Systems and methods for thresholding for residual cancer cell detection - Patents.com

Abstract

Embodiments are disclosed relating to methods of using an image analysis system to identify residual cancer cells after a surgical procedure. In some embodiments, a patient-specific threshold value for use in detecting abnormal cells in a surgical site can be determined. A medical imaging device can be configured to generate a set of images of a patient's anatomy. An image analysis system can include one or more processors and be configured to receive the set of images and, by analyzing the set of images, determine a patient-specific threshold value for use in detecting abnormal tissue in the patient.

Description

[Technical field]

[0001] Related Applications This application claims the benefit of priority under 35 U.S.C. 119(e) to U.S. Provisional Application No. 62 / 786,657, filed December 31, 2018, the disclosure of which is incorporated herein by reference in its entirety. [Background technology]

[0002] background According to the National Cancer Institute's Surveillance Epidemiology and End Results report, over one million cancer surgeries are performed annually in the United States, and roughly 40% of these fail to remove the entire tumor. The occurrence of cancer cells remaining in the patient can lead to secondary surgical procedures. For example, in breast cancer lumpectomy, failure to remove all of the cancer cells during the primary surgical procedure has been reported in the literature to occur 15-40% of the time, necessitating secondary surgical procedures. Such secondary or multiple surgical procedures cause an increased rate of cancer recurrence and reduce the long-term survival rate of patients.

[0003] Typically, after a solid tumor resection, the surgeon removes the bulk of the tumor and sends the resected tumor tissue to a pathologist for post-operative evaluation to determine whether residual cancer was left behind in the patient. However, this pathology evaluation is a very time-consuming procedure, often taking several days to obtain a final result that is sent to the surgeon. If the pathologist indicates in the pathology report that the removed tissue has cancer cells bordering its area (a diagnostic term known as "positive margins"), the patient may require additional resection to complete the removal of the remaining cancer cells, or in the event that the patient has already completed the initial surgical procedure, this finding may require the surgeon to perform a secondary surgical procedure. Summary of the Invention

[0004] overview Various aspects of the present application relate to the in situ observation of residual cancer cells in a tumor resection bed during cancer surgery. Imaging techniques exist, including probes that fluoresce upon activation by target cells and handheld fluorescent imagers, to reduce the need for surgical follow-up and allow intraoperative identification of residual cancer to minimize the risk of recurrence. See, for example, U.S. Patent Application Nos. 14 / 211,201 and 14 / 211,259, the disclosures of which are incorporated herein by reference in their entirety. In situ observation techniques may typically provide intraoperative images of the surgical site in which cancer cells are marked with a contrast agent against a background. The background of the surgical site image may include healthy tissue. Imaging healthy tissue in the field of view of the surgical site image may facilitate the surgeon to locate the surgical site during surgery, while cancer cells highlighted with a contrast agent in the image may help the surgeon identify residual cancer to be removed. It is understood that there remains a need to detect residual cancer cells during the surgical procedure to ensure that all of the cancer has been removed from the tumor bed.

[0005] The present application is directed to methods and systems for performing image analysis for the identification of residual cancer cells in a tumor resection bed. Additionally, the systems and methods described herein may also be used to detect precancerous conditions. In particular, the present techniques are directed to determining patient-specific thresholds used to determine whether an image contains (or is likely to contain) healthy or abnormal and / or cancerous tissue.

[0006] In accordance with some embodiments, techniques provide a system for determining a patient-specific threshold for use in detecting abnormal cells in a surgical procedure site. The system includes a medical imaging device configured to produce a set of images of a patient's anatomy. The system includes an image analysis system including one or more processors configured to receive the set of images and to analyze the set of images to determine a patient-specific threshold for use in detecting abnormal tissue in the patient.

[0007] In accordance with some aspects, the techniques provide a computer-implemented method for determining a patient-specific threshold value to be used in detecting abnormal cells in a surgical procedure site, the method comprising receiving a set of images of a patient's anatomy and analyzing the set of images to determine a patient-specific threshold value to be used in detecting abnormal tissue in the patient.

[0008] In accordance with some aspects, the technique provides at least one non-transitory computer-readable storage medium including computer-executable instructions that, when executed by at least one processor of an image analysis system configured to collect surgical site images, perform a method to calculate a threshold value used to detect abnormal cells in the surgical site, the method including receiving a set of images of a patient's anatomy and determining a patient-specific threshold value to use to detect abnormal tissue in the patient. [Brief description of the drawings]

[0009] BRIEF DESCRIPTION OF THE DRAWINGS The accompanying drawings are not intended to be drawn to scale. In the drawings, each identical or nearly identical component illustrated in various figures is represented by a like numeral. For clarity, every component may not be labeled in every drawing. In the drawings:

[0010] [Figure 1]FIG. 1 is a schematic diagram showing an overview of the components of an exemplary system for identifying residual cancer cells after a surgical procedure according to some embodiments;

[0011] [Diagram 2] FIG. 2 is a block diagram of another exemplary cancer cell detection system according to some embodiments;

[0012] [Diagram 3] FIG. 3 is a flow chart illustrating an exemplary process for detecting residual cancer cells according to some embodiments;

[0013] [Figure 4] FIG. 4 is a flow chart illustrating an example process for calculating a patient-specific threshold according to some embodiments;

[0014] [Diagram 5] FIG. 5 is a flowchart illustrating an example process for selecting a comparator in accordance with some aspects; and

[0015] [Figure 6] FIG. 6 is a plot of example parametric and non-parametric receiver operating characteristics (ROC) for the lowest 2 mean comparator in accordance with some embodiments. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0016] Detailed Description The present inventors have recognized and appreciated that to improve the effectiveness of cancer surgery, it is desirable to have a system that can identify abnormal cells from surgical site images, such as residual cancer features that are less than mm to mm in size and / or that are not easily identified by existing visual inspection. Such techniques may use thresholding to determine whether one or more images collected from a patient's surgical site contain abnormal and / or cancerous cells and therefore may require further resection or removal. An initial set of images of the surgical site may be acquired and used to generate a custom patient threshold. As described herein, for example, the initial images may be aggregated into a representative intensity value and used to calculate a patient-specific threshold. Prior techniques have not allowed image analysis systems to be fine-tuned specifically to a particular patient. As further described herein, various pixel analysis techniques are provided that analyze pixels of surgical site images and utilize statistical learning to further fine-tune the threshold calculation, in accordance with various aspects.

[0017] 1 is a schematic diagram showing an overview of components of an exemplary system 100 for identifying residual cancer cells after a surgical procedure according to an embodiment of the present application. An operator 180 may issue commands and / or queries related to the image analysis system 101 using one or more input devices 105 that interact with a user interface (UI) 109. The image analysis system may be implemented using computer-based hardware elements 103 that are specifically programmed to process one or more images to identify residual cancer features in the images. The output of the image analysis system may be presented to an operator in a surgical room environment on a display device 107.

