System and method for interactive automated analysis of bone scan image for detecting metastasis

Abstract

To provide a system and a method for interactive automated analysis of a bone scan image to detect metastasis.SOLUTION: A system and a method that enable improved computer-aided display and analysis of a nuclear medicine image are presented herein. Specifically, in a certain embodiment, the system and the method described herein provide an improvement to several image processing steps used in automated analysis of a bone scan image to assess a cancer status of a patient. For example, an improved method is provided for image segmentation, hotspot detection, automated classification of a hotspot as indicative of metastases, and calculation of risk indicators such as a bone scan index (BSI) value.SELECTED DRAWING: None

Description

[Technical field]

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS This application claims priority to and the benefit of U.S. Provisional Application No. 62 / 837,955, filed April 24, 2019, the contents of which are incorporated by reference in their entirety.

[0002] Technical Field The present invention relates generally to systems and methods for the generation, analysis, and / or presentation of medical image data, and more particularly, in certain embodiments, the present invention relates to systems and methods for improved computer-assisted display and analysis of nuclear medicine images. [Background technology]

[0003] background Nuclear medicine imaging involves the use of radiolabeled compounds called radiopharmaceuticals. Radiopharmaceuticals are administered to a patient and accumulate in various areas within the body in a manner that depends on and is therefore indicative of the biophysical and / or biochemical properties of the internal tissues, such as properties that are affected by the presence and / or condition of a disease, such as cancer. For example, certain radiopharmaceuticals, after administration to a patient, accumulate in areas of abnormal bone formation associated with malignant bone lesions, indicative of metastases. Other radiopharmaceuticals may bind to specific receptors, enzymes, and proteins in the body that undergo alterations during disease progression. After administration to a patient, the molecules circulate in the blood until they find their intended target. The bound radiopharmaceutical remains at the site of the disease, while the remainder of the drug leaves the body.

[0004] Nuclear medicine imaging techniques capture images by detecting radiation emitted from the radioactive moiety of the radiopharmaceutical. The deposited radiopharmaceutical acts as a landmark so that images can be obtained depicting the location and density of disease using commonly available nuclear medicine modalities. Examples of nuclear medicine imaging modalities include bone scan imaging (also called scintigraphy), single photon emission computed tomography (SPECT), and positron emission tomography (PET). Bone scan, SPECT, and PET imaging systems can be found in most hospitals around the world. The choice of a particular imaging modality depends on and / or determines the particular radiopharmaceutical used. For example, technetium-99m ( 99m Tc) labeled compounds are compatible with bone scan and SPECT imaging, whereas PET imaging often uses 18F labeled fluorinated compounds. 99m Tc methylenediphosphonic acid ( 99m Tc MDP) is a common radiopharmaceutical used in bone scan imaging to detect metastatic cancer. 99m Radiolabeled prostate-specific membrane antigen (PSMA)-targeting compounds, such as Tc-labeled 1404 and PyL™ (also called [18F]DCFPyL), may be used in SPECT and PET imaging, respectively, offering the potential for highly specific prostate cancer detection.

[0005] Nuclear medicine imaging is therefore a valuable technique for providing physicians with information that can be used to determine the presence and extent of disease within a patient's body, which the physician can then use to provide recommended courses of treatment to the patient and track the progression of the disease.

[0006] For example, an oncologist may use nuclear medicine images as input from a patient study in assessing whether the patient has a particular disease, e.g., prostate cancer, what stage of the disease is evident, what the recommended course of treatment (if any) would be, whether surgical intervention is required, and the likely prognosis. The oncologist may use a radiologist report in this assessment. A radiologist report is a technical evaluation of the nuclear medicine images prepared by a radiologist for the physician who ordered the imaging study, including, for example, the type of study performed, the clinical history, comparisons between images, the technique used to perform the study, the radiologist's findings and findings, and any overall impression and recommendations the radiologist may have based on the imaging study findings. The signed radiologist report is sent to the physician who ordered the study for review, followed by a discussion between the physician and the patient regarding the results and treatment recommendations.

[0007] Thus, the process requires a radiologist to perform an imaging study on the patient, analyze the resulting images, prepare a radiologist's report, forward the report to the requesting internist, have the internist formulate an assessment and treatment recommendation, and have the internist communicate the results, recommendations, and risks to the patient. The process may also require repeating the imaging study for inconclusive results or ordering further tests based on the initial results. If the imaging study indicates that the patient has a particular disease or condition (e.g., cancer), the internist will consider various treatment options, including surgery, as well as the risks of doing nothing or adopting a watchful waiting or active surveillance approach rather than performing surgery. Summary of the Invention [Means for solving the problem]

[0008] Therefore, the process of reviewing and analyzing multiple patient images over time plays a critical role in cancer diagnosis and treatment. Thus, there is a great need for improved tools that facilitate and improve the accuracy of reviewing and analyzing images for cancer diagnosis and treatment. Such improvements to the toolkit utilized by physicians, radiologists, and other healthcare professionals would result in significant improvements in standards of care and patient experience.

[0009] Summary of the Invention Presented herein are systems and methods that allow for improved computer-assisted display and analysis of nuclear medicine images. Specifically, in certain embodiments, the systems and methods described herein provide improvements to several image processing steps used in the automated analysis of bone scan images to assess a patient's cancer status.

[0010] For example, improved techniques are provided for image segmentation, hotspot detection, automated classification of hotspots as indicative of metastasis, and calculation of risk indicators such as bone scan index (BSI) values. Thanks to these improved image processing techniques, the systems and methods described herein can be used for accurate and reliable image-based lesion detection and quantification to assess various types of metastatic bone cancers (e.g., any cancer that has metastasized to bone), including metastases associated with prostate cancer, breast cancer, lung cancer, and various other metastatic cancers.

[0011] Bone scan imaging is widely used for the diagnosis and evaluation of metastatic cancer. Patients are injected with radiopharmaceuticals that emit nuclear radiation that can be detected to image the spatial distribution of the radiopharmaceutical within the patient's body. Radiopharmaceuticals can be selected to selectively accumulate in types of tissue associated with cancer lesions, such as areas of abnormal bone formation.

[0012] Although this technique allows lesions to be visualized as bright spots in bone scan images, pinpointing image regions that represent true metastatic lesions is by no means trivial. Radiopharmaceuticals can also accumulate in cancer-free anatomical regions, such as the patient's bladder, and physicians and specialists must carefully distinguish hot spots representing lesions from such regions, as well as from noise and artifacts. This task is time-consuming, error-prone, and subject to significant inter-operator variability.

[0013] Computer-automated lesion detection and analysis paves the way to address these challenges and can dramatically increase the accuracy and repeatability of lesion detection and cancer diagnosis. However, tools for automated lesion detection and analysis rely on a complex combination of imaging processing and artificial intelligence steps. For example, image segmentation to identify skeletal regions can be used to focus the analysis on bone regions. Filtering and thresholding steps can be used to automatically detect hotspots, and machine learning techniques such as artificial neural networks (ANNs) can be used to quantitatively assess the likelihood that a detected hotspot represents a metastasis based on features such as hotspot size, shape, and intensity. Finally, in certain embodiments, a set of detected hotspots representing metastases is used to calculate a patient's overall risk index, which represents the patient's overall likelihood of having and / or developing metastases or having a particular cancer condition. One such risk index is the Bone Scan Index (BSI), which provides an estimated mass fraction of a patient's skeleton occupied by metastases.

[0014] The accuracy of any one step can have a large impact on downstream steps and the overall lesion detection and analysis process. The systems and methods described herein provide several specific improvements to various steps within the automated lesion detection and analysis workflow, thereby increasing the accuracy of results across a wider range of patient types and cancer stages.

[0015] First, in certain embodiments, the improved image analysis techniques described herein include an improved skeletal segmentation technique in which the entire humerus and / or femur region (e.g., more than three-quarters length) is identified in a bone scan image. Previous techniques identified only a limited percentage of the femur and humerus. Now, segmenting a larger portion of these bones allows lesions located toward the distal ends of the arms and legs to be identified, whereas such lesions previously escaped detection. Furthermore, while the reduced uptake of radiopharmaceuticals in the arms and legs makes lesion identification therein difficult, the techniques described herein utilize region-dependent thresholding techniques that increase detection sensitivity in the femur and humerus region to overcome this problem.

[0016] Second, the present disclosure also provides a global thresholding technique that improves hotspot detection accuracy, especially in high disease burden (e.g., when patients have many lesions). This approach detects a priori set of potential hotspots and then adjusts the threshold used for hotspot detection based on a scaling factor calculated from the priori set. This improved hotspot detection provides benefits to downstream calculations and improves the linearity of the BSI values ​​calculated for patients with high levels of metastasis.

[0017] Third, in certain embodiments, the systems and methods described herein improve the accuracy with which automated decisions are made as to whether a hotspot represents a metastasis. Specifically, in certain embodiments, the techniques described herein leverage clinical experience showing that the selection of a hotspot as a potential metastasis depends not only on image features of the hotspot itself, but also on information from the entire image. As such, the techniques described herein may use global features, such as the total number of hotspots, as inputs to the automated decision-making step for lesion identification (e.g., as inputs to an ANN).

[0018] Fourth, in certain embodiments, the techniques described herein also provide an improvement to techniques for calculating risk index values ​​based on bone invasion by using a correction factor that accounts for potential errors in the accuracy with which hotspots can be automatically localized to specific skeletal regions. This is particularly important for hotspots located in or near the sacral region, which is a complex three-dimensional structure that can be difficult to identify in two-dimensional bone scan images. This technique improves the accuracy of the BSI calculation and limits the sensitivity to errors in hotspot localization.

[0019] As such, the systems and methods described herein include several improved image analysis techniques for lesion identification and quantification that improve the accuracy and robustness with which bone scan images can be analyzed. As described herein, they can be used as part of a cloud-based system that facilitates review and reporting of patient data and enables improved disease detection, treatment, and monitoring.

[0020] In one aspect, the present invention is directed to a method for lesion marking and quantitative analysis (e.g., user assisted / reviewed automated or semi-automated lesion marking and quantitative analysis) of nuclear medicine images (e.g., a bone scan image set) of a human subject, the method comprising: (a) accessing (e.g., and / or receiving), by a processor of a computing device, a bone scan image set (e.g., a set of one, two or more images) of the human subject, wherein the bone scan image set is associated with administration of a pharmaceutical agent (e.g., a radiopharmaceutical agent) to the human subject; (b) automatically segmenting, by a processor, each image in the bone scan image set to correspond to a particular anatomical region of the human subject's skeleton (e.g., a particular bone and / or set of one or more bones, such as cervical spine, clavicle, ribs, lumbar spine, pelvis, sacrum, scapula, skull, thoracic spine, sternum, femur, humerus, etc.) of the human subject's skeleton (e.g., a particular bone and / or set of one or more bones, such as cervical spine, clavicle, ribs, lumbar spine, pelvis, sacrum, scapula, skull, thoracic spine, sternum, femur, humerus, etc.). identifying corresponding one or more skeletal regions of interest, thereby obtaining a set of annotated images, the one or more skeletal regions of interest including at least one of: (i) a femur region corresponding to a portion of a femur of the human subject, the portion of the femur encompassing at least three-quarters [e.g., more than about three-quarters (e.g., approximately all)] of the femur along a length of the femur, and (ii) a humerus region corresponding to a portion of a humerus of the human subject, the portion of the humerus encompassing at least three-quarters [e.g., more than about three-quarters (e.g., approximately all)] of the humerus along a length of the humerus; and (c) automatically detecting, by a processor, an initial set of one or more hotspots, each hotspot corresponding to an area of ​​high intensity in the set of annotated images, the automatically detecting including identifying the one or more hotspots using intensities of pixels in the set of annotated images and using one or more region-dependent thresholds (e.g., each region-dependent threshold being(e) for each hot spot in the initial set of hot spots, the processor extracts, by the processor, a set (e.g., one or more sets) of hot spot features associated with the hot spot; and (f) for each hot spot in the initial set of hot spots, the processor extracts, by the processor, a set (e.g., one or more sets) of hot spot features associated with the hot spot; and (g) for each hot spot in the initial set of hot spots, the processor extracts, by the processor, a set (e.g., one or more sets) of hot spot features associated with the hot spot; (f) calculating a metastasis likelihood value corresponding to the likelihood that the hotspot represents metastasis (e.g., for a particular hotspot, using one or more machine learning modules (e.g., a pre-trained machine learning module; e.g., an artificial neural network (ANN)) that receives as input at least a portion of the hotspot features and outputs a metastasis likelihood value for the hotspot); and (f) causing the processor to render a graphical representation of at least a portion of the initial set of hotspots (e.g., visual indications of the hotspots (e.g., points, boundaries) overlaid on one or more members of the set of bone scan images and / or the set of annotated images; e.g., a table listing the identified hotspots along with additional information for each hotspot (e.g., location; e.g., metastasis likelihood value)) for display within a graphical user interface (GUI) (e.g., a cloud-based GUI).

[0021] In one particular embodiment, step (b) includes comparing each member of the bone scan image set with a corresponding atlas image of the atlas image set, each atlas image including one or more identifications (e.g., graphic identifications superimposed on the atlas image) of one or more skeletal regions of interest, the skeletal regions including the femur region and / or humerus region, and aligning, for each image of the bone scan image set, the corresponding atlas image with an image of the bone scan image set such that the identifications of the one or more skeletal regions of interest in the atlas image are applied to (e.g., superimposed on) the image of the bone scan image set.

[0022] In certain embodiments, each atlas image includes identification of (i) a femur region that includes at least a portion of the knee region of the human subject, and / or (ii) a humerus region that includes at least a portion of the elbow region of the human subject, and for each image in the set of bone scan images, registering the corresponding atlas image to the bone scan image includes using the identified knee region and / or the identified elbow region as a landmark within the image (e.g., identifying the knee region in the bone scan image, matching it with the identified knee region in the corresponding atlas image, and then adjusting the atlas image (e.g., calculating a coordinate transformation) to register the corresponding atlas image with the bone scan image).

[0023] In one particular embodiment, the location of at least one detected hotspot in the initial set of hotspots corresponds to a physical location in or on the femur that is more than three-quarters of the distance along the femur from the end of the femur that faces toward the human subject's buttocks to the end of the femur that faces toward the human subject's knee.

[0024] In a particular embodiment, the location of at least one detected hotspot in the initial set of hotspots corresponds to a physical location in or on the humerus that is more than three-quarters of the way along the humerus from the end of the humerus pointing toward the human subject's shoulder to the end of the humerus pointing toward the human subject's elbow.

[0025] In one particular embodiment, step (c) includes (e.g., iteratively) identifying healthy tissue regions within the bone scan image set that are determined to not contain any hot spots (e.g., localized regions of relatively high intensity), calculating by the processor a normalization factor such that the product of the normalization factor and the average intensity of the identified healthy tissue regions is a predetermined intensity level, and normalizing the images of the bone scan image set by the normalization factor.

[0026] In certain embodiments, the method further includes (g) calculating, by the processor, one or more risk index values ​​for the human subject based at least in part on the calculated percentage (e.g., area percentage) of the human subject's skeleton that is occupied by the initial set of hotspots (e.g., the calculated percentage is the ratio of the total area of ​​the initial set of hotspots divided by the total area of ​​all identified skeletal regions).

[0027] In certain embodiments, the method includes: (h) selecting, by the processor, a first subset of (e.g., up to all of) the initial set of hotspots based at least in part on the likelihood of metastasis values ​​(e.g., determining whether to include a particular hotspot from the initial set of hotspots in the subset based on the calculated likelihood of metastasis value for the particular hotspot exceeding a threshold); and (i) causing the processor to render, for display within a graphical user interface (GUI) (e.g., a cloud-based GUI), a graphical representation of the first subset (e.g., visual indications of the hotspots (e.g., points, boundaries) overlaid on one or more members of the set of bone scan images and / or the set of annotated images; e.g., a table listing the identified hotspots along with additional information for each hotspot (e.g., location; e.g., likelihood value)).

[0028] In certain embodiments, the method further includes (j) calculating, by the processor, one or more risk index values ​​for the human subject based at least in part on the calculated percentage (e.g., area percentage) of the human subject's skeleton occupied by the first subset of hotspots (e.g., the calculated percentage is the total area of ​​the initial set of hotspots divided by the total area of ​​all identified skeletal regions).

[0029] In certain embodiments, the method includes (k) receiving, by the processor, via the GUI, a user selection of a second subset of the initial set of hotspots; and (l) calculating, by the processor, one or more risk index values ​​for the human subject based at least in part on the calculated percentage (e.g., area percentage) of the human subject's skeleton occupied by the second subset of hotspots (e.g., the calculated percentage is the total area of ​​the second subset of hotspots divided by the total area of ​​all identified skeletal regions).

[0030] In certain embodiments, at least one of the risk index values ​​indicates the risk of the human subject having and / or developing metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancers).

[0031] In certain embodiments, the metastatic cancer is metastatic prostate cancer.

[0032] In certain embodiments, at least one of the risk index values ​​indicates that the human subject has a particular state of metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancer).

[0033] In one particular embodiment, the processor is a processor of a cloud-based system.

[0034] In certain embodiments, the GUI is part of a general Picture Archiving and Communications System (PACS) (eg, as well as clinical applications for oncology including lesion marking and quantitative analysis).

[0035] In certain embodiments, the agent (e.g., radiopharmaceutical) is technetium 99m methylenediphosphonate ( 99m Tc-MDP).

[0036] In another aspect, the present invention is directed to a method for lesion marking and quantitative analysis (e.g., user assisted / reviewed automated or semi-automated lesion marking and quantitative analysis) of nuclear medicine images (e.g., a bone scan image set) of a human subject, the method comprising: (a) accessing (e.g., and / or receiving), by a processor of a computing device, a bone scan image set (e.g., a set of one, two or more images) of the human subject, wherein the bone scan image set is associated with administration of an agent (e.g., a radioactive (b) automatically segmenting, by a processor, each image in the bone scan image set into a particular anatomical region of the human subject's skeleton (e.g., a particular bone and / or set of one or more bones, such as cervical spine, clavicle, ribs, lumbar spine, pelvis, sacrum, scapula, skull, thoracic spine, sternum, femur, humerus, etc.) (e.g., the bone scan image set includes a ventral bone scan image and a dorsal bone scan image) (e.g., each image in the bone scan image set comprises a plurality of pixels, each pixel having a value corresponding to an intensity) after administration of a pharmaceutical agent (e.g., a pharmaceutical agent); and (b) automatically segmenting, by a processor, each image in the bone scan image set into a particular anatomical region of the human subject's skeleton (e.g., a particular bone and / or set of one or more bones, such as cervical spine, clavicle, ribs, lumbar spine, pelvis, sacrum, scapula, skull, thoracic spine, sternum, femur, humerus, etc.). identifying one or more skeletal regions of interest corresponding to each of the annotated images, thereby obtaining a set of annotated images; (c) automatically detecting, by the processor, an initial set of one or more hotspots, each hotspot corresponding to an area of ​​high intensity in the set of annotated images, said automatically detecting including: detecting a set of potential hotspots using (i) the intensities of pixels in the set of annotated images and (ii) a plurality of pre-thresholds (e.g., the plurality of pre-thresholds are region-dependent thresholds that depend on the identified skeletal region of interest in which a particular pixel is located); calculating a global threshold scaling factor using the set of potential hotspots; adjusting the plurality of pre-thresholds using the global threshold scaling factor, thereby obtaining a plurality of adjusted thresholds; and identifying the initial set of hotspots using (i) the intensities of pixels in the set of annotated images and (ii) the plurality of adjusted thresholds; (d) for each hotspot in the initial set of hotspots,(e) for each hotspot in the initial set of hotspots, the processor calculates, based on the set of hotspot features associated with the hotspot, a metastasis likelihood value corresponding to the likelihood that the hotspot represents a metastasis (e.g., using, for a particular hotspot, one or more machine learning modules (e.g., a pre-trained machine learning module; e.g., an artificial neural network (ANN)) that receives as input at least a portion of the hotspot features and outputs a metastasis likelihood value for the hotspot); and (f) causing the processor to render, for display within a graphical user interface (GUI) (e.g., a cloud-based GUI), a graphical representation of at least a portion of the initial set of hotspots (e.g., visual indications of the hotspots (e.g., points, boundaries) overlaid on one or more images of the set of bone scan images and / or the set of annotated images; e.g., a table listing the identified hotspots along with additional information for each hotspot (e.g., location; e.g., metastasis likelihood value).

[0037] In certain embodiments, the global threshold scaling rate is a function of a measure of disease burden in the human subject (e.g., the percentage area of ​​the subject's skeleton occupied by metastases (e.g., hotspots); e.g., a risk index value), and adjusting the plurality of pre-thresholds performed in step (c) comprises decreasing the adjusted thresholds (e.g., relative to the pre-thresholds) as disease burden increases (e.g., as measured by the global threshold scaling rate) (e.g., such that the total number and / or size of hotspots increases with decreasing adjusted thresholds) to compensate for underestimation of hotspot area that occurs with increasing disease burden.

[0038] In certain embodiments, the global threshold scaling rate is a function (e.g., a nonlinear function) of the percentage (e.g., area percentage) of the identified skeletal regions occupied by the set of potential hotspots (e.g., the global threshold scaling rate is a function of the total area of ​​all hotspots in the pre-set divided by the total area of ​​all identified skeletal regions).

[0039] In one particular embodiment, the global threshold scaling rate is based on (eg, calculated as a function of) a risk index value calculated using a set of potential hotspots.

[0040] In one particular embodiment, step (c) includes (e.g., iteratively) identifying healthy tissue regions within the bone scan image set that are determined to not contain any hot spots (e.g., localized regions of relatively high intensity), calculating by the processor a normalization factor such that a product of the normalization factor and an average intensity of the identified healthy tissue regions is a predetermined intensity level, and normalizing (e.g., iteratively) the images of the bone scan image set by the normalization factor.

[0041] In certain embodiments, the method further includes (g) calculating, by the processor, one or more risk index values ​​for the human subject based at least in part on the calculated percentage (e.g., area percentage) of the human subject's skeleton that is occupied by the initial set of hotspots (e.g., the calculated percentage is the ratio of the total area of ​​the initial set of hotspots divided by the total area of ​​all identified skeletal regions).

