System and method for automated interactive analysis of bone scan images to detect 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-references to related applications This application claims priority and benefits under U.S. Provisional Application No. 62 / 837,955, filed on April 24, 2019, the contents of which are incorporated in their entirety by reference.

[0002] Technical field The present invention relates, in general, to systems and methods for creating, analyzing, and / or presenting medical image data. More specifically, in certain embodiments, the present invention relates to systems and methods for improved computer-aided 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 of the body, depending on and thus reflecting the biophysical and / or biochemical properties of the internal tissues, such as properties influenced by the presence and / or state of disease, including cancer. For example, certain radiopharmaceuticals accumulate after administration to a patient in areas of abnormal bone formation associated with malignant bone lesions, indicating metastasis. Other radiopharmaceuticals may bind to specific receptors, enzymes, and proteins in the body that undergo alteration during disease progression. After administration to a patient, these molecules circulate in the bloodstream until they find their intended target. The bound radiopharmaceutical remains at the site of disease, while the rest of the drug exits the body.

[0004] Nuclear medicine imaging technology captures images by detecting radiation emitted from the radioactive portion of radiopharmaceuticals. The accumulated radiopharmaceuticals serve as markers so that images depicting the location and density of diseases can be obtained 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 are found in most hospitals around the world. The selection of a particular imaging modality depends on and / or determines the particular radiopharmaceutical used. For example, compounds labeled with technetium 99m ( 99m Tc) are compatible with both bone scan imaging and SPECT imaging, whereas PET imaging often uses fluorinated compounds labeled with 18F. The compound 99m Tc methylene diphosphonate ( 99m Tc MDP) is a common radiopharmaceutical used in bone scan imaging to detect metastatic cancer. 99m Compounds targeting radiolabeled prostate specific membrane antigen (PSMA), such as 1404 and PyL (trademark) (also called [18F]DCFPyL) labeled with 99m Tc, may be used in SPECT and PET imaging respectively, providing the potential for highly specific prostate cancer detection.<(

[0005] Therefore, nuclear medicine imaging is a useful technique for providing physicians with information that can be used to determine the presence and extent of diseases in a patient's body. Physicians can use this information to provide patients with the recommended treatment process and track the progression of the disease.

[0006] For example, a tumor specialist may use nuclear medicine images as input obtained from a patient's examination to assess whether the patient has a specific disease, such as prostate cancer, which stage of the disease is evident, (if present) what the recommended course of treatment would be, whether surgical intervention is required, and the likely prognosis. The tumor specialist may use the report of a radiologist during this assessment. The radiologist's report is a technical evaluation of the nuclear medicine images created by a radiologist for the physician who requested the imaging examination and includes, for example, the type of examination performed, clinical history, comparison between images, the technique used to perform the examination, the radiologist's findings and observations, and the overall impression and recommendations the radiologist may have based on the imaging examination results. The signed radiologist's report is sent to the physician who ordered the examination for the physician's review, followed by discussions between the physician and the patient regarding the results and treatment recommendations.

[0007] Thus, this process requires having the radiologist perform the patient's imaging examination, analyze the resulting images, create the radiologist's report, transfer the report to the requesting physician, have the physician formulate the assessment and treatment recommendations, and have the physician convey the results, recommendations, and risks to the patient. This process may also require repeating the imaging examination for inconclusive results or ordering additional tests based on the initial results. If the imaging examination indicates that the patient has a specific disease or condition (such as cancer), the physician considers various treatment options, including surgery, as well as the risks of doing nothing, or adopting a watchful waiting or active surveillance approach instead of 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 crucial role in cancer diagnosis and treatment. Thus, there is a great need for improved tools that facilitate and enhance the accuracy of image review and analysis for cancer diagnosis and treatment. Improving toolkits used by physicians, radiologists, and other healthcare professionals in this way would result in a significant improvement in the standards of treatment and patient experience.