[0018] In some embodiments, images of the patient's anatomy are captured by one or more surgeons 182 during or after a surgical procedure on a patient on a surgical bed 184. The patient's anatomy may include, for example, both internal and / or external structures of the patient's head, neck, chest, abdomen, arms, hands, legs, feet, another suitable structure, or a portion thereof, or a combination thereof. The patient's anatomy may be an area where the surgeon determines whether there may be a tumor, cancerous tissue, abnormal tissue such as a skin lesion, wound, and / or any other structure that may contain abnormal tissue. The images may be captured in-situ at the surgical site 186 where the tumor is excised using medical imaging equipment and transmitted in real time to the image analysis system 101 via the data connection 111. The system 100 used in this scenario may provide real-time feedback regarding the size and location of residual cancer cells remaining at the surgical site after the excision surgical procedure has been performed. In some cases, the image analysis results provided to the operator 180 may facilitate the surgeon 182 performing a residual cancer cell removal procedure on the patient, and may provide further confirmation from the image analysis system that all remaining cancer cells at the surgical site 186 have been completely removed, without the need for a secondary, separate residual cancer cell removal surgical procedure.

[0019] FIG. 2 is a block diagram of a cancer cell detection system 200, which is an example of the system 100 of FIG. 1. Some of the components are the same, and therefore the same reference numbers are used. In the example in FIG. 2, the image analysis system 201 includes an interconnect 220 that is used to interface and cooperate with various components in the system. The interconnect 220 may include one or more digital buses or any other type of suitable interconnect that allows high-speed communication of digital data between the interconnected components. In some embodiments, the interconnect 220 interacts with a processor 231 for processing image data received from the image receiving unit 213, and obtains image analysis results from the processor 231 to provide image analysis results to the external display 107 for the operator 180. In some embodiments, the interconnect 220 may communicate non-image data to the processor. For example, the processor 231 may receive programming instructions from the interconnect 220 to execute a series of methods on the processor. In some embodiments, the interconnect 220 communicates with a memory 233 to store and read image data and non-image data to the processor. For example, memory 233 may be a computer memory that stores program instructions executable by processor 231 that, when executed by processor 231, implement one or more methods for identifying features of residual cancer cells from an image. It should be understood that any suitable computer-readable storage medium may be used as memory 233 for storing image data and program instructions in the image analysis system.

[0020] According to some embodiments, the processor 231 may additionally communicate with an image analysis engine 240 to process image data and identify residual cancer cell features. The image analysis engine 240 may include one or more units, each of which is specialized to execute instructions for image enhancement and analysis. Image enhancement may also help to remove noise and artifacts not associated with cancer cell features and improve the signal-to-noise ratio of cancer cell signals to background. Image enhancement may be performed using one or more methods selected from the group including, for example, Canny filtering, Sobel filtering, Kayyali filtering, Principle Curvature-Based Region Detection, Features from Accelerated Segment Testing, Forstner filtering, Laplacian of Gaussians, Tophat filtering, Difference of Gaussians, maximally stable extremum region, and determinant of Hessian.

[0021] In some embodiments, the units in the image analysis engine 240 may be integrated on a single chip together with the processor 231 to reduce packaging size, power consumption, and manufacturing costs. In some other embodiments, the image analysis engine 240 may be implemented on a discrete processing chip specialized for performing high-speed image processing, such as a graphics coprocessor or a field programmable gate array (FPGA). Additionally, one or more image analysis features may be incorporated within a medical imaging device (e.g., device 110).

[0022] 2, the display controller 250 receives image analysis results from the processor 231 via the interconnect 220 and provides one or more processed images indicating the size and location of the detected cancerous features to the external display 107 for the operator 180. In some embodiments, an overlay generation unit 253 generates an overlay and / or highlighting based on the image analysis results to be provided to the display controller such that an overlay and / or highlighting is presented to the operator 180 on the display 107 to highlight the identified cancerous features.

[0023] In the system in FIG. 2, the medical imaging device 110 is used to capture intraoperative images of the patient's surgical site and transmit the captured image data to the image receiving unit 213 via the data connection 111. The captured image data may be stored in the memory 233 and subsequently processed by the processor 231 and the image analysis engine 240 to identify residual cancer features in the image data. It should be understood that multiple image data may be captured by the medical imaging device, stored in the memory 233, and processed by the processor 231 and the image analysis engine 240 at substantially the same time for output to the display 107 to provide real-time feedback to the operator of the size and location of the residual cancer cells. For example, the medical imaging device 110 may capture images at a particular video rate (e.g., 10 frames per second, 12 frames per second, 24 frames per second, or 30 frames per second) and transfer the video rate images at substantially the same rate as real-time video features on the display 107 for processing and output to facilitate the surgeon's operation to remove the identified cancer cells.

[0024] In some embodiments, the medical imaging device 110 may be a handheld imaging device. In one non-limiting example, the handheld imaging device may be a handheld fluorescent imaging device that includes a photosensitive detector that is sensitive to a fluorescent signal corresponding to photons emitted from the fluorescence of a certain fluorescent imaging agent with which cancer cells are labeled. In some embodiments, the imaging device may also include an excitation source configured to emit an excitation wavelength of a specific fluorescent imaging agent into the object being imaged or into the surgical bed. A description of an example of a handheld fluorescent imaging device may be found in U.S. Patent Application No. 14 / 211,201, filed March 14, 2014, entitled "MEDICAL IMAGING DEVICE AND METHODS OF USE," the disclosure of which is incorporated herein by reference in its entirety. Of course, although a handheld device is mentioned above, the systems and methods described herein are not limited to use only with handheld devices. Instead, they may be implemented on any suitable imaging and / or analysis system, regardless of size scale.

[0025] The image captured by the handheld imaging device according to some embodiments may include pixels with brightness or intensity levels that substantially correspond to the number of fluorescent photons emitted from the tissue portion being imaged at the surgical site. The intensity value of each pixel on the captured image is substantially proportional to the number of fluorescently labeled cancer cells in the tissue portion within the total field of view of the imaging device. The size and location of the pixel on the captured image corresponds to the size and location of the tissue portion relative to the total field of view of the imaging device. In some embodiments, the handheld imaging device may include a field of view of 2.5 cm, 1.3 cm, and / or other fields of view, as can be understood by those skilled in the art. The image transmitted from the medical imaging device 110 to the image receiving unit 213 may include a high number of pixels in each of two orthogonal directions in the array, such that the size of the tissue portion corresponding to one pixel, or in other words the field of view corresponding to a single pixel, is not larger than the size that represents the desired spatial resolution of the imaging device. Without wishing to be bound by theory, a typical cancer cell may be on the order of approximately 15 μm in diameter. The medical imaging device may be configured such that the field of view of each pixel may be equal to or greater than about 1 μm, 2 μm, 3 μm, 4 μm, 5 μm, 10 μm, 15 μm, 30 μm, or any other desired size. In addition, the field of view of each pixel may be less than about 100 μm, 50 μm, 40 μm, 30 μm, 20 μm, 10 μm, or any other desired scale.