[0042] In certain embodiments, the method includes: (h) selecting, by the processor, a first subset of (e.g., up to all of) the initial set of hotspots based at least in part on the likelihood of metastasis values ​​(e.g., determining whether to include a particular hotspot from the initial set of hotspots in the subset based on the calculated likelihood of metastasis value for the particular hotspot exceeding a threshold); and (i) causing the processor to render, for display within a graphical user interface (GUI) (e.g., a cloud-based GUI), a graphical representation of the first subset (e.g., visual indications of the hotspots (e.g., points, boundaries) overlaid on one or more members of the set of bone scan images and / or the set of annotated images; e.g., a table listing the identified hotspots along with additional information for each hotspot (e.g., location; e.g., likelihood value)).

[0043] In certain embodiments, the method further includes (j) calculating, by the processor, one or more risk index values ​​for the human subject based at least in part on the calculated percentage (e.g., area percentage) of the human subject's skeleton occupied by the first subset of hotspots (e.g., the calculated percentage is the total area of ​​the initial set of hotspots divided by the total area of ​​all identified skeletal regions).

[0044] In certain embodiments, the method includes (k) receiving, by the processor, via the GUI, a user selection of a second subset of the initial set of hotspots; and (l) calculating, by the processor, one or more risk index values ​​for the human subject based at least in part on the calculated percentage (e.g., area percentage) of the human subject's skeleton occupied by the second subset of hotspots (e.g., the calculated percentage is the total area of ​​the second subset of hotspots divided by the total area of ​​all identified skeletal regions).

[0045] In certain embodiments, at least one of the risk index values ​​indicates the risk of the human subject having and / or developing metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancers).

[0046] In certain embodiments, the metastatic cancer is metastatic prostate cancer.

[0047] In certain embodiments, at least one of the risk index values ​​indicates that the human subject has a particular state of metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancer).

[0048] In one particular embodiment, the processor is a processor of a cloud-based system.

[0049] In certain embodiments, the GUI is part of a general picture archiving and communication system (PACS) (eg, as well as clinical applications for oncology including lesion marking and quantitative analysis).

[0050] In certain embodiments, the agent (e.g., radiopharmaceutical) is technetium 99m methylenediphosphonate ( 99m Tc-MDP).

[0051] In another aspect, the present invention is directed to a method for lesion marking and quantitative analysis (e.g., user assisted / reviewed automated or semi-automated lesion marking and quantitative analysis) of nuclear medicine images (e.g., a bone scan image set) of a human subject, the method comprising: (a) accessing (e.g., and / or receiving), by a processor of a computing device, a bone scan image set (e.g., a set of one, two or more images) of the human subject, wherein the bone scan image set is associated with administration of a pharmaceutical agent (e.g., a radiopharmaceutical agent) to the human subject; (b) automatically segmenting, by a processor, each image in the bone scan image set to correspond to a particular anatomical region of the human subject's skeleton (e.g., a particular bone and / or set of one or more bones, such as cervical spine, clavicle, ribs, lumbar spine, pelvis, sacrum, scapula, skull, thoracic spine, sternum, femur, humerus, etc.) (e.g., each image in the bone scan image set comprises a plurality of pixels, each pixel having a value corresponding to an intensity) after administration of a drug (e.g., a drug drug). (c) automatically detecting, by the processor, an initial set of one or more hotspots, each hotspot corresponding to an area of ​​high intensity in the set of annotated images (e.g., detecting one or more hotspots in the initial set of hotspots includes comparing pixel intensities to one or more thresholds (e.g., the one or more thresholds vary depending on the identified skeletal region of interest in which a particular pixel is located)); (d) for each hotspot in the initial set of hotspots, extracting, by the processor, a set of hotspot features associated with the hotspot (e.g., a set of one or more hotspot features); and (e) for each hotspot in the initial set of hotspots, calculating, by the processor, a metastasis likelihood value corresponding to the likelihood that the hotspot represents a metastasis based on the set of hotspot features associated with the hotspot (e.g., for a particular hotspot,(f) selecting, by the processor, a first subset of the initial set of hotspots (e.g., up to all of them), wherein the selection of a particular hotspot to include in the first subset is based on (i) a transition likelihood value calculated for the particular hotspot (e.g., a comparison of the likelihood value calculated for the particular hotspot with a likelihood threshold (e.g., if the particular hotspot has a likelihood value greater than the likelihood threshold, then include it in the first subset)); and (ii) one or more global hotspot features, each of which is a transition likelihood value calculated for the particular hotspot (e.g., a transition likelihood value calculated for the particular hotspot with a likelihood threshold (e.g., if the particular hotspot has a likelihood value greater than the likelihood threshold, then include it in the first subset)). (g) causing the processor to render, at least in part, a graphical representation of at least a portion of the first subset of hotspots (e.g., visual indications of the hotspots (e.g., points, boundaries) overlaid on one or more images of the set of bone scan images and / or the set of annotated images; e.g., a table listing the identified hotspots along with additional information (e.g., location; e.g., likelihood value) for each hotspot) for display within a graphical user interface (GUI) (e.g., a cloud-based GUI).

[0052] In one particular embodiment, the one or more global hotspot features include a total number of hotspots in the initial hotspot set.

[0053] In certain embodiments, step (f) includes adjusting criteria for selecting hotspots to be included in the first subset based on the total number of hotspots in the initial hotspot set (e.g., by relaxing the criteria as the total number of hotspots in the initial hotspot set increases (e.g., by reducing the metastasis likelihood value to which the metastasis likelihood value of each hotspot is compared; e.g., by scaling the metastasis likelihood value based on the total number of hotspots in the initial hotspot set)

[0054] In certain embodiments, step (f) includes selecting the first subset using a machine learning module (e.g., an ANN module) (e.g., the machine learning module receives, for each hotspot, at least the metastasis likelihood value calculated for that hotspot and one or more global hotspot features, and outputs (i) an adjusted metastasis likelihood value that takes into account the global hotspot features (e.g., a scaled value comparable to a threshold for selection of hotspots in the first subset), and / or (ii) a binary (e.g., 0 or 1; e.g., Boolean true or false) value representing whether the hotspot should be included in the first subset).

[0055] In one particular embodiment, step (c) includes (e.g., iteratively) identifying healthy tissue regions within the bone scan image set that are determined to not contain any hot spots (e.g., localized regions of relatively high intensity), calculating by the processor a normalization factor such that a product of the normalization factor and an average intensity of the identified healthy tissue regions is a predetermined intensity level, and normalizing (e.g., iteratively) the images of the bone scan image set by the normalization factor.

[0056] In certain embodiments, the method further includes (g) calculating, by the processor, one or more risk index values ​​for the human subject based at least in part on the calculated percentage (e.g., area percentage) of the human subject's skeleton that is occupied by the initial set of hotspots (e.g., the calculated percentage is the ratio of the total area of ​​the initial set of hotspots divided by the total area of ​​all identified skeletal regions).

[0057] In certain embodiments, the method includes: (h) selecting, by the processor, a first subset of (e.g., up to all of) the initial set of hotspots based at least in part on the likelihood of metastasis values ​​(e.g., determining whether to include a particular hotspot from the initial set of hotspots in the subset based on the calculated likelihood of metastasis value for the particular hotspot exceeding a threshold); and (i) causing the processor to render, for display within a graphical user interface (GUI) (e.g., a cloud-based GUI), a graphical representation of the first subset (e.g., visual indications of the hotspots (e.g., points, boundaries) overlaid on one or more members of the set of bone scan images and / or the set of annotated images; e.g., a table listing the identified hotspots along with additional information for each hotspot (e.g., location; e.g., likelihood value)).

[0058] In certain embodiments, the method further includes (j) calculating, by the processor, one or more risk index values ​​for the human subject based at least in part on the calculated percentage (e.g., area percentage) of the human subject's skeleton occupied by the first subset of hotspots (e.g., the calculated percentage is the total area of ​​the initial set of hotspots divided by the total area of ​​all identified skeletal regions).

[0059] In certain embodiments, the method includes (k) receiving, by the processor, via the GUI, a user selection of a second subset of the initial set of hotspots; and (l) calculating, by the processor, one or more risk index values ​​for the human subject based at least in part on the calculated percentage (e.g., area percentage) of the human subject's skeleton occupied by the second subset of hotspots (e.g., the calculated percentage is the total area of ​​the second subset of hotspots divided by the total area of ​​all identified skeletal regions).

[0060] In certain embodiments, at least one of the risk index values ​​indicates the risk of the human subject having and / or developing metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancers).

[0061] In certain embodiments, the metastatic cancer is metastatic prostate cancer.

[0062] In certain embodiments, at least one of the risk index values ​​indicates that the human subject has a particular state of metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancer).

[0063] In one particular embodiment, the processor is a processor of a cloud-based system.

[0064] In certain embodiments, the GUI is part of a general picture archiving and communication system (PACS) (eg, as well as clinical applications for oncology including lesion marking and quantitative analysis).

[0065] In certain embodiments, the agent (e.g., radiopharmaceutical) is technetium 99m methylenediphosphonate ( 99m Tc-MDP).

[0066] In another aspect, the present invention is directed to a method for lesion marking and quantitative analysis (e.g., user assisted / reviewed automated or semi-automated lesion marking and quantitative analysis) of nuclear medicine images (e.g., a set of bone scan images) of a human subject, the method comprising: (a) accessing (e.g., and / or receiving), by a processor of a computing device, a set of bone scan images (e.g., a set of one, two or more images) of the human subject (e.g., the bone scan image set includes a ventral bone scan image and a dorsal bone scan image) (e.g., each image of the set of bone scan images comprises a plurality of pixels, each pixel having a value corresponding to an intensity); and (b) automatically segmenting, by the processor, each image in the set of bone scan images to identify one or more skeletal regions of interest each corresponding to a particular anatomical region of the human subject's skeleton (e.g., a particular bone and / or set of one or more bones, such as cervical vertebrae, clavicle, ribs, lumbar vertebrae, pelvis, sacrum, scapula, skull, thoracic vertebrae, sternum, femur, humerus, etc.), thereby obtaining a set of annotated images. (c) automatically detecting, by the processor, an initial set of one or more hotspots, each hotspot corresponding to an area of ​​high intensity in the set of annotated images (e.g., detecting one or more hotspots in the initial set of hotspots includes comparing pixel intensities to one or more thresholds (e.g., the one or more thresholds vary depending on the identified skeletal region of interest in which a particular pixel is located)); (d) for each hotspot in the initial set of hotspots, extracting, by the processor, a set of hotspot features associated with the hotspot (e.g., a set of one or more hotspot features); and (e) for each hotspot in the initial set of hotspots, calculating, by the processor, a likelihood value corresponding to the likelihood that the hotspot represents a metastasis based on the set of hotspot features associated with the hotspot (e.g., for a particular hotspot, receiving as input at least a portion of the hotspot features and outputting a likelihood value for the hotspot).(f) selecting, by the processor, a first subset of hotspots (e.g., up to all of them) of the hotspots of the initial set of hotspots based at least in part on the likelihood value calculated for each hotspot of the initial set of hotspots (e.g., determining whether to include a particular hotspot of the initial set of hotspots in the preselected set of hotspots based on the likelihood value calculated for the particular hotspot (e.g., by comparing it to a likelihood threshold)); and (g) calculating, by the processor, one or more risk index values ​​(e.g., bone scan index values) using at least a portion (e.g., up to all of them) of the first subset of hotspots, wherein said calculating includes determining whether to include a particular hotspot of the initial set of hotspots in the preselected set of hotspots based on the likelihood value calculated for the particular hotspot (e.g., by comparing it to a likelihood threshold). For a given hotspot, calculating a bone invasion rate based on a ratio of (i) the size (e.g., area) of the particular hotspot to (ii) the size (e.g., area) of a particular skeletal region to which the particular hotspot has been assigned (e.g., by a processor) based on its location within the set of annotated images, thereby determining one or more bone invasion rates; adjusting the bone invasion rate using one or more region-dependent correction factors (e.g., each region-dependent correction factor associated with one or more skeletal regions; e.g., the value of the region-dependent correction factor is selected to reduce the degree to which assigning a particular hotspot to a particular skeletal region (e.g., of a plurality of adjacent or nearby skeletal regions, such as the sacrum, pelvis, and lumbar regions) causes variability in the calculated bone invasion rate), thereby obtaining one or more adjusted bone invasion rates; and summing the adjusted bone invasion rates to determine one or more risk index values.

[0067] In certain embodiments, for each particular hotspot, the calculated bone infiltration rate estimates the proportion of the total skeletal mass occupied by the physical volume associated with the particular hotspot.

[0068] In certain embodiments, the step of calculating the bone infiltration rate includes the steps of: calculating, by the processor, a ratio of the area of ​​the particular hotspot to the area of ​​the corresponding skeletal region of interest, thereby calculating an area percentage of the particular hotspot; and scaling (e.g., multiplying) the area percentage by a density factor associated with the skeletal region of interest to which the particular hotspot is assigned (e.g., thereby taking into account the weight and / or density of bonds in the corresponding skeletal region of interest (e.g., the density factor is the weight percentage of the corresponding skeletal region of interest relative to the total skeleton (e.g., of an average human)) thereby calculating the bone infiltration rate corresponding to the particular hotspot.

[0069] In a particular embodiment, at least a portion of the hotspots of the first subset are assigned to a skeletal region of interest that is a member selected from the group consisting of a pelvic region (e.g., corresponding to the pelvis of a human subject), a lumbar region (e.g., corresponding to the lumbar column of a human subject), and a sacral region (e.g., corresponding to the sacrum of a human subject).

[0070] In certain embodiments, the one or more region-dependent correction factors include a sacral region correction factor associated with the sacral region that is used to adjust the bone infiltration rate of hotspots identified (e.g., by a processor) as being located within the sacral region, where the sacral region correction factor has a value less than 1 (e.g., less than 0.5).

[0071] In certain embodiments, the one or more region-dependent correction factors include one or more correction factor pairs, each correction factor pair associated with a particular skeletal region of interest and including a first member and a second member (of the pair), where the first member of the pair is a ventral image correction factor that is used to adjust the bone infiltration rate calculated for a hotspot detected in an annotated ventral bone scan image of the annotated image set, and the second member of the pair is a dorsal image correction factor that is used to adjust the bone infiltration rate calculated for a hotspot detected in an annotated dorsal bone scan image of the annotated image set.

[0072] In one particular embodiment, step (c) includes (e.g., iteratively) identifying healthy tissue regions within the bone scan image set that are determined to not contain any hot spots (e.g., localized regions of relatively high intensity); calculating, by the processor, a normalization factor such that the product of the normalization factor and the average intensity of the identified healthy tissue regions is a predetermined intensity level; and normalizing, by the processor, the images of the bone scan image set by the normalization factor.

[0073] In certain embodiments, the method further includes (g) calculating, by the processor, one or more risk index values ​​for the human subject based at least in part on the calculated percentage (e.g., area percentage) of the human subject's skeleton that is occupied by the initial set of hotspots (e.g., the calculated percentage is the ratio of the total area of ​​the initial set of hotspots divided by the total area of ​​all identified skeletal regions).

[0074] In certain embodiments, the method includes: (h) selecting, by the processor, a first subset of (e.g., up to all of) the initial set of hotspots based at least in part on the likelihood of metastasis values ​​(e.g., determining whether to include a particular hotspot from the initial set of hotspots in the subset based on the calculated likelihood of metastasis value for the particular hotspot exceeding a threshold); and (i) causing the processor to render, for display within a graphical user interface (GUI) (e.g., a cloud-based GUI), a graphical representation of the first subset (e.g., visual indications of the hotspots (e.g., points, boundaries) overlaid on one or more members of the set of bone scan images and / or the set of annotated images; e.g., a table listing the identified hotspots along with additional information for each hotspot (e.g., location; e.g., likelihood value)).

[0075] In certain embodiments, the method further includes (j) calculating, by the processor, one or more risk index values ​​for the human subject based at least in part on the calculated percentage (e.g., area percentage) of the human subject's skeleton occupied by the first subset of hotspots (e.g., the calculated percentage is the total area of ​​the initial set of hotspots divided by the total area of ​​all identified skeletal regions).

[0076] In certain embodiments, the method includes (k) receiving, by the processor, via the GUI, a user selection of a second subset of the initial set of hotspots; and (l) calculating, by the processor, one or more risk index values ​​for the human subject based at least in part on the calculated percentage (e.g., area percentage) of the human subject's skeleton occupied by the second subset of hotspots (e.g., the calculated percentage is the total area of ​​the second subset of hotspots divided by the total area of ​​all identified skeletal regions).

[0077] In certain embodiments, at least one of the risk index values ​​indicates the risk of the human subject having and / or developing metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancers).

[0078] In certain embodiments, the metastatic cancer is metastatic prostate cancer.

[0079] In certain embodiments, at least one of the risk index values ​​indicates that the human subject has a particular state of metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancer).

[0080] In one particular embodiment, the processor is a processor of a cloud-based system.

[0081] In certain embodiments, the GUI is part of a general picture archiving and communication system (PACS) (eg, as well as clinical applications for oncology including lesion marking and quantitative analysis).

[0082] In certain embodiments, the agent (e.g., radiopharmaceutical) is technetium 99m methylenediphosphonate ( 99m Tc-MDP).

[0083] In another aspect, the present invention is directed to a system for lesion marking and quantitative analysis (e.g., user assisted / reviewed automated or semi-automated lesion marking and quantitative analysis) of nuclear medicine images (e.g., a bone scan image set) of a human subject, the system comprising a processor and a memory having instructions that, when executed by the processor, cause the processor to (a) access (e.g., and / or receive) a bone scan image set (e.g., a set of one, two, or more images) of the human subject; (b) automatically segmenting each image in the bone scan image set to identify specific bones and / or bone structures of a particular anatomical region of the human subject's skeleton (e.g., a particular bone, such as the cervical vertebrae, clavicle, ribs, lumbar vertebrae, pelvis, sacrum, scapula, skull, thoracic vertebrae, sternum, femur, humerus, etc.); identifying one or more skeletal regions of interest each corresponding to a set of one or more bones of a human subject, thereby obtaining a set of annotated images, the one or more skeletal regions of interest including at least one of: (i) a femur region corresponding to a portion of a femur of the human subject, the portion of the femur including at least three-quarters [e.g., more than about three-quarters (e.g., approximately all)] of the femur along a length of the femur, and (ii) a humerus region corresponding to a portion of a humerus of the human subject, the portion of the humerus including at least three-quarters [e.g., more than about three-quarters (e.g., approximately all)] of the humerus along a length of the humerus; and (c) automatically detecting an initial set of one or more hotspots, each hotspot corresponding to an area of ​​high intensity in the set of annotated images, the automatically detecting including identifying the one or more hotspots using intensities of pixels in the set of annotated images and using one or more region-dependent thresholds (e.g., each region-dependent threshold is(d) for each hot spot in the initial set of hot spots, extracting a set of hot spot features associated with the hot spot (e.g., a set of one or more hot spot features); and (e) for each hot spot in the initial set of hot spots, extracting a set of hot spot features associated with the hot spot based on the set of hot spot features associated with the hot spot. (f) calculating a metastasis likelihood value corresponding to the likelihood that the hotspot represents a metastasis (e.g., for a particular hotspot, using one or more machine learning modules (e.g., a pre-trained machine learning module; e.g., an artificial neural network (ANN)) that receives as input at least a portion of the hotspot features and outputs a metastasis likelihood value for the hotspot); and (f) rendering for display within a graphical user interface (GUI) (e.g., a cloud-based GUI) a graphical representation of at least a portion of the initial set of hotspots (e.g., visual indications of the hotspots (e.g., points, boundaries) overlaid on one or more members of the set of bone scan images and / or the set of annotated images; e.g., a table listing the identified hotspots along with additional information for each hotspot (e.g., location; e.g., metastasis likelihood value).

[0084] In one particular embodiment, in step (b), the instructions cause the processor to compare each member of the bone scan image set to a corresponding atlas image in the atlas image set, each atlas image including one or more identifications (e.g., graphical identifications superimposed on the atlas image) of one or more skeletal regions of interest, the skeletal regions including the femur region and / or humerus region, and to align, for each image in the bone scan image set, the corresponding atlas image with an image in the bone scan image set such that the identifications of the one or more skeletal regions of interest in the atlas image are applied to (e.g., superimposed on) the image in the bone scan image set.

[0085] In one particular embodiment, each atlas image includes identification of (i) a femur region that includes at least a portion of the knee region of the human subject, and / or (ii) a humerus region that includes at least a portion of the elbow region of the human subject, and for each image in the set of bone scan images, the instructions cause the processor to align the corresponding atlas image with the bone scan image using the identified knee region and / or the identified elbow region as a landmark within the image (e.g., align the corresponding atlas image with the bone scan image by identifying the knee region in the bone scan image, matching it with the identified knee region in the corresponding atlas image, and then adjusting the atlas image (e.g., calculating a coordinate transformation)).

[0086] In one particular embodiment, the location of at least one detected hotspot in the initial set of hotspots corresponds to a physical location in or on the femur that is more than three-quarters of the distance along the femur from the end of the femur that faces toward the human subject's buttocks to the end of the femur that faces toward the human subject's knee.

[0087] In a particular embodiment, the location of at least one detected hotspot in the initial set of hotspots corresponds to a physical location in or on the humerus that is more than three-quarters of the way along the humerus from the end of the humerus pointing toward the human subject's shoulder to the end of the humerus pointing toward the human subject's elbow.

[0088] In one particular embodiment, in step (c), the instructions cause the processor to (e.g., iteratively) identify healthy tissue regions in images of the bone scan image set that are determined to not contain any hot spots (e.g., localized regions of relatively high intensity), calculate a normalization factor such that the product of the normalization factor and the average intensity of the identified healthy tissue regions is a predetermined intensity level, and normalize the images of the bone scan image set by the normalization factor.

[0089] In certain embodiments, the instructions further cause the processor to: (g) calculate one or more risk index values ​​for the human subject based at least in part on the calculated percentage (e.g., area percentage) of the human subject's skeleton that is occupied by the initial set of hotspots (e.g., the calculated percentage is the ratio of the total area of ​​the initial set of hotspots divided by the total area of ​​all identified skeletal regions).

[0090] In a particular embodiment, the instructions cause the processor to (h) select a first subset of the initial set of hotspots (e.g., up to all of them) based at least in part on the likelihood of metastasis values ​​(e.g., determining whether to include a particular hotspot in the initial set of hotspots in the subset based on the calculated likelihood of metastasis value for the particular hotspot exceeding a threshold value), and (i) render a graphical representation of the first subset (e.g., visual indications of the hotspots (e.g., points, boundaries) overlaid on one or more members of the set of bone scan images and / or the set of annotated images; e.g., a table listing the identified hotspots along with additional information for each hotspot (e.g., location; e.g., likelihood value)) for display within a graphical user interface (GUI) (e.g., a cloud-based GUI).