[0009] Summary of the Invention This specification presents systems and methods that enable improved computer-aided 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 for assessing a patient's cancer status.

[0010] For example, improved techniques are provided for image segmentation, hotspot detection, automated classification of hotspots as indicators 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 for assessing various metastatic bone cancers (e.g., any cancer that has metastasized to the bone). These include 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. A radiopharmaceutical emitting nuclear radiation is injected into the patient, and its detection allows for imaging of the spatial distribution of the radiopharmaceutical within the patient's body. The radiopharmaceutical can be selected to selectively accumulate in types of tissue associated with cancerous lesions, such as areas of abnormal bone formation.

[0012] While this technique allows lesions to be visualized as bright spots in bone scan images, accurately identifying image regions representing true metastatic lesions is by no means straightforward. Radiopharmaceuticals can accumulate in anatomical areas without cancer, such as the patient's bladder, and physicians and specialists must carefully distinguish hotspots representing lesions from such areas, as well as from noise and artifacts. This process is time-consuming, error-prone, and subject to significant inter-labor variability.

[0013] Computer-assisted lesion detection and analysis can pave the way for addressing these challenges and 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, analysis can be focused on the bone region using image segmentation to identify skeletal regions. Hotspots can be automatically detected using filtering and thresholding steps, and machine learning techniques such as artificial neural networks (ANNs) can be used to quantitatively assess the likelihood that detected hotspots represent metastases based on features such as size, shape, and intensity of the hotspots. Finally, in certain embodiments, a set of detected hotspots representing metastases can be used to calculate a comprehensive risk index for a patient, which represents the overall likelihood that the patient has and / or develops or has a particular cancerous condition. One such risk index is the Bone Scan Index (BSI), which provides an estimated percentage of the patient's skeleton occupied by metastases.

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

[0015] Firstly, in certain embodiments, the improved image analysis techniques described herein include an improved skeletal segmentation method in which the entire humerus and / or femur region (e.g., more than three-quarters of its length) is identified within the bone scan image. Previous methods identified only a limited proportion of the femur and humerus. Here, by segmenting larger portions of these bones, it becomes possible to identify lesions located at the extremities of the arm and leg, whereas previously such lesions would have gone undetected. Furthermore, reduced uptake of radiopharmaceuticals in the arm and leg makes lesion identification in these regions difficult, but the techniques described herein utilize region-dependent thresholding techniques to overcome this problem by increasing detection sensitivity in the femoral and humerus regions.

[0016] Secondly, the disclosure also provides a global thresholding technique that improves the accuracy of hotspot detection, particularly in high disease burden situations (e.g., when a patient has many lesions). This technique detects a pre-set of potential hotspots and then adjusts the threshold used for hotspot detection based on a scaling rate calculated from that pre-set. This improvement in hotspot detection benefits downstream calculations and improves the linearity of BSI values ​​calculated for patients with high levels of metastasis.

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

[0018] Fourth, in certain embodiments, the method described herein also provides an improvement to a method for calculating risk index values ​​based on bone infiltration by using a correction factor that takes into account the potential error in accuracy of automatically localizing hotspots 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 method improves the accuracy of BSI calculations and limits sensitivity to errors in hotspot localization.

[0019] Therefore, the systems and methods described herein include several improved image analysis techniques for the identification and quantification of lesions. These techniques 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 the review and reporting of patient data and enables improved disease detection, treatment, and monitoring.