[0026] In one particular embodiment, the field of view per pixel may be between about 5 μm and 100 μm, inclusive. In another embodiment, the field of view per pixel may be between about 5 μm and 50 μm, inclusive. Having a field of view for each pixel that is as large as or smaller than each cancer cell leads to enhanced resolution and contrast for the features of labeled cancer cells in the surgical site image, since each cancer cell is covered by one or more full pixels with high fluorescent intensity against a background of much lower intensity. Such high resolution and contrast allows the system to identify small features with only a few cancer cells from the surgical site image that are not easily identified by existing means of visual inspection. Of course, pixels with both larger and smaller fields of view than those noted above are also contemplated, as the disclosure is not so limited. Furthermore, such imaging capabilities may also enable the detection of certain types of cancer cells with specific shapes and / or geometries. It is understood that since particular cancer cells have particular fundamental geometries, such geometries may be stored in memory, compared to the sampled geometries, and detected by the system.

[0027] In some embodiments, each image captured by the medical imaging device 110 includes an array of pixels, with each pixel having a digitized intensity value representing the concentration of contrast agent in the pixel's corresponding field of view. In one example, the intensity value of each pixel is sampled and digitized with a resolution of 12 bits. In other examples, the digitization resolution may be other resolutions such as 8 bits, 14 bits, 16 bits, or any resolution suitable for recording and transmitting images. The captured images are transmitted from the medical imaging device 110 to the image receiving unit 213 via a data connection 111. In some embodiments, raw uncompressed image data is transmitted to ensure image integrity, although it should be understood that any suitable data compression technique and format may be used to enhance the image data transmission rate from the medical imaging device 110.

[0028] In addition to image data, the data connection 111 may also transmit non-image data from the medical imaging device 110 to the image analysis system 201. In one embodiment, image scaling information may be transmitted along with each surgical site image for the image analysis system to convert pixel sizes into length units such as mm or μm. In some embodiments, the converted size of one or more groups of pixels may be compared to a known threshold based on the size and / or geometry of cancer cells to identify one or more groups of pixels as cancer cells in the surgical site image.

[0029] Additionally, the data connection 111 may transmit non-image data from the image analysis system 201 to the medical imaging device 110. Such non-image data may be used by the operator 180, for example, to control the medical imaging device performing image capture, transmission, adjustment of magnification / field of view, and / or depth of focus of an optical system within the medical imaging device to provide a full survey of the surgical bed for residual cancer cells after the initial tumor resection.

[0030] In some embodiments, an operator 180 as shown in FIG. 1 may use one or more input devices 105 to interact with the image analysis system. In the embodiment in FIG. 2, the input device 105 is coupled with an interconnect 220 that sends commands and queries from the operator to various components of the system. The input device may include a keyboard for typing text, a mouse for interacting with one or more screen objects on the UI 109 on the display 107 as shown in FIG. 1. For example, the operator may type in annotation text that accompanies the received surgical site image or the analyzed residual cancer cell results. The operator may use the input device to reconfigure and interact with the UI 109 to adjust settings on the medical imaging device and / or image analysis settings, such as one or more thresholds depending on the type of cancer cell being treated. In some embodiments, the input device may also include a stylus-sensitive or finger-sensitive touch screen that is external to or integrated with the display 107. It should be understood that any suitable human interface device may be used by the operator 180 to provide commands and queries to the system. In some embodiments, the input device may include gesture control or voice control. In other embodiments, the input device may additionally provide audio, visual, or tactile feedback to the operator indicating one or more aspects related to the input command and / or query.

[0031] 2 may include a UI generation engine 251 employed to generate screen objects to a display controller 250 that are presented on a display 107 as a UI for operator interaction. The UI may be generated based on instructions stored in memory 233 and may be dynamically adjusted based on user input from input device 105.

[0032] 2 may also include an external storage 235 connected to the interconnect 220 that is configured to store both image data and non-image data, e.g., for archival purposes. In some embodiments, program instructions may be read from or stored on the external storage 235.

[0033] As explained above, aspects of the present application relate to methods for detecting cancer cells, such as cancer cells remaining at a surgical site after removal of a large portion of a tumor to facilitate complete removal of residual cancer cells. In particular, aspects relate to thresholding algorithms that determine a threshold value used in a cancer cell detection process that is tailored to each patient. As a general introduction that is not intended to be limiting, the technique may include one or more of the following steps. The technique is described in an exemplary order for clarity, but may be performed using a different order and / or combination of steps. First, an image of the surgical site may be processed (e.g., background correction and / or flattening). Second, the technique may search for areas above a threshold based on normal tissue. Third, it may be determined whether a feature(s) exceeds a smaller size threshold. If the feature satisfies the second and third steps, the feature is highlighted. Fourth, if no features are present, contrast enhancement may be used to modify all or part of the image. Fifth, the technique may search for features above a certain threshold in the contrast-enhanced view. Sixth, it may be determined whether the feature meets one or more criteria, such as a size criterion and a contrast criterion. If the feature meets the fifth and sixth steps, the feature is highlighted.

[0034] 3 is a flow chart illustrating an example process 300 for detecting residual cancer cells according to some embodiments. After the process begins at block 302 in FIG. 3, images are captured at the surgical site at block 304, for example, using a handheld imaging device as described above. In some embodiments, the surgical site images may be still images and / or still videos containing images captured at a specific video rate. At block 306, an image enhancement process may optionally and additionally be performed to enhance the images collected at block 304 in process 300.

[0035] At block 308, the collected image portions are selected based on a comparison between the characteristics of the pixels in the image portions, where a characteristic threshold is predefined. In one non-limiting example, the image portions may be selected for further analysis after a determination that the intensity parameters of each pixel in the image portion exceed a threshold corresponding to a simple brightness threshold expected for a given type of cancer cells labeled with a contrast agent. One exemplary goal for performing a selection process based on thresholding may be to exclude background pixels representing light emission from healthy tissue from further analysis. The inventors have recognized that in some embodiments, even if a fluorescent contrast agent is configured to selectively bind to cancer cells, a finite fluorescent intensity may still appear in pixels corresponding to non-cancerous cell regions in the surgical site. As further described herein, the technique may include one or more thresholding steps that apply a threshold to the pixels. For example, feature detection (e.g., small cell feature detection) may involve a first thresholding step that applies a threshold to an image(s) (e.g., processed image(s)), and a second thresholding step that applies a threshold to (e.g., local) features in the original image(s) that exceed the threshold of the first thresholding step.