[0091] In certain embodiments, the instructions cause the processor to: (j) calculate one or more risk index values ​​for the human subject based at least in part on the calculated percentage (e.g., area percentage) of the human subject's skeleton that is occupied by the first subset of hotspots (e.g., the calculated percentage is the total area of ​​the initial set of hotspots divided by the total area of ​​all identified skeletal regions).

[0092] In certain embodiments, the instructions cause the processor to: (k) receive a user selection of a second subset of the initial set of hotspots via the GUI; and (l) calculate one or more risk index values ​​for the human subject based at least in part on the calculated percentage (e.g., area percentage) of the human subject's skeleton occupied by the second subset of hotspots (e.g., the calculated percentage is the total area of ​​the second subset of hotspots divided by the total area of ​​all identified skeletal regions).

[0093] In certain embodiments, at least one of the risk index values ​​indicates the risk of the human subject having and / or developing metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancers).

[0094] In certain embodiments, the metastatic cancer is metastatic prostate cancer.

[0095] In certain embodiments, at least one of the risk index values ​​indicates that the human subject has a particular state of metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancer).

[0096] In certain embodiments, the system is a cloud-based system. In certain embodiments, the processor is a processor of a cloud-based system.

[0097] In certain embodiments, the GUI is part of a general picture archiving and communication system (PACS) (eg, as well as clinical applications for oncology including lesion marking and quantitative analysis).

[0098] In certain embodiments, the agent (e.g., radiopharmaceutical) is technetium 99m methylenediphosphonate ( 99m Tc-MDP).

[0099] In another aspect, the present invention is directed to a system for lesion marking and quantitative analysis (e.g., user assisted / reviewed automated or semi-automated lesion marking and quantitative analysis) of nuclear medicine images (e.g., bone scan image sets) of a human subject, the system comprising a processor and a memory having instructions which, when executed by the processor, cause the processor to: (a) access (e.g., and / or receive), by the processor of a computing device, a bone scan image set (e.g., a set of one, two or more images) of the human subject, the bone scan image set having been obtained following administration of a pharmaceutical agent (e.g., a radiopharmaceutical) to the human subject (e.g., the bone scan image set includes a ventral bone scan image and a dorsal bone scan image) (e.g., each image of the bone scan image set comprises a plurality of pixels, each pixel having a value corresponding to an intensity); and (b) automatically segment each image in the bone scan image set to identify a particular anatomical region of the human subject's skeleton (e.g., cervical vertebrae, clavicle, ribs, lumbar vertebrae, pelvis, sacrum, scapula, skull, pelvis, scapula ... (c) automatically detecting an initial set of one or more hotspots, each hotspot corresponding to an area of ​​high intensity in the set of annotated images, said automatically detecting including: detecting a set of potential hotspots using (i) the intensities of pixels in the set of annotated images and (ii) a plurality of pre-thresholds (e.g., the pre-thresholds are region-dependent thresholds that depend on the identified skeletal region of interest in which a particular pixel is located); calculating a global threshold scaling factor using the set of potential hotspots; adjusting the pre-thresholds using the global threshold scaling factor, thereby obtaining a plurality of adjusted thresholds; and identifying the initial set of hotspots using (i) the intensities of pixels in the set of annotated images and (ii) the plurality of adjusted thresholds;(d) for each hotspot in the initial set of hotspots, extracting a set of hotspot features associated with the hotspot (e.g., a set of one or more hotspot features); (e) for each hotspot in the initial set of hotspots, calculating a metastasis likelihood value corresponding to the likelihood that the hotspot represents a metastasis based on the set of hotspot features associated with the hotspot (e.g., using one or more machine learning modules (e.g., a pre-trained machine learning module; e.g., an artificial neural network (ANN)) that receives as input at least a portion of the hotspot features and outputs a metastasis likelihood value for the hotspot for a particular hotspot); and (f) rendering for display within a graphical user interface (GUI) (e.g., a cloud-based GUI) a graphical representation of at least a portion of the initial set of hotspots (e.g., visual indications of the hotspots (e.g., points, boundaries) overlaid on one or more images of the set of bone scan images and / or the set of annotated images; e.g., a table listing the identified hotspots along with additional information for each hotspot (e.g., location; e.g., metastasis likelihood value).

[0100] In certain embodiments, the instructions cause the processor to calculate a global threshold scaling factor as a function of a measure of disease burden in the human subject (e.g., the percentage area of ​​the subject's skeleton occupied by metastases (e.g., hotspots); e.g., a risk index value), and in step (c), adjust the multiple pre-thresholds by decreasing the adjusted thresholds (e.g., relative to the pre-thresholds) as disease burden increases (e.g., as measured by the global threshold scaling factor) (e.g., such that the total number and / or size of hotspots increases as the adjusted thresholds decrease) to compensate for underestimation of hotspot area that occurs with increasing disease burden.

[0101] In one particular embodiment, the instructions cause the processor to calculate a global threshold scaling factor as a function (e.g., a nonlinear function) of the percentage (e.g., area percentage) of the identified skeletal region occupied by the set of potential hotspots (e.g., the global threshold scaling factor is a function of the total area of ​​all hotspots in the pre-set divided by the total area of ​​all identified skeletal regions).

[0102] In one particular embodiment, the instructions cause the processor to calculate a global threshold scaling factor based on (eg, as a function of) a risk index value calculated using the set of potential hotspots.

[0103] In one particular embodiment, in step (c), the instructions cause the processor to (e.g., iteratively) identify healthy tissue regions in images of the bone scan image set that are determined to not contain any hot spots (e.g., localized regions of relatively high intensity), calculate a normalization factor such that the product of the normalization factor and the average intensity of the identified healthy tissue regions is a predetermined intensity level, and normalize the images of the bone scan image set by the normalization factor.

[0104] In certain embodiments, the instructions further cause the processor to: (g) calculate one or more risk index values ​​for the human subject based at least in part on the calculated percentage (e.g., area percentage) of the human subject's skeleton that is occupied by the initial set of hotspots (e.g., the calculated percentage is the ratio of the total area of ​​the initial set of hotspots divided by the total area of ​​all identified skeletal regions).

[0105] In a particular embodiment, the instructions cause the processor to (h) select a first subset of the initial set of hotspots (e.g., up to all of them) based at least in part on the likelihood of metastasis values ​​(e.g., determining whether to include a particular hotspot in the initial set of hotspots in the subset based on the calculated likelihood of metastasis value for the particular hotspot exceeding a threshold value), and (i) render a graphical representation of the first subset (e.g., visual indications of the hotspots (e.g., points, boundaries) overlaid on one or more members of the set of bone scan images and / or the set of annotated images; e.g., a table listing the identified hotspots along with additional information for each hotspot (e.g., location; e.g., likelihood value)) for display within a graphical user interface (GUI) (e.g., a cloud-based GUI).

[0106] In certain embodiments, the instructions cause the processor to: (j) calculate one or more risk index values ​​for the human subject based at least in part on the calculated percentage (e.g., area percentage) of the human subject's skeleton that is occupied by the first subset of hotspots (e.g., the calculated percentage is the total area of ​​the initial set of hotspots divided by the total area of ​​all identified skeletal regions).

[0107] In certain embodiments, the instructions cause the processor to: (k) receive a user selection of a second subset of the initial set of hotspots via the GUI; and (l) calculate one or more risk index values ​​for the human subject based at least in part on the calculated percentage (e.g., area percentage) of the human subject's skeleton occupied by the second subset of hotspots (e.g., the calculated percentage is the total area of ​​the second subset of hotspots divided by the total area of ​​all identified skeletal regions).

[0108] In certain embodiments, at least one of the risk index values ​​indicates the risk of the human subject having and / or developing metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancers).

[0109] In certain embodiments, the metastatic cancer is metastatic prostate cancer.

[0110] In certain embodiments, at least one of the risk index values ​​indicates that the human subject has a particular state of metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancer).

[0111] In certain embodiments, the system is a cloud-based system. In certain embodiments, the processor is a processor of a cloud-based system.

[0112] In certain embodiments, the GUI is part of a general picture archiving and communication system (PACS) (eg, as well as clinical applications for oncology including lesion marking and quantitative analysis).

[0113] In certain embodiments, the agent (e.g., radiopharmaceutical) is technetium 99m methylenediphosphonate ( 99m Tc-MDP).

[0114] In another aspect, the present invention is directed to a system for lesion marking and quantitative analysis (e.g., user assisted / reviewed automated or semi-automated lesion marking and quantitative analysis) of nuclear medicine images (e.g., a bone scan image set) of a human subject, the system comprising a processor and a memory having instructions that, when executed by the processor, cause the processor to (a) access (e.g., and / or receive) a bone scan image set (e.g., a set of one, two, or more images) of the human subject; (b) automatically segmenting each image in the bone scan image set to identify particular anatomical regions of the human subject's skeleton (e.g., particular bones such as the cervical vertebrae, clavicle, ribs, lumbar vertebrae, pelvis, sacrum, scapula, skull, thoracic vertebrae, sternum, femur, humerus, etc.); and (c) automatically segmenting each image in the bone scan image set to identify particular anatomical regions of the human subject's skeleton (e.g., particular bones such as the cervical vertebrae, clavicle, ribs, lumbar vertebrae, pelvis, sacrum, scapula, skull, thoracic vertebrae, sternum, femur, humerus, etc.). (c) automatically detecting an initial set of one or more hotspots, each hotspot corresponding to an area of ​​high intensity in the set of annotated images (e.g., detecting one or more hotspots in the initial set of hotspots includes comparing pixel intensities to one or more thresholds (e.g., the one or more thresholds vary depending on the identified skeletal region of interest in which a particular pixel is located)); (d) for each hotspot in the initial set of hotspots, extracting a set of hotspot features associated with the hotspot (e.g., a set of one or more hotspot features); and (e) for each hotspot in the initial set of hotspots, calculating a metastasis likelihood value corresponding to the likelihood that the hotspot represents a metastasis based on the set of hotspot features associated with the hotspot (e.g., for a particular hotspot,(f) automatically selecting a first subset of the initial set (e.g., up to all) of hotspots, where the selection of a particular hotspot to include in the first subset is based on (i) a transition likelihood value calculated for the particular hotspot (e.g., a comparison of the likelihood value calculated for the particular hotspot with a likelihood threshold (e.g., if the particular hotspot has a likelihood value greater than the likelihood threshold, then include it in the first subset)); and (ii) one or more global hotspot features, each of which is a transition likelihood value calculated for the particular hotspot (e.g., a comparison of the likelihood value calculated for the particular hotspot with a likelihood threshold (e.g., if the particular hotspot has a likelihood value greater than the likelihood threshold, then include it in the first subset)). (g) rendering a graphical representation of at least a portion of the first subset of hotspots (e.g., visual indications of hotspots (e.g., points, boundaries) overlaid on one or more images of the set of bone scan images and / or the set of annotated images; e.g., a table listing the identified hotspots along with additional information (e.g., location; e.g., likelihood value) for each hotspot) for display within a graphical user interface (GUI) (e.g., a cloud-based GUI).

[0115] In one particular embodiment, the one or more global hotspot features include a total number of hotspots in the initial hotspot set.

[0116] In one particular embodiment, in step (f), the instructions cause the processor to adjust criteria for selecting hotspots to be included in the first subset based on the total number of hotspots in the initial hotspot set (e.g., by relaxing the criteria as the total number of hotspots in the initial hotspot set increases (e.g., by reducing the transition likelihood value to which the transition likelihood value of each hotspot is compared; e.g., by scaling the transition likelihood value based on the total number of hotspots in the initial hotspot set)).

[0117] In one particular embodiment, in step (f), the instructions cause the processor to select the first subset using a machine learning module (e.g., an ANN module) (e.g., the machine learning module receives, for each hotspot, at least the transition likelihood value calculated for that hotspot and one or more global hotspot features, and outputs (i) an adjusted transition likelihood value (e.g., a scaled value comparable to a threshold for selection of hotspots in the first subset) that takes into account the global hotspot features, and / or (ii) a binary (e.g., 0 or 1; e.g., Boolean true or false) value representing whether the hotspot should be included in the first subset).

[0118] In one particular embodiment, in step (c), the instructions cause the processor to (e.g., iteratively) identify healthy tissue regions in images of the bone scan image set that are determined to not contain any hot spots (e.g., localized regions of relatively high intensity), calculate a normalization factor such that the product of the normalization factor and the average intensity of the identified healthy tissue regions is a predetermined intensity level, and normalize the images of the bone scan image set by the normalization factor.

[0119] In certain embodiments, the instructions further cause the processor to: (g) calculate one or more risk index values ​​for the human subject based at least in part on the calculated percentage (e.g., area percentage) of the human subject's skeleton that is occupied by the initial set of hotspots (e.g., the calculated percentage is the ratio of the total area of ​​the initial set of hotspots divided by the total area of ​​all identified skeletal regions).

[0120] In a particular embodiment, the instructions cause the processor to (h) select a first subset of the initial set of hotspots (e.g., up to all of them) based at least in part on the likelihood of metastasis values ​​(e.g., determining whether to include a particular hotspot in the initial set of hotspots in the subset based on the calculated likelihood of metastasis value for the particular hotspot exceeding a threshold value), and (i) render a graphical representation of the first subset (e.g., visual indications of the hotspots (e.g., points, boundaries) overlaid on one or more members of the set of bone scan images and / or the set of annotated images; e.g., a table listing the identified hotspots along with additional information for each hotspot (e.g., location; e.g., likelihood value)) for display within a graphical user interface (GUI) (e.g., a cloud-based GUI).

[0121] In certain embodiments, the instructions cause the processor to: (j) calculate one or more risk index values ​​for the human subject based at least in part on the calculated percentage (e.g., area percentage) of the human subject's skeleton that is occupied by the first subset of hotspots (e.g., the calculated percentage is the total area of ​​the initial set of hotspots divided by the total area of ​​all identified skeletal regions).

[0122] In certain embodiments, the instructions cause the processor to: (k) receive a user selection, via the GUI, of a second subset of the initial set of hotspots; and (l) calculate one or more risk index values ​​for the human subject based at least in part on the calculated percentage (e.g., area percentage) of the human subject's skeleton occupied by the second subset of hotspots (e.g., the calculated percentage is the total area of ​​the second subset of hotspots divided by the total area of ​​all identified skeletal regions).

[0123] In certain embodiments, at least one of the risk index values ​​indicates the risk of the human subject having and / or developing metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancers).

[0124] In certain embodiments, the metastatic cancer is metastatic prostate cancer.

[0125] In certain embodiments, at least one of the risk index values ​​indicates that the human subject has a particular state of metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancer).

[0126] In certain embodiments, the system is a cloud-based system. In certain embodiments, the processor is a processor of a cloud-based system.

[0127] In certain embodiments, the GUI is part of a general picture archiving and communication system (PACS) (eg, as well as clinical applications for oncology including lesion marking and quantitative analysis).

[0128] In certain embodiments, the agent (e.g., radiopharmaceutical) is technetium 99m methylenediphosphonate ( 99m Tc-MDP).

[0129] In another aspect, the present invention is directed to a system for lesion marking and quantitative analysis (e.g., user assisted / reviewed automated or semi-automated lesion marking and quantitative analysis) of nuclear medicine images (e.g., a bone scan image set) of a human subject, the system comprising a processor and a memory having instructions that, when executed by the processor, cause the processor to (a) access (e.g., and / or receive) (e.g., a set of one, two or more images) of a bone scan image set of the human subject; (b) automatically segmenting each image in the bone scan image set to identify one or more skeletal regions of interest, each corresponding to a particular anatomical region of the human subject's skeleton (e.g., a particular bone and / or set of one or more bones, such as the cervical vertebrae, clavicle, ribs, lumbar vertebrae, pelvis, sacrum, scapula, skull, thoracic vertebrae, sternum, femur, humerus, etc.), thereby annotating the human subject's skeleton. (c) automatically detecting an initial set of one or more hotspots, each hotspot corresponding to an area of ​​high intensity in the set of annotated images (e.g., detecting the one or more hotspots in the initial set of hotspots includes comparing pixel intensities to one or more thresholds (e.g., the one or more thresholds vary depending on the identified skeletal region of interest in which a particular pixel is located)); (d) for each hotspot in the initial set of hotspots, extracting a set of hotspot features associated with the hotspot (e.g., the set of one or more hotspot features); and (e) for each hotspot in the initial set of hotspots, calculating a likelihood value corresponding to the likelihood that the hotspot represents a metastasis based on the set of hotspot features associated with the hotspot (e.g., for a particular hotspot, one or more machine learning modules (e.g.,(f) selecting a first subset of hotspots (e.g., up to all of them) of the hotspots of the initial set of hotspots based at least in part on the likelihood value calculated for each hotspot in the initial set of hotspots (e.g., determining whether to include a particular hotspot in the initial set of hotspots in the set of pre-selected hotspots based on the likelihood value calculated for the particular hotspot (e.g., by comparing it to a likelihood threshold)); and (g) calculating one or more risk index values ​​(e.g., bone scan index values) using at least a portion (e.g., up to all of them) of the first subset of hotspots, wherein the calculating includes, for each particular hotspot in the first subset, (ii) determining whether or not the particular hotspot in the calculating a bone invasion rate based on a ratio of (i) a size (e.g., area) of a particular hotspot of the bone invasion rate to a size (e.g., area) of a particular skeletal region to which the particular hotspot is assigned (e.g., by a processor) based on its location within the set of annotated images, thereby determining one or more bone invasion rates; adjusting the bone invasion rate using one or more region-dependent correction factors (e.g., each region-dependent correction factor associated with one or more skeletal regions; e.g., the value of the region-dependent correction factor is selected to reduce the degree to which assigning a particular hotspot to a particular skeletal region (e.g., of a plurality of adjacent or nearby skeletal regions, such as the sacrum, pelvis, and lumbar regions) causes variability in the calculated bone invasion rate), thereby obtaining one or more adjusted bone invasion rates; and summing the adjusted bone invasion rates to determine one or more risk index values.

[0130] In certain embodiments, for each particular hotspot, the calculated bone infiltration rate estimates the proportion of the total skeletal mass occupied by the physical volume associated with the particular hotspot.

[0131] In certain embodiments, the instructions cause the processor to calculate a bone infiltration rate by calculating a ratio of an area of ​​a particular hotspot to an area of ​​a corresponding skeletal region of interest, thereby calculating an area percentage of the particular hotspot, and scaling (e.g., multiplying) the area percentage by a density factor associated with the skeletal region of interest to which the particular hotspot is assigned (e.g., thereby taking into account the weight and / or density of bonds in the corresponding skeletal region of interest (e.g., the density factor is the weight percentage of the corresponding skeletal region of interest relative to the total skeleton (e.g., of an average human))), thereby calculating a bone infiltration rate for the particular hotspot.

[0132] In a particular embodiment, at least a portion of the hotspots of the first subset are assigned to a skeletal region of interest that is a member selected from the group consisting of a pelvic region (e.g., corresponding to the pelvis of a human subject), a lumbar region (e.g., corresponding to the lumbar column of a human subject), and a sacral region (e.g., corresponding to the sacrum of a human subject).

[0133] In certain embodiments, the one or more region-dependent correction factors include a sacral region correction factor associated with the sacral region that is used to adjust the bone infiltration rate of hotspots identified (e.g., by a processor) as being located within the sacral region, where the sacral region correction factor has a value less than 1 (e.g., less than 0.5).

[0134] In certain embodiments, the one or more region-dependent correction factors include one or more correction factor pairs, each correction factor pair associated with a particular skeletal region of interest and including a first member and a second member (of the pair), where the first member of the pair is a ventral image correction factor that is used to adjust the bone infiltration rate calculated for a hotspot detected in an annotated ventral bone scan image of the annotated image set, and the second member of the pair is a dorsal image correction factor that is used to adjust the bone infiltration rate calculated for a hotspot detected in an annotated dorsal bone scan image of the annotated image set.

[0135] In one particular embodiment, in step (c), the instructions cause the processor to (e.g., iteratively) identify healthy tissue regions in images of the bone scan image set that are determined to not contain any hot spots (e.g., localized regions of relatively high intensity), calculate a normalization factor such that the product of the normalization factor and the average intensity of the identified healthy tissue regions is a predetermined intensity level, and normalize the images of the bone scan image set by the normalization factor.

[0136] In certain embodiments, the instructions further cause the processor to: (g) calculate one or more risk index values ​​for the human subject based at least in part on the calculated percentage (e.g., area percentage) of the human subject's skeleton that is occupied by the initial set of hotspots (e.g., the calculated percentage is the ratio of the total area of ​​the initial set of hotspots divided by the total area of ​​all identified skeletal regions).

[0137] In a particular embodiment, the instructions cause the processor to (h) select a first subset of the initial set of hotspots (e.g., up to all of them) based at least in part on the likelihood of metastasis values ​​(e.g., determining whether to include a particular hotspot in the initial set of hotspots in the subset based on the calculated likelihood of metastasis value for the particular hotspot exceeding a threshold value), and (i) render a graphical representation of the first subset (e.g., visual indications of the hotspots (e.g., points, boundaries) overlaid on one or more members of the set of bone scan images and / or the set of annotated images; e.g., a table listing the identified hotspots along with additional information for each hotspot (e.g., location; e.g., likelihood value)) for display within a graphical user interface (GUI) (e.g., a cloud-based GUI).

[0138] In certain embodiments, the instructions cause the processor to: (j) calculate one or more risk index values ​​for the human subject based at least in part on the calculated percentage (e.g., area percentage) of the human subject's skeleton that is occupied by the first subset of hotspots (e.g., the calculated percentage is the total area of ​​the initial set of hotspots divided by the total area of ​​all identified skeletal regions).

[0139] In certain embodiments, the instructions cause the processor to: (k) receive a user selection of a second subset of the initial set of hotspots via the GUI; and (l) calculate one or more risk index values ​​for the human subject based at least in part on the calculated percentage (e.g., area percentage) of the human subject's skeleton occupied by the second subset of hotspots (e.g., the calculated percentage is the total area of ​​the second subset of hotspots divided by the total area of ​​all identified skeletal regions).

[0140] In certain embodiments, at least one of the risk index values ​​indicates the risk of the human subject having and / or developing metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancers).

[0141] In certain embodiments, the metastatic cancer is metastatic prostate cancer.

[0142] In certain embodiments, at least one of the risk index values ​​indicates that the human subject has a particular state of metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancer).

[0143] In certain embodiments, the system is a cloud-based system. In certain embodiments, the processor is a processor of a cloud-based system.

[0144] In certain embodiments, the GUI is part of a general picture archiving and communication system (PACS) (eg, as well as clinical applications for oncology including lesion marking and quantitative analysis).