[0020] In one embodiment, the present invention relates to a method for lesion marking and quantitative analysis of nuclear medicine images (e.g., a set of bone scan images) of a human subject (e.g., automated or semi-automated lesion marking and quantitative analysis assisted / checked by a user), the method comprising the steps of (a) accessing (e.g., and / or receiving) a set of bone scan images (e.g., a set of one, two, or more images) of a human subject by a processor of a computing device, wherein the set of bone scan images is used to mark a drug (e.g., a radiopharmaceutical drug) of a human subject. (b) The images are obtained after administration of a drug (for example, the bone scan image set includes ventral and dorsal bone scan images) (for example, each image in the bone scan image set has multiple pixels, and each pixel has a value corresponding to a certain intensity), and (b) the processor automatically segments each image in the bone scan image set into specific anatomical regions of the human skeleton (for example, specific bones and / or sets of bones such as the cervical vertebrae, clavicle, ribs, lumbar vertebrae, pelvis, sacrum, scapula, skull, thoracic vertebrae, sternum, femur, humerus). Steps include: (c) Identifying one or more corresponding skeletal regions of interest to obtain a set of annotated images, wherein the one or more skeletal regions of interest include at least one of: (i) a femoral region corresponding to a portion of the femur of a human subject, wherein the portion of the femur encompasses at least three-quarters of the femur along its length (e.g., more than about three-quarters (e.g., approximately all)); and (ii) a humeral region corresponding to a portion of the humerus of a human subject, wherein the portion of the humerus encompasses at least three-quarters of the humerus along its length (e.g., more than about three-quarters (e.g., approximately all)); and (c) a processor automatically detecting an initial set of one or more hotspots, wherein each hotspot corresponds to a high-intensity area in the set of annotated images, and the automatic detection includes identifying one or more hotspots using the intensity 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 hotspot in an initial set of hotspots, the processor extracts a set of hotspot features associated with the hotspot (e.g., one or more sets) for each hotspot in an initial set of hotspots, based on the set of hotspot features associated with the hotspot. (f) a step of calculating a transfer probability value corresponding to the likelihood of a hotspot representing a transfer [for example, using one or more machine learning modules (e.g., pre-trained machine learning modules; e.g., artificial neural networks (ANNs)) that take at least a portion of the hotspot features as input for a particular hotspot and output a transfer probability value for that hotspot], and (f) a step of having the processor 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) superimposed on one or more members of a set of bone scan images and / or a set of annotated images; e.g., a table listing the identified hotspots along with additional information for each hotspot (e.g., location; e.g., transfer probability value)] for display in a graphical user interface (GUI) (e.g., cloud-based GUI).

[0021] In a particular embodiment, step (b) is a step of comparing each member of the bone scan image set with a corresponding atlas image of an atlas image set, wherein each atlas image includes one or more identifications of one or more skeletal regions of interest (e.g., identifications by graphics superimposed on the atlas image), wherein the skeletal regions of interest include a femoral region and / or a humeral region; and for each image of the bone scan image set, a step of aligning the corresponding atlas image with the image of the bone scan image set such that the identifications of one or more skeletal regions of interest of the atlas image are applied to (e.g., superimposed on) the image of the bone scan image set.

[0022] In a particular embodiment, each atlas image includes the identification of (i) a femoral region including at least a portion of the knee region of a human subject, and / or (ii) a humeral region including at least a portion of the elbow region of a human subject, and for each image in the bone scan image set, the alignment of the corresponding atlas image to the bone scan image includes using the identified knee region and / or identified elbow region as landmarks in the image [for example, 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) to align the corresponding atlas image with the bone scan image].

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

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

[0025] In a particular embodiment, step (c) includes (e.g., iteratively) the steps of: the processor identifying healthy tissue regions in a set of bone scan images that it has determined do not contain any hotspots (e.g., localized areas of relatively high intensity); the processor calculating a normalization coefficient such that the product of the normalization coefficient 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 coefficient.

[0026] In a particular embodiment, the method further includes the step of (g) a processor calculating one or more risk index values ​​for a human subject based at least in part on a calculated proportion (e.g., area proportion) of the human subject's skeleton occupied by an initial set of hotspots [e.g., the calculated proportion 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 a particular embodiment, the method includes (h) a step of having a processor select a first subset (e.g., up to all of) an initial set of hotspots based at least in part on the likelihood of displacement values ​​[e.g., deciding whether to include a particular hotspot in the subset from the initial set of hotspots based on whether the likelihood of displacement calculated for that particular hotspot exceeds a threshold], and (i) a step of having the processor render a graphical representation of the first subset [e.g., visual indications of the hotspots (e.g., points, boundaries); e.g., a table listing the identified hotspots along with additional information for each hotspot (e.g., location; e.g., likelihood value)] for display in a graphical user interface (GUI) (e.g., a cloud-based GUI).