[0036] In some cases, the finite background intensity may be due to background noise in the photodetector or in the circuitry of the medical imaging device. The finite background intensity may also arise from the fluorescent signal emitted from a small amount of contrast agent due to non-specific binding in healthy, non-cancer cell areas. Therefore, in some embodiments, relying solely on simple threshold determination may not be sufficient to prevent false positive identification of cancer cell areas, to eliminate background tissue, and to make only cancer cells stand out from the image.

[0037] Subsequently, at block 310, the collected image portion selected at block 308 is analyzed to identify features of one or more groups of remaining cancer cells based on size against the background. For example, features in the selected image portion may be discarded or eliminated if they are determined to be outside of a size range associated with the cancer cell type targeted for removal. It should be understood that feature size, as used herein, refers to the correctly scaled size dimension of the area of ​​tissue on the surgical site as measured, for example, in mm or μm, that corresponds to the feature on the surgical site image.

[0038] At block 312, one or more groups of remaining cancer cells identified at block 310 are further filtered using a local based contrast method, described in detail below. In some embodiments, a local contrast is determined for each group of remaining cancer cells features based on a comparison between the signal intensities of pixels substantially within and outside the feature. Only features with contrast above a predetermined threshold are identified as cancer cells to distinguish against background and reduce false positive identifications.

[0039] The cancer cells identified according to blocks 308 , 310 and 312 are presented to operator 180 on a display (such as display 107 in FIG. 1) in block 314 before the residual cancer cell identification process ends in block 316 .

[0040] In some embodiments, each of the processing blocks 308-312 and 306 as illustrated in the block diagram in FIG. 3 may be applied individually or selectively in any suitable ordered combination group depending on the type of tissue being targeted and / or the residual cancer cell type.

[0041] In some embodiments, the example process 300 shown in FIG. 3 may be implemented by the system 200 in FIG. 2 to continuously process images received by the image receiving unit 213 from a medical imaging device.

[0042] For example, according to one embodiment, the system may be employed to detect specific cancer and precancerous cells. In one embodiment, the system may store parameters indicative of the specific size and geometry of an indicated cancer, as well as precancerous indications, including, but not limited to, invasive ductal carcinoma, invasive lobular carcinoma, ductal carcinoma in situ, ovarian cancer, metastatic ovarian cancer, brain metastasis, peritonitis carcinomatosa, esophageal cancer, Barrett's esophagus, colorectal polyps, colon cancer, melanoma, basal cell skin cancer, squamous cell skin cancer, prostate cancer, lung cancer, sarcoma, endometrial hyperplasia, endometrial cancer, and cervical dysplasia, among other types.

[0043] As shown in block 312 of method 300, the threshold value can be an important aspect of the detection process, since the image analysis system uses the threshold value to determine whether the image contains residual cancer cells. In some embodiments, the image portion selected in block 308 can be based on the brightest continuous feature in the image. For example, the image analysis system can select a specific diameter circle (e.g., 500um, 1000um) around the area with the brightest continuous feature. The image analysis system can use the pixels in the selected area to compare the associated feature to a threshold value. For example, the image analysis system can compare the lowest pixel intensity value in the selected area to a threshold value to make an automatic determination of whether the associated feature in the image contains abnormal tissue (e.g., cancerous tissue) that requires further resection. This is referred to as TB1000um below.

[0044] FIG. 4 is a flow diagram illustrating an example process 400 for calculating a patient-specific threshold value according to some embodiments.

[0045] Process 400 begins at block 402. At block 404, a medical imaging device (e.g., medical imaging device 110 in FIG. 1) produces a set of surgical site images for a patient. In general, at blocks 406-412, an image analysis system (e.g., image analysis system 101 in FIG. 1) collects and analyzes the set of surgical site images collected at block 404 to determine patient-specific thresholds that may be used to detect abnormal tissue in the patient (e.g., using new images captured of the patient's surgical site using the medical imaging device). In some embodiments, the set of surgical site images are images of healthy tissue. However, embodiments in which the images are images of abnormal tissue and / or images that include both healthy and abnormal tissue are also contemplated.

[0046] At block 406, the image analysis system uses a pixel comparator to calculate a representative intensity metric for each surgical site image collected at step 404 to generate a set of representative intensity metrics. The representative intensity metric represents pixel intensity values ​​of the associated images. The pixel comparator may be, for example, a mean comparator, a percentile comparator, a fixed comparator, and / or other statistical comparators, as further described herein. For example, in some embodiments, the comparator may be a mean comparator and the representative intensity metric may be the average of pixel intensity values ​​of the associated images, as further described herein.

[0047] At block 408, the image analysis system selects a portion of the representative intensity set of metrics based on a predetermined criterion. In some embodiments, the predetermined criterion is used to eliminate one or more of the representative intensity metrics from the representative intensity set of metrics. For example, the image analysis system may be configured to select a certain minimum value of the representative intensity metric for a particular comparator, as described further herein.

[0048] At block 410, the image analysis system averages a selected portion of the representative intensity metrics to calculate an average intensity value.

[0049] At block 412, the image analysis system calculates a threshold value for the patient based on the average intensity value and an adjustment factor associated with the pixel comparator. For example, the adjustment factor may be a multiplier used to multiply the average intensity value to determine the threshold value. Processing ends at block 414. Once calculated, the threshold value may be used by the image analysis system to detect abnormal cells for a particular patient.

[0050] Referring to block 406, the image analysis system uses a pixel comparator to calculate a representative intensity metric for each surgical site image. In some embodiments, the comparator is a maximum comparator that calculates a representative intensity value for each surgical site image by calculating a maximum pixel intensity value for each surgical site image. In some embodiments, the comparator is a minimum comparator that calculates a representative intensity value for each surgical site image by calculating a minimum pixel intensity value for each surgical site image. In some embodiments, the comparator is an average comparator that calculates a representative intensity value for each surgical site image by calculating an average of the pixel intensity values ​​for the surgical site images. In some embodiments, the comparator is a 25th percentile comparator that calculates a representative intensity value for each surgical site image by calculating the 25th (twenty-fifth) percentile of the pixel intensity values ​​for the surgical site images. In some embodiments, the comparator is an invariant comparator that uses an invariant threshold. The threshold may be calculated, for example, using the Youden point on a ROC (e.g., a ROC that does not use any patient-specific comparator statistics). In some embodiments, the invariant threshold may be the same for all patients and / or a subset of patients. In some embodiments, the comparator is a median comparator that calculates a representative intensity value for each surgical site image by calculating the median pixel intensity value for each image.