[0145] In certain embodiments, the agent (e.g., radiopharmaceutical) is technetium 99m methylenediphosphonate ( 99m Tc-MDP).

[0146] In another aspect, the present invention is directed to a computer-aided image analysis device (e.g., a computer-aided detection (CADe) device; e.g., a computer-aided diagnosis (CADx) device), comprising a system of any one of the aspects and embodiments described herein (e.g., paragraphs

[0083] to

[0145] ).

[0147] In certain embodiments, the device is programmed for use by trained healthcare professionals and / or researchers (e.g., for receipt, transfer, storage, image display, manipulation, quantification, and reporting of digital medical images acquired using nuclear medicine imaging; e.g., the device provides general picture archiving and communication system (PACS) tools, and / or clinical applications for oncology including lesion marking and quantitative analysis).

[0148] In certain embodiments, the device is programmed for use in analyzing bone scan images to evaluate and / or detect metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancers).

[0149] In certain embodiments, the device is programmed for use in analyzing bone scan images to assess and / or detect prostate cancer.

[0150] In certain embodiments, the device is provided with a label specifying that the device is intended for use by trained healthcare professionals and / or researchers [e.g., for receipt, transfer, storage, image display, manipulation, quantification, and reporting of digital medical images acquired using nuclear medicine imaging; e.g., the device provides general picture archiving and communication system (PACS) tools and / or clinical applications for oncology including lesion marking and quantitative analysis].

[0151] In certain embodiments, the label further specifies that the device is used for analyzing bone scan images to evaluate and / or detect metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancers).

[0152] In certain embodiments, the label further specifies that the device is used for analyzing bone scan images to evaluate and / or detect prostate cancer.

[0153] An embodiment described with respect to one aspect of the invention may be applied to another aspect of the invention (e.g., it is contemplated that features of an embodiment described with respect to one independent claim, e.g. a method claim, are applicable to other embodiments of other independent claims, e.g. an apparatus claim, and vice versa).

[0154] The foregoing and other objects, aspects, features, and advantages of the present disclosure will become more apparent and will be better understood by referring to the following description taken in conjunction with the accompanying drawings. [Brief description of the drawings]

[0155] [Figure 1] FIG. 1 is a block flow diagram showing a quality control and reporting workflow for generating a BSI report, according to an illustrative embodiment.

[0156] [Diagram 2]FIG. 2 is a screenshot of a graphical user interface (GUI) for selecting patient data for review, used with a software-based implementation of the quality control and reporting workflow shown in FIG. 1, according to an illustrative embodiment.

[0157] [Diagram 3] FIG. 3 is a screenshot of a graphical user interface (GUI) for reviewing patient information used with a software-based implementation of the quality control and reporting workflow shown in FIG. 1, according to an illustrative embodiment.

[0158] [Figure 4] FIG. 4 is a screenshot of a graphical user interface (GUI) for reviewing patient image data and editing hotspot selections used in conjunction with a software-based implementation of the quality control and reporting workflow shown in FIG. 1, according to an illustrative embodiment.

[0159] [Diagram 5] FIG. 5 is a screenshot of an automatically generated report generated by a user following a software-based implementation of the quality control and reporting workflow shown in FIG. 1, according to an illustrative embodiment.

[0160] [Figure 6] FIG. 6 is a block flow diagram of a process for processing whole body bone scan images and determining a Bone Scan Index (BSI) value, according to an illustrative embodiment.

[0161] [Figure 7] FIG. 7 is a diagram of a whole body bone scan image set showing a skeletal atlas overlaid on ventral and dorsal bone scan images for skeletal segmentation, in accordance with an illustrative embodiment.

[0162] [Figure 8]FIG. 8 is a schematic diagram illustrating construction of a skeletal atlas from multiple patient images, according to an illustrative embodiment.

[0163] [Figure 9-1] FIG. 9A is a screenshot of a GUI window displaying a list of patients presented in a predicate device, according to an illustrative embodiment.

[0164] [Figure 9-2] FIG. 9B is a screenshot of a GUI window displaying a list of patients presented with proposed new devices, according to an illustrative embodiment.

[0165] [Figure 10A-1] FIG. 10A is a screenshot of a GUI window for displaying and inspecting bone scan images and calculated BSI values, according to an illustrative embodiment. [Figure 10A-2] FIG. 10A is a screenshot of a GUI window for displaying and inspecting bone scan images and calculated BSI values, according to an illustrative embodiment. [Figure 10A-3] FIG. 10A is a screenshot of a GUI window for displaying and inspecting bone scan images and calculated BSI values, according to an illustrative embodiment.

[0166] [Figure 10B-1] FIG. 10B is a screenshot of a GUI window for displaying and inspecting bone scan images and calculated BSI values, according to an illustrative embodiment. [Figure 10B-2] FIG. 10B is a screenshot of a GUI window for displaying and inspecting bone scan images and calculated BSI values, according to an illustrative embodiment. [Figure 10B-3] FIG. 10B is a screenshot of a GUI window for displaying and inspecting bone scan images and calculated BSI values, according to an illustrative embodiment.

[0167] [Figure 10C-D] FIG. 10C is a screenshot of a portion of a GUI window showing a color map option for displaying a bone scan image, according to an illustrative embodiment.

[0168] FIG. 10D is a screenshot of a portion of a GUI window showing a color map option for displaying a bone scan image, according to an illustrative embodiment.

[0169] [Figure 11-1] FIG. 11A is a screenshot of a portion of a GUI window for viewing a bone scan image, according to an illustrative embodiment.

[0170] [Figure 11-2] FIG. 11B is a screenshot of a portion of a GUI window for viewing a bone scan image, according to an illustrative embodiment.

[0171] FIG. 11C is a screenshot of a portion of a GUI window for viewing a bone scan image, illustrating the zoom functionality, according to an illustrative embodiment.

[0172] FIG. 11D is a screenshot of a portion of a GUI window for viewing a bone scan image, illustrating the zoom functionality, according to an illustrative embodiment.

[0173] [Figure 12-1] FIG. 12A is a screenshot of a GUI window displaying a bone scan image using an intensity window that covers only a limited range of intensity values.

[0174] FIG. 12B is a screenshot of a GUI window displaying a bone scan image using an intensity window ranging up to a maximum intensity value, according to an illustrative embodiment.

[0175] FIG. 12C is a screenshot of a portion of the GUI of FIG. 12A showing graphical controls for adjusting the intensity window threshold.

[0176] FIG. 12D is a screenshot of a portion of the GUI of FIG. 12B showing additional graphical controls for adjusting the intensity window threshold.

[0177] [Figure 12-2] FIG. 12E is a screenshot of a GUI displaying ventral and dorsal images of a bone scan image set, with each image displayed using a separate intensity window.

[0178] FIG. 12F is a screenshot of a GUI displaying ventral and dorsal images of a bone scan image set, with the same intensity window used for both images.

[0179] [Figure 13] FIG. 13A is a screenshot of a portion of a GUI showing local intensity values ​​displayed at the location of a mouse pointer, according to an illustrative embodiment.

[0180] FIG. 13B is a screenshot of a GUI showing local intensity values ​​at the location of the mouse pointer displayed in the corner (bottom left) of the GUI, according to an illustrative embodiment.

[0181] [Figure 14-1] FIG. 14A is a GUI screenshot showing bone scan images from different studies that are viewable and selectable via different GUI tabs, according to an illustrative embodiment.

[0182] FIG. 14B is a screenshot of a GUI showing bone scan images from different studies displayed simultaneously next to each other, in accordance with an illustrative embodiment.

[0183] [Figure 14-2]FIG. 14C is a GUI screenshot showing ventral and dorsal images displayed and selectable via different GUI tabs, according to an illustrative embodiment.

[0184] FIG. 14D is a screenshot of a GUI showing ventral and dorsal images displayed simultaneously next to each other, with visibility of the various images selectable via checkboxes in the GUI, according to an illustrative embodiment.

[0185] [Figure 15] FIG. 15A is a screenshot of a GUI showing a display of total image intensity and total skeleton intensity, according to an illustrative embodiment.

[0186] FIG. 15B is a screenshot of a GUI showing a display of total image intensity, according to an illustrative embodiment.

[0187] [Figure 16] FIG. 16A is a screenshot of a GUI showing identified hotspots displayed as highlighted regions overlaid on a bone scan image, according to an illustrative embodiment.

[0188] FIG. 16B is a screenshot of a GUI showing identified hotspots displayed as highlighted regions overlaid on a bone scan image, according to an illustrative embodiment.

[0189] [Figure 17-1] FIG. 17A is a screenshot of a GUI showing a hotspot table listing hotspots identified in a bone scan image set, according to an illustrative embodiment.

[0190] [Figure 17-2] FIG. 17B is a screenshot of a GUI showing identified hotspots and updated BSI values ​​after filtering out automatically detected hotspots based on user input via the GUI, in accordance with an illustrative embodiment.

[0191] FIG. 17C is a screenshot of a GUI showing identified hotspots and updated BSI values ​​after including automatically detected hotspots that were previously excluded based on user input via the GUI, in accordance with an illustrative embodiment.

[0192] [Figure 18] FIG. 18A is a GUI screenshot showing pop-up graphical controls for including and / or excluding automatically identified hotspots, according to an illustrative embodiment.

[0193] FIG. 18B is a screenshot of a GUI showing toggle graphical controls for enabling user selection of hotspots to include and / or exclude, according to an illustrative embodiment.

[0194] FIG. 18C is a GUI screenshot showing a window prompting a user to follow a quality control workflow before generating a report, according to an illustrative embodiment.

[0195] [Figure 19] FIG. 19A is a GUI screenshot showing a tabular display of calculated BSI values ​​and included hotspots for various studies, according to an illustrative embodiment.

[0196] FIG. 19B is a screenshot of a GUI showing the display of calculated BSI values ​​as a header above a window showing bone scan images for various studies, according to an illustrative embodiment.

[0197] [Figure 20] FIG. 20A is a screenshot of a GUI displaying a graph showing changes in calculated BSI values ​​over time, according to an illustrative embodiment.

[0198] FIG. 20B is a screenshot of a GUI displaying a graph showing changes in calculated BSI values ​​over time, according to an illustrative embodiment.

[0199] [Figure 21] FIG. 21 is a screenshot of a GUI providing a table listing the number of hot spots in specific anatomical regions used to calculate BSI values ​​in various studies, according to an illustrative embodiment.

[0200] [Figure 22-1] FIG. 22A is a screenshot of an automatically generated report based on a technique for automated analysis of bone scan images and calculation of BSI values, according to an illustrative embodiment.

[0201] [Figure 22-2] FIG. 22B is a screenshot of an automatically generated report based on a technique for automated analysis of bone scan images and calculation of BSI values, according to an illustrative embodiment.

[0202] [Figure 23] FIG. 23 is a diagram of a set of bone scan images overlaid with atlas-based skeletal segmentation comparing the use of a limited atlas (left) with a full-length atlas (right), in accordance with an illustrative embodiment.

[0203] [Figure 24] FIG. 24 is a block flow diagram of an exemplary process for improved image analysis via segmentation of the full length humerus and / or femur regions and enhanced hotspot detection therein, in accordance with an illustrative embodiment.

[0204] [Figure 25-1] FIG. 25A is a graph showing the variation of global threshold scaling ratio as a function of disease burden measure, according to an illustrative embodiment.

[0205] [Figure 25-2] FIG. 25B is a block flow diagram of an exemplary process for improved image analysis using a global threshold scaling approach, in accordance with an illustrative embodiment.

[0206] [Figure 26] FIG. 26 is a block flow diagram of an example process for pre-selecting a first subset of hotspots using global hotspot features, according to an illustrative embodiment.

[0207] [Figure 27] FIG. 27 is an illustration of a set of two bone scan images showing areas around the sacrum, pelvis, and lumbar region, according to an illustrative embodiment.

[0208] [Figure 28] FIG. 28 is a block flow diagram of an exemplary process for calculating a risk index value based on bone invasion rate and a region-dependent correction factor that accounts for potential errors in hotspot localization, according to an illustrative embodiment.

[0209] [Figure 29] FIG. 29 is a graph comparing BSI values ​​calculated via an automated software-based approach according to aspects and embodiments described herein with known analytical standards.

[0210] [Diagram 30] FIG. 30 is a graph showing the reproducibility of automated BSI values ​​between scans for 50 simulated phantoms.

[0211] [Diagram 31] FIG. 31 is a graph showing the reproducibility of automated BSI values ​​between scans for 35 metastatic patients.

[0212] [Figure 32-1]FIG. 32A is a graph showing the performance of a current image analysis software version using an embodiment according to the improvements described herein.

[0213] [Figure 32-2] FIG. 32B is a graph showing the performance of a previous image analysis software version, according to an illustrative embodiment.

[0214] [Figure 33-1] FIG. 33A illustrates an exemplary cloud platform architecture.

[0215] [Figure 33-2] FIG. 33B illustrates an example microservices architecture.

[0216] [Diagram 34] FIG. 34 is a block diagram of an exemplary cloud computing environment for use in certain embodiments.

[0217] [Diagram 35] FIG. 35 is a block diagram of an example computing device and an example mobile computing device for use in certain embodiments. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0218] Features and advantages of the present disclosure will become more apparent from the following detailed description taken in conjunction with the drawings, in which like reference numbers identify corresponding elements throughout and generally indicate identical, functionally similar, and / or structurally similar elements.

[0219] Detailed Description The systems, devices, methods, and processes of the claimed invention are intended to encompass variations and adaptations developed using information from the embodiments described herein. Adaptations and / or modifications of the systems, devices, methods, and processes described herein may be made by one skilled in the art.

[0220] Throughout this description, where articles, devices, and systems are described as having, including, or comprising certain components, or where processes and methods are described as having, including, or comprising certain steps, it is additionally contemplated that there are articles, devices, and systems of the invention that consist essentially of or consist of the recited components, and that there are processes and methods of the invention that consist essentially of or consist of the recited processing steps.

[0221] It should be understood that the order of steps or order for performing certain actions is immaterial so long as the invention remains operable. Moreover, two or more steps or actions may be conducted simultaneously.

[0222] The reference to any document herein, for example in the Background section, is not an admission that the document is prior art with respect to any of the claims presented herein. The Background section is provided for purposes of clarity and is not intended as a description of prior art with respect to any claim.

[0223] Headings are provided for the convenience of the reader, and the presence and / or placement of headings is not intended to limit the scope of the subject matter described herein.

[0224] In this application, the use of "or" means "and / or" unless otherwise specified. As used in this application, the term "comprise" and variations of that word, such as "comprising" and "comprises," are not intended to exclude other additions, components, integers, or steps. As used in this application, the terms "about" and "approximately" are used synonymously. All numbers used in this application with or without "about / approximately" are intended to cover normal variations understood by one of ordinary skill in the art. In certain embodiments, the term "approximately" or "about" refers to a range of values ​​that falls within 25%, 20%, 19%, 18%, 17%, 16%, 15%, 14%, 13%, 12%, 11%, 10%, 9%, 8%, 7%, 6%, 5%, 4%, 3%, 2%, 1%, or less in either direction (greater or less than the reference value) of a stated reference value, unless otherwise specified or otherwise clear from the context (except where such number exceeds 100% of possible values).

[0225] As used herein, the articles "a" and "an" refer to one or to more than one (i.e., to at least one) of the grammatical object of the article. By way of example, "an element" means one element or more than one element. Thus, in this application and the appended claims, the singular forms "a," "an," and "the" include plural references unless the context clearly dictates otherwise. Thus, for example, a pharmaceutical composition containing "an agent" includes references to two or more agents.

[0226] The systems and methods described herein are directed to improved computer-assisted display and analysis of nuclear medicine images. Specifically, in certain embodiments, the systems and methods described herein provide improvements to several image processing steps used in automated analysis of bone scan images to assess a patient's cancer status. For example, improved techniques are provided for image segmentation, hotspot detection, automated classification of hotspots as representative of metastasis, and calculation of risk indicators such as bone scan index (BSI) values. Automated BSI calculation techniques are described in detail in U.S. Patent Application No. 15 / 282,422, filed September 30, 2016, U.S. Patent No. 8,855,387, issued October 7, 2014 (of which U.S. Patent Application No. 15 / 282,422 is a reissue), and PCT Application No. PCT / US17 / 58418, filed October 26, 2017, the contents of each of which are incorporated by reference in their entirety. PCT Application No. PCT / US2017 / 058418, filed October 26, 2017, which is also incorporated herein in its entirety, describes a cloud-based platform that can serve as a platform for providing image analysis and BSI calculation tools according to the techniques described herein.

[0227] Specifically, in certain embodiments, a bone scan image is obtained after administering a drug, such as a radiopharmaceutical, to a human subject. The administered drug accumulates in cancerous bone lesions as a result of the physical properties of the underlying tissue (e.g., increased vasculature, abnormal bone formation) or due to the recognition (by the drug) of a specific biomolecule that is selectively expressed or overexpressed in tumors, such as prostate-specific membrane antigen (PSMA). The drug contains a radionuclide that emits nuclear radiation, which can be detected and therefore used to image the spatial distribution of the drug within the subject.

[0228] For example, in certain embodiments, a gamma camera is used to obtain bone scan images as two-dimensional scans. For example, two images, a ventral image and a dorsal image, are obtained to form a bone scan image set. The body area where the drug accumulates in high concentration appears in the bone scan image as a high intensity area, i.e., a bright spot. The drug may accumulate in other areas, such as cancerous bone lesions, as described above, and in the subject's bladder.

[0229] A series of image processing steps are performed to accurately identify regions of the bone scan image that represent lesions and generate a quantitative estimate of tumor burden. Specifically, the bone scan image is segmented to identify regions that correspond to the bones of the skeleton of the subject, forming a set of annotated images. Regions of high intensity relative to the surroundings are identified within the skeleton regions and compared to a threshold to detect an initial set of hotspots. Features of the initial hotspots, such as hotspot size (e.g., area), hotspot shape (e.g., described by various measures such as radius, eccentricity, etc.), and / or a measure of hotspot intensity (e.g., peak intensity, average intensity, integrated intensity, etc.), are extracted and used to determine, for each hotspot, a metastasis likelihood value that represents the likelihood that the hotspot represents a metastasis. For example, in one particular embodiment, the metastasis likelihood value is calculated using an artificial neural network (ANN) that receives as input a set of hotspot features for each hotspot and outputs a metastasis likelihood value.

[0230] The metastasis likelihood value can be used to automatically filter an initial set of hotspots to determine a subset used to calculate a risk index indicative of the risk of a subject having and / or developing metastatic cancer. By filtering the hotspots in this manner, only those determined to be likely to represent metastasis are included in the risk index calculation. In certain embodiments, a graphical representation of the hotspots and / or their likelihood values ​​is rendered for display to the user, for example as markings and / or a table of information overlaid on the annotated image, allowing the user to select a subset of hotspots to use in the calculation of the risk index. This allows the user to augment the automatic hotspot selection for calculating the risk index with their own input.

[0231] The techniques described herein include several improvements to the image processing steps described above, resulting in improved accuracy of lesion detection and risk index calculation. For example, the present disclosure includes an improved segmentation technique whereby the entire humerus region (e.g., more than three-quarters length) and / or the entire femur region (e.g., more than three-quarters length) are identified. Previously, only limited percentages of the femur and humerus were identified. Segmenting larger (e.g., entire) portions of these bones allows lesions located toward the ends of the subject's arms and legs to be identified. To account for the reduced drug uptake at such ends, the techniques described herein also utilize region-dependent thresholds in the hotspot detection step. The region-dependent thresholds vary for different skeletal regions, with lower values ​​for the femur and humerus regions to increase detection sensitivity in those regions.

[0232] In another improved approach, the systems and methods described herein may use a global threshold scaling technique to detect hot spots. In this approach, an initial set of hot spots is detected by first identifying a set of potential hot spots using multiple prior region-dependent thresholds. The set of potential hot spots is used to calculate a global threshold scaling factor based on the area percentage of the subject's skeleton that is occupied by the set of potential hot spots. The global threshold scaling factor is then used to adjust the prior thresholds, and the adjusted thresholds are used to detect the initial set of hot spots. This approach has been found to ultimately increase the linearity of the risk index calculated using the initial set of hot spots, especially in cases of high disease burden, e.g., when the subject suffers from a large number of lesions and a large percentage of the skeleton is occupied by hot spots.

[0233] The present disclosure also includes improvements regarding hotspot selection and risk index calculation. For example, the techniques described herein may use hotspot pre-selection techniques that filter hotspots based not only on the calculated metastasis likelihood value, but also on global hotspot features that measure the overall properties of the initial set of hotspots, such as the total number of hotspots in the set. Other examples of global hotspot features include other measures of total hotspot count, such as the average number of hotspots per region, measures of overall hotspot intensity, such as peak or average hotspot intensity, measures of total hotspot size, such as total hotspot area or average hotspot size, etc. This allows the processing techniques to take advantage of clinical experience that shows that hotspot selection depends on the rest of the image. Specifically, the probability of a hotspot being selected is higher if there are many other hotspots and lower if it is the only hotspot. Thus, selecting or filtering hotspots based solely on their respective metastasis likelihood values ​​may result in subjects with many hotspots having an underestimated calculated risk index value. Incorporating global features as described herein can improve outcomes in patients with many hotspots.

[0234] Finally, the systems and methods described herein also provide improved techniques for calculating risk index values ​​based on bone invasion, such as the bone scan index (BSI). For example, the BSI is a risk index value that provides an estimate of the percentage of a subject's total skeletal mass that is occupied by cancerous lesions. Calculating the BSI involves calculating the bone invasion rate for each particular hotspot based on the ratio of the area of ​​that particular hotspot to the area of ​​the skeletal region in which it is located. The scaled versions (e.g., to convert area ratios to relative mass) are summed to calculate the subject's BSI value. However, difficulties in pinpointing the specific skeletal region in which a particular hotspot is located can lead to errors in the BSI value. Although bone scan images are two-dimensional images, the underlying subject's skeleton is a three-dimensional structure. As such, hotspots may be erroneously identified as being located in one region when in fact they represent lesions located in a different bone. This is particularly challenging in the case of the sacrum and adjacent pelvic and lumbar regions. To account for this challenge, the present disclosure includes a modified risk index calculation approach that uses a region-dependent correction factor to scale the bone invasion rate in a manner that accounts for potential errors in localizing hot spots. This approach improves the accuracy of the BSI calculation and limits its sensitivity to hot spot localization.

[0235] As such, the systems and methods described herein include several improved image analysis techniques for lesion identification and quantification that improve the accuracy and robustness with which bone scan images can be analyzed. As described herein, they can be used as part of a cloud-based system that facilitates review and reporting of patient data and enables improved disease detection, treatment, and monitoring. A. Nuclear Medicine Imaging

[0236] Nuclear medicine images are obtained using nuclear imaging modalities, such as bone scan imaging, positron emission tomography (PET) imaging, and single photon emission computed tomography (SPECT) imaging.

[0237] As used herein, an "image", e.g., a 3D image of a mammal, includes any visual representation, such as a photograph, a video frame, a streaming video, and any electronic, digital, or mathematical analogue of a photograph, a video frame, or a streaming video. Any device described herein, in certain embodiments, includes a display for displaying an image or any other result produced by a processor. Any method described herein, in certain embodiments, includes a step of displaying an image or any other result produced using the method. As used herein, "3D" or "three-dimensional" with respect to an "image" means conveying information about three dimensions. A 3D image may be rendered as a three-dimensional data set and / or displayed as a set of two-dimensional representations or as a three-dimensional representation.

[0238] In certain embodiments, nuclear medicine images use imaging agents that include radiopharmaceuticals. Nuclear medicine images are obtained after administration of the radiopharmaceutical to a patient (e.g., a human subject) and provide information regarding the distribution of the radiopharmaceutical within the patient's body. A radiopharmaceutical is a compound that includes a radionuclide.

[0239] As used herein, "administering" an agent means introducing a substance (e.g., an imaging agent) into a subject. Generally, any route of administration may be used, including, for example, parenteral (e.g., intravenous), oral, topical, subcutaneous, peritoneal, intra-arterial, inhalation, vaginal, rectal, nasal, introduction into the spinal fluid, or injection into a body compartment.

[0240] As used herein, "radionuclide" refers to a moiety that comprises a radioactive isotope of at least one element. Exemplary suitable radionuclides include, but are not limited to, those described herein. In some embodiments, the radionuclide is a radionuclide used in positron emission tomography (PET). In some embodiments, the radionuclide is a radionuclide used in single photon emission computed tomography (SPECT). In some embodiments, a non-limiting list of radionuclides includes: 99m Tc, 111 In, 64 Cu, 67 Ga, 68 Ga, 186 Re, 188 Re, 153 Sm, 177 Lu, 67 Cu, 123 I, 124 I, 125 I, 126 I, 131 I, 11 C. 13 N, 15 O. 18 F, 153 Sm, 166 Ho, 177 Lu, 149 Pm, 90 Y, 213 Bi, 103 Pd, 109 Pd, 159 Gd, 140 La, 198 Au, 199 Au, 169 Yb, 175 Yb, 165 Dy, 166 Dy, 105 Rh, 111 Ag, 89 Zr, 225 Ac, 82 Rb, 75 Br, 76 Br, 77 Br, 80 Br, 80m Br, 82 Br, 83 Br, 211 At, and 192Contains Ir.

[0241] As used herein, the term "radiopharmaceutical" refers to a compound that contains a radionuclide. In certain embodiments, radiopharmaceuticals are used for diagnostic and / or therapeutic purposes. In certain embodiments, radiopharmaceuticals include small molecules labeled with one or more radionuclides, antibodies labeled with one or more radionuclides, and antigen-binding portions of antibodies labeled with one or more radionuclides.

[0242] Nuclear medicine images (e.g., PET scans; e.g., SPECT scans; e.g., whole body bone scans; e.g., composite PET-CT images; e.g., composite SPECT-CT images) detect radiation emitted from radionuclides in radiopharmaceuticals to form images. The distribution of a particular radiopharmaceutical within a patient's body can be determined by biological mechanisms such as blood flow and perfusion, as well as by specific enzyme or receptor binding interactions. Different radiopharmaceuticals are designed to exploit different biological mechanisms and / or specific unique enzyme or receptor binding interactions, and thus, when administered to a patient, may preferentially concentrate within certain types of tissues and / or regions within the patient's body. Regions within the patient's body that have a higher concentration of the radiopharmaceutical will emit more radiation than other regions, and those regions will appear brighter in the nuclear medicine images. Thus, the variation in intensity within the nuclear medicine images can be used to map the distribution of the radiopharmaceutical within the patient's body. This mapped distribution of the radiopharmaceutical within the patient's body can be used to infer, for example, the presence of cancerous tissue within various regions of the patient's body.

[0243] For example, when administered to a patient, technetium 99m methylenediphosphonate ( 99m Tc The radiopharmaceutical (MDP) selectively accumulates in the skeletal regions of a patient, particularly at sites of abnormal bone formation associated with malignant bone lesions. The selective concentration of the radiopharmaceutical at such sites creates identifiable hot spots, i.e., localized areas of high intensity in nuclear medicine images. Thus, by identifying such hot spots in a patient's whole body scan, the presence of malignant bone lesions associated with metastatic prostate cancer can be inferred. As explained below, risk indicators are provided to patients that correlate with the patient's overall survival and other prognostic measures indicative of disease status, progression, treatment response, etc. 99m The dose can be calculated based on an automated analysis of the intensity variability within a whole-body scan obtained after administration of the Tc MDP. In certain embodiments, other radiopharmaceuticals can also be used. 99m It can be used in the same manner as Tc MDP.

[0244] In certain embodiments, the particular radiopharmaceutical used will depend on the particular nuclear medicine imaging modality being used. For example, 18F sodium fluoride (NaF) can also be used. 99m It accumulates in bone lesions in the same way as Tc MDP, but can be used in PET imaging. In certain embodiments, PET imaging may use a radioactive form of the vitamin choline, which is readily taken up by prostate cancer cells.

[0245] In certain embodiments, radiopharmaceuticals may be used that selectively bind to specific proteins or receptors of interest, particularly those whose expression is increased in cancer tissue, including, but not limited to, tumor antigens such as CEA expressed in colorectal carcinoma, Her2 / neu expressed in several cancers, BRCA1 and BRCA2 expressed in breast and ovarian cancers, and TRP-1 and -2 expressed in melanoma.

[0246] For example, human prostate-specific membrane antigen (PSMA) is upregulated in prostate cancer, including metastatic disease. PSMA is expressed by virtually all prostate cancers. and its expression is further increased in poorly differentiated metastatic carcinomas and hormone refractory carcinomas. Therefore, a radiopharmaceutical corresponding to a PSMA binding agent (e.g., a compound having high affinity for PSMA) labeled with one or more radionuclides can be used to obtain nuclear medicine images of a patient, from which the presence and / or state of prostate cancer in various regions of the patient (e.g., including, but not limited to, the skeletal region) can be assessed. In certain embodiments, when the disease is localized, nuclear medicine images obtained using a PSMA binding agent are used to identify the presence of cancer tissue within the prostate. In certain embodiments, as is the case when the disease is metastatic, nuclear medicine images obtained using a radiopharmaceutical containing a PSMA binding agent are used to identify the presence of cancer tissue in various regions, including not only the prostate but also other organ and tissue regions such as the lungs, lymph nodes, and bones.

[0247] Specifically, when administered to a patient, the radionuclide-labeled PSMA-binding agent selectively accumulates in cancer tissue based on its affinity for PSMA. 99m Similar to that described above with respect to Tc MDP, selective concentration of the radionuclide-labeled PSMA binding agent at specific sites within the patient's body creates detectable hot spots within nuclear medicine images. As the PSMA binding agent concentrates within various cancer tissues and areas of the body that express PSMA, localized cancer within the patient's prostate and / or metastatic cancer in various areas of the patient's body can be detected and evaluated. Risk indices that correlate with the patient's overall survival and other prognostic measures indicative of disease state, progression, treatment response, etc., can be calculated based on automated analysis of intensity variability within nuclear medicine images obtained after administration of the PSMA binding agent radiopharmaceutical to the patient.

[0248] To detect and evaluate prostate cancer, various radionuclide-labeled PSMA binding agents may be used as radiopharmaceutical imaging agents for nuclear medicine imaging. In certain embodiments, the specific radionuclide-labeled PSMA binding agent used depends on factors such as the specific imaging modality (e.g., PET; e.g., SPECT) and the specific region (e.g., organ) of the patient to be imaged. For example, certain radionuclide-labeled PSMA binding agents are suitable for PET imaging, while others are suitable for SPECT imaging. For example, some radionuclide-labeled PSMA binding agents facilitate imaging of the patient's prostate and are used primarily when the disease is localized, while other PSMA binding agents facilitate imaging of organs and regions throughout the patient's body and are useful for evaluating metastatic prostate cancer.

[0249] Various PSMA binding agents and radionuclide-labeled versions thereof are described in U.S. Patent Nos. 8,778,305, 8,211,401, and 8,962,799, each of which is incorporated herein by reference in its entirety. Several PSMA binding agents and radionuclide-labeled versions thereof are also described in PCT Application No. PCT / US2017 / 058418, filed October 26, 2017, which is incorporated herein by reference in its entirety. B. Bone-scanning imaging device for image analysis

[0250] In certain embodiments, the computer-assisted image analysis devices described herein are intended for use by trained health care professionals and researchers for the receipt, transfer, storage, image display, manipulation, quantification, and reporting of digital medical images acquired using nuclear medicine (NM) imaging. In certain embodiments, such devices provide general picture archiving and communication system (PACS) tools as well as clinical applications for oncology, including lesion marking and quantitative analysis. C. Analysis of bone scan images and calculation of bone scan index values

[0251] Bone scintigraphy (also called bone scan imaging) is a widely used imaging modality to assess skeletal disease burden. The current standard for assessing disease progression based on bone scan images is based on the semiquantitative modified Prostate Cancer Working Group 2 and Prostate Cancer Working Group 3 (PCWG) criteria. The definition of this criterion relies on the appearance of new lesions interpreted by a trained reader as either: (i) two new lesions on the first and second post-scans compared to the pre-treatment scan followed by two additional lesions (known as the 2+2 rule), or (ii) two new confirmed lesions on the first subsequent post-scan. However, this semiquantitative approach to lesion counting has several limitations. Specifically, it is subject to manual variability, is limited in its assessment of disease progression, and cannot assess the exact burden of confluent / spread disease, e.g., associated with metastatic prostate cancer.

[0252] Therefore, there is a high unmet need for an automated quantitative assessment of bone scans. The automated Bone Scan Index (BSI), developed by EXINI Diagnostics AB, Lund, Sweden, is a fully quantitative assessment of a patient's skeletal disease on bone scan as a percentage of total skeletal mass. The BSI has undergone rigorous pre-analytical and post-analytical validation as an objective measure of quantitative change in disease burden bone scans. In a recent phase 3 prospective study, assessment of the BSI was demonstrated to stratify patients with metastatic prostate cancer.

[0253] The systems and methods described herein are directed to improved computer-assisted techniques for analyzing bone scan images and calculating BSI in an automated and semi-automated, user-guided manner. Also described herein are GUI tools that facilitate user review and automated analysis of bone scan images used to determine BSI values. D. Device Description of Exemplary aBSI Platform

[0254] In certain embodiments, the systems and methods described herein can be implemented as a cloud-based platform for automated and semi-automated image analysis for detection and evaluation of cancer status in a patient.An exemplary device described herein is an automated BSI device (aBSI), which is a cloud-based software platform with a web interface that allows users to upload bone scan image data in the form of a specific image file, such as a DICOM file.The software complies with the Digital Imaging and Communications in Medicine (DICOM 3) standard.

[0255] In certain embodiments, a device (e.g., a computer-assisted image analysis tool) according to the systems and methods described in the present invention is programmed for an intended user, typically a health care professional who uses the software to view patient images and analyze the results. The user operates the service within a web browser (such as Google's Chrome browser) on a computer running an operating system such as Microsoft Windows or OSX. The software may be configured to occupy a single application window. The service is web-based and accessed via a specific URL. Keyboard and mouse controls may be used to operate the software.

[0256] Multiple scans can be uploaded for each patient, and the system provides a separate automated analysis based on the images for each. The automated analyses are reviewed by a physician, who may be guided through a quality control and reporting workflow. Once the automated assessment quality control is certified, a report may be generated and signed. The service may be configured to be HIPAA and 21 CFR part 11 compliant. i. Access to the Services

[0257] In certain embodiments, access to software tools in accordance with the systems and methods described herein is restricted and protected by security measures. For example, access to aBSI, a cloud-based implementation of the systems and methods described herein, is protected by multi-factor authentication in the form of a username, password, and a verification code sent as a text message to a phone number associated with the account. ii. System Requirements

[0258] In one particular embodiment, the requirements enforced by the software include one or more of the following: A computer with Windows or OS X with Internet access; Chrome browser, ·Available personal mobile phone (for multi-factor authentication only).

[0259] In one particular embodiment, the requirements enforced by the user include one or more of the following: Chrome Browser a. At least version 54 b. JavaScript must be enabled c. HTML5 is required d. Writing to local storage and session storage is required Display resolution of at least 1280x960 iii. Image Requirements

[0260] In one particular embodiment, the requirements enforced by the software include one or more of the following: · Images must be decompressed in DICOM3 format. Modality (0008, 0060) must be "NM" Image type (0008,0008) must be "ORIGINAL\PRIMARY\WHOLE BODY\EMISSION" Survey date (0008, 0020) must contain a valid date - Number of frames (0028, 0008) must be 1 or 2 The number of slices (0054, 0081) must be 1 or 2 Pixel spacing (0028, 0030) must be ≥ 1.8mm / pixel and ≤ 2.8mm / pixel The image must be shaped such that the number of rows is greater than or equal to the number of columns. · Patient gender (0010, 0040) must be M The ventral and dorsal images may be stored as two different series (two files with different Series Instance UIDs) or as one multi-frame series containing the two frames (one file). Image pixel data should be in the 16-bit range. Images with pixels in the range 0-255 (8-bit) are not sufficient.

[0261] In one particular embodiment, the requirements enforced by the user include one or more of the following: · Ventral and dorsal images should cover at least the area from the scalp to the top of the tibia and top of the forearm of each arm. No filtering or other post-processing techniques may be applied to the images.

[0262] In certain embodiments, whole body bone scintigraphy images are obtained in compliance with relevant guidelines such as the "EANM Bone Scintigraphy: Procedural Guidelines for Tumor Imaging" and the "ACR-SPR Practice Parameters for Performing Bone Scintigraphy (Bone Scan)." iv. Workflow

[0263] 1 is a block flow diagram showing a quality control and reporting workflow 100 for generating a BSI report, according to an illustrative embodiment. In certain embodiments, the systems and methods described herein include a GUI to guide a user (e.g., a healthcare professional) through the review and analysis of patient image data to calculate an automated BSI value and generate a report. The workflow 100 allows the user to select 102 and upload 104 an image file of patient data, which is then analyzed by the software and reviewed by the user.

[0264] The user may be presented with a first GUI window, such as window 200 shown in Figure 2, in which the user can select a particular patient from a list for review / analysis. Upon selecting a row in the list of Figure 2 corresponding to a particular patient, a new and / or updated window 300, shown in Figure 3, is displayed showing patient information.

[0265] Turning to FIG. 4, the user may then access a guided analysis software review page 400. Review page 400 provides the user (e.g., a healthcare professional such as a physician) with a GUI where the user can review the image data and the automated identification of hotspots performed by the software backend to calculate an automated BSI index for the patient. The automated BSI calculation technique is described in further detail herein, as well as in earlier versions (without the improvements of this disclosure) in U.S. Patent Application No. 15 / 282,422, filed September 30, 2016, U.S. Patent No. 8,855,387, issued October 7, 2014 (of which U.S. Patent Application No. 15 / 282,422 is a reissue), and PCT Application No. PCT / US17 / 58418, filed October 26, 2017, the contents of each of which are incorporated herein by reference in their entirety.

[0266] The review page allows a user to review 106 hotspots representing cancerous lesions that have been automatically identified in the image by the software. The user can use the review page GUI to edit the set of regions identified as hotspots 108 and must confirm that the image quality, bone segmentation (depicted by contour lines in the screenshot), and set of identified hotspots have been reviewed and accepted 110 to proceed to report generation. Once the user review and quality control are confirmed 112, a report such as report 500 shown in FIG. 5 can be generated 114, including the final BSI calculation. v. Image processing

[0267] As shown in the block flow diagram of FIG. 6, the systems and methods described herein may be used in a process 600 for automated detection and pre-selection of hotspots, receiving user verification of a set of pre-selected hotspots, and calculating a risk index value, such as a Bone Scan Index (BSI) value, among others.

[0268] In particular, in a particular embodiment, as shown in FIG. 6, in a segmentation step 602, the software tool described herein uses an image registration algorithm to place the patient images into a reference coordinate frame. This is done by non-rigidly adjusting the atlas images to each patient image. By fitting the atlas images to the patient, the patient images can be segmented into skeletal and background regions. The skeletal regions can be further divided into smaller skeletal regions of interest, also referred to herein as localization. In a normalization step 604, the whole body bone scan images are normalized to provide the user with a standardized intensity range when viewing images with different contrast and brightness levels. In a hotspot detection step 606, hotspots are detected. In a particular embodiment, hotspots are detected using thresholding rules. In another step, hotspot preselection 608 is performed. In a particular embodiment, hotspot preselection is based on image analysis and machine learning techniques aimed at preselecting important hotspots for inclusion in a set of preselected hotspots that can be inspected by the user in a hotspot validation step 610. This pre-selection is based on a range of characteristics such as hotspot size, location, orientation, shape, and texture. In certain embodiments, the pre-selection is a user convenience tool, aimed at reducing the number of manual clicks a user must make (e.g., to select hotspots to use in calculating a risk index such as a BSI value). In certain embodiments, the pre-selection step may be followed by a mandatory hotspot validation step, in which the user must review and approve the hotspot pre-selection, or, if required, manually include and / or exclude hotspots to be able to generate a report.

[0269] In another step 612, a specific risk indicator called the Bone Scan Index (BSI) is calculated using the set of validated hotspots. In a specific embodiment, the Bone Scan Index (BSI) is defined as the sum of the bone involvement of all included hotspots. Involvement is an estimate of the proportion of the total skeletal mass contributed by the volume corresponding to one hotspot, expressed as a percentage. Involvement may be calculated using the following formula, where C is the anatomical area factor related to the density of bone:

number

[0270] In one particular embodiment, skeletal segmentation is performed using a segmentation by registration approach, where an atlas image is registered with the bone scan image to be segmented. In this approach, the device automatically contours the skeleton into separate skeletal regions by elastically fitting a manually contoured annotated template image set to each analyzed image set. This annotated template image set is known as a skeletal atlas image set. This atlas image set is structured the same as any patient image, i.e., it looks like a normal bone scan and includes one ventral image and one dorsal image. The atlas image provides a fixed reference for analyzing the scans. The atlas is manually annotated with regions of interest (skeletal regions) that can be transferred to new scans to accurately calculate the BSI. An exemplary atlas image set 700 with 31 manually drawn skeletal regions is shown in FIG. 7. As shown in FIG. 7, like the bone scan image set, the atlas image set 700 can include a ventral image 702 and a dorsal image 704.

[0271] When the bone scan image is analyzed, the atlas image is elastically deformed to resemble the bone scan image. The same transformation is then applied to the contours of the atlas, thereby producing a contour delineation / segmentation of each skeletal region of interest in the patient's bone scan image. Additional details regarding the construction of a suitable atlas are provided at the end of this section.

[0272] Deforming the skeletal atlas to the patient scan. In one particular embodiment, the deformation of the skeletal atlas to the patient's bone scan follows an iterative approach. The segmentation algorithm proceeds in an iterative manner, and at each iteration, for each pixel, a vector is estimated that describes how that pixel should be displaced to a corresponding position in the target image. With individual displacements for each pixel, the possibility is that the displacement vector will intersect or share the target position, resulting in holes and / or cracks in the deformed image. To avoid this, a filtering approach is used to smooth the vector field. The displacements are estimated by applying complex-valued filters to the atlas and the target image. The complex filter response can be expressed as a pixel-by-pixel amplitude and phase. It can be shown that the local phase difference, i.e. the phase difference between pixels within a short distance from each other, is proportional to the size of the displacement required to align them. To obtain an estimate of the direction of the displacement, this process is repeated several times for different filter angles. Knowing the angle of each filter and the resulting magnitude of the displacement makes it possible to infer the direction in which the largest displacements are likely to be observed. While the technique works for small displacements, it must also be applicable when the atlas and target images are far apart. To achieve this, a subsampling approach is taken, where the algorithm is first applied to a subsampled (reduced size) version of the image. This approach treats large displacements as local differences. The algorithm then proceeds to increasingly more detailed (less subsampled) images, adding more detail and variability to the resulting displacement field. The algorithm is run on a fixed pyramid of subsampled images, for a fixed number of iterations at each pyramid level, and with a predefined level of smoothing.

[0273] Skeletal Atlas Construction. The exemplary aBSI device described herein relies on a single atlas image. The contouring algorithm is driven by structural information in the atlas image and the target image, and attempts to deform the atlas image such that the distance between similar structures in the two images is minimized. Structures are defined by edges and ridges (lines) in each image. Thus, global intensity differences and texture patterns are ignored by the algorithm. As a result, a good atlas image exhibits two important properties: Show edges and ridges of the same pattern, and The (elastic) transformation required to align the images is generally minimized over the set of expected anatomical variations between the images being analyzed.

[0274] In one particular embodiment, to meet these requirements, an atlas image based on a database of actual normal (free of metastases or other visible medical conditions) bone scan images is used. A contouring algorithm can be used to align all images in the database with each other. An average transformation is then calculated from all the resulting transformations. All images are then transformed to this anatomical average to represent the average anatomical structure in the database. In this process, the intensity is also normalized to create a representative bone scan image suitable as an anatomical reference. The schematic diagram shown in FIG. 8 conceptually illustrates this idea, i.e., how the average anatomical structure and intensity (middle image 802) can be inferred from multiple bone scan images 804a, 804b, 804c, 804d in the database.

[0275] In certain embodiments, the average anatomical structures used in the atlas image set rapidly converge to a stable estimate as the number of included scans is increased. A relatively small number of scans (e.g., 30) may be sufficient to create a representative reference image. Furthermore, because the algorithm is driven by the dominant structures in the image and is insensitive to differences in shape and / or size, a single atlas may be applied to any bone scan image for skeletal segmentation. Intensity normalization and hotspot detection

[0276] One of the challenges when reading scintigraphy images such as bone scans is that intensity levels between scans may vary due to various parameters such as injection dose, time from injection to scan, scan duration, body type, camera hardware and settings, etc. In one particular embodiment, to facilitate reading for the user and as part of the quantification pipeline (e.g. as shown in FIG. 6), the input bone scan is normalized such that the average intensity of healthy bone tissue is scaled to a predefined reference level. At this stage of the quantification pipeline, skeleton segmentation has already been performed and the pixels that belong to the skeleton are known. However, in order to measure the average intensity of healthy tissue and thus be able to normalize the scan, regions of high intensity must be identified and excluded. Detecting this hotspot in the skeleton is easy if the image is already normalized. This is a chicken-and-egg kind of problem, where hotspot detection is needed to normalize the image, which in turn depends on the normalized image. Therefore, an iterative approach such as the steps listed below can be used to address this challenge. (1) Estimate normalization assuming all bone tissue is healthy (no hot spots). (2) Detect hotspots in light of the current normalization. (3) Estimate the normalization given the current set of hotspots. (4) Iterate steps (2) and (3) until convergence.

[0277] This iterative process converges to stable values ​​for both normalization and hotspot detection within three or four iterations. Hotspots are detected using a simple thresholding technique, where the image is filtered using a difference-of-Gaussians bandpass filter that emphasizes small regions of high intensity relative to their surroundings. This filtered image is then thresholded to a certain level based on the region.

[0278] In certain embodiments, different thresholds are used for different skeletal regions of interest, for example, the threshold levels used in an exemplary embodiment of cloud-based aBSI are 650 for the cervical spine, clavicle, ribs, lumbar spine, pelvis, sacrum, scapula, skull, thoracic spine, sternum, and 500 for the femur and humerus.

[0279] The output of the hotspot detection is a set of ROIs (regions of interest) that represent the hotspots in the image, as well as normalization coefficients that are used to set initial maximum and minimum thresholds for windowing of the image. Hotspot preselection

[0280] Hot spots may be classified as either included or excluded for pre-selection using a learning (data-driven) approach based on an artificial neural network (ANN). The ANN may be trained / trained on a training database containing patients ranging from normal bone scans to bones with multiple widespread hot spots.

[0281] Each hotspot in the training database is characterized using a set of features (measurements) relating to their size, location, orientation, shape, and texture. These features are first fed to the ANN during a training phase where the parameters of the ANN are set to maximize classification performance in a cross-validation study, and then in the actual software to classify the hotspot as inclusive or exclusive.

[0282] The preselection training analyzes hotspots and their immediate neighborhood in the image, making the classifier robust to large scale differences in the training material. Thus, the classifier is applicable to a wide range of input data. However, small differences in performance between cohorts can be expected. To avoid the influence of localization on the ANN parameters, separate networks can be built for different bone localizations. The set of input features to each of these ANNs will then be somewhat different. For example, symmetric features will only apply to localizations that have a natural symmetric counterpart.

[0283] In addition, hot spots in the training set are usually manually marked as included or excluded by medical experts trained to read bone scans. The target labels can be verified by a second medical expert. In one exemplary approach, the US and European procedural guidelines for bone scintigraphy are aligned, and the equipment used to obtain bone scans is the same in the US and Europe. Furthermore, the standards used to decode bone scans for clinical trials, such as the Prostate Working Group 2 standards, are used worldwide. This is based on the common knowledge in the nuclear medicine community that the variability in the appearance of bone scans due to cancer disease is much more pronounced compared to the small variability that can be measured within normal bone density, for example, between different races. The parameters of the ANN are optimized such that the resulting classifier mimics the selection of medical experts. To avoid bias towards the training set, cross-validation techniques can be used.

[0284] In one particular embodiment, while pre-selection saves the reader time, all hotspots are reviewed and approved by the reader before the report is generated.

[0285] The ANN described herein may be implemented via one or more machine learning modules. As used herein, the term "machine learning module" refers to a computer-implemented process (e.g., function) that implements one or more specific machine learning algorithms to determine one or more output values ​​for a given input (an image (e.g., a 2D image; e.g., a 3D image), a dataset, etc.). For example, a machine learning module may receive as input a 3D image (e.g., a CT image; e.g., an MRI) of a subject and determine, for each voxel of the image, a value that represents the likelihood that the voxel is in a region of the 3D image that corresponds to a representation of a particular organ or tissue of the subject. In certain embodiments, two or more machine learning modules may be combined and implemented as a single module and / or a single software application. In certain embodiments, two or more machine learning modules may be implemented separately, e.g., as separate software applications. The machine learning modules may be software and / or hardware. For example, the machine learning modules may be implemented entirely as software, or certain functions of the CNN module may be performed via specialized hardware (e.g., an application specific integrated circuit (ASIC)). E. Graphical User Interfaces and Image Displays

[0286] In certain embodiments, the systems and methods described herein include a graphical user interface for reviewing patient data and images. The GUI allows a user to review a list of patients and select a patient for reviewing and analyzing images. FIGS. 9A and 9B show an exemplary GUI window that provides a list of patients from which a particular patient can be selected. Once a particular patient is selected, the user can view the bone scan images of that patient and review the automated analysis (e.g., hotspot detection and pre-selection) using a GUI such as the GUI shown in FIGS. 10A and 10B. The bone scan images may be colored using various color maps. FIGS. 10C and 10D show exemplary color maps that may be used.

[0287] In certain embodiments, a GUI system according to the techniques described herein facilitates viewing of images. For example, automatic adjustment of the image to the screen size may be provided. FIG. 11A shows a GUI embodiment without this functionality, where the user clicks on a zoom icon to change the size of the image. FIG. 11B shows a screenshot of a GUI where the image is automatically resized to fill the screen vertically. FIGS. 11C and 11D show two techniques for providing zoom functionality for detailed viewing of an image. In the embodiment shown in FIG. 11C, zooming in around the mouse pointer occurs when the scroll wheel button is clicked and held. In the embodiment shown in FIG. 11D, zoom and pan functionality is provided using the mouse scroll wheel and click and drag actions.

[0288] Figures 12A-12F show intensity windowing techniques for displaying images. In particular, Figures 12B and 12D show GUI embodiments that implement a custom intensity windowing slider that facilitates user adjustment of the intensity window threshold.

[0289] 13A and 13B show screenshots of an exemplary GUI window in which local intensity is displayed. Specifically, in one particular embodiment, local intensity is displayed when the user hovers the mouse pointer over the image. In the embodiment shown in FIG. 13A, the local intensity value is displayed at the location of the mouse pointer. In the embodiment shown in FIG. 13B, the intensity is displayed in the bottom left corner of the image instead of next to the mouse pointer and obscuring part of the image (e.g., providing increased security accordingly).

[0290] 14A-14D show exemplary GUI embodiments for displaying multiple bone scan images. The embodiment shown in FIG. 14A and FIG. 14C uses a tabular approach to allow the user to switch between different sets of bone scan images for different studies, or between ventral and dorsal images. The embodiment shown in FIG. 14B and FIG. 14C uses checkboxes to toggle the visibility of various images, allowing multiple images from different studies to be displayed side-by-side with each other.

[0291] In certain embodiments, embodiments of the GUI tools for image analysis described herein may provide information useful for quality control. For example, total image intensity may be displayed. Figure 15A is a screenshot of a GUI showing the display of total image intensity and total skeleton intensity according to an illustrative embodiment. In certain embodiments, such as the GUI shown in Figure 15B, only total image intensity (and not total skeleton intensity) is displayed to provide a simpler user interface (e.g., to avoid confusion).

[0292] A GUI tool for inspecting the bone scan image may show a graphical indication of the detected (e.g., pre-selected) hotspots superimposed on the bone scan image, as shown, for example, in FIG. 16A and FIG. 16B. In a particular embodiment, a table of hotspots is also displayed, as shown in FIG. 17A. In a particular embodiment, the GUI tool described herein allows the user to select or deselect pre-selected hotspots for inclusion in the final set of selected hotspots used to calculate the BSI value. The GUI may display the obtained BSI value 1602 above the image, as shown in FIG. 17B and FIG. 17C, allowing the user to observe the changes (compare 1704a to 1704b) as the user selects and / or deselects various hotspots (1702a and 1702b). Various graphical control techniques may be used to allow the user to select pre-selected hotspots for inclusion or exclusion. For example, FIG. 18A shows a pop-up control that appears after a user's right-click action. FIG. 18B shows a toggle switch 1802 ("Edit Hotspots") that can be turned on and off. When the toggle switch is turned on, the user clicks on various hotspots to select or deselect them. In certain embodiments, the GUI tool may include a safety / quality control feature, such as the pop-up window shown in FIG. 18C that prompts the user to verify quality control requirements before a report is generated.

[0293] In certain embodiments, as the BSI values ​​for the various studies are calculated, they are displayed for user inspection. For example, the calculated BSI values ​​may be displayed in a table, as shown in FIG. 19A. In certain embodiments, the calculated BSI values ​​for the different studies are displayed above the displayed bone scan images for their corresponding studies, as shown in FIG. 19B. A chart showing the time evolution of the calculated BSI values ​​may also be displayed, as shown in FIGS. 20A and 20B. Additional information regarding the calculated BSI values ​​may be displayed. FIG. 21 shows a screenshot of a GUI providing a table listing the number of hot spots in a particular anatomical region used to calculate the BSI values ​​for different studies, according to an illustrative embodiment. In certain embodiments, the systems and methods described herein allow for automated report generation. FIGS. 22A and 22B show examples of automatically generated reports. F. Improved Image Processing Methods

[0294] In certain embodiments, the systems and methods described herein include improvements to one or more of the image processing steps shown in FIG. i. Skeleton Segmentation

[0295] As described herein, a skeletal atlas image set may be used for image segmentation. The skeletal atlas image set includes a pair of template bone scan images (ventral and dorsal) representing a typical normal bone scan, and manual contouring of 31 skeletal regions. The regions are warped to fit the current patient image to be analyzed. In one particular embodiment, a restricted atlas is used with skeletal identification covering only three-quarters of the femur and humerus. In one particular embodiment, an improved full-length atlas is used, including identification of skeletal regions covering the entire femur and humerus. FIG. 23 shows a set of bone scan images overlaid with atlas-based skeletal segmentation, comparing the use of a restricted atlas 2300a (left) and a full-length atlas 2300b (right). As shown in FIG. 23, the limited atlas 2300a includes only three-quarters or less of the humerus 2302a and femur 2304a, whereas the full-length atlas includes well over three-quarters of the length of the humerus 2302b and femur 2304b. The use of a full-length atlas can increase the stability of skeletal segmentation. Specifically, the augmented atlas allows the knee and elbow to be utilized as references during alignment. Such reference points are beneficial for image analysis because of their clear contrast, providing increased safety. The augmented atlas also allows detection of hot spots further down the limbs. ii. Hotspot detection threshold

[0296] As described herein, an initial set of candidate hotspots is found using image intensity thresholding. In one particular embodiment, a global fixed threshold is used such that the same value is used across all skeletal regions of interest and all images. Another improved approach sets a variable region threshold across different skeletal regions of interest. For example, this approach allows a reduced threshold to be used for the femur and humerus regions (e.g., lowered from 650 to 500) to increase detection sensitivity. The femur and humerus regions show less uptake in the bone scan image than other skeletal regions. Therefore, a lower threshold can be used for those regions to achieve a similar level of sensitivity as the rest of the body. Setting individual thresholds for different skeletal regions enables this functionality, which leads to increased detection of lower intensity hotspots within those skeletal regions.

[0297] 24 illustrates an exemplary process 2400 for lesion marking and analysis using the improved segmentation and region-dependent thresholding approach described above. In the exemplary process 2400, a set of bone scan images of a human subject is accessed 2410. Each member image of the bone scan image set is automatically segmented 2420 to identify skeletal regions of interest, including a femur region 2422 corresponding to more than three-quarters of the length of the subject's femur and / or a humerus region 2424 corresponding to more than three-quarters of the length of the subject's humerus. The identified skeletal regions of interest are then analyzed to automatically detect 2430 an initial set of hot spots. As described herein, this step of automatically detecting hot spots may include applying a thresholding operation to each skeletal region of interest using a region-dependent threshold, i.e., a threshold whose value is not uniform but is different for each skeletal region. Specifically, lower threshold values ​​for the femur and / or humerus regions (2432 and 2434, respectively) may be used to increase detection sensitivity in the femur and / or humerus regions, thereby accounting for reduced uptake of the agent (e.g., radiopharmaceutical) therein.

[0298] Once an initial set of hotspots is detected, a set of hotspot features is extracted 2440 for each hotspot, which is used to calculate a metastasis likelihood value for each hotspot 2450. The detected hotspots may be rendered for graphical display to a user 2460, in certain embodiments, along with additional information, such as the calculated likelihood value. The hotspots may be filtered (preselected) based on the calculated likelihood value for inclusion in a first subset for either being presented to the user and / or being used to calculate a risk index, such as a BSI value. The user may inspect the detected hotspots, e.g., the initial hotspot set or the first subset including the filtered hotspots, via a graphical display, and confirm or reject the hotspots for inclusion in a second, final subset. This final subset may then be used to calculate a risk index value, thereby incorporating the user's expertise into the decision-making process.

[0299] In one particular embodiment, a global dynamic threshold adjustment is used. This method examines the resulting BSI values ​​and fine-tunes the global threshold scaling factor to accommodate high burden disease. The scaling is calculated according to the following formula:

number

[0300] Specifically, in one particular embodiment, the global scaling rate approach is a data-driven approach that takes into account errors that may lead to underestimation of BSI values ​​at high levels of disease, i.e., high levels of metastasis. Such errors were discovered using a simulation system that allows bone scan images to be simulated for patients with any (e.g., selected) degree of disease. The simulation system creates realistic bone scan images and also takes into account specific camera and examination parameters. Thus, realistic bone scan images can be created such that for known specific input parameters, the ground truth in terms of lesion volume, skeletal volume, and therefore BSI value, is known. Thus, this approach allows BSI values ​​calculated via the image analysis approach described herein to be checked against and compared to known ground truth BSI values ​​determined from the input parameters of the image simulation. By running multiple simulations with varying degrees of disease burden, it was demonstrated that previous systems that do not use the global threshold scaling approach described herein nonlinearly underestimate BSI values ​​for higher disease burdens. The form of the nonlinear function used for the global threshold scaling rate in Equation 1 was based on the pattern of errors observed in the simulation study and is intended to correct for the observed nonlinear underestimation of the calculated BSI values.

[0301] As shown in the graph, which is a graph of the scaling factor as a function of b, the a priori threshold is multiplied by the global threshold scaling factor to reduce the threshold as disease burden increases as measured by hotspot area percentage b. The reduced (adjusted) threshold results in larger areas being identified as hotspots, thereby increasing the calculated BSI value, which measures the total percentage of the patient's skeleton occupied by hotspots. In this way, the global threshold scaling factor is used to adjust the threshold used for hotspot detection to correct for the observed underestimation of the calculated BSI value at high levels of disease.

[0302] FIG. 25B illustrates an exemplary process 2500 for using a global threshold scaling approach to detect hotspots. In a first step 2510, a set of bone scan images of a human subject is accessed. The bone scan image set is automatically segmented 2520 to produce a set of annotated images, including identification of skeletal regions of interest. In a particular embodiment, the segmentation may include identification of full length femurs and / or humeri, as described with respect to FIG. 24 above. In another step 2530, an initial set of hotspots is automatically detected. In this hotspot detection step 2530, a set of potential hotspots is initially detected 2532 using a pre-threshold. The set of potential hotspots is then used to calculate 2534 a global threshold scaling factor, for example as described herein, which is then used to adjust 2536 the pre-threshold. The adjusted thresholds are then used to automatically detect 2538 hotspots for inclusion in the initial hotspot set. Similar to process 2400, once an initial set of hotspots is detected, a set of hotspot features are extracted 2540 for each hotspot and used to calculate a transition likelihood value for each hotspot 2550. The detected hotspots may be rendered 2560 for graphical display to a user. iii. Hotspot Pre-selection

[0303] As described herein, hotspots are classified to determine whether they should be preselected. In certain embodiments, the classification of hotspots is performed via a two-step process. In certain embodiments, in a first step, the local features of a particular hotspot are used to classify that particular hotspot, but the remainder of the image is not classified, for example, if the patient does not have any other hotspots or does not have many other hotspots. In certain embodiments, a second step is included to incorporate global information about hotspots in the remainder of the image.

[0304] Clinical experience shows that the selection of hot spots depends on the rest of the image. The probability of a hot spot being selected is higher if there are many other hot spots and lower if it is the only hot spot. Thus, using only a single-step process may result in an underestimation of BSI values ​​in patients with many hot spots. The use of a two-step process can improve outcomes in patients with many hot spots and high BSI. The selection of hot spots using global hot spot features may be performed using a machine learning module. For example, in certain embodiments, a first machine learning module is used to calculate the metastasis likelihood value of each hot spot, while a second machine learning module (e.g., implementing a different ANN) can receive the calculated likelihood value together with the global hot spot feature and determine whether the hot spot should be included in the pre-selected hot spot subset.

[0305] 26 illustrates an exemplary process 2600 for pre-selection of hotspots using global hotspot features. The process 2600 begins by accessing 2610 a set of bone scan images, automatically segmenting the images (members) of the set of bone scan images to identify skeletal regions 2620, and automatically detecting 2630 an initial set of hotspots. The segmentation step 2620 and hotspot detection step 2630 may utilize the improved segmentation technique of process 2400 and / or the global threshold scaling technique of process 2500. For each hotspot, a hotspot feature is extracted 2640 and used to calculate 2650 a likelihood of metastasis value for each hotspot. The process 2600 uses the likelihood of metastasis value in conjunction with the global hotspot features to pre-select 2662 a first subset of initial hotspots. This first subset thereby filters the hotspots, allowing a smaller, more targeted set of hotspots that are automatically determined to be likely cancer lesions to be displayed to the user 2664. iv. Atlas Weight

[0306] In certain embodiments, the correction factors for the sacral, pelvic, and lumbar regions are adjusted so that hot spots of equal area correspond to a more uniform measurement of BSI involvement. In certain embodiments, no such adjustment is made and the sacral region differs significantly from the adjacent pelvic and lumbar regions. Figure 27 shows these regions as outlined in a skeletal atlas: sacral region 2706, pelvic region 2702, and lumbar region 2704.

[0307] To calculate the BSI value, the percentage of the total skeleton for each selected hotspot is calculated, and the BSI value is calculated as the sum of all such percentages. For each hotspot, the percentage is calculated as follows: the hotspot size is divided by the size of the corresponding skeletal region (e.g., skull, ribs, lumbar spine, pelvis) obtained from the skeleton segmentation, and multiplied by the weight percentage constant of the current skeletal region relative to the total skeleton weight. These constants, one for each skeletal region, can be determined based on International Commission on Radiological Contamination Prevention (ICRP) publication 23.

[0308] Concomitant occurrence is,

number

[0309] In a particular embodiment, this basic method works well for most skeletal regions, but not for the sacrum region, which is surrounded by the pelvis and lumbar areas. The sacrum is a complex three-dimensional structure, and two-dimensional bone scans make it difficult to separate hotspots from different regions and localize the hotspots to the correct region. Depending on the assigned hotspot localization (e.g., to the pelvis, lumbar column, or sacrum), similarly sized hotspots may have large differences in their contribution to the calculated BSI percentage. To reduce such differences, the coefficient c in the above formula is adjusted for the sacrum, such that the difference from the pelvic region to the lumbar region is more gradual. Specifically, the correction factor is the ratio

number

[0310] 28 illustrates an exemplary process 2800 for utilizing the correction factors described herein to adjust the skeletal involvement rate and calculate a risk index value therefrom. The process 2800 includes accessing 2810 a bone scan image, automatically segmenting 2820 the bone scan image to identify skeletal regions, automatically detecting 2830 an initial set of hotspots, and extracting 2840 hotspot features and calculating 2850 a likelihood of metastasis value for each hotspot. The segmentation step 2820 and the hotspot detection step 2830 may utilize the improved segmentation technique of process 2400 and / or the global threshold scaling technique of process 2500. A first subset of the initial set of hotspots is automatically selected 2860 based at least on the likelihood of metastasis values ​​calculated in step 2850. Global hotspot features as described with respect to process 2600 may be used, or user input received, for example, via interaction with a GUI, may be used to select the first subset. The first subset of hotspots are then used to calculate a risk index value for the subject 2870. As described herein, calculating the risk index may include calculating 2872 bone infiltration rates, adjusting 2874 the bone infiltration rates using a region-dependent correction factor, and summing 2876 the adjusted bone infiltration rates. EXAMPLES

[0311] G. Example: BSI Computational Performance This example demonstrates the linearity, precision, and reproducibility of the calculated BSI values. i. Linearity and Accuracy

[0312] The automated BSI as the dependent variable was determined from two sets of simulated bone scans and measured in comparison to the known phantom BSI, which was considered as the independent variable. In the first set of 50 simulated bone scans, the Shapiro-Wilk test confirmed that the residuals of the dependent variable were normally distributed (p=0.850). Also, the mean residual value of 0.00 with a standard deviation of 0.25 confirmed homoscedasticity and indicated constant variation across all values ​​of the independent variables. Given that the residuals showed normality and homoscedasticity, the model was considered to be linear. Figure 25 shows a scatter plot of the linear fit line and the relevant parameters of the linear regression in the range of BSI from 0.10 to 13.0 are presented in Table 1 below. Table 1. First set of 50 phantoms with a predefined BSI range of 0.10 to 13.0. Parameters of the linear regression model in [Table 1] ii. Precision of BSI calculation

[0313] The coefficient of variation and standard deviation of the automated BSI values ​​at each of five predefined tumor burdens with different localizations was determined for a second set of 50 simulated bone scans. The coefficient of variation at each of the five predefined phantom BSIs was less than or equal to 30%. The results are shown in Table 2 below. Table 2. Variation and standard deviations of automated BSI values ​​in each of five predefined tumor burdens. Coefficient of standard deviation [Table 2] iii. Reproducibility using different cameras

[0314] Table 3 below shows the simulation results of the calculated BSI values ​​for five disease burdens and different cameras. Different collimator settings of the cameras had minimal impact on the reproducibility of the BSI values. The standard deviation corresponding to each disease burden was <10%. Table 3. Calculated BSI values ​​for simulations using different cameras [Table 3] iv. Reproducibility with different image counts

[0315] Figure 26 shows a Bland-Altman diagram assessing the repeatability of the automated BSI readings from 50 simulated phantoms. The mean and median BSI difference was 0.0 (solid horizontal line) with a standard deviation of 0.20, with a coefficient of repeatability (2xSD) of 0.40 and -0.40 (dotted horizontal line). Descriptive statistics are presented in Table 4 below. A paired t-test demonstrated a p-value of 0.15, suggesting that there was no statistically significant difference between the two aBSI values ​​obtained in the repeat scans. Table 4. Descriptive statistics showing the reproducibility of automated BSI calculations from 50 simulated phantoms. [Table 4-1] [Table 4-2] v. Reproducibility through repeated scanning of patients

[0316] Figure 27 shows a Bland-Altman plot assessing the reproducibility of automated BSI readings from repeat bone scans of 35 metastatic patients. The mean and median BSI difference was 0.0 (solid horizontal line) with a standard deviation of 0.18, with a coefficient of repeatability (2xSD) of 0.36 and -0.36 (dotted horizontal line). Descriptive statistics are presented in Table 5 below. A paired t-test demonstrated a p-value of 0.09, suggesting no statistically significant difference between the two aBSI values ​​obtained on repeat scans. Table 5. Descriptive statistics showing the reproducibility of automated BSI calculations from repeat bone scans of 35 metastatic patients. [Table 5-1] [Table 5-2] vi. Comparison with approved devices

[0317] 28A and 28B compare Bland-Altman plots of BSI calculations for an automated BSI device (software aBSI version 3.4) utilizing the improvements of the present disclosure with an approved device (software EXINI version 1.7) implementing a prior version of the software that does not utilize the improvements of the present disclosure. Each plot shows the difference between the known BSI values ​​of a simulated phantom and the values ​​of the phantom calculated using one of the two BSI software versions. FIG. 28A shows the difference between the known phantom values ​​and the values ​​calculated with aBSI3.4, while FIG. 28B shows the difference between the known phantom values ​​and the values ​​calculated using EXINI1.7. The average BSI difference (between the phantom and calculated BSI values) was 0.14 for aBSI3.4 and -0.89 for EXINI1.7, as indicated by the solid horizontal lines in FIG. 28A and FIG. 28B. The standard deviation of aBSI3.4 (SD=0.30) was observed to be significantly lower than the standard deviation of EXINI1.7 (SD=0.88). Descriptive statistics are presented in Tables 6A and 6B below. Table 6A. Descriptive statistics of BSI values ​​calculated with the automated BSI (software version 3.4) using the methods described herein. [Table 6A] Table 6B. Descriptive statistics of BSI values ​​calculated with approved software (EXINI 1.7). [Table 6B-1] [Table 6B-2] H. Computer System and Network Environment

[0318] In certain embodiments, the systems and methods described herein are implemented using a cloud-based microservices architecture. Figure 33A shows an example cloud platform architecture, and Figure 33B shows an example microservices communication design chart.

[0319] FIG. 34 illustrates an illustrative network environment 3400 for use with the methods and systems described herein. As a brief overview, referring now to FIG. 34, a block diagram of an exemplary cloud computing environment 3400 is shown and described. The cloud computing environment 3400 may include one or more resource providers 3402a, 3402b, 3402c (collectively 3402). Each resource provider 3402 may include computing resources. In some implementations, the computing resources may include any hardware and / or software used to process data. For example, the computing resources may include hardware and / or software capable of executing algorithms, computer programs, and / or computer applications. In some implementations, exemplary computing resources may include application servers and / or databases with storage and retrieval capabilities. Each resource provider 3402 may be connected to other resource providers 3402 in the cloud computing environment 3400. In some implementations, the resource providers 3402 may be connected through a computer network 3408. Each resource provider 3402 may be connected to one or more computing devices 3404 a , 3404 b , 3404 c (collectively 3404 ) through a computer network 3408 .

[0320] The cloud computing environment 3400 may include a resource manager 3406. The resource manager 3406 may be connected to the resource providers 3402 and the computing devices 3404 through a computer network 3408. In some implementations, the resource manager 3406 may facilitate the provision of computing resources by one or more resource providers 3402 to one or more computing devices 3404. The resource manager 3406 may receive a request for a computing resource from a particular computing device 3404. The resource manager 3406 may identify one or more resource providers 3402 capable of providing the computing resource requested by the computing device 3404. The resource manager 3406 may select a resource provider 3402 that provides the computing resource. The resource manager 3406 may facilitate a connection between the resource provider 3402 and the particular computing device 3404. In some implementations, the resource manager 3406 may establish a connection between a particular resource provider 3402 and a particular computing device 3404. In some implementations, the resource manager 3406 may redirect a particular computing device 3404 to a particular resource provider 3402 that has the requested computing resource.

[0321] FIG. 35 illustrates examples of computing devices 3500 and mobile computing devices 3550 that may be used in the methods and systems described in this disclosure. Computing device 3500 is intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other suitable computers. Mobile computing device 3550 is intended to represent various forms of mobile devices, such as personal digital assistants, mobile phones, smartphones, and other similar computing devices. The components illustrated herein, their connections and relationships, and their functions are intended to be merely exemplary and not limiting.

[0322] The computing device 3500 includes a processor 3502, a memory 3504, a storage device 3506, a high-speed interface 3508 connecting to the memory 3504 and a number of high-speed expansion ports 3510, and a low-speed interface 3512 connecting to a low-speed expansion port 3514 and the storage device 3506. Each of the processor 3502, the memory 3504, the storage device 3506, the high-speed interface 3508, the high-speed expansion port 3510, and the low-speed interface 3512 are connected to each other using various buses and may be mounted on a common motherboard or in other manners as appropriate. The processor 3502 is capable of processing instructions for execution within the computing device 3500, including instructions stored in the memory 3504 or the storage device 3506 for displaying graphical information of a GUI on an external input / output device, such as a display 3516 coupled to the high-speed interface 3508. In other implementations, multiple processors and / or multiple buses may be used as appropriate, along with multiple memories and types of memories. Also, multiple computing devices may be connected, with each device providing a portion of the required operations (e.g., as a bank of servers, a group of blade servers, or a multiprocessor system). Thus, as the term is used herein, when functions are described as being performed by a "processor," this encompasses embodiments in which the functions are performed by any number of processor(s) in any number of computing device(s). Furthermore, when a function is described as being performed by a "processor," this encompasses embodiments in which the function is performed by any number of processor(s) in any number of computing device(s) (e.g., in a distributed computer system).

[0323] The memory 3504 stores information within the computing device 3500. In some implementations, the memory 3504 is one or more volatile memory units. In some implementations, the memory 3504 is one or more non-volatile memory units. The memory 3504 may be another form of computer-readable medium, such as a magnetic disk or optical disk.

[0324] The storage device 3506 is capable of providing mass storage for the computing device 3500. In some implementations, the storage device 3506 may be or include a computer-readable medium such as a floppy disk device, a hard disk device, an optical disk device, or a tape device, a flash memory, or other similar solid-state memory device, or an array of devices including devices in a storage area network or other configuration. The instructions may be stored on an information carrier. The instructions, when executed by one or more processing devices (e.g., the processor 3502), perform one or more methods such as those described above. The instructions may also be stored in one or more storage devices, such as a computer or machine-readable medium (e.g., the memory 3504, the storage device 3506, or a memory on the processor 3502).

[0325] The high-speed interface 3508 manages the bandwidth-intensive operations of the computing device 3500, while the low-speed interface 3512 manages the less bandwidth-intensive operations. Such an allocation of functions is merely an example. In some implementations, the high-speed interface 3508 is coupled to the memory 3504, the display 3516 (e.g., through a graphics processor or accelerator), and to a high-speed expansion port 3510, which may accept various expansion cards (not shown). In this implementation, the low-speed interface 3512 is coupled to the storage device 3506 and to a low-speed expansion port 3514. The low-speed expansion port 3514 may include various communication ports (e.g., USB, Bluetooth, Ethernet, wireless Ethernet), and may be coupled to one or more input / output devices, such as a keyboard, a pointing device, a scanner, or a networking device, such as a switch or router, for example, through a network adapter.

[0326] The computing device 3500 may be implemented in several different forms as shown in the figure. For example, it may be implemented as a standard server 3520 or multiple times among a group of such servers. It may also be implemented as a personal computer, such as a laptop computer 3522. It may also be implemented as part of a rack server system 3524. Alternatively, components from the computing device 3500 may be combined with other components in a mobile device (not shown), such as a mobile computing device 3550. Each such device may include one or more of the computing device 3500 and the mobile computing device 3550, and the entire system may consist of multiple computing devices in communication with each other.

[0327] The mobile computing device 3550 includes, among other components, a processor 3552, a memory 3564, input / output devices such as a display 3554, a communication interface 3566, and a transceiver 3568. The mobile computing device 3550 may also include storage devices such as a microdrive or other devices to provide additional storage. Each of the processor 3552, memory 3564, display 3554, communication interface 3566, and transceiver 3568 are interconnected using various buses, and some of these components may be mounted on a common motherboard or in other manners as appropriate.

[0328] The processor 3552 can execute instructions within the mobile computing device 3550, including instructions stored in the memory 3564. The processor 3552 may be implemented as a chipset of chips including separate analog and digital processors. The processor 3552 may enable coordination of other components of the mobile computing device 3550, such as, for example, control of a user interface, applications executed by the mobile computing device 3550, and wireless communication by the mobile computing device 3550.

[0329] The processor 3552 may communicate with a user through a control interface 3558 and a display interface 3556 coupled to a display 3554. The display 3554 may be, for example, a TFT (thin film transistor liquid crystal display) display, or an OLED (organic light emitting diode) display, or other suitable display technology. The display interface 3556 may comprise suitable circuitry for driving the display 3554 to present graphical and other information to the user. The control interface 3558 may receive commands from the user and convert them for sending to the processor 3552. An external interface 3562 may also provide communication with the processor 3552 to enable near area communication between the mobile computing device 3550 and other devices. The external interface 3562 may, for example, enable wired communication in some implementations or wireless communication in other implementations, and multiple interfaces may be used.

[0330] The memory 3564 stores information within the mobile computing device 3550. The memory 3564 may be implemented as one or more of one or more computer readable media, one or more volatile memory units, or one or more non-volatile memory units. An expansion memory 3574 may be provided and connected to the mobile computing device 3550 through an expansion interface 3572, which may include, for example, a SIMM (single in-line memory module) card interface. The expansion memory 3574 may provide additional storage space for the mobile computing device 3550 or may store applications or other information for the mobile computing device 3550. In particular, the expansion memory 3574 may include instructions for performing or supplementing the processes described above and may also include secure information. Thus, for example, the expansion memory 3574 may be provided as a security module for the mobile computing device 3550 and may be programmed with instructions that enable secure use of the mobile computing device 3550. Also, secure applications can be provided via the SIMM card along with additional information, such as placing identifying information on the SIMM card in an unhackable form.

[0331] The memory may include, for example, flash memory and / or NVRAM memory (non-volatile random access memory), as discussed below. In some implementations, the instructions are stored on an information carrier and, when executed by one or more processing devices (e.g., processor 3552), perform one or more methods, such as those described above. The instructions may also be stored by one or more storage devices, such as one or more computer or machine-readable media (e.g., memory 3564, expansion memory 3574, or memory on processor 3552). In some implementations, the instructions may be received in a propagated signal, for example through transceiver 3568 or external interface 3562.

[0332] The mobile computing device 3550 can communicate wirelessly through a communication interface 3566, which may include digital signal processing circuitry, if necessary. The communication interface 3566 allows for communication under various modes or protocols, such as Global System for Mobile Communication (GSM) voice calls, Short Message Service (SMS), Enhanced Messaging Service (EMS), or Multimedia Messaging Service (MMS), Code Division Multiple Access (CDMA), Time Division Multiple Access (TDMA), Personal Digital Cellular (PDC), Wideband Code Division Multiple Access (WCDMA), CDMA2000, or General Packet Radio Service (GPRS), among others. Such communication may occur through a transceiver 3568, for example, using radio frequencies. Also, short-range communication may occur using Bluetooth, Wi-Fi, or other such transceivers (not shown). A GPS (Global Positioning System) receiver module 3570 may also provide additional navigation- and location-related wireless data to the mobile computing device 3550, which may be used as appropriate by applications running on the mobile computing device 3550.

[0333] The mobile computing device 3550 may communicate vocally using an audio codec 3560, which may receive vocal information from a user and convert it into usable digital information. The audio codec 3560 may also generate audible sounds for the user, such as through a speaker in a handset of the mobile computing device 3550. Such sounds may include sounds from a voice call, may include recorded sounds (e.g., voice messages, music files, etc.), and may include sounds generated by applications running on the mobile computing device 3550.

[0334] The mobile computing device 3550 may be implemented in a number of different forms as shown in the figure, for example as a mobile phone 3580, or as part of a smartphone 3582, personal digital assistant, or other similar mobile device.

[0335] Various implementations of the systems and techniques described herein may be realized as digital electronic circuitry, integrated circuits, specially designed ASICs (application-specific integrated circuits), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include implementation as one or more computer programs executable and / or interpretable on a programmable system that includes at least one programmable processor, which may be specialized or general-purpose, coupled to receive data and instructions from and transmit data and instructions to a storage system, at least one input device, and at least one output device.

[0336] These computer programs (also known as programs, software, software applications or codes) contain machine instructions for a programmable processor and may be implemented in high-level procedural and / or object-oriented programming languages, and / or assembly and / or machine languages. As used herein, the terms machine-readable medium and computer-readable medium refer to any computer program product, apparatus and / or device used to provide machine instructions and / or data to a programmable processor (e.g., magnetic disks, optical disks, memory, programmable logic devices (PLDs)), including machine-readable media that receive machine instructions as machine-readable signals. The term machine-readable signal refers to any signal used to provide machine instructions and / or data to a programmable processor.

[0337] To enable interaction with a user, the systems and techniques described herein can be implemented on a computer having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal disk) monitor) for displaying information to the user, as well as a keyboard and pointing device (e.g., a mouse or trackball) by which the user can provide input to the computer. Other types of devices can also be used to enable interaction with the user, for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, audible feedback, or tactile feedback), and input from the user can be received in any form, including acoustic, speech, or tactile input.

[0338] The systems and techniques described herein can be implemented within a computing system that includes back-end components (e.g., as data servers), or includes middleware components (e.g., application servers), or includes front-end components (e.g., client computers having a graphical user interface or web browser through which a user can interact with an implementation of the systems and techniques described herein), or any combination of such back-end, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communications network). Examples of communications networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0339] A computer system may include clients and servers. Clients and servers are generally remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. In some implementations, the modules and / or services described herein may be separated, combined, or integrated into a single or combined module and / or service. The modules and / or services depicted in the figures are not intended to limit the systems described herein to the software architectures shown therein.

[0340] While the present invention has been shown and described in detail with respect to certain preferred embodiments, it should be understood by those skilled 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 by the appended claims. The present invention provides, for example, the following items. (Item 1) 1. A method for lesion marking and quantitative analysis of nuclear medicine images of a human subject, comprising: (a) accessing, by a processor of a computing device, a set of bone scan images of the human subject, the set of bone scan images being obtained after administration of a pharmaceutical agent to the human subject; (b) automatically segmenting, by the processor, each image in the set of bone scan images to identify one or more skeletal regions of interest, each corresponding to a particular anatomical region of the human subject's skeleton, thereby obtaining a set of annotated images, the one or more skeletal regions of interest comprising: (i) a femoral region corresponding to a portion of a femur of the human subject, the portion of the femur encompassing at least three-quarters of the femur along a length of the femur; and (ii) a humerus region corresponding to a portion of a humerus of the human subject, the portion of the humerus encompassing at least three-quarters of the humerus along a length of the humerus; and (c) automatically detecting, by the processor, an initial set of one or more hotspots, each hotspot corresponding to an area of ​​high intensity in the set of annotated images, the automatically detecting including identifying the one or more hotspots using intensities of pixels in the set of annotated images and using one or more region-dependent thresholds, the one or more region-dependent thresholds including one or more values ​​associated with the femur and / or humerus regions that provide increased hotspot detection sensitivity in the femur and / or humerus regions to compensate for reduced uptake of the agent in those regions; (d) for each hotspot in the initial set of hotspots, extracting, by the processor, a set of hotspot features associated with the hotspot; (e) for each hotspot in the initial set of hotspots, calculating, by the processor, a transition likelihood value corresponding to a likelihood that the hotspot represents a transition based on the set of hotspot features associated with the hotspot; (f) causing the processor to render a graphical representation of at least a portion of the initial set of hotspots for display within a graphical user interface (GUI). (Item 2) Step (b) comparing each member of the set of bone scan images to a corresponding atlas image of a set of atlas images, each atlas image including one or more identifications of the one or more skeletal regions of interest, the skeletal regions of interest including the femur region and / or the humerus region; and for each image of the bone scan image set, aligning the corresponding atlas image with the image of the bone scan image set such that the identification of the one or more skeletal regions of interest in the atlas image is applied to the image of the bone scan image set. (Item 3) 3. The method of claim 2, wherein each atlas image includes identification of (i) the femur region that includes at least a portion of the knee region of the human subject, and / or (ii) the humerus region that includes at least a portion of the elbow region of the human subject, and for each image of the set of bone scan images, the registration of the corresponding atlas image to the bone scan image includes using the identified knee region and / or the identified elbow region as a landmark within the image. (Item 4) 2. The method of any one of the preceding items, wherein a location of at least one detected hotspot of the initial set of hotspots corresponds to a physical location in or on the femur that is more than three-quarters of the way along the femur from the end of the femur that faces toward the human subject's buttocks to the end of the femur that faces toward the human subject's knee. (Item 5) 20. The method of any one of the preceding claims, wherein a location of at least one detected hotspot of the initial set of hotspots corresponds to a physical location in or on the humerus that is more than three-quarters of the way along the humerus from the end of the humerus pointing towards the human subject's shoulder to the end of the humerus pointing towards the human subject's elbow. (Item 6) Step (c) identifying, by the processor, healthy tissue regions within the set of bone scan images determined to contain no hot spots; calculating, by the processor, a normalization factor such that the product of the normalization factor and an average intensity of the identified healthy tissue region is a predetermined intensity level; and normalizing, by the processor, the images of the bone scan image set by the normalization factor. (Item 7) (g) calculating, by the processor, one or more risk index values ​​for the human subject based at least in part on the calculated percentage of the skeleton of the human subject that is occupied by the initial set of hotspots. (Item 8) (h) selecting, by the processor, a first subset of the initial set of hotspots based at least in part on the transition likelihood values; (i) causing said processor to render a graphical representation of said first subset for display within a graphical user interface (GUI); The method according to any one of the preceding items, comprising: (Item 9) (j) calculating, by the processor, one or more risk index values ​​for the human subject based at least in part on the calculated percentage of the skeleton of the human subject that is occupied by the first subset of hotspots. (Item 10) (k) receiving, by the processor, via the GUI, a user selection of a second subset of the initial set of hotspots; (l) calculating, by the processor, one or more risk index values ​​for the human subject based at least in part on the calculated percentage of the skeleton of the human subject that is occupied by the second subset of hotspots; The method according to any one of the preceding items, comprising: (Item 11) 11. The method of any one of items 7 to 10, wherein at least one of the risk index values ​​is indicative of the risk of the human subject having and / or developing metastatic cancer. (Item 12) 12. The method of claim 11, wherein the metastatic cancer is metastatic prostate cancer. (Item 13) 11. The method of any one of items 7 to 10, wherein at least one of the risk index values ​​indicates that the human subject has a particular state of metastatic cancer. (Item 14) 13. The method of any one of the preceding items, wherein the processor is a processor of a cloud-based system. (Item 15) 2. The method of any one of the preceding claims, wherein the GUI is part of a general Picture Archiving and Communication System (PACS). (Item 16) The drug is technetium 99m methylenediphosphonate ( 99m 3. The method of any one of the preceding claims, comprising administering to the subject a therapeutically effective amount of Tc-MDP. (Item 17) 1. A method for lesion marking and quantitative analysis of nuclear medicine images of a human subject, comprising: (a) accessing, by a processor of a computing device, a set of bone scan images of the human subject, the set of bone scan images being obtained after administration of a pharmaceutical agent to the human subject; (b) automatically segmenting, by the processor, each image in the set of bone scan images to identify one or more skeletal regions of interest, each corresponding to a particular anatomical region of the human subject's skeleton, thereby obtaining a set of annotated images; (c) automatically detecting, by the processor, an initial set of one or more hotspots, each hotspot corresponding to an area of ​​high intensity within the set of annotated images, the automatically detecting including: Detecting a set of potential hotspots using (i) intensities of pixels in the set of annotated images, and (ii) a number of pre-thresholds; calculating a global threshold scaling factor using the set of potential hotspots; adjusting the plurality of pre-thresholds using the global threshold scaling factor, thereby obtaining a plurality of adjusted thresholds; and identifying the initial set of hotspots using (i) intensities of pixels in the set of annotated images, and (ii) the plurality of adjusted thresholds; (d) for each hotspot in the initial set of hotspots, extracting, by the processor, a set of hotspot features associated with the hotspot; (e) for each hotspot in the initial set of hotspots, calculating, by the processor, a transition likelihood value corresponding to a likelihood that the hotspot represents a transition based on the set of hotspot features associated with the hotspot; (f) causing the processor to render a graphical representation of at least a portion of the initial set of hotspots for display within a graphical user interface (GUI). (Item 18) 18. The method of claim 17, wherein the global threshold scaling factor is a function of a measure of disease burden in the human subject, and the adjusting of the plurality of pre-thresholds performed in step (c) comprises decreasing the adjusted thresholds as disease burden increases to compensate for underestimation of hotspot area that occurs with increasing disease burden. (Item 19) 19. The method of claim 17 or 18, wherein the global threshold scaling ratio is a function of the percentage of the identified skeletal region that is occupied by the set of potential hotspots. (Item 20) 20. The method of any one of items 17 to 19, wherein the global threshold scaling rate is based on a risk index value calculated using the set of potential hotspots. (Item 21) 1. A method for lesion marking and quantitative analysis of nuclear medicine images of a human subject, comprising: (a) accessing, by a processor of a computing device, a set of bone scan images of the human subject, the set of bone scan images being obtained after administration of a pharmaceutical agent to the human subject; (b) automatically segmenting, by the processor, each image in the set of bone scan images to identify one or more skeletal regions of interest, each corresponding to a particular anatomical region of the human subject's skeleton, thereby obtaining a set of annotated images; (c) automatically detecting, by the processor, an initial set of one or more hotspots, each hotspot corresponding to an area of ​​high intensity within the set of annotated images; (d) for each hotspot in the initial set of hotspots, extracting, by the processor, a set of hotspot features associated with the hotspot; (e) for each hotspot in the initial set of hotspots, calculating, by the processor, a transition likelihood value corresponding to a likelihood that the hotspot represents a transition based on the set of hotspot features associated with the hotspot; (f) selecting, by the processor, a first subset of the initial set of hotspots, wherein selecting particular hotspots for inclusion in the first subset comprises: (i) the calculated metastasis likelihood value for the particular hotspot; and (ii) one or more global hotspot features, each determined using a plurality of hotspots in the initial set of hotspots; Based at least in part on (g) causing the processor to render a graphical representation of at least a portion of the first subset of hotspots for display within a graphical user interface (GUI). (Item 22) 22. The method of claim 21, wherein the one or more global hotspot characteristics include a total number of hotspots in the initial hotspot set. (Item 23) 23. The method of claim 22, wherein step (f) includes adjusting criteria for selecting hotspots to include in the first subset based on the total number of hotspots in the initial set of hotspots. (Item 24) 24. The method of any one of claims 21 to 23, wherein step (f) comprises using a machine learning module to select the first subset. (Item 25) 1. A method for lesion marking and quantitative analysis of nuclear medicine images of a human subject, comprising: (a) accessing, by a processor of a computing device, a set of bone scan images of the human subject; (b) automatically segmenting, by the processor, each image in the set of bone scan images to identify one or more skeletal regions of interest, each corresponding to a particular anatomical region of the human subject's skeleton, thereby obtaining a set of annotated images; (c) automatically detecting, by the processor, an initial set of one or more hotspots, each hotspot corresponding to an area of ​​high intensity within the set of annotated images; (d) for each hotspot in the initial set of hotspots, extracting, by the processor, a set of hotspot features associated with the hotspot; (e) for each hotspot in the initial set of hotspots, calculating, by the processor, a likelihood value corresponding to a likelihood that the hotspot represents a metastasis based on the set of hotspot features associated with the hotspot; (f) selecting, by the processor, a first subset of the hotspots of the initial set of hotspots based at least in part on the likelihood value calculated for each hotspot of the initial set of hotspots; (g) calculating, by the processor, one or more risk index values ​​using at least a portion of the first subset of hotspots, wherein the calculating includes: for each particular hotspot of the portion of the first subset, calculating a bone infiltration rate based on the ratio of (i) the size (e.g., area) of the particular hotspot to (ii) the size of a particular skeletal region to which the particular hotspot was assigned based on its location within the set of annotated images, thereby determining one or more bone infiltration rates; adjusting said bone infiltration rates using one or more region-dependent correction factors, thereby obtaining one or more adjusted bone infiltration rates; and summing the adjusted bone invasion rates to determine the one or more risk index values. (Item 26) 26. The method of claim 25, wherein for each particular hotspot, the calculated bone invasion rate estimates the proportion of total skeletal mass occupied by the physical volume associated with said particular hotspot. (Item 27) The step of calculating the bone invasion rate comprises: calculating, by the processor, a ratio of an area of ​​the particular hotspot to an area of ​​the corresponding skeletal region of interest, thereby calculating an area percentage of the particular hotspot; and scaling the area percentage by a density factor associated with the skeletal region of interest to which the particular hotspot is assigned, thereby calculating the bone invasion rate of the particular hotspot. (Item 28) 28. The method of any one of claims 25 to 27, wherein at least a portion of the hotspots of the first subset are assigned to a skeletal region of interest that is a member selected from the group consisting of a pelvic region, a lumbar region, and a sacral region. (Item 29) 29. The method of any one of items 25 to 28, wherein the one or more region-dependent correction factors include a sacral region correction factor associated with a sacral region that is used to adjust the bone infiltration rate of hotspots identified as being located within the sacral region, wherein the sacral region correction factor has a value less than 1. (Item 30) the one or more region-dependent correction factors include one or more correction factor pairs, each correction factor pair being associated with a particular skeletal region of interest and including a first member and a second member (of the pair); the first member of the pair is a ventral image correction factor, which is used to adjust the calculated bone invasion rate for hotspots detected in annotated ventral bone scan images of the annotated image set; 30. The method of any one of items 25 to 29, wherein the second member of the pair is a dorsal image correction factor, which is used to adjust the calculated bone invasion rate for hotspots detected in annotated dorsal bone scan images of the annotated image set. (Item 31) 1. A system for lesion marking and quantitative analysis of nuclear medicine images of a human subject, comprising: A processor; and a memory having instructions that, when executed by the processor, cause the processor to: (a) accessing a set of bone scan images of the human subject, the set of bone scan images being obtained after administration of a pharmaceutical agent to the human subject; (b) automatically segmenting each image in the set of bone scan images to identify one or more skeletal regions of interest, each corresponding to a particular anatomical region of a skeleton of the human subject, thereby obtaining a set of annotated images, the one or more skeletal regions of interest comprising: (i) a femoral region corresponding to a portion of a femur of the human subject, the portion of the femur encompassing at least three-quarters of the femur along a length of the femur; and (ii) a humerus region corresponding to a portion of a humerus of the human subject, the portion of the humerus encompassing at least three-quarters of the humerus along a length of the humerus; and (c) automatically detecting an initial set of one or more hot spots, each hot spot corresponding to an area of ​​high intensity in the set of annotated images, said automatically detecting including identifying the one or more hot spots using intensities of pixels in the set of annotated images and using one or more region-dependent thresholds, the one or more region-dependent thresholds including one or more values ​​associated with the femur and / or humerus regions that provide increased hot spot detection sensitivity in the femur and / or humerus regions to compensate for reduced uptake of the agent in those regions; (d) for each hotspot in the initial set of hotspots, extracting a set of hotspot features associated with the hotspot; (e) calculating, for each hotspot in the initial set of hotspots, a transition likelihood value corresponding to a likelihood that the hotspot represents a transition based on the set of hotspot features associated with the hotspot; (f) rendering a graphical representation of at least a portion of the initial set of hotspots for display within a graphical user interface (GUI). (Item 32) 1. A system for lesion marking and quantitative analysis of nuclear medicine images of a human subject, comprising: A processor; and a memory having instructions that, when executed by the processor, cause the processor to: (a) accessing, by a processor of a computing device, a set of bone scan images of the human subject, the set of bone scan images being obtained after administration of a pharmaceutical agent to the human subject; and (b) automatically segmenting each image in the set of bone scan images to identify one or more skeletal regions of interest, each corresponding to a particular anatomical region of the human subject's skeleton, thereby obtaining a set of annotated images; (c) automatically detecting an initial set of one or more hotspots, each hotspot corresponding to an area of ​​high intensity within the set of annotated images, said automatically detecting including: Detecting a set of potential hotspots using (i) intensities of pixels in the set of annotated images, and (ii) a number of pre-thresholds; calculating a global threshold scaling factor using the set of potential hotspots; adjusting the plurality of pre-thresholds using the global threshold scaling factor, thereby obtaining a plurality of adjusted thresholds; and identifying the initial set of hotspots using (i) intensities of pixels in the set of annotated images, and (ii) the plurality of adjusted thresholds; (d) for each hotspot in the initial set of hotspots, extracting a set of hotspot features associated with the hotspot; (e) calculating, for each hotspot in the initial set of hotspots, a transition likelihood value corresponding to a likelihood that the hotspot represents a transition based on the set of hotspot features associated with the hotspot; (f) rendering a graphical representation of at least a portion of the initial set of hotspots for display within a graphical user interface (GUI). (Item 33) 1. A system for lesion marking and quantitative analysis of nuclear medicine images of a human subject, comprising: A processor; and a memory having instructions that, when executed by the processor, cause the processor to: (a) accessing a set of bone scan images of the human subject, the set of bone scan images being obtained after administration of a pharmaceutical agent to the human subject; (b) automatically segmenting each image in the set of bone scan images to identify one or more skeletal regions of interest, each corresponding to a particular anatomical region of the human subject's skeleton, thereby obtaining a set of annotated images; (c) automatically detecting an initial set of one or more hotspots, each hotspot corresponding to an area of ​​high intensity within the set of annotated images; (d) for each hotspot in the initial set of hotspots, extracting a set of hotspot features associated with the hotspot; (e) calculating, for each hotspot in the initial set of hotspots, a transition likelihood value corresponding to a likelihood that the hotspot represents a transition based on the set of hotspot features associated with the hotspot; (f) automatically selecting a first subset of the initial set of hotspots, wherein selecting particular hotspots for inclusion in the first subset comprises: (i) the calculated metastasis likelihood value for the particular hotspot; and (ii) one or more global hotspot features, each determined using a plurality of hotspots in the initial set of hotspots; Based at least in part on (g) rendering a graphical representation of at least a portion of the first subset of hotspots for display within a graphical user interface (GUI). (Item 34) 1. A system for lesion marking and quantitative analysis of nuclear medicine images of a human subject, comprising: A processor; and a memory having instructions that, when executed by the processor, cause the processor to: (a) accessing a set of bone scan images of the human subject; (b) automatically segmenting each image in the set of bone scan images to identify one or more skeletal regions of interest, each corresponding to a particular anatomical region of the human subject's skeleton, thereby obtaining a set of annotated images; (c) automatically detecting an initial set of one or more hotspots, each hotspot corresponding to an area of ​​high intensity within the set of annotated images; (d) for each hotspot in the initial set of hotspots, extracting a set of hotspot features associated with the hotspot; (e) calculating, for each hotspot in the initial set of hotspots, a likelihood value corresponding to a likelihood that the hotspot represents a metastasis based on the set of hotspot features associated with the hotspot; (f) selecting a first subset of the hotspots of the initial set of hotspots based at least in part on the likelihood value calculated for each hotspot of the initial set of hotspots; and (g) calculating one or more risk index values ​​using at least a portion of the first subset of hotspots, wherein said calculating comprises: for each particular hotspot of the portion of the first subset, (ii) calculating a bone infiltration rate based on a ratio of (i) a size of the particular hotspot to a size of a particular skeletal region to which the particular hotspot was assigned based on its location within the set of annotated images, thereby determining one or more bone infiltration rates; adjusting said bone infiltration rates using one or more region-dependent correction factors, thereby obtaining one or more adjusted bone infiltration rates; and summing the adjusted bone invasion rates to determine the one or more risk index values. (Item 35) A computer-assisted image analysis device comprising the system of any one of items 31 to 34. (Item 36) 36. The device of item 35, wherein the device is programmed for use by trained health care professionals and / or researchers. (Item 37) 37. The device of item 36, wherein the device is programmed for use in analyzing bone scan images to assess and / or detect metastatic cancer. (Item 38) 38. The device of claim 36 or 37, wherein the device is programmed for use in analyzing bone scan images to assess and / or detect prostate cancer. (Item 39) 39. The device of any one of items 35 to 38, comprising a label specifying that the device is intended for use by trained health care professionals and / or researchers. (Item 40) 40. The device of claim 39, wherein the label further specifies that the device is intended for use in analyzing bone scan images to evaluate and / or detect metastatic cancer. (Item 41) 41. The device of claim 39 or 40, wherein the label further specifies that the device is intended for use in analyzing bone scan images to evaluate and / or detect prostate cancer.

Claims

1. 1. A method for lesion marking and quantitative analysis of nuclear medicine images of a human subject, comprising: (a) accessing, by a processor of a computing device, a set of bone scan images of the human subject, the set of bone scan images being obtained after administration of a pharmaceutical agent to the human subject; (b) automatically segmenting, by the processor, each image in the bone scan image set to identify one or more skeletal regions of interest, each corresponding to a particular anatomical region of the human subject's skeleton, thereby obtaining a set of annotated images; (c) automatically detecting, by the processor, an initial set of one or more hotspots, each hotspot corresponding to an area of ​​high intensity within the set of annotated images, the automatically detecting step comprising: Detecting a set of potential hotspots using (i) pixel intensities in the set of annotated images and (ii) a number of pre-thresholds; calculating a global threshold scaling factor using the set of potential hotspots; adjusting the plurality of pre-thresholds using the global threshold scaling factor, thereby obtaining a plurality of adjusted thresholds; and Identifying the initial set of hotspots using (i) intensities of pixels in the set of annotated images and (ii) the plurality of adjusted thresholds. and (d) for each hotspot in the initial set of hotspots, extracting, by the processor, a set of hotspot features associated with the hotspot; (e) for each hotspot in the initial set of hotspots, calculating, by the processor, a transition likelihood value corresponding to a likelihood that the hotspot represents a transition based on the set of hotspot features associated with the hotspot; (f) causing the processor to render a graphical representation of at least a portion of the initial set of hotspots for display within a graphical user interface (GUI); A method comprising:

2. The method of claim 1, wherein the global threshold scaling rate is a function of a measure of disease burden in the human subject, and adjusting the multiple pre-thresholds performed in step (c) includes decreasing the adjusted thresholds as disease burden increases to compensate for underestimation of hotspot area that occurs with increasing disease burden.

3. A method according to any one of claims 1 or 2, wherein the global threshold scaling rate is a function of the percentage of the identified skeletal region that is occupied by the set of potential hotspots.

4. A method according to any one of claims 1 to 3, wherein the global threshold scaling rate is based on a risk index value calculated using the set of potential hotspots.

5. A method for lesion marking and quantitative analysis of nuclear medicine images of a human subject, comprising: (a) accessing, by a processor of a computing device, a set of bone scan images of the human subject, the set of bone scan images being obtained after administration of a pharmaceutical agent to the human subject; (b) automatically segmenting, by the processor, each image in the bone scan image set to identify one or more skeletal regions of interest, each corresponding to a particular anatomical region of the human subject's skeleton, thereby obtaining a set of annotated images; (c) automatically detecting, by the processor, an initial set of one or more hotspots, each hotspot corresponding to an area of ​​high intensity within the set of annotated images; (d) for each hotspot in the initial set of hotspots, extracting, by the processor, a set of hotspot features associated with the hotspot; (e) for each hotspot in the initial set of hotspots, calculating, by the processor, a transition likelihood value corresponding to a likelihood that the hotspot represents a transition based on the set of hotspot features associated with the hotspot; (f) selecting, by the processor, a first subset of the initial set of hotspots, the selection of particular hotspots for inclusion in the first subset comprising: (i) a calculated metastasis likelihood value for the particular hotspot; and (ii) one or more global hotspot features, each determined using a plurality of hotspots in the initial set of hotspots; Based at least in part on (g) causing the processor to render a graphical representation of at least a portion of the first subset of hotspots for display within a graphical user interface (GUI); A method comprising:

6. The method of claim 5, wherein the one or more global hotspot features include a total number of hotspots in the initial hotspot set.

7. The method of claim 6, wherein step (f) includes adjusting criteria for selecting hotspots to include in the first subset based on the total number of hotspots in the initial hotspot set.

8. A method according to any one of claims 5 to 7, wherein step (f) includes selecting the first subset using a machine learning module.

9. A method for lesion marking and quantitative analysis of nuclear medicine images of a human subject, comprising: (a) accessing, by a processor of a computing device, a set of bone scan images of the human subject; (b) automatically segmenting, by the processor, each image in the bone scan image set to identify one or more skeletal regions of interest, each corresponding to a particular anatomical region of the human subject's skeleton, thereby obtaining a set of annotated images; (c) automatically detecting, by the processor, an initial set of one or more hotspots, each hotspot corresponding to an area of ​​high intensity within the set of annotated images; (d) for each hotspot in the initial set of hotspots, extracting, by the processor, a set of hotspot features associated with the hotspot; (e) for each hotspot in the initial set of hotspots, calculating, by the processor, a likelihood value corresponding to a likelihood that the hotspot represents a metastasis based on the set of hotspot features associated with the hotspot; (f) selecting, by the processor, a first subset of the hotspots of the initial set of hotspots based at least in part on the likelihood value calculated for each hotspot of the initial set of hotspots; (g) calculating, by the processor, one or more risk index values ​​using at least a portion of the first subset of the hotspots, wherein said calculating includes: for each particular hotspot of the portion of the first subset, (ii) calculating a bone infiltration rate based on a ratio of (i) a size of the particular hotspot to a size of a particular skeletal region to which the particular hotspot was assigned based on its location within the set of annotated images, thereby determining one or more bone infiltration rates; adjusting said bone infiltration rates using one or more region-dependent correction factors, thereby obtaining one or more adjusted bone infiltration rates; and summing said adjusted bone invasion rates to determine said one or more risk index values. and A method comprising:

10. The method of claim 9, wherein for each particular hotspot, the calculated bone infiltration rate estimates the proportion of total skeletal mass occupied by the physical volume associated with that particular hotspot.

11. The step of calculating the bone infiltration rate comprises: calculating, by the processor, a ratio of an area of ​​the particular hotspot to an area of ​​a corresponding skeletal region of interest, thereby calculating an area percentage of the particular hotspot; scaling the area percentage by a density factor associated with the skeletal region of interest to which the particular hotspot is assigned, thereby calculating a bone invasion rate for the particular hotspot; The method of claim 10, comprising:

12. A method according to any one of claims 9 to 11, wherein at least a portion of the hotspots of the first subset are assigned to a skeletal region of interest that is a member selected from the group consisting of a pelvic region, a lumbar region, and a sacral region.

13. A method according to any one of claims 9 to 12, wherein the one or more region-dependent correction factors include a sacral region correction factor associated with a sacral region that is used to adjust the bone infiltration rate of hotspots identified as being located within the sacral region, the sacral region correction factor having a value less than 1.

14. The one or more region-dependent correction factors include one or more correction factor pairs, each correction factor pair being associated with a particular skeletal region of interest and including a first member and a second member (of the pair); the first member of the pair is a ventral image correction factor, which is used to adjust the calculated bone invasion rate for hotspots detected in annotated ventral bone scan images of the annotated image set; 14. The method of any one of claims 9 to 13, wherein the second member of the pair is a dorsal image correction factor, which is used to adjust the calculated bone infiltration rate for hotspots detected in annotated dorsal bone scan images of the annotated image set.

15. 1. A system for lesion marking and quantitative analysis of nuclear medicine images of a human subject, comprising: A processor; and a memory having instructions that, when executed by the processor, cause the processor to: (a) accessing, by a processor of a computing device, a set of bone scan images of the human subject, the set of bone scan images being obtained after administration of a pharmaceutical agent to the human subject; and (b) automatically segmenting each image in the set of bone scan images to identify one or more skeletal regions of interest, each corresponding to a particular anatomical region of the human subject's skeleton, thereby obtaining a set of annotated images; (c) automatically detecting an initial set of one or more hotspots, each hotspot corresponding to an area of ​​high intensity within the set of annotated images, said automatically detecting comprising: Detecting a set of potential hotspots using (i) pixel intensities in the set of annotated images and (ii) a number of pre-thresholds; calculating a global threshold scaling factor using the set of potential hotspots; adjusting the plurality of pre-thresholds using the global threshold scaling factor, thereby obtaining a plurality of adjusted thresholds; and Identifying the initial set of hotspots using (i) intensities of pixels in the set of annotated images and (ii) the plurality of adjusted thresholds. and (d) for each hotspot in the initial set of hotspots, extracting a set of hotspot features associated with the hotspot; (e) calculating, for each hotspot in the initial set of hotspots, a transition likelihood value corresponding to a likelihood that the hotspot represents a transition based on the set of hotspot features associated with the hotspot; (f) rendering a graphical representation of at least a portion of the initial set of hotspots for display within a graphical user interface (GUI).

16. A system for lesion marking and quantitative analysis of nuclear medicine images of a human subject, comprising: A processor; and a memory having instructions that, when executed by the processor, cause the processor to: (a) accessing a set of bone scan images of the human subject, the set of bone scan images being obtained after administration of a pharmaceutical agent to the human subject; (b) automatically segmenting each image in the set of bone scan images to identify one or more skeletal regions of interest, each corresponding to a particular anatomical region of the human subject's skeleton, thereby obtaining a set of annotated images; (c) automatically detecting an initial set of one or more hotspots, each hotspot corresponding to an area of ​​high intensity within the set of annotated images; (d) for each hotspot in the initial set of hotspots, extracting a set of hotspot features associated with the hotspot; (e) calculating, for each hotspot in the initial set of hotspots, a transition likelihood value corresponding to a likelihood that the hotspot represents a transition based on the set of hotspot features associated with the hotspot; (f) automatically selecting a first subset of the initial set of hotspots, the selection of particular hotspots for inclusion in the first subset comprising: (i) a calculated metastasis likelihood value for the particular hotspot; and (ii) one or more global hotspot features, each determined using a plurality of hotspots in the initial set of hotspots; Based at least in part on (g) rendering a graphical representation of at least a portion of the first subset of hotspots for display within a graphical user interface (GUI); A system that allows the user to:

17. A system for lesion marking and quantitative analysis of nuclear medicine images of a human subject, comprising: A processor; and a memory having instructions that, when executed by the processor, cause the processor to: (a) accessing a set of bone scan images of the human subject; (b) automatically segmenting each image in the set of bone scan images to identify one or more skeletal regions of interest, each corresponding to a particular anatomical region of the human subject's skeleton, thereby obtaining a set of annotated images; (c) automatically detecting an initial set of one or more hotspots, each hotspot corresponding to an area of ​​high intensity within the set of annotated images; (d) for each hotspot in the initial set of hotspots, extracting a set of hotspot features associated with the hotspot; (e) for each hotspot in the initial set of hotspots, calculating a likelihood value corresponding to a likelihood that the hotspot represents a metastasis based on the set of hotspot features associated with the hotspot; (f) selecting a first subset of the hotspots of the initial set of hotspots based at least in part on the likelihood value calculated for each hotspot of the initial set of hotspots; and (g) calculating one or more risk index values ​​using at least a portion of the first subset of the hotspots, said calculating comprising: for each particular hotspot of the portion of the first subset, (ii) calculating a bone infiltration rate based on a ratio of (i) a size of the particular hotspot to a size of a particular skeletal region to which the particular hotspot was assigned based on its location within the set of annotated images, thereby determining one or more bone infiltration rates; adjusting said bone infiltration rates using one or more region-dependent correction factors, thereby obtaining one or more adjusted bone infiltration rates; and summing said adjusted bone invasion rates to determine said one or more risk index values. and A system that allows the user to:

18. A computer-aided image analysis device comprising a system according to any one of claims 15 to 17.

19. 20. The device of claim 18, wherein the device is programmed for use by trained health care professionals and / or researchers.

20. 20. The device of claim 19, wherein the device is programmed for use in analyzing bone scan images to assess and / or detect metastatic cancer.

21. 21. A device according to claim 19 or 20, wherein the device is programmed for use in analysing bone scan images for the assessment and / or detection of prostate cancer.

22. 22. The device of any one of claims 18 to 21, comprising a label specifying that the device is intended for use by trained health care professionals and / or researchers.

23. 23. The device of claim 22, wherein the label further specifies that the device is intended for use in analyzing bone scan images to evaluate and / or detect metastatic cancer.

24. 24. The device of claim 22 or 23, wherein the label further specifies that the device is intended for use in analyzing bone scan images to evaluate and / or detect prostate cancer.

Images