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

[0029] In a particular embodiment, the method includes the steps of (k) a processor receiving a user selection of a second subset of an initial set of hotspots via a GUI, and (l) a processor calculating one or more risk index values ​​for a human subject based at least in part on a 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 indicator values ​​indicates the risk that a human subject has and / or develops metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancer).

[0031] In a particular embodiment, metastatic cancer is metastatic prostate cancer.

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

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

[0034] In certain embodiments, the GUI is part of a general picture archiving and communications system (PACS) (for example, a clinical application for oncology, including lesion marking and quantitative analysis).

[0035] In a particular embodiment, the drug (e.g., radiopharmaceutical) is technetium-99m methylenediphosphonate. 99m Includes Tc-MDP).

[0036] In another aspect, the present invention relates to a method for lesion marking and quantitative analysis of nuclear medicine images (e.g., a set of bone scan images) of a human subject (e.g., automated or semi-automated lesion marking and quantitative analysis assisted / checked by a user), the method comprising the steps of (a) accessing (e.g., and / or receiving) a set of bone scan images (e.g., a set of one, two, or more images) of a human subject by a processor of a computing device, wherein the set of bone scan images is a drug (e.g., radioactive) to the human subject (b) The images are obtained after administration of a drug (for example, a bone scan image set including ventral and dorsal bone scan images) (for example, each image in the bone scan image set has multiple pixels, and each pixel has a value corresponding to a certain intensity), (b) the processor automatically segments each image in the bone scan image set into specific anatomical regions of the human skeleton (for example, specific bones and / or sets of bones such as the cervical vertebrae, clavicle, ribs, lumbar vertebrae, pelvis, sacrum, scapula, skull, thoracic vertebrae, sternum, femur, humerus). (c) The steps of (i) identifying one or more corresponding skeletal regions of interest and thereby obtaining a set of annotated images, and (d) the processor automatically detecting an initial set of one or more hotspots, each hotspot corresponding to a high-intensity area in the set of annotated images, wherein the automatic detection includes detecting a set of potential hotspots using (i) the intensity of pixels in the set of annotated images and (ii) a set of pre-thresholds (for example, the pre-thresholds are region-dependent thresholds that depend on a identified skeletal region of interest where a particular pixel is located), calculating a global threshold scaling rate using the set of potential hotspots, adjusting the pre-thresholds using the global threshold scaling rate to obtain a set of adjusted thresholds, and identifying an initial set of hotspots using (i) the intensity of pixels in the set of annotated images and (ii) a set of adjusted thresholds, and (d) for each hotspot in the initial set of hotspots,The process includes the steps of: (e) the processor extracting a set of hotspot features associated with a hotspot (e.g., a set of one or more hotspot features); (a) for each hotspot in the initial set of hotspots, the processor calculating a likely transfer value corresponding to the likelihood that the hotspot will represent a transfer, based on the set of hotspot features associated with the hotspot [e.g., using one or more machine learning modules (e.g., pre-trained machine learning modules; e.g., artificial neural networks (ANNs)) that take at least a portion of the hotspot features as input for a particular hotspot and output a likely transfer value for that hotspot]; and (f) the processor rendering 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); e.g., a table listing the identified hotspots along with additional information for each hotspot (e.g., location; e.g., likely transfer value)] for display in a graphical user interface (GUI) (e.g., a cloud-based GUI).