[0051] Referring to block 408, the image analysis system selects a portion of the representative intensity set of metrics based on predetermined criteria. For example, the image analysis system may select a certain representative intensity metric, such as a certain highest metric or a certain lowest metric. For example, if block 404 collects six images of the surgical site, then in block 406 the image analysis system may select six different representative intensity metrics (e.g., average, or 25% confidence interval). thpercentiles), one for each image. The predetermined criteria may cause the image analysis system to select, for example, a certain lowest representative intensity value. For example, if block 404 collects six images, the predetermined criteria may cause the image analysis system to select all six representative intensity values, five of the representative intensity values, four of the values, etc. (up to two of the representative intensity values, or even one). For example, a minimum 6 average value comparator may relate to a predetermined criterion that selects the lowest six of the representative intensity values ​​calculated for the set of collected images using an average value function. As another example, a minimum 2 average value comparator may relate to a predetermined criterion that selects the lowest two of the representative intensity values ​​calculated for the set of collected images using an average value function. As a further example, the predetermined criteria may cause the image analysis system to perform an averaging of various comparators, such as an averaging of the average values ​​of the second and third lowest cavity initialization images. In some embodiments, the predetermined criteria is associated with a comparator (e.g., a specific comparator, such as a mean value comparator, is associated with a predetermined criteria that identifies how to select a portion of the representative intensity metrics).

[0052] Referring to block 410, the image analysis system averages a selected portion of the representative intensity metrics to calculate an average intensity value. In some examples, the image analysis system calculates the average by summing the representative intensity metrics and dividing the sum by the number of representative intensity metrics. Those skilled in the art will appreciate that other techniques may be used to combine the representative intensity metrics into an average intensity value. For example, root mean square, geometric mean, median, sum, and / or the like may be used to combine the representative intensity metrics.

[0053] Referring to block 412, the image analysis system calculates a threshold value for the patient based on the average intensity values ​​calculated in block 410. The image analysis system may also use an adjustment factor associated with the pixel comparator to calculate the threshold value. For example, the adjustment factor may be a multiplier associated with the pixel comparator (e.g., a multiplier with an integer or decimal value ranging from 1.0 to 25, such as, for example, from 1.0 to 3.5, from 18 to 21, and / or the like). Each pixel comparator may be associated with the same and / or different multipliers. As described above in conjunction with step 408, each comparator may be associated with a predetermined criterion. Those skilled in the art will appreciate that the comparators and adjustment factors described herein are for illustrative purposes only and are not intended to be limiting.

[0054] Once the image analysis system calculates the threshold, the image analysis system may use the threshold to identify residual abnormal cells in an image of a patient's surgical site. For example, process 300 described in conjunction with FIG. 3 may be used to analyze a patient's surgical site, where the calculated threshold is used in block 312 to identify one or more groups of residual cancer cells.

[0055] FIG. 5 is a flow chart illustrating an example process 500 for selecting a comparator to use for threshold calculation according to some embodiments. Process 500 begins at block 502. At block 504, a medical imaging device collects a set of training images of a patient's surgical site. At block 506, an image analysis system runs a plurality of comparators (e.g., mean comparators, percentile comparators, invariant comparators, etc.) on the training set of surgical site images and calculates a set of result values ​​for each of the comparators. At block 508, the image analysis system analyzes the set of result values ​​for each of the plurality of comparators. At block 510, the image analysis system selects a comparator from the plurality of comparators based on the analysis performed at block 508. Process 500 concludes at block 512. As described in conjunction with FIG. 4, the selected comparator may be used to calculate a patient threshold value.

[0056] Referring to block 504, in some embodiments, the set of training images may be of a particular patient's surgical site and / or of a group of patients. For example, surgical site images may be collected from a patient population. In some embodiments, patients may be pre-processed so that patients are excluded for various reasons such as due to imaging times outside the recommended imaging window. For each patient, a predetermined number (e.g., 2, 6, 10, etc.) of cavity initialization images may be generated after removal of the gross lump. Additional images, image processing, and / or surgery may be performed to refine the data set. For example, the image analysis system may generate an oriented image of the first surgical cavity. Standard of care shaves may be removed after each orientation is imaged, and additional images may be generated and used to assess whether additional tissue should be removed. To assess whether additional tissue may be used, the normal tissue threshold may be periodically re-optimized during the data collection procedure. After surgery, histopathology can be used to determine whether the image actually contains residual cancer.

[0057] Referring to block 506, once the full set of data is collected in block 504, the image analysis system runs a number of comparators against the surgical site image training set and calculates a set of result values ​​for each of the comparators. The system may be configured to run one or more analyses of the comparators to determine the result values. The set of result values ​​may include, for example, a receiver operating characteristic (ROC) curve. Those skilled in the art will appreciate that a ROC curve is a graphical plot that illustrates the diagnostic ability (e.g., abnormal cells and / or cancerous cells, or not) of a classifier system as its discrimination threshold (e.g., thresholds described herein) is varied. ROC statistics may include, for example, the total area under the ROC curve (AUC), AUC after leave one out (LOO) cross-validation (e.g., referred to as AUC LOO), and / or the like. AUC is generally a summary statistic that represents the probability that a classifier ranks a randomly selected positive instance higher than a randomly selected negative instance. Those skilled in the art will understand that LOO cross-validation is a model validation technique used to assess how well statistical analysis results generalize to independent data sets, where certain findings are used as a validation set and the remaining findings are used as a training set. Although the techniques described herein can be configured to use ROC and LOO cross-validation, those skilled in the art will understand that other statistical techniques can also be used without departing from the spirit of the techniques described herein.

[0058] The set of result values ​​may include result values ​​determined based on the Youden index. The Youden index is defined for every point of the ROC curve with a value ranging from -1 to 1 (where 0 is a useless test and 1 is an ideal test). For example, the result values ​​may include the optimized Youden points for each AUC. Additionally or alternatively, the result values ​​may include Wilcoxon results. Those skilled in the art will understand that the Wilcoxon rank sum test is a test that can be used to determine whether two independent samples are selected from populations with the same distribution. For a given comparator, the technique may obtain the ROC by scanning the values ​​multiplied by that comparator. The multiplier may be calculated, for example, by determining which gives the highest Youden point.

[0059] In some embodiments, the set of result values ​​may include, for each comparator, the number of true negative results, true positive results, false negative results, false positive results, and / or some combination thereof. Exemplary result values ​​may include a sensitivity value, a specificity value, or both. In some examples, for each comparator, a result value (e.g., a statistical value) may be calculated for that comparator at its Youden point.

[0060] In some embodiments, the full set of data used to calculate the result value can be divided into a cancer positive group and a cancer negative group, for example after normalization by a predefined thresholding value.These two separate positive and negative groups can be evaluated for normality by assessing the goodness of fit to a log-normal distribution.The log-normal distribution can be used to generate a parameterized ROC curve for the normal tissue-containing image and the cancer tissue-containing image.The best optimization Youden point from the parameterized ROC curve can be used to select the comparator and related aspects (for example, predetermined criteria and adjustment factors).

[0061] In some embodiments, other comparators may be used in addition to or instead of the mean and percentile comparators. The mean and percentiles of the images may be chosen to represent normal tissue because they are robust and / or statistically efficient. For example, the mean comparator may incorporate more pixel values ​​from each image, which may make it more likely (e.g., compared to other comparators) to arrive at a number representative of normal tissue for a given patient. As another example, percentiles (e.g., 25 th percentiles) may avoid pixels that may contain abnormal tissue or tumors, which may artificially raise the threshold and lead to false negatives. In some instances, an invariant comparator that uses an invariant threshold may not account for patient to patient variations in the image signal.

[0062] In some embodiments, cross-validation is used to determine the result value. For example, cross-validation can be used to assess how well a statistical model translates to an independent dataset (e.g., a different dataset other than the dataset used to train the model, such as a real-time dataset captured during surgery). For example, cross-validation can be used to define a dataset and test the model during the training phase to avoid problems such as overfitting. Such validation can be used for predictive modeling. In some embodiments, LOO cross-validation can be performed for each comparator as described above, where one of the findings (e.g., an image and / or associated value) is used as a validation set and the remaining findings are used as a training set. For example, each value in a set of images (e.g., a single image) can be iteratively eliminated, and the remaining values ​​are used as a training set. In some examples, these values ​​can be fitted using logistic regression, and the output of the logistic regression can be used to determine the probability that the eliminated value is either positive or negative for pathology. Thus, for each iteration, the eliminated value can serve as a validation set and receive a probability value. The array of probabilities of rejected values ​​along with the corresponding pathology values ​​can be used to generate the ROC and their AUC values.

[0063] Referring to block 508, in some embodiments, the image analysis system may analyze the set of result values ​​by comparing them to the values ​​of AUC for the full set of data, LOO cross-validation, and / or both. For example, if the image analysis system determines that a particular percentage decrease is observed for a comparator compared to AUC (e.g., a decrease of more than 5% in AUC and / or AUC LOO), that comparator may be excluded from use in calculating the threshold. For example, such a comparator may be excluded based on a belief that the AUC for the full set of data is a poor predictor of future performance.

[0064] In some embodiments, the data can be evaluated for normality to refine the comparator result value. For example, after an initial investigation of thresholding approaches is performed on a data set (e.g., a set of images for different patients), the data can be evaluated for normality, such as by fitting the data to a log-normal distribution. The normality of a data set can be assessed based on the linearity of a normal probability plot.

[0065] In some embodiments, a result value may be calculated for the comparator without analyzing the dataset for outliers. In some embodiments, a normal probability plot of the natural logarithm of the ratio of TB1000um (the minimum value in a 1000um diameter circle around the brightest continuous feature in the image, as explained above) to the normalized set of values ​​may be calculated to search for outliers in the training dataset for both normal and cancerous tissue-containing images. In some embodiments, after removal of image outliers, the comparator search may be recalculated to determine an updated result value. In some embodiments, the image analysis system evaluates the comparator with the best Youden score. In some embodiments, the comparator may be evaluated using a parametric model.

[0066] 6 is a plot of example parametric and non-parametric ROCs according to some embodiments for a comparator. In FIG. 6, the x-axis is false positives ranging from 0 to 1 and the y-axis is sensitivity ranging from 0 to 1. The parametric ROC is shown as 602 and the non-parametric ROC is shown as 604. In some instances, such similarity can be seen because the fit is visually good.

[0067] Referring to block 510, the image analysis system selects a comparator from the plurality of comparators based on the analysis performed in block 508. After analyzing the result values, as described in conjunction with block 506, the system may select the top performing comparator. In some embodiments, the image analysis system may perform further statistical analysis of the selected comparator. For example, in some embodiments, the image analysis system uses the selected comparator to perform a Wilcoxon rank sum test on a data set (e.g., the data set used to generate the result values ​​for the comparator).

[0068] Depending on the nature of the computing device, one or more additional elements may be present. For example, a smartphone or other portable electronic device may include a camera and be able to capture still or video images. In some embodiments, the computing device may include sensors such as a global positioning system (GPS) to sense location, and inertial sensors such as a compass, an inclinometer, and / or an accelerometer. An operating system may include utilities to control these devices, capture data from them, and make the data available to applications running on the computing device.

[0069] As another example, in some embodiments, a computing device may include a network interface for implementing a personal area network. Such an interface may operate according to any suitable technology, including, for example, Bluetooth, Zigbee, or 802.11 ad-hoc modes.

[0070] Having thus described several aspects of at least one embodiment of this invention, it is to be appreciated that various alterations, modifications, and improvements will readily occur to those skilled in the art.

[0071] Such changes, modifications, and improvements are intended to be part of this disclosure and are intended to be within the spirit and scope of the invention. Moreover, although advantages of the present application are indicated, it should be understood that not every aspect of the invention necessarily encompasses every advantage described. Some aspects may not implement every feature described herein, and in some cases, as advantageous. Consequently, the above description and drawings are by way of example only.

[0072] The above-described aspects of the present application may be accomplished in any of a myriad of ways. For example, the disclosed methods may be applied to imaging methodologies beyond simple fluorescence, including MRI, ultrasound, mammography and other X-ray techniques, Raman, two-photon microscopy, and others.

[0073] For example, the embodiments may be implemented using hardware, software, or a combination thereof. When implemented in software, the software code may be executed on any suitable processor or group of processors, whether provided on a single computer or distributed among multiple computers. Such processors may be implemented as integrated circuits, or one or more processors may be implemented in integrated circuit components, including commercially available integrated circuit components known in the art under names such as CPU chips, GPU chips, microprocessors, microcontrollers, or coprocessors. Alternatively, the processor may be implemented in custom circuitry, such as an ASIC, or in semi-custom circuitry resulting from constructing a programmable logic device. Yet as a further alternative, the processor may be part of a larger circuit or semiconductor device, whether commercially available, semi-custom, or custom. As a specific example, some commercially available microprocessors have multiple cores, such that one or several of the multiple cores may constitute a processor. However, the processor may be implemented using circuitry in any suitable format.

[0074] Further, it should be understood that the computer may be integrated into any of a number of forms, such as a rack-mounted computer, a desktop computer, a laptop computer, or a tablet computer, etc. In addition, the computer may be embedded in devices not generally considered to be computers but having suitable processing capabilities, including a personal digital assistant (PDA), a smart phone, or any other suitable portable or fixed electronic device.

[0075] A computer may also have one or more input and output devices. These devices may be used, among others, to present a user interface. Examples of output devices that may be used to provide a user interface include a printer or display screen for visual presentation of output, and a speaker or other sound generating device for audible presentation of output. Examples of input devices that may be used in a user interface include keyboards and pointing devices, such as mice, touchpads, and digitizer tablets. As another example, a computer may receive input information through speech recognition or other audible format input information. In the described embodiments, the input / output devices are described as being physically separate from the computing device. However, in some embodiments, the input and / or output devices may be physically integrated into the same unit as the processor or other elements of the computing device. For example, the keyboard may be implemented as a soft keyboard on a touch screen. Alternatively, the input / output devices may be entirely disconnected from the computing device and functionally integrated through a wireless connection.

[0076] Such computers may be interconnected in any suitable manner by one or more networks, including as a local area network or a wide area network, such as a corporate network or the Internet, etc. Such networks may be based on any suitable technology and operate according to any suitable protocol, and may include wireless networks, wired networks, or fiber optic networks.

[0077] Also, the various methods or processes outlined herein may be coded as software executable on one or more processors employing any one of a variety of operating systems or platforms. In addition, such software may be written using any of a number of suitable programming languages ​​and / or programming or scripting tools, and may also be compiled as executable machine code or intermediate code that runs on a framework or virtual machine.

[0078] In this regard, the present invention may be embodied as a computer-readable storage medium (or a plurality of computer-readable media) (e.g., a computer memory, one or more floppy disks, compact disks (CDs), optical disks, digital video disks (DVDs), magnetic tapes, flash memories, circuit configurations in field programmable gate arrays or other semiconductor devices, or other tangible computer storage media) encoded with one or more programs that, when executed on one or more computers or other processors, perform the methods for performing the various aspects of the present invention as described above. As is evident from the above examples, a computer-readable storage medium may retain information for a sufficient time to provide computer-executable instructions in a non-transitory form. Such computer-readable storage medium(s) may be transportable such that the program(s) stored thereon may be loaded onto one or more different computers or other processors that perform the various aspects of the present application as described above. As used herein, the term "computer-readable storage medium" covers only computer-readable media that may be considered to be an article of manufacture (i.e., product) or machine. Alternatively, or in addition, the present invention may be embodied as a computer-readable medium other than a computer-readable storage medium, such as a propagating signal.

[0079] The terms "code," "program," or "software" are used herein in a general sense to refer to any type of computer code or set of computer-executable instructions that may be employed to program a computer or other processor to perform various aspects of the present application as described above. In addition, in accordance with one aspect of this embodiment, it should be understood that one or more computer programs that, when executed, perform the methods of the present application need not reside on a single computer or processor, but may be distributed in a modular manner among a number of different computers or processors that perform various aspects of the present application.

[0080] Computer-executable instructions may be executed by one or more computers or other devices in many forms, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. Typically the functionality of the program modules may be combined or distributed as desired in various embodiments.

[0081] Also, the data structures may be stored in the computer-readable medium in any suitable form. For simplicity of explanation, the data structures may be shown as having fields that are related through locations in the data structure. Such relationships may also be achieved by specifying storage for the fields at locations in the computer-readable medium that convey the relationships between the fields. However, any suitable mechanism may be used to establish relationships between information in the fields of the data structures, for example through the use of pointers, tags, or other mechanisms that establish relationships between data elements.

[0082] Various aspects of the present application may be used alone, in combination, or in various arrangements not specifically described in the above described embodiments, and therefore are not limited in their application to the details and arrangements of components set forth in the above description or illustrated in the drawings. For example, aspects described in one embodiment may be combined in any manner with aspects described in other embodiments.

[0083] The present invention may also be embodied as a method, for which examples are provided. The acts performed as part of the method may be ordered in any suitable manner. Consequently, embodiments may be constructed in which acts are performed in a different order than as described, and may include performing some of the acts simultaneously, even though the illustrative embodiment shows them as sequential acts.

[0084] The indefinite articles "a" and "an" as used herein and in the claims, unless clearly indicated to the contrary, should be understood to mean "at least one."

[0085] The phrase "and / or" as used herein and in the claims should be understood to mean "either or both" of the elements so conjoined, i.e., elements that are conjunctively present in some cases and disjunctively present in other cases. Multiple elements listed with "and / or" should be interpreted similarly, i.e., as "one or more" of the elements so conjoined. Other elements may optionally be present other than the elements explicitly identified by the "and / or" clause, whether related or unrelated to those elements explicitly identified. Thus, as a non-limiting example, a reference to "A and / or B," when used in conjunction with open-ended language such as "comprising," can refer in one embodiment to A only (optionally including elements other than B); in another embodiment to B only (optionally including elements other than A); in another embodiment to both A and B (optionally including other elements); and so forth.

[0086] The phrase "at least one" as used herein in the specification and claims in connection with a list of one or more elements is understood to mean at least one element selected from any one or more of the elements in the list of elements, but should not necessarily include at least one of each and every element specifically listed within the list of elements, and should not exclude any combination of elements in the list of elements. This definition also allows that elements other than the elements specifically identified may optionally be present within the list of elements to which the phrase "at least one" refers, whether related or unrelated to those elements specifically identified. Thus, as a non-limiting example, "at least one of A and B" (or in other words "at least one of A or B" or in other words "at least one of A and / or B") can refer, in one embodiment, to at least one (optionally including more than one) A and no B (and optionally including elements other than B); in another embodiment, to at least one (optionally including more than one) B and no A (and optionally including elements other than A); in another embodiment, to at least one (optionally including more than one) A and at least one (optionally including more than one) B (and optionally including other elements); and so forth.

[0087] The use in the claims of ordinal terms, e.g., "first," "second," "third," etc., to modify claim elements does not, by itself, imply any priority, sequence, or order of one claim element over another or the chronological order in which the acts of a method are performed, but is used merely to distinguish between claim elements as a label that distinguishes one claim element having a certain name from another element that would have the same name (in the absence of the ordinal term).

[0088] Also, the phraseology and terminology used herein is for the purpose of description and should not be regarded as limiting. The use of "including," "comprising," or "having," "containing," "involving," and variations thereof herein are intended to cover the items listed thereafter and equivalents thereof, as well as additional items. The following is claimed:

Claims

1. 1. A system for determining a patient specific threshold used to detect abnormal cells, the system comprising: a medical imaging device configured to generate a set of fluorescent images of an anatomy of a patient; and 1. An image analysis system, comprising: receiving a set of fluorescence images; and The set of fluorescence images is analyzed to determine a patient-specific threshold value used to detect abnormal tissue in the patient. the image analysis system including one or more processors configured Including, The image analysis system analyzing the set of fluorescent images by calculating, using a pixel comparator, one or more representative intensity values ​​for each fluorescent image in the set of fluorescent images representing one or more pixel intensity values ​​of each fluorescent image, thereby generating a set of representative intensity values; selecting a portion of the set of representative intensity values ​​based on a predetermined criterion used to remove one or more of the representative intensity values ​​in the set of representative intensity values; calculating an average intensity value by averaging a selected portion of the representative intensity values; and configured to calculate a patient-specific threshold for the patient based on the average intensity value and an adjustment factor associated with the pixel comparator; the fluorescence image is a surgical site fluorescence image of a surgical site on a patient; The system.

2. A medical imaging device is configured to generate a novel surgical site fluorescence image; The image analysis system is further configured to identify one or more groups of abnormal cells in the new surgical site fluorescent images based on the calculated patient-specific threshold; and the system includes a display configured to indicate one or more locations of at least one of the one or more populations of identified abnormal cells; The system of claim 1 .

3. The pixel comparator a maximum value comparator that calculates, for each image in the set of images, one or more representative intensity values ​​by calculating a maximum pixel intensity value in each image; a minimum comparator that calculates, for each image in the set of images, one or more representative intensity values ​​by calculating a minimum pixel intensity value in each image; an average value comparator that calculates, for each image in the set of images, one or more representative intensity values ​​by calculating an average of the pixel intensity values ​​in each image; a 25% percentile comparator that calculates, for each image in the set of images, one or more representative intensity values ​​by calculating the 25% percentile of the pixel intensity values ​​for each image; and a median comparator for calculating, for each image in the set of images, one or more representative intensity values ​​by calculating a median pixel intensity value for each image; The system of claim 1, wherein the system is selected from the group consisting of:

4. An image analysis system configured to select a pixel comparator from the plurality of pixel comparators, the image analysis system comprising: operating a plurality of pixel comparators on a set of training images to calculate a set of result values ​​for each pixel comparator in the plurality of pixel comparators; analyzing the set of result values ​​for each pixel comparator in the plurality of pixel comparators; and Selecting a pixel comparator from the plurality of pixel comparators based on the analysis The system of claim 1 , comprising:

5. 1. A computer-implemented method for determining a patient-specific threshold value used to detect abnormal cells, the method comprising: receiving a set of fluorescence images of the patient's anatomy; and Analyzing the set of fluorescent images to determine a patient-specific threshold value for use in detecting abnormal tissue in the patient. Including, Analyzing the set of fluorescent images: calculating, using a pixel comparator, one or more representative intensity values ​​for each fluorescent image of the set of fluorescent images representing the one or more pixel intensity values ​​of each fluorescent image to generate a set of representative intensity values; selecting a portion of the set of representative intensity values ​​based on a predetermined criterion used to remove one or more of the representative intensity values ​​in the set of representative intensity values; calculating an average intensity value by averaging a selected portion of the representative intensity values; and calculating a patient-specific threshold value based on the average intensity value and an adjustment factor associated with the pixel comparator; the fluorescence image is a surgical site fluorescence image of a surgical site on a patient; The method.

6. receiving new surgical site fluorescence images; Identifying one or more groups of abnormal cells in the new surgical site fluorescent image based on the calculated patient-specific threshold; and 6. The method of claim 5, further comprising indicating via a display one or more locations of at least one of the one or more populations of identified abnormal cells.

7. The pixel comparator is: a maximum value comparator that calculates, for each image in the set of images, one or more representative intensity values ​​by calculating a maximum pixel intensity value in each image; a minimum comparator that calculates, for each image in the set of images, one or more representative intensity values ​​by calculating a minimum pixel intensity value in each image; an average value comparator that calculates, for each image in the set of images, one or more representative intensity values ​​by calculating an average of the pixel intensity values ​​in each image; a 25% percentile comparator that calculates, for each image in the set of images, one or more representative intensity values ​​by calculating the 25% percentile of the pixel intensity values ​​for each image; and a median comparator for calculating, for each image in the set of images, one or more representative intensity values ​​by calculating a median pixel intensity value for each image; The method of claim 5, wherein the compound is selected from the group consisting of:

8. selecting a pixel comparator from the plurality of pixel comparators, comprising: operating a plurality of pixel comparators on a set of training images to calculate a set of result values ​​for each pixel comparator in the plurality of pixel comparators; analyzing the set of result values ​​for each pixel comparator in the plurality of pixel comparators; and Selecting a pixel comparator from the plurality of pixel comparators based on the analysis The method of claim 5 , comprising:

9. At least one non-transitory computer readable storage medium comprising computer executable instructions that, when executed by at least one processor of an image analysis system configured to collect one or more fluorescent images, calculate a threshold value used to detect abnormal cells by performing a method, the method comprising: receiving a set of fluorescence images of the patient's anatomy; and Determining a patient-specific threshold value used to detect abnormal tissue in a patient by analyzing the set of fluorescent images. Including, Analyzing the set of fluorescent images: calculating, using a pixel comparator, one or more representative intensity values ​​for each fluorescent image of the set of fluorescent images representing the one or more pixel intensity values ​​of each fluorescent image to generate a set of representative intensity values; selecting a portion of the set of representative intensity values ​​based on a predetermined criterion used to remove one or more of the representative intensity values ​​in the set of representative intensity values; calculating an average intensity value by averaging a selected portion of the representative intensity values; and calculating a patient-specific threshold value based on the average intensity value and an adjustment factor associated with the pixel comparator; the fluorescence image is a surgical site fluorescence image of a surgical site on a patient; said at least one non-transitory computer-readable storage medium.

10. Here's how: receiving new surgical site fluorescence images; Identifying one or more groups of abnormal cells in the new surgical site fluorescent image based on the calculated patient-specific threshold; and displaying, via the display, one or more locations of at least one of the one or more populations of identified abnormal cells; 10. The at least one non-transitory computer-readable storage medium of claim 9, further comprising:

11. The pixel comparator is: a maximum value comparator that calculates, for each image in the set of images, one or more representative intensity values ​​by calculating a maximum pixel intensity value in each image; a minimum comparator that calculates, for each image in the set of images, one or more representative intensity values ​​by calculating a minimum pixel intensity value in each image; an average value comparator that calculates, for each image in the set of images, one or more representative intensity values ​​by calculating an average of the pixel intensity values ​​in each image; a 25% percentile comparator that calculates, for each image in the set of images, one or more representative intensity values ​​by calculating the 25% percentile of the pixel intensity values ​​for each image; and a median comparator for calculating, for each image in the set of images, one or more representative intensity values ​​by calculating a median pixel intensity value for each image; 10. At least one non-transitory computer-readable storage medium according to claim 9, selected from the group consisting of:

12. The method further includes selecting a pixel comparator from the plurality of pixel comparators, comprising: operating a plurality of pixel comparators on a set of training images to calculate a set of result values ​​for each pixel comparator in the plurality of pixel comparators; analyzing the set of result values ​​for each pixel comparator in the plurality of pixel comparators; and Selecting a pixel comparator from the plurality of pixel comparators based on the analysis 10. At least one non-transitory computer-readable storage medium according to claim 9, comprising:

Images