may be used instead of k-means. In some embodiments, each cluster may be represented by a centroid (e.g., FIG. 3E ). Two-dimensional scatter plots do not show curves; rather, they serve only as a visualization aid. Depending on the application, one or more such clustering techniques may be used. For example, a hierarchical clustering method may first be used to separate subjects into distinct attributes, and then density-based clustering may be applied to the perfusion data derived from such subjects.

[0144] One of the challenges of unsupervised learning is that it does not use labels in the dataset, which differs from supervised learning approaches, which allow for the evaluation of model performance. Therefore, we can compare the performance of different k-means clustering methods by using intrinsic measures to quantify the quality of the clustering. Graphical tools (e.g., the so-called elbow method) can be used to estimate the optimal number of clusters k for a given task. As k increases, distortion will decrease because samples will be closer to their assigned centroids. The idea behind the elbow method is to identify the value of k where distortion begins to increase most rapidly, as revealed by plotting distortion against different values ​​of k. This is shown, for example, in Figure 3F, where the cumulative classification error (distortion) is calculated for a range of cluster numbers from 1 to 10 and plotted as a graph for easy visualization to determine the optimal number of classes in the curve. The graph in Figure 3F shows that the distortion curve flattens out after reaching 5-6 clusters. Therefore, in this particular example data context, it can be inferred that all of the pixel-based intensity versus time curves are grouped into roughly seven different categories, with minimal impact on overall accuracy.

[0145] After determining the optimal number of clusters, the algorithm may be applied to the training set again using this number as an input parameter. The output of the algorithm is a trained model that can predict the label (i.e., cluster ID) of any feature vector with the same attributes as the feature vector used in the training dataset. The model may also be polled to output the centroids used for labeling. After successfully generating the trained model, it may be used to label pixel curves in a new sequence, thus facilitating the generation of a false color space map (clusters) that represents the curve distribution in the imaged tissue. Deriving clinically relevant information about the tissue from the classified clusters

[0146] In some embodiments, the clusters themselves may provide useful information about the tissue. For example, the clusters may characterize the tissue based on their spatial distribution, their properties, their data, or a combination thereof. In some embodiments, the properties of the clusters include their shape.

[0147] In certain embodiments, the classified clusters may be converted into a spatial map 116a (FIG. 1) that shows the distribution of the clusters, thereby visualizing any relative differences between subregions or calculation regions in the time series of fluorescence images, and representing differences in blood flow, perfusion patterns, or a combination thereof, between multiple subregions in the time series of fluorescence images. Thus, the classified clusters may indicate relative differences between different portions of the imaged tissue with respect to one or more specified attributes of the data related to the clinical characterization of the tissue. This may facilitate highlighting different characteristics (e.g., physiological properties) of the tissue in an objective and easily understandable manner. As described in more detail below, the resulting classified clusters may facilitate more effective and unified clinical evaluation and decision-making.

[0148] In some embodiments, the cluster centroid values ​​may be mapped to grayscale or color scale values, for example, to 8-bit grayscale display values ​​(e.g., 0 to 255), thereby enabling a grayscale image representation of the centroids. In some embodiments, a color scheme may be applied to the grayscale image representation, with different grayscale value ranges represented in appropriately contrasting colors (e.g., false or pseudocolor) to optimize visual reception. Other scales may additionally or alternatively be applied to convert the centroids to pixel values ​​in the spatial map image 116a, where differences in pixel values ​​reflect relative differences between different regions of the imaged tissue from which the data is derived.

[0149] In further embodiments, the classified cluster data may be compiled into other forms, including graphical and mathematical characterization, calculation of the proportion of curves with a particular cluster label, statistical calculations for spatial maps (cluster maps) built, including, for example, histograms, standard deviations for the labels, or combinations thereof. In some embodiments, the centroids themselves may represent particular clinical conditions (e.g., venous occlusion) and may be used by medical personnel to diagnose the clinical condition of a particular subject that correlates with a particular centroid and its data. Viewing the spatial map of clusters and other steps

[0150] In certain embodiments, as shown in FIG. 1 , the method may further include displaying the spatial map image 116b on a display. For example, the spatial map image may be displayed within a user interface on a video monitor of the fluorescence imaging system or on another suitable display. The spatial map image may be displayed alone or in combination with other images (e.g., overlaid with or on an anatomical image) or other data. Such other data may, for example, relate to the general or local condition of the subject or subject population, providing a particular clinical context for the subject and / or subject population. Such conditions may include comorbid conditions, such as hypertension, dyslipidemia, diabetes, chronic obstructive pulmonary disease, coronary artery disease, chronic kidney disease, or a combination thereof. In certain embodiments, the spatial map image may be displayed with other data or metadata about the subject, subject population, tissue, or a combination thereof, as further described below.

[0151] In certain aspects, the method may further include correlating the clusters and / or spatial map with a risk assessment of a clinically relevant (e.g., tissue perfusion-related) condition. Such assessment may be performed pre-interventionally, during treatment / procedure, and post-intervention. The method may also include defining a consultation based on the clusters to identify and characterize the subject's clinically relevant (e.g., tissue perfusion-related) condition pre-interventionally, during treatment / procedure, and post-intervention. In other aspects, the method may exclude the correlation step and the consultation step. Use of clusters to characterize target time series of fluorescence images or other data of tissues of interest

[0152] In some embodiments, the method may further include training a machine learning model based on the classified data. In some embodiments, the machine learning model may be trained with a supervised machine learning algorithm. As shown in FIG. 1 , following clustering, the method may further include receiving 118 data for the target time series of fluorescence images of the target, associating 120 each of a plurality of subregions in the target time series of fluorescence images with a corresponding cluster, and generating 122 a target space map based on the clusters associated with the plurality of subregions in the target time series of fluorescence images.

[0153] Generating the object spatial map may be performed in a manner similar to that described above in connection with generating spatial map 116a. For example, generating the object spatial map may include assigning at least one of an intensity value and a color to each subregion in the object time series of fluorescence images based on the associated cluster.

[0154] Unlike raw data of a time series of fluorescent images with a wide continuous range of intensity / color values, object space maps (e.g., 122 in Figure 1 and 422 in Figure 4) are based on a highly structured, discrete set of parameters. As a result, clinically relevant flow and / or perfusion patterns can be more easily detected by trained neural networks, which are often used in image classification tasks. The flow and / or perfusion patterns revealed by object space maps can predict various clinical conditions that would otherwise be imperceptible to human observers. By training specially designed neural networks on a large number of labeled object space maps as input, a predictive machine learning framework can be established that can automatically identify clinically relevant conditions in imaged tissues. For example, various learning models, including information-based learning (decision trees and their ensembles), similarity-based learning (k-nearest neighbors), probability-based learning (Bayesian networks), error-based learning (logistic regression, support vector machines, artificial neural networks), or combinations thereof, can be used for predictive analysis of tissues (e.g., wound healing duration predictors).

[0155] In one aspect, for example, as shown in Figure 4, an exemplary method 400 may be used to predict clinical data, where the method 400 includes generating object space maps based on the associated clusters (e.g., steps 410 through 422 of Figure 4, which may generally correspond to steps 110 through 122 of Figure 1), receiving metadata associated with each object space map 424, and storing each object space map and its associated metadata in a database record 426. The method may further include using the database record as input for a machine learning algorithm, e.g., a supervised machine learning algorithm, to generate a predictive model 428.

[0156] In some embodiments, the metadata may include clinical data, non-clinical data, or a combination thereof. Clinical data may include, for example, the subject's health history (e.g., comorbidities, smoking, etc.), the subject's vital statistics (e.g., blood pressure, temperature, etc.), a diagnosis of a tissue abnormality, a predicted wound healing time, a proposed treatment plan, mechanical measures associated with the size / shape of the wound, the presence and characteristics of granulation tissue formation, the oxygenation status of the wound and / or surrounding area, the infection status of the wound and / or surrounding area, or a combination thereof. Non-clinical data may include the subject's age, wealth, visit number, or a combination thereof. In some embodiments, the metadata may be weighted relative to other factors (e.g., depending on the importance of each parameter). Furthermore, in some embodiments, the weighting applied may be modified to better understand each input.

[0157] In certain embodiments, a method may be used to predict clinical data, as illustrated in the exemplary method 500 of FIG. 5A. The method 500 may include receiving 510 data of a subject's time series of fluorescence images, which may be generated and processed as described in connection with various embodiments above, and predicting 514 clinical data associated with the subject's time series of fluorescence images by using a predictive model generated 512 according to the method. FIG. 5B illustrates the use of a spatial map generated according to various methods described herein in combination with subject metadata for the generation of a database or registry, and further for the generation of a neural network classification model. Thus, as illustrated schematically in FIG. 5C, a new subject may be evaluated as follows: That is, during imaging, a subject time series of fluorescent images of the tissue being evaluated is generated, a subject spatial map is generated as described herein, such map is stored in a database or registry, and various data derived from the map (e.g., statistical data derived from the map, such as the proportion of each cluster in the map, its mean / median / standard deviation, map histogram, or a combination thereof), one or more of which may be used as inputs to a previously generated / trained classification neural network model, which then suggests possible predicted outputs (e.g., diagnoses) for consideration by a medical professional to help facilitate a diagnosis by the medical professional. In various embodiments, such systems do not provide a diagnosis but rather provide possible suggested outputs, while in other aspects, such systems do provide a diagnosis. In various other aspects, such systems are not used to facilitate a diagnosis, but rather to create a database or registry of spatial maps, data derived from the spatial maps, or a combination thereof. The database or registry may contain data organized, for example, by tissue type, modality, and clinical condition, which, when accessed by a user (e.g., a medical professional), helps facilitate diagnosis.

[0158] Thus, in one embodiment, as illustrated in exemplary method 600 of FIG. 6 , a method for characterizing tissue of a subject may include receiving 612 data from a plurality of time series of fluorescence images, selecting 614 feature vectors for the data, each feature vector characterizing one or more features of the data, generating 616 a dataset including the feature vectors, classifying the dataset to generate a labeled dataset 618, and generating 620 a plurality of centroids. In one embodiment, the output centroids may be further used to create a spatial (cluster) map for new subject data, as described above. A further embodiment of a method for characterizing tissue of a subject may include receiving a training dataset including a plurality of feature vectors characterizing one or more features of a plurality of data entries, each data entry being at least a portion of a time-intensity curve for a training subregion in the training time series of fluorescence images.

[0159] In some embodiments, tissue may include, for example, healthy tissue, unhealthy tissue, wound tissue, or a combination thereof. Wounds may include any type of chronic or acute wound to tissue, such as, for example, an incision, a pressure ulcer, a venous ulcer, an arterial ulcer, a diabetic leg ulcer, a laceration, an abrasion, a puncture wound, a contusion, an abrasion, a cavity, a burn, a combination thereof, and / or the like. Furthermore, wounds may be caused by one or more of a variety of traumatic events and / or medical conditions. Such traumatic events and / or medical conditions may be, for example, a crush wound, a battle wound (e.g., a bullet wound / explosion), or a wound resulting from gangrene, inflammation, venous congestion, lymphedema, or the like.

[0160] One challenge in wound management is that different practitioners may have different perspectives on the medical condition or nature of the wound, depending, for example, on the practitioner's skill and experience. Traditional wound management techniques can provide information about the pathological history of a wound, but cannot provide reliable indicators of viability and / or likelihood of recovery (e.g., whether the wound and / or its surrounding area is likely to cause complications, whether it is healable, how healing will progress, and whether the administered treatment is effective and when it can be stopped). Furthermore, there are wounds for which pathology cannot be demonstrated by traditional techniques.

[0161] In an attempt to address some of these challenges, some fluorescence imaging techniques can generate metrics from video data to numerically characterize blood flow and / or perfusion in and around a wound, in addition to providing a visual display, thus reducing subjectivity and cognitive bias when assessing tissue blood flow and / or perfusion status. However, such numerical characterizations lack an understanding of the biological mechanisms underlying wound healing, which are necessary to convey information that enables clinicians to make meaningful assessments. More specifically, understanding blood flow and / or tissue perfusion dynamics during the wound healing process would be beneficial for such image data to yield accurate interpretations of wound healing status. Existing fluorescence imaging techniques do not incorporate such knowledge, support standardized protocols for assessing blood flow and / or tissue perfusion, or provide accurate characterization and classification of blood flow / perfusion behavior within tissues that is consistent across clinicians, patients, and multiple imaging sessions.

[0162] In certain aspects, methods described herein relate to medical imaging techniques for characterizing a wound (e.g., wound, wound periphery) in a target tissue region. Spatial maps and / or object spatial maps (cluster maps) generated using the methods described herein embody both ease of interpretation and overall accuracy in characterizing the tissue, which is due to the quality of the measurement signal rather than subjective human selection of relevant parameters. The methods may provide enhanced diagnostic power by minimizing dilution of information of interest. Furthermore, the methods may provide a unified, objective representation of the condition of the target tissue (e.g., wound or wound periphery). Such a representation is not subject to the cognitive and / or skill biases of a medical professional. Furthermore, the methods may provide a reliable and consistent approach for comparing and tracking a subject's wound healing status (e.g., based on blood flow and / or perfusion) over time and across multiple imaging sessions. Thus, the methods may not only allow for more accurate and uniform assessment of target tissue regions, but also for targeted formulation of medical care strategies (e.g., recommending treatment, monitoring treatment efficacy, determining if / when to stop treatment, and formulating surgical strategies). Finally, the methods may also facilitate reduced patient risk for drug-sensitive patients and reduced overall costs of treatment and / or therapy.

[0163] Evaluating wounds according to various embodiments includes assessing perfusion dynamics. For example, the methods and systems described herein are applicable to other medical applications, such as preoperative assessment of patients undergoing plastic reconstructive surgery, general surgeries involving tissue reapproximation with vascular anastomosis (e.g., skin flap transfer, colon reconstruction, etc.), or assessment of cardiac tissue viability and function during cardiac surgery. Furthermore, the methods and systems described herein are also applicable to clinical assessment of any dynamic process, such as tissue perfusion or other dynamic behavior of imaging agents within tissue. A dynamic process may be represented by a spatial map of image data generated from a time series of input data (e.g., image frames) representing the process.

[0164] Data obtained from performing the methods and using the systems described herein further facilitates distinguishing between multiple wound regions in a target tissue, which may develop, progress, and / or heal according to different timelines.

[0165] Additionally, although various methods are described herein in the context of time series of fluorescence images, the methods are applicable to other sources of input data generated as time series of the dynamic behavior of an imaging agent within tissue, and for other medical purposes where the target tissue includes regions with different tissue properties. Examples may include other sources of input data, such as a time series of images generated by detecting fluorescence from an excited imaging agent and detecting absorption associated with the imaging agent. Quantification of clusters, spatial maps, target spatial maps, or a combination thereof

[0166] The methods of the present invention may further include quantification of the classified clusters, spatial maps generated from the clusters, spatial maps of interest generated from a time series of fluorescent images of interest, or combinations thereof. Quantification may include generating a numerical value (quantifier) ​​for a region of interest within the map or for the entire map.

[0167] The generated values ​​may provide a quantitative representation of the tissue (e.g., a wound). According to embodiments, the values ​​may represent tissue activity (e.g., wound activity values). The values ​​may be tracked over time and represented in the form of a graph that facilitates deriving information about rates and slopes. The graphical representation of the values ​​over time facilitates assessment of changes in the values ​​over time and, in some embodiments, may indicate changes in the state or activity of the tissue (e.g., a wound) over time. Examples of tissue state or activity include tissue characteristics, tissue condition, and tissue healing states (e.g., inflammation, malignancy, abnormalities, disease). Tracking the values ​​over time facilitates tracking the rate of change, which may be correlated, for example, with the stage of tissue healing (e.g., wound healing). Tracking the values ​​over time may further relate to angiogenesis and the healing stage the patient is currently in. Furthermore, information about changes in the values ​​over time may provide predictive information about when treatments, such as hyperbaric oxygen therapy, negative pressure therapy, or other known wound treatment methods, can be stopped without sacrificing the healing process. As a result, the numerical value provides an objective, standardized protocol for assessing tissue blood flow and / or tissue perfusion, which may facilitate a method for reliably and consistently comparing and tracking a subject's blood flow and / or perfusion status over time and across multiple imaging sessions, independent of the medical professional performing the assessment. In certain aspects, the numerical value (quantifier) ​​itself may be complex, for example, when derived from statistics regarding the various types of categories of curves present in the spatial map and / or the distribution of clusters in the spatial map, or other parameters.

[0168] In certain embodiments, the method may further include displaying a numerical value (quantifier) ​​on a display. For example, the numerical value may be displayed within a user interface on a video monitor of the fluorescence imaging system or on another suitable display. In certain embodiments, the numerical value may be used alone or in combination with visualization of other steps of the methods described herein to enhance the information provided to the medical practitioner (which may facilitate enhanced diagnosis). The numerical value may also be overlaid on an anatomical image and / or correlated with other data or information regarding the subject (e.g., the patient's overall condition). For example, in certain embodiments, the numerical value may be displayed alone or in combination with a subject spatial map (e.g., 122, 422). As another example, the numerical value may be displayed in combination with a spatial (cluster) map and / or other suitable maps or images. In certain embodiments, the numerical value may be correlated with a risk assessment of a clinically relevant (e.g., perfusion-related) condition. Such an assessment may be performed pre-interventionally, during treatment / procedure, and / or post-intervention. The method may further include defining an examination to identify and characterize a subject's clinically relevant (e.g., perfusion-related) condition pre-interventionally, during treatment / operation, and post-intervention. In various other embodiments, the method may exclude the correlation step and / or the examination step.

[0169] Various aspects of the method are further described in the Examples section, along with applications to various clinical contexts. System for characterizing tissue and / or predicting clinical data

[0170] According to one embodiment, a system for characterizing tissue of a subject and / or predicting clinical data and / or outcome includes an imaging system that acquires a time series of images of the tissue (e.g., a time series of fluorescence images), one or more processors, and a memory having instructions stored thereon that, when executed by the one or more processors, cause the system to perform a method for characterizing tissue and / or predicting clinical data, substantially as described above.

[0171] In certain embodiments, a system for generating a time series of fluorescence images / object time series and / or for characterizing tissue of a target and / or predicting clinical data, as described in connection with various embodiments herein, is a fluorescence imaging system. Figure 7 is a schematic example of a fluorescence imaging system 710. The fluorescence imaging system 710 includes a light source 712 that illuminates the tissue of the target to stimulate fluorescence emission from a fluorescent imaging agent 714 in the tissue of the target (e.g., in blood), an image acquisition assembly 716 configured to generate a time series of fluorescence images and / or object time series from the fluorescence emission, and a processor assembly 718 configured to process the generated time series of fluorescence images / object time series according to any of the method embodiments described herein. The processor assembly 718 may include a memory 768 with instructions, a processor module 762 configured to execute the instructions on the memory 768 to process the time series of fluorescence images and / or the object time series as described herein in connection with various embodiments of the methods, and a data storage module 764 that holds raw and / or processed time series of fluorescence images and / or the object time series. In some embodiments, the memory 768 and the data storage module 764 may be embodied in the same storage medium, while in other embodiments, the memory 768 and the data storage module 764 may be embodied in different storage mediums. The system may further include a display 766 on which images and other data may be displayed, such as some or all of the time series / object time series of fluorescence images or other input data, the spatial map, the object spatial map, and / or tissue numerical values ​​(quantifiers).

[0172] In some embodiments, the light source 712 includes, for example, an illumination module 720. The illumination module 720 may include a fluorescence excitation source configured to generate excitation light having an intensity and wavelength suitable for exciting the fluorescent imaging agent 714. As shown in FIG. 8 , the illumination module 720 may include a laser diode 822 (which may include, for example, one or more fiber-coupled diode lasers) configured to provide excitation light for exciting a fluorescent imaging agent (not shown) in the tissue of interest. Examples of other sources of excitation light that may be used in various embodiments include one or more LEDs, arc lamps, or other illumination technologies of sufficient intensity and suitable wavelength to excite the fluorescent imaging agent in the tissue. For example, excitation of a fluorescent imaging agent in blood (which may be a fluorescent dye with near-infrared excitation and emission characteristics) may be performed using one or more 793 nm, conduction-cooled, single-bar, fiber-coupled laser diode modules available from DILAS Diode Laser Co., Germany.

[0173] Referring again to FIG. 7 , in some embodiments, the optical output from the light source 712 may be projected through one or more optical elements to shape and direct the output used to illuminate the tissue region of interest. The optical elements may include one or more lenses, light guides, and / or diffractive elements to achieve a flat field across substantially the entire field of view of the image acquisition assembly 716. The fluorescent excitation source may be selected to emit at a wavelength near the maximum absorption of the fluorescent imaging agent 714 (e.g., ICG, etc.). For example, as shown in FIG. 8 , output 824 from a laser diode 822 may pass through one or more converging lenses 826 and then through a homogenizing light pipe 828, such as a light pipe commonly available from Newport Corporation, USA. Finally, the light may pass through an optical diffractive element 832 (i.e., one or more optical diffusers), such as a ground-glass diffractive element available from Newport Corporation, USA. Power to the laser diode 822 may be provided by a high-current laser driver, such as a high-current laser driver available from Lumina Power Inc., USA. The laser may optionally be operated in a pulsed mode during the image acquisition process. An optical sensor, such as a solid-state photodiode 830, may be incorporated into the illumination module 720 to sample the illumination intensity produced by the illumination module 720 via scattering or diffuse reflection from various optical elements. In some embodiments, an additional illumination source may be used to provide guidance when aligning and positioning the module over the region of interest.

[0174] Referring again to FIG. 7 , in some embodiments, image acquisition assembly 716 may be a component of fluorescence imaging system 710 and may be configured to acquire a time series of fluorescence images and / or a time series of interest from the fluorescence emission from the fluorescence imaging agent 714. Image acquisition assembly 716 may include a camera module 740. As shown in FIG. 9 , camera module 740 may acquire images of the fluorescence emission 942 from the fluorescence imaging agent in the tissue by using a system of imaging optics (e.g., 946a, 946b, 948, and 950) to focus the fluorescence emission onto image sensor assembly 944. Image sensor assembly 944 may include at least one two-dimensional solid-state image sensor. The solid-state image sensor may be a charge-coupled device (CCD), a CMOS sensor, a CID, or similar two-dimensional sensor technology. The electrical charges resulting from the optical signals converted by image sensor assembly 944 are converted into electronic video signals, including both digital and analog video signals, by appropriate readout and amplification electronics within camera module 940.

[0175] According to an exemplary variation of the fluorescence imaging system, the light source provides an excitation wavelength of about 800 nm + / - 10 nm, and the image acquisition assembly uses an emission wavelength of >820 nm, e.g., with NIR-compatible optics for ICG fluorescence imaging. In an exemplary embodiment, the NIR-compatible optics may include a CCD monochrome image sensor with a GigE standard interface and a lens compatible with the sensor in terms of optical format and mounting format (e.g., C / CS mount).

[0176] In one aspect, processor module 762 includes any computer or computing means, such as a tablet, laptop, desktop, network computer, or dedicated standalone microprocessor. For example, processor module 762 may include one or more central processing units (CPUs). In an exemplary embodiment, processor module 762 is a quad-core, 2.5 GHz processor with four CPUs, each of which is a 64-bit microprocessor (e.g., sold as an INTEL Core i3, i5, or i7, or AMD Core FX series). However, in other embodiments, processor module 762 may be any suitable number of CPUs and / or any other suitable processor with a suitable clock speed.

[0177] Input to the processor module 762 may be obtained, for example, from the image sensor 944 of the camera module 740 shown in FIG. 9, from the solid-state photodiode 830 of the illumination module 720 of FIG. 8, and / or from external control hardware such as a foot switch or remote control. Output is provided to a laser diode driver and optical aligner. As shown in FIG. 7, in certain embodiments, the processor assembly 718 may include a data storage module 764 with the capability to store image / object time series, or data representing the image / object time series, or other input data, in a tangible, non-transitory, computer-readable medium such as internal memory (e.g., hard disk or flash memory), thereby enabling recording and processing of acquired data. In certain embodiments, the processor module 762 may include an internal clock, which allows control of various elements and ensures proper timing of illumination and sensor shutters. In certain embodiments, the processor module 762 may provide a graphical display of user inputs and outputs. The fluorescence imaging system may optionally include a video display 766 or other monitor for displaying the time series of fluorescence images as they are being acquired or as they are played back after recording. The video display 766 may additionally or alternatively visualize data generated during the performance of the methods described herein, such as spatial maps, object spatial maps, and / or tissue values.

[0178] In operation of the exemplary system shown in FIGS. 7-9 , a subject is positioned relative to the fluorescence imaging system 710 such that the area of ​​interest (e.g., a target tissue region) is positioned beneath the light source 712 and image acquisition assembly 716. As a result, the illumination module 720 of the light source 712 generates a substantially uniform illumination field across substantially the entire area of ​​interest. In certain embodiments, an image of the area of ​​interest may be acquired for background subtraction prior to administering the fluorescent imaging agent 714 to the subject. To acquire a fluorescence image / object fluorescence image, an operator of the fluorescence imaging system 710 may initiate acquisition of a time series of fluorescence images / object time series by depressing a remote switch or foot control or via a keyboard (not shown) connected to the processor assembly 718. As a result, the light source 712 is turned on, and the processor assembly 718 begins recording the fluorescence image data / object fluorescence image data provided by the image acquisition assembly 716. When operating in pulsed mode in an embodiment, the image sensor 944 of the camera module 740 is synchronized to collect the fluorescence emission following a laser pulse generated by the diode laser 822 of the illumination module 720. In this manner, maximum fluorescence emission intensity is recorded and the signal-to-noise ratio is optimized. In this embodiment, the fluorescence imaging agent 714 is administered to the subject and transported to the area of ​​interest via the arterial flow. Acquisition of a time series of fluorescence images / object time series begins, for example, immediately after administration of the fluorescence imaging agent 714, and a time series of fluorescence images from substantially the entire area of ​​interest is acquired through the ingress of the fluorescence imaging agent 714. Fluorescence emission from the region of interest is collected by the collection optics of the camera module 740. Residual ambient excitation light and reflected excitation light are attenuated by subsequent optical elements of the camera module 740 (e.g., optical element 950 of FIG. 9, which may be a filter). As a result, fluorescence emission can be acquired by the image sensor assembly 944 with minimal interference with light from other sources.

[0179] In certain embodiments, following acquisition or generation of the fluorescence image time series / object time series, processor assembly 718 (e.g., processor module 762 or another processor) may be activated to execute instructions stored in memory 768 and perform one or more of the methods described herein. System 710 may visualize on display 766 the spatial map and / or object spatial map and / or medical correlations and / or diagnostic results derived therefrom, which may be displayed to a user, for example, as grayscale or false color images, and / or may be retained for later use. Additionally or alternatively, system 710 may display tissue values ​​on display 766.

[0180] In certain aspects, a system for predicting clinical data and / or outcomes or characterizing tissue includes a user interface, a processor configured to communicate with the user interface, and a non-transitory computer-readable storage medium having instructions stored thereon that, when executed by the processor, cause the processor to perform one or more of the methods for characterizing tissue and / or predicting clinical data described herein. In certain aspects, the processor may be a component of an imaging system. In other aspects, the processor may be remote from and in communication with the imaging system. In this case, the imaging system may be a fluorescence imaging system as described above or any other suitable imaging system.

[0181] A tangible, non-transitory computer-readable medium having computer-executable (readable) program code embodied thereon may provide instructions that, when executed, cause one or more processors to perform one or more of the methods for characterizing tissue and / or predicting clinical data described herein. The program code may be written in any suitable programming language and may be provided to a processor in many forms, including, but not limited to, information permanently stored on a non-writable storage medium (e.g., a read-only memory device such as a ROM or CD-ROM disk), information reversibly stored on a writable storage medium (e.g., a hard drive), or information conveyed to the processor over a communications medium, such as a local area network, a public network such as the Internet, or any type of medium suitable for carrying electronic instructions. When carrying computer-readable instructions that implement various embodiments of the methods of the present invention, such computer-readable media represent examples of various embodiments of the present invention. In various embodiments, tangible non-transitory computer readable medium includes all computer readable media, and the scope of the present invention encompasses computer readable media, both tangible and non-transitory.

[0182] A kit may include any of the parts of the system described herein and a fluorescent imaging agent, such as a fluorescent dye such as ICG or any suitable fluorescent imaging agent. In a further embodiment, a kit may include a tangible, non-transitory computer-readable medium having computer-executable (readable) program code embodied therein, the medium may provide instructions that, when executed by one or more processors, cause one or more of the methods for characterizing tissue and / or predicting clinical data described herein. The kit may include instructions for use of at least some of its components (e.g., for using the fluorescent imaging agent, for installing computer-executable (readable) program code embodied therein, etc.). In yet another embodiment, a fluorescent imaging agent, such as a fluorescent dye, for use with the methods and systems described herein is provided. In a further variation, a kit may include all or any part of the system described herein and a fluorescent agent, such as a fluorescent dye such as ICG or any other suitable fluorescent agent or combination of fluorescent agents.

[0183] example Application of the method and system in wound management

[0184] One challenge in wound management, including chronic wound management, is that healthcare professionals may have different perspectives on the medical condition and nature of a wound. While conventional techniques can provide information about a wound's pathological history, they cannot provide reliable indicators of viability and / or likelihood of recovery, e.g., whether the wound and / or its surrounding area are likely to develop complications, whether it is healable, or how healing will progress (e.g., the time it will take to reach an acceptable stage of healing). Furthermore, there are wounds for which pathology cannot be demonstrated using conventional diagnostic techniques. Various embodiments of the methods and systems described herein facilitate the generation of a unified (free from cognitive bias) representation of the condition of a particular tissue area (e.g., wound, wound surroundings), thereby facilitating more accurate assessment and formulation of subsequent treatment strategies (e.g., recommendation and assessment of efficacy care, such as topical treatments, hyperbaric therapy, pre- and post-operative tissue assessment, formulation of surgical strategies, and recommendations regarding the time it will take for tissue to reach various healing stages).

[0185] Training Set 1 - Breast Tissue in Reconstructive Surgery

[0186] 10 and 11 illustrate the application of various embodiments of the method and system to breast reconstruction surgery. Data was collected during a mastectomy procedure. The patient was a 46-year-old woman who underwent a bilateral mastectomy with immediate reconstruction. 48 hours after surgery, she was deemed to have ischemic deterioration in the lower pole of her right breast. HBOT treatment was recommended. A time series (moving) of fluorescence angiography images was recorded using a SPY® Elite Fluorescence Imaging System (available from NOVAQ® Technologies Inc.). Three types of recordings were made for each breast treatment being performed: pre-incision baseline, post-mastectomy, and post-reconstruction. Additionally, color snapshots were taken one week after surgery as a means of assessing clinical outcomes.

[0187] The first data set described in connection with methods and systems according to various embodiments was generated by combining pixel intensity curves from three different sequences of the chest. A k-means algorithm was then trained on this data set to generate a model with seven centroids, which is shown in FIG. 10.

[0188] Applying this trained model to the pixels successively labeled the two training sequences and one new sequence. As a final step, visual spatial maps were generated for the three sequences by assigning each pixel a color corresponding to the color of its associated centroid (as shown in the legend of the centroid graph in Figure 10). Figures 11A, 11B, and 11C show color images of the wound during the initial consultation (Figure 11A), followed by images taken one week (Figure 11B) and three weeks (Figure 11C) after the initial consultation as treatment continued. Figures 11D, 11E, and 11F are the corresponding spatial (cluster) maps generated according to the methods and systems described herein.

[0189] This case demonstrates the healing of a hyperperfused wound. As shown in Figures 11D, 11E, and 11F, the spatial (cluster) map provides details about blood flow and / or perfusion that are not apparent from the visible light images in Figures 11A, 11B, and 11C. The spatial (cluster) map image identified an area adjacent to the nipple (indicated by the arrow). The tissue in this area was significantly different (damaged) compared to the adjacent tissue.

[0190] HBOT treatment triggered the angiogenesis process, which initially caused increased blood flow activity around the hyperperfused area of ​​tissue (Figure 11D, arrow). As healing progressed, the dark blue areas collapsed over time, and increased flow spread into the wound, as indicated by increased blood flow and / or perfusion (Figures 11E, 11F). The progression of healing was represented in the spatial (cluster) map by how the intensity curve gradually changed from the center of the wound outward—from dark blue to light blue to green to yellow—with the dark blue areas eventually collapsing as healing progressed. The spatial (cluster) map indicated that healing did not occur suddenly, but rather gradually and symmetrically around the wound. Such information would not have been apparent from examining the color images (i.e., Figures 11A, 11B, and 11C).

[0191] Training Set 2 - Legs

[0192] A time series (moving images) of fluorescence angiography images was recorded using a LUNA® fluorescence imaging system (available from NOVAQ® Technologies Inc.). A time series of fluorescence images of the paw and a paw dataset were generated in a similar manner to the breast tissue example. More specifically, the paw dataset was generated by combining pixel intensity data over time from three different paw sequences and then trained using seven clusters and k-means. The resulting centroids are shown in Figure 12A, and the resulting spatial maps showing the wound status are shown in Figures 12B and 12C. Application of cluster analysis to the universal perfusion-based wound scale for tissue classification

[0193] There are many existing wound classification systems, including for example: (i) the Wagner classification of neuropathic ulcers, which grades wounds by their depth and the presence or absence of infection, and has five numerical grades; (ii) the University of Texas scheme used for neuropathic ulcers, which grades wounds by their depth and the presence or absence of infection, with four numerical grades for depth and four letter grades for infection and ischemia; (iii) the National Pressure Ulcer Advisory Panel Classification, which grades pressure ulcers by their color and the presence or absence of tissue loss and scabs, defining six numerical stages; (iv) the Rutherford and Fontaine scheme used for arterial insufficiency ulcers, which grades wounds according to their clinical presentation and has descriptive stages from 4 to 6; (v) The CEAP classification for venous insufficiency ulcers, which consists of two separately scored parts, with four letter grades for Part I and three numerical grades for Part II.

[0194] There are also special grading systems for burns, which rank wounds according to their depth and the area affected, and the PEDIS system, DEPA score and SAD score for diabetic foot ulcers.

[0195] Existing wound classification systems are primarily based on grading the wound's surface appearance, its texture, and morphology. As a result, different systems have evolved for different wound etiologies to efficiently capture the broad spectrum of aspects of damaged tissue. With so many options available to medical professionals, the question arises as to which system they should use. Having several different systems for describing similar types of wounds has obvious disadvantages, and therefore, a well-designed universal wound classification scheme would be advantageous.

[0196] The methods and systems described herein facilitate identifying unique blood flow patterns and correlating them with corresponding wound types, thus creating a universal wound classification system based on a common perfusion profile that is applicable to different wound etiologies and severities. Wound grades based on such a scale can be correlated to their etiology, healing potential, and optimal treatment.

[0197] Twenty patients undergoing treatment for various chronic wounds (DFUs, trauma, surgery, arterial ulcers) were imaged weekly for five consecutive weeks (average) using the LUNA® Imaging System (available from Novadaq Technologies® Inc.). Maximum intensity maps were generated from the NIR video sequences recorded during the imaging sessions ("maximum intensity map" refers to a map generated by assigning to each pixel in the computational domain of the time series of fluorescent input images its maximum intensity value reached during the entire measurement period). At the end of the patient's treatment, the examining physician recorded the date of wound healing. The duration between the date of a particular imaging session and the date of wound healing was then calculated and associated with every individual maximum intensity map. To generate a sufficient number of training and test samples, the continuous labels representing healing duration in days were replaced by discrete categories of "healing bins": "A" - healing duration from 0 to 20 days, "B" - healing duration from 21 to 80 days, and "C" - healing duration greater than 80 days. The resulting sample dataset of approximately 100 each contained a maximum intensity map image labeled with the associated "healing bin" grade (A, B, or C).

[0198] In this example, the Microsoft CustomVision cloud-based service (customvision.ai) was selected as the training platform for the predictor. This tool allows for the creation of custom image classifiers with approximately 20–30 training images per category. The following criteria were used to select training samples: representative types of imaged body parts (i.e., leg, foot, heel, hand, abdomen), presence of noise in a given image, and representative types of wound etiology (e.g., diabetic foot ulcer (DFU), trauma, surgery). Because maps were generated using the same false color scheme, fewer training samples were required for the image classifier in this example to identify relevant blood flow patterns associated with healing duration. In this example, the training procedure was repeated twice. First, a selected number of images (e.g., 76) from the maximum intensity maps were uploaded to the cloud-based service and tagged with their corresponding "healing grades": 11A-s, 45B-s, and 20C-s. After training, the classifier's performance was automatically evaluated on the training set using k-fold cross-validation, generating a precision / recall metric as a measure of the classifier's predictive ability. As shown in Figure 13, the classifier performed best at identifying grade B and scored worst for grade A. These results directly correlated with the number of training samples for each category: highest for Bs and lowest for As. Subsequently, additional tagged images (e.g., 10 additional images) were uploaded to the training platform. The resulting new training set contained a total of 86 images: 13 A-s, 49 B-s, and 24 C-s, and the classifier was retrained.

[0199] The evaluation results for the second iteration are shown in Figure 14, which shows an improvement in the overall scores, with particularly notable changes for the Grade A predictions.

[0200] To test the trained classifier from iteration 2 in Figure 14, a set of five images from a single patient was used, with a known "days to heal" metric associated with each image. These images had never been "seen" by the classifier before, thus allowing us to measure how well it can generalize to new data. Figure 15 (shown with a "healing bin" scale) shows the images submitted for prediction, with observed labels associated with the images and tags predicted by the classifier along with their probabilities. Correctly predicted observed labels are shown in green (labeled "correct prediction"), and incorrect predictions are shown in red (labeled "incorrect prediction"). As shown by the results in Figure 15, the classifier correctly predicted all but one label. Furthermore, as shown in Figure 15, both the probability and label of the second choice change consistently as healing progresses along the timeline.

[0201] For example, the first sample in Figure 15 was marked as 76 days to heal, which places it in the B-bin (80-21 days), but very close to the boundary of the C-bin (>80 days). Although the classifier correctly predicted B as the most likely category, it also assigned a 31% probability to the C label.

[0202] The second sample in Figure 15 (46 days since healing) is roughly in the middle of the B category, which is correctly reflected by the classifier by assigning 99.9% to the B label and a much lower but roughly equal probability to being either A or C (9.4% and 7.3%, respectively).

[0203] The third sample in Figure 15 (39 days after healing) was misclassified as a C grade, but it also assigned a relatively high probability (74.2%) of the correct grade, B.

[0204] The fourth sample in Figure 15 (20 days after healing) is located exactly on the dividing boundary between the A and B categories, and the classifier correctly assigned equally high probabilities to both grades (95.2% to A and 94.6% to B).

[0205] Finally, the last sample in Figure 15 shows a nearly completely healed wound, and the classifier correctly assigned a very high probability (99.6%) to grade A, as well as very low probabilities to B and C (2.7% and 0.3%, respectively).

[0206] The training and prediction trends described herein show that by increasing the number and variety of training samples, and by introducing more labels representing narrower time periods, the accuracy and consistency of predictions of healing grade for new data can be increased.

[0207] FIG. 16 schematically illustrates an exemplary method 2000 for training a classifier on fluorescence image data and using the trained classifier to predict clinical data. As shown in FIG. 16, the classifier may be trained 2010 using the custom vision cloud service described herein. Once the trained classifier's performance reaches an acceptable level, the trained model may be output 2020 as a REST API service. Using the public URL of the prediction endpoint, a client application can submit REST API requests to the server to predict labels for new images and receive a response with the resulting tags 2030, as described herein in various embodiments. The wound classification scale (wound grade) output is generated, for example, based on automatically classifying perfusion patterns in tissue and assigning specific grades to associated clinical observations, according to the methods and systems described herein. The wound classification scale (wound grade) illustrated herein facilitates eliminating the observer / healthcare professional subjectivity inherent in all conventional wound classification schemes. In addition to the wound classification scale, suggested treatment options based on the classification may be provided to the healthcare professional (e.g., wound classification scale number / letter; etiology DFU (80% confidence) and arterial tumor (65% confidence); suggested treatment HBOT (40 dives - twice weekly), Dermacell® (80%), amputation surgery (50%), nothing (10%)).

[0208] The examples demonstrate a unique set of advantages achievable in practice by employing machine learning algorithms in applications to blood flow and / or perfusion analysis of tissue (e.g., wound tissue). In certain embodiments, input data to the algorithm does not rely on preprocessing or a detailed understanding of blood flow dynamics. As a result, the accuracy of the analysis depends primarily on the quality of the measured signal, rather than on subjective human selection of relevant parameters. Furthermore, machine learning classification and characterization results are less susceptible to noise in the input signal, due to the benefits of "big data" processing. Furthermore, spatial maps generated according to the machine learning-based methods and systems described herein exhibit both simplicity of interpretation and overall accuracy of results. They can be used as a realistic replacement for and / or augmentation of currently implemented but yet-to-be-conceptualized visual maps and / or images. Because the color scheme of the spatial map can be easily associated with centroids representing different angiographic curve classes, manual region-of-interest (ROI) selection and subsequent graph generation are not required. By simply viewing the color legend of the spatial map and its corresponding centroids, users can immediately assess blood flow patterns across the entire image area. Furthermore, as described in connection with the methods and systems, once the clustering model is trained on a relevant dataset, it can be stored on any computing platform. The model is highly scalable and can be easily extended to other modalities (i.e., orthopedics, MIS, pressure ulcers, etc.).

[0209] Although exemplary embodiments and optional variations thereof have been disclosed herein and specific terms have been used, they are used and should be construed as being general and descriptive only, and not for purposes of limitation. In certain instances, as would be apparent to one of ordinary skill in the art as of the filing of this application, features, characteristics, and / or elements described in connection with particular embodiments may be used alone or in combination with features, characteristics, and / or elements described in connection with other embodiments, unless otherwise specified. Accordingly, it will be apparent to one of ordinary skill in the art that various changes in form and detail may be made therein without departing from the spirit and scope of the invention as defined below.

[0210] While the present disclosure has been described in connection with various detailed embodiments, it is not intended to be limited to the details shown, as various modifications and structural changes can be made without departing from the scope of the present disclosure. Various changes in the form, component arrangement, step, detail, and sequence of operations of the described embodiments can be made, and other embodiments of the present disclosure can be made, which will be apparent to those skilled in the art having access to this disclosure, without departing from the scope of the present disclosure. Accordingly, it is intended that the appended claims will cover such modifications and embodiments as fall within the scope of the present disclosure. For purposes of brevity and clarity, features have been described herein as part of the same or different embodiments. However, it will be understood that the scope of the present disclosure includes embodiments having all or any combination of the described features. The terms "for example" and "such as," and their grammatical equivalents, unless expressly stated otherwise, are to be understood as followed by the phrase "and without limitation." As used herein, the singular forms "a," "an," and "the" include plural referents unless the context clearly dictates otherwise.

Claims

1. 1. A method for characterizing tissue of a subject, comprising: receiving data of a plurality of time series of fluorescence images of the object, the time series of fluorescence images being or having been acquired by an image acquisition system; identifying one or more attributes of said data related to a clinical characterization of said tissue; classifying the data into a plurality of clusters based on the one or more attributes of the data such that the data within the same cluster are more similar to each other than the data within different clusters, the clusters characterizing the organization; generating a characterization output representative of the tissue based on the classified clusters; and displaying the characterization output.

2. The method of claim 1 , wherein the data for the plurality of time series of fluorescence images of the object comprises raw data, pre-processed data, or a combination thereof.

3. The method of claim 2 , wherein the preprocessed data is preprocessed by applying data compression, principal component analysis, auto-encoding, or a combination thereof.

4. 4. The method of claim 1, wherein the one or more attributes of the data relating to the clinical characterization of the tissue are identified for a plurality of sub-regions within the time series of fluorescence images of the subject.

5. The method of claim 4 , wherein at least one of the subregions is a pixel or voxel in the time series of fluorescence images.

6. The method of claim 4 , wherein at least one of the sub-regions is a group of pixels or a group of voxels in the time series of fluorescence images of the object.

7. 7. The method of any one of claims 1 to 6, wherein the one or more attributes of the data of the plurality of time series of fluorescence images of the object comprise a time-intensity curve, a coefficient, a spatial location, an onset time, a time to flush, a maximum fluorescence intensity, blood inflow, blood outflow, or a combination thereof.

8. The method of claim 1 , wherein the clusters characterize the tissue based on the spatial distribution of the clusters, the properties of the clusters, cluster data, or a combination thereof.

9. The method of claim 8 , wherein the properties of the clusters include the shape of the clusters.

10. 10. The method of claim 1, wherein each cluster is represented by a centroid.

11. The method of claim 10 , wherein the cluster centroids indicate which of the one or more attributes of the data in the plurality of time series of fluorescence images of the subject contribute to data classification.

12. 12. The method of claim 1, wherein classifying the data of the plurality of time series of fluorescence images of the object into the plurality of clusters comprises classifying the data into 10 or fewer clusters.

13. 12. The method of claim 1, wherein classifying the data of the plurality of time series of fluorescence images of the object into the plurality of clusters comprises classifying the data into seven clusters.

14. 14. The method of claim 1, wherein classifying the data of the plurality of time series of fluorescence images of the object into the plurality of clusters comprises applying an unsupervised clustering algorithm.

15. The method of claim 14 , wherein the clustering algorithm is a k-means algorithm.

16. The method of claim 1 , further comprising generating a spatial map based on the plurality of clusters.

17. 17. The method of claim 16, wherein the spatial map represents differences in blood flow, perfusion patterns, or a combination thereof, among multiple subregions in the time series of fluorescence images.

18. 18. The method of any one of claims 1 to 17, further comprising training a machine learning model based on the classified data.

19. 20. The method of claim 18, wherein the machine learning model is trained with a supervised machine learning algorithm.

20. receiving data of a target time series of fluorescence images of the target, associating each of a plurality of sub-regions in the target time series of fluorescence images with a corresponding cluster; generating an object spatial map based on the clusters associated with the plurality of sub-regions in the object time series of fluorescence images; The method of claim 1 , further comprising: displaying the spatial map.

21. 21. The method of claim 20, wherein generating the object spatial map comprises assigning at least one of an intensity value and a color to each subregion in the object time series of fluorescence images based on the associated cluster.

22. 1. A method for predicting clinical data of a tissue of a subject, comprising: receiving a plurality of object space maps generated according to claim 19 or 20 and receiving metadata associated with each object space map; storing each spatial map of interest and its associated metadata in a database record; and using the records of the database as input to a supervised machine learning algorithm to generate a predictive model for characterizing the tissue.

23. 23. The method of claim 22, wherein the metadata comprises clinical data, non-clinical data, or a combination thereof.

24. 24. The method of claim 23, wherein the clinical data includes a diagnosis of a tissue abnormality, a predicted healing time for a wound, a proposed treatment plan, or a combination thereof.

25. 1. A method for predicting clinical data characterizing a tissue of a subject, comprising: receiving data of a target time series of fluorescence images of the target, the target time series of fluorescence images of the target being or having been acquired by an image acquisition device; characterizing tissue of the subject by using the predictive model generated according to the method of any one of claims 22 to 24 to predict clinical data associated with the subject time series of fluorescence images of the subject; generating a characterization output representative of the tissue.

26. 26. Use of the database of any one of claims 22 to 25 for predicting clinical data associated with said subject time series of fluorescence images of said subject.

27. 1. A method for characterizing tissue of a subject, comprising: receiving data of a target time series of fluorescence images of the target, the target time series of fluorescence images of the target being or having been acquired by an image acquisition device; Associating each of a plurality of sub-regions in the target time series of fluorescence images with a corresponding category, the categories being defined based on one or more attributes that characterize the tissue and relate to a clinical characterization of the tissue such that data within the same category are more similar to each other than data within different categories; generating a spatial map representing the tissue based on the categories associated with the plurality of sub-regions in the subject time series of fluorescence images; and displaying the spatial map.

28. 1. A method for characterizing tissue of a subject, comprising: receiving data of a plurality of time series of fluorescent images, the plurality of time series of fluorescent images being or having been acquired by an image acquisition system; selecting a plurality of feature vectors of the data, each feature vector characterizing one or more features of the data; generating a dataset including the feature vector; classifying the dataset to generate a labeled dataset; generating a plurality of centroids representing the tissue characterization; and displaying the tissue characterization output based on the plurality of centroids.

29. 1. A method for characterizing tissue of a subject, comprising:

1. A method comprising: receiving a training dataset including a plurality of feature vectors characterizing one or more features of a plurality of data entries, each data entry being at least a portion of a time-intensity curve of a training subregion in a training time series of fluorescence images, the time series of fluorescence images being or having been acquired by an image acquisition system.

30. 1. A system comprising: an image acquisition device configured to acquire a time series of fluorescence images; and one or more processors configured to cause the system to perform the method of any one of claims 1 to 29.

31. 31. The system of claim 30, further comprising a display for displaying the space map image, the object space map image, or both.

32. 32. The system of claim 31, wherein the one or more processors are further configured to overlay the spatial map image, the object map image, or both, on the anatomical image of the tissue.

33. 33. The system of any one of claims 30 to 32, further comprising a light source that provides excitation light to induce fluorescent emission from the fluorescent imaging agent in the tissue.

34. 34. The system of any one of claims 30 to 33, further comprising an image acquisition assembly that generates the time series of fluorescence images, the object time series of fluorescence images, or both based on the fluorescence emission.

35. 1. A system for processing a time series of images of tissue of a target, comprising: A user interface; a processor configured to communicate with the user interface; a non-transitory computer-readable storage medium having instructions stored thereon; 30. A system in which the instructions, when executed by the processor, cause the processor to perform the method of any one of claims 1 to 29.

36. 36. The system of claim 35, wherein the processor is in communication with an imaging system.

37. 36. The system of claim 35, further comprising an imaging system.

38. 36. The system of claim 35, wherein the processor is a component of the imaging system.

39. 39. The system of any one of claims 37 to 38, wherein the processor is configured to control operation of the imaging system.

40. 40. The system of any one of claims 37 to 39, wherein the imaging system is a fluorescence imaging system and the time series of images is a time series of fluorescence images.

41. The fluorescence imaging system includes: an illumination module configured to illuminate the tissue of the subject to stimulate fluorescent emission from a fluorescent imaging agent in the tissue of the subject; a camera assembly configured to acquire the time series of fluorescent images.

42. A non-transitory tangible computer readable medium having computer executable program code means embodied thereon for performing the method of any one of claims 1 to 29.

43. 35. A kit for processing a time series of fluorescent images of tissue of a subject, the kit comprising the system of any one of claims 30 to 34 and a fluorescent imaging agent.

44. 35. A fluorescent imaging agent for use in a method according to any one of claims 1 to 29 or in a system according to any one of claims 30 to 34.

45. 35. A fluorescent imaging agent for use in a method according to any one of claims 1 to 29 or in a system according to any one of claims 30 to 34 for wound management.

46. 46. ​​The fluorescent imaging agent of claim 45, wherein the wound management comprises chronic wound management.

47. 46. ​​The fluorescent imaging agent of claim 44 or 45, wherein the fluorescent imaging agent comprises ICG.

48. 46. ​​The fluorescent imaging agent of claim 44 or 45, wherein the fluorescent imaging agent is ICG.

49. 1. A method for visualizing angiographic data, comprising: a) receiving at least one temporal image sequence being or having been acquired by an image acquisition system; b) dividing said at least one temporal image sequence into a plurality of temporal sequences of spatial regions of said temporal image sequence; c) automatically dividing the plurality of time sequences in the spatial domain into a plurality of clusters such that the sequences within the same cluster are more similar to each other than to sequences from different clusters; d) receiving an angiographic image sequence to be visualized; e) determining for each pixel of the angiographic image sequence to which cluster the time sequence of that pixel corresponds; f) generating an image, wherein each pixel of the image is assigned a pixel value according to the cluster, and the position of the pixel in the angiographic image sequence is determined to correspond to the cluster.

50. 50. The method of claim 49, wherein step b) comprises determining, for each time sequence in a spatial domain, a feature vector representing temporal image change in that spatial domain.

51. 50. The method of claim 49, wherein the feature vector is determined using a dimensionality-reducing machine learning algorithm.

52. 52. The method of claim 51 , wherein the dimensionality reduction machine learning algorithm is based on principal component analysis, an autoencoding neural network, or a combination thereof.

53. 53. A method according to any one of claims 49 to 52, wherein in step b) said time sequence in the spatial domain is a time sequence of individual pixels of said images of said temporal image sequence.

54. 54. A method according to any one of claims 49 to 53, wherein step c) is performed using an unsupervised clustering algorithm.

55. 55. The method of claim 54, wherein the unsupervised clustering algorithm comprises a k-means algorithm.

56. Step c) automatically dividing the plurality of time sequences in the spatial domain into a plurality of clusters using an unsupervised clustering algorithm; Partitioning the plurality of time sequences in the spatial domain into a training data set and a test data set; using the training dataset as an input for a supervised machine learning algorithm to generate a predictive model; testing the predictive model against the test dataset; 54. A method according to any one of claims 49 to 53, wherein step e) comprises using the predictive model to determine which cluster the time sequence of pixels corresponds to.

57. 57. The method of any one of claims 49 to 56, wherein step c) comprises automatically dividing the plurality of time sequences of spatial domains into a plurality of clusters based on the time dependence of intensities of the spatial domains.

58. 58. The method of any one of claims 49 to 57, wherein step c) comprises determining the plurality of clusters based on a cumulative classification error.

59. 1. A method for visualizing angiographic data, comprising: a) obtaining a plurality of masks representing different time dependencies of intensity in spatial regions of an image; b) receiving an angiographic image sequence to be visualized; c) for each pixel of the angiographic image sequence, determining with which mask the time sequence of that pixel best corresponds; d) generating an image, each pixel of which is assigned a pixel value according to the mask, the position of the pixel in the angiographic image sequence being determined to correspond to the mask.

60. the plurality of masks e) receiving at least one temporal image sequence; f) dividing said at least one temporal image sequence into a plurality of temporal sequences of spatial domains of said images of said temporal image sequence; g) automatically dividing the plurality of time sequences in the spatial domain into a plurality of clusters such that the sequences within the same cluster are more similar to each other than sequences from different clusters; h) generating for each cluster a mask representing the time dependence of the intensity of the spatial domain of that cluster.

61. 61. The method of claim 60, wherein each mask represents the time dependence of the intensity of the centroid of the corresponding cluster.

62. 1. A method for predicting clinical data, comprising: a) receiving a plurality of angiographic image visualizations generated in accordance with any one of claims 49 to 61; b) for each angiographic image visualization, storing data representing it in a record in a database; c) for each angiographic image visualization, storing its associated clinical data in the corresponding record in the database; d) using the records of the database as input to a supervised machine learning algorithm to generate a predictive model; e) receiving an angiographic image sequence to be analyzed; f) visualizing said angiographic image sequence according to any one of claims 49 to 61; g) using the prediction model to predict clinical data associated with the angiographic image sequence.

63. 1. A method for predicting clinical data, comprising: a) receiving an angiographic image sequence to be analyzed, the time series of fluorescence images being or having been acquired by an image acquisition system; b) visualizing said angiographic image sequence according to any one of claims 49 to 61; c) using the prediction model to predict clinical data associated with the angiographic image sequence.

64. The predictive model d) receiving a plurality of angiographic image visualizations generated in accordance with any one of claims 49 to 61; and e) for each angiographic image visualization, storing data representing it in a record in a database; f) for each angiographic image visualization, storing its associated clinical data in the corresponding record in the database; g) using the records of the database as input to a supervised machine learning algorithm to generate the predictive model.

65. 65. Use of the database of any one of claims 62 to 64 for predicting clinical data associated with said angiographic image sequence.

66. 65. Use of the predictive model of any one of claims 62 to 64 for predicting clinical data associated with said angiographic image sequence.

67. 1. A method for generating a plurality of masks, comprising: a) receiving at least one temporal image sequence being or having been acquired by an image acquisition system; b) dividing said at least one temporal image sequence into a plurality of temporal sequences of spatial regions of said images of said temporal image sequence; c) automatically dividing the plurality of time sequences in the spatial domain into a plurality of clusters such that the sequences within the same cluster are more similar to each other than to sequences from different clusters; d) for each cluster, generating a mask representing the time dependence of the intensity of the spatial region of that cluster.

68. 68. The method of claim 67, wherein each mask represents the time dependence of the intensity of the centroid of the corresponding cluster.

69. 69. Use of a plurality of masks obtained by the method of claim 67 or 68 for visualizing angiographic data.

70. 1. A system for visualizing angiographic data, comprising: a) a first receiver for receiving at least one temporal image sequence being or having been acquired by an image acquisition system; b) a segmentation unit for segmenting said at least one temporal image sequence into a plurality of temporal sequences of spatial domains of said images of said temporal image sequence; c) a clustering unit that automatically divides the plurality of time sequences in the spatial domain into a plurality of clusters such that the sequences within the same cluster are more similar to each other than sequences from different clusters; d) a second receiver for receiving an angiographic image sequence to be visualized; e) a determination unit for determining, for each pixel of the angiographic image sequence, to which cluster the time sequence of that pixel corresponds; f) an image generator that generates an image, wherein each pixel of the image is assigned a pixel value according to the cluster, and the position of the pixel in the angiographic image sequence is determined to correspond to the cluster.

71. 1. A system for visualizing angiographic data, comprising: a) an acquisition unit for acquiring a plurality of masks representing a plurality of different time dependencies of the intensity of a spatial domain of an image; b) a receiver for receiving an angiographic image sequence to be visualized; c) a decision unit for deciding, for each pixel of the angiographic image sequence, which mask the time sequence of that pixel best corresponds to; d) an image generator that generates an image, wherein each pixel of the image is assigned a pixel value according to the mask, and the position of the pixel in the angiographic image sequence is determined to correspond to the mask.

72. 72. Use of a plurality of masks obtained by the method of claim 67 or 68 in the system of claim 71.

73. 1. A system for generating a plurality of masks, comprising: a) a receiver for receiving at least one temporal image sequence being or having been acquired by an image acquisition system; b) a segmentation unit for segmenting said at least one temporal image sequence into a plurality of temporal sequences of spatial domains of said images of said temporal image sequence; c) a clustering unit that automatically divides the plurality of time sequences in the spatial domain into a plurality of clusters such that the sequences within the same cluster are more similar to each other than sequences from different clusters; d) a generator for generating, for each cluster, a mask representing the time dependence of the intensity of the spatial region of that cluster.

Images