[0037] In a particular embodiment, the global threshold scaling rate is a function of a measure of disease burden in a human subject [e.g., the percentage of the subject's skeleton occupied by metastases (e.g., hotspots); e.g., a risk index value], and adjusting a plurality of prior thresholds performed in step (c) includes reducing the adjusted threshold (e.g., with respect to the prior threshold) as the disease burden increases (e.g., as measured by the global threshold scaling rate) to compensate for underestimation of the hotspot area that occurs with increasing disease burden (e.g., so that the total number and / or size of hotspots increases with the reduction of the adjusted threshold).

[0038] In a particular embodiment, the global threshold scaling rate is a function of the proportion (e.g., area proportion) of the identified skeleton region occupied by a set of potential hotspots (e.g., a nonlinear function) (for example, the global threshold scaling rate is a function of the total area of ​​all hotspots in the preset divided by the total area of ​​all identified skeleton regions).

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

[0040] In a particular embodiment, step (c) includes (e.g., iteratively) the steps of: the processor identifying healthy tissue regions in a set of bone scan images that the processor has determined do not contain any hotspots (e.g., localized areas of relatively high intensity); the processor calculating a normalization coefficient such that the product of the normalization coefficient and the average intensity of the identified healthy tissue regions is a predetermined intensity level; and the processor normalizing the images of the bone scan image set by the normalization coefficient.

[0041] In a particular embodiment, the method further includes the step of (g) a processor calculating one or more risk index values ​​for a human subject based at least in part on a calculated proportion (e.g., area proportion) of the human subject's skeleton occupied by an initial set of hotspots [e.g., the calculated proportion 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 a particular embodiment, the method includes (h) a step of having a processor select a first subset (e.g., up to all of) an initial set of hotspots based at least in part on the likelihood of displacement values ​​[e.g., deciding whether to include a particular hotspot in the subset from the initial set of hotspots based on whether the likelihood of displacement calculated for that particular hotspot exceeds a threshold], and (i) a step of having the processor render a graphical representation of the first subset [e.g., visual indications of the hotspots (e.g., points, boundaries); e.g., a table listing the identified hotspots along with additional information for each hotspot (e.g., location; e.g., likelihood value)] for display in a graphical user interface (GUI) (e.g., a cloud-based GUI).

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

[0044] In a particular embodiment, the method includes the steps of (k) a processor receiving a user selection of a second subset of an initial set of hotspots via a GUI, and (l) a processor calculating one or more risk index values ​​for a human subject based at least in part on a 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 indicator values ​​indicates the risk that a human subject has and / or develops metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancer).

[0046] In a particular embodiment, metastatic cancer is metastatic prostate cancer.

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

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

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

[0050] In a particular embodiment, the drug (e.g., radiopharmaceutical) is technetium-99m methylenediphosphonate. 99m Includes Tc-MDP).

[0051] In another aspect, the present invention relates to a method for lesion marking and quantitative analysis of nuclear medicine images (e.g., a set of bone scan images) of a human subject (e.g., automated or semi-automated lesion marking and quantitative analysis assisted / checked by a user), the method comprising the steps of (a) accessing (e.g., and / or receiving) a set of bone scan images (e.g., a set of one, two, or more images) of a human subject by a processor of a computing device, wherein the set of bone scan images is used to mark a drug (e.g., a radiopharmaceutical drug) of a human subject. (b) The images are obtained after administration of a drug (for example, a bone scan image set including ventral and dorsal bone scan images) (for example, each image in the bone scan image set has multiple pixels, and each pixel has a value corresponding to a certain intensity), and (b) the processor automatically segments each image in the bone scan image set into specific anatomical regions of the human skeleton (for example, specific bones and / or sets of bones such as the cervical vertebrae, clavicle, ribs, lumbar vertebrae, pelvis, sacrum, scapula, skull, thoracic vertebrae, sternum, femur, humerus). (c) The process involves identifying one or more corresponding skeletal regions of interest and thereby obtaining a set of annotated images; (d) the processor automatically detecting an initial set of one or more hotspots, each hotspot corresponding to a high-intensity area in the set of annotated images [for example, detecting one or more hotspots in the initial set of hotspots involves comparing the pixel intensity to one or more thresholds (for example, these one or more thresholds vary depending on the identified skeletal region of interest where a particular pixel is located)]; (e) for each hotspot in the initial set of hotspots, the processor extracts a set of hotspot features associated with the hotspot (for example, a set of one or more hotspot features); and (f) for each hotspot in the initial set of hotspots, the processor calculates a transition probability value corresponding to the likelihood that the hotspot represents a transition, based on the set of hotspot features associated with the hotspot [for example, for a particular hotspot,(f) a step of using one or more machine learning modules (e.g., pre-trained machine learning modules; e.g., artificial neural networks (ANNs)) that take at least a portion of the hotspot features as input and output the transfer likelihood values ​​for the hotspots; and (f) a step of the processor selecting a first subset of an initial set of hotspots (e.g., up to all of them), wherein the selection of specific hotspots to be included in the first subset is based on (i) a calculated transfer likelihood value for the specific hotspot [e.g., a comparison of the calculated likelihood value for the specific hotspot with an likelihood threshold (e.g., if the specific hotspot has a likelihood value greater than the likelihood threshold, it will be included in the first subset)], and (ii) one or more global hotspot features, each within the initial set of hotspots The process includes (g) a step at least partially based on global hotspot features (e.g., the total number of hotspots in the initial hotspot set, the average intensity of the hotspots in the initial hotspot set, the peak intensity of the hotspots in the initial hotspot set, etc.) determined using multiple hotspots, and (g) a step causing a processor to render a graphical representation of at least a portion of a first subset of hotspots [e.g., a visual indication of the hotspots (e.g., points, boundaries); e.g., a table listing the identified hotspots along with additional information for each hotspot (e.g., location; e.g., estimated value)] for display in a graphical user interface (GUI) (e.g., a cloud-based GUI).

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

[0053] In a particular embodiment, step (f) includes adjusting the selection criteria for hotspots to include in a first subset based on the total number of hotspots in the initial hotspot set [for example, by relaxing the criteria as the total number of hotspots in the initial hotspot set increases (for example, by reducing the transfer likelihood value of each hotspot compared to the transfer likelihood value; for example, by scaling the transfer likelihood value based on the total number of hotspots in the initial hotspot set)].

[0054] In a particular embodiment, step (f) includes selecting a first subset using a machine learning module (e.g., an ANN module) [for example, for each hotspot, the machine learning module receives at least a calculated transfer likelihood value for that hotspot and one or more global hotspot features, and outputs (i) an adjusted transfer likelihood value that takes the global hotspot features into account (e.g., a scale value that can be compared to a threshold for selecting hotspots in the first subset), and / or (ii) a binary value (e.g., 0 or 1; e.g., Boolean true or false) indicating whether or not that hotspot should be included in the first subset].

[0055] In a particular embodiment, step (c) includes (e.g., iteratively) the steps of: the processor identifying healthy tissue regions in a set of bone scan images that the processor has determined do not contain any hotspots (e.g., localized areas of relatively high intensity); the processor calculating a normalization coefficient such that the product of the normalization coefficient and the average intensity of the identified healthy tissue regions is a predetermined intensity level; and the processor normalizing the images of the bone scan image set by the normalization coefficient.

[0056] In a particular embodiment, the method further includes the step of (g) a processor calculating one or more risk index values ​​for a human subject based at least in part on a calculated proportion (e.g., area proportion) of the human subject's skeleton occupied by an initial set of hotspots [e.g., the calculated proportion 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 a particular embodiment, the method includes (h) a step of having a processor select a first subset (e.g., up to all of) an initial set of hotspots based at least in part on the likelihood of displacement values ​​[e.g., deciding whether to include a particular hotspot in the subset from the initial set of hotspots based on whether the likelihood of displacement calculated for that particular hotspot exceeds a threshold], and (i) a step of having the processor render a graphical representation of the first subset [e.g., visual indications of the hotspots (e.g., points, boundaries); e.g., a table listing the identified hotspots along with additional information for each hotspot (e.g., location; e.g., likelihood value)] for display in a graphical user interface (GUI) (e.g., a cloud-based GUI).

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

[0059] In a particular embodiment, the method includes the steps of (k) a processor receiving a user selection of a second subset of an initial set of hotspots via a GUI, and (l) a processor calculating one or more risk index values ​​for a human subject based at least in part on a 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 indicator values ​​indicates the risk that a human subject has and / or develops metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancer).

[0061] In a particular embodiment, metastatic cancer is metastatic prostate cancer.

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

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

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

[0065] In a particular embodiment, the drug (e.g., radiopharmaceutical) is technetium-99m methylenediphosphonate. 99m Includes Tc-MDP).

[0066] In another aspect, the present invention relates to a method for lesion marking and quantitative analysis of nuclear medicine images (e.g., a set of bone scan images) of a human subject (e.g., automated or semi-automated lesion marking and quantitative analysis assisted / checked by a user), the method comprising: (a) a step of accessing (e.g., and / or receiving) a set of bone scan images of a human subject (e.g., a set of one, two, or more images) by a processor of a computing device (e.g., the set of bone scan images includes ventral bone scan images and dorsal bone scan images) (e.g., each image in the set of bone scan images comprises multiple pixels, each pixel having a value corresponding to a certain intensity); and (b) a step of the processor automatically segmenting each image in the set of bone scan images to identify one or more skeletal regions of interest, each corresponding to a specific anatomical region of the human subject's skeleton (e.g., a specific 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 obtaining a set of annotated images. (c) The processor automatically detects an initial set of one or more hotspots, each hotspot corresponding to a high-intensity area in a set of annotated images [for example, detecting one or more hotspots in the initial set of hotspots includes comparing pixel intensity to one or more thresholds (for example, these one or more thresholds vary depending on a specific skeletal region of interest where a particular pixel is located)]; (d) For each hotspot in the initial set of hotspots, the processor extracts a set of hotspot features associated with the hotspot (for example, a set of one or more hotspot features); and (e) For each hotspot in the initial set of hotspots, the processor calculates an expected value corresponding to the likelihood that the hotspot represents a transition based on the set of hotspot features associated with the hotspot [for example, for a particular hotspot, it takes at least a portion of the hotspot features as input and outputs an expected value for that hotspot].The process includes the steps of: (f) using one or more machine learning modules (e.g., pre-trained machine learning modules; e.g., artificial neural networks (ANNs)); (g) the processor selects a first subset of hotspots from an initial set of hotspots (e.g., up to all of them) based at least in part on the expected value calculated for each hotspot in the initial set of hotspots [e.g., determining whether to include a particular hotspot from the initial set of hotspots in the pre-selected set of hotspots based on the expected value calculated for that particular hotspot (e.g., by comparing it to an expected threshold)]; and (g) the processor calculates one or more risk index values ​​(e.g., bone scan index values) using at least a portion (e.g., up to all) of the first subset of hotspots, wherein the calculation is performed for each portion of the first subset. The method includes the steps of: calculating a bone invasion rate for a given hotspot based on the ratio of (i) the size (e.g., area) of the given hotspot to (ii) the size (e.g., area) of a given skeletal region to which the given hotspot is assigned (e.g., by a processor) based on its location within a 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 values ​​of the region-dependent correction factors are selected such that assigning the given hotspot to a given skeletal region (e.g., one of several adjacent or nearby skeletal regions such as the sacrum, pelvis, and lumbar region) reduces the degree to which assigning the given hotspot to a given skeletal region causes variation 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 a particular embodiment, for each specific hotspot, the calculated bone infiltration rate estimates the proportion of the total skeletal mass that is occupied by the physical volume associated with that particular hotspot.

[0068] In a particular embodiment, the step of calculating the bone infiltration rate includes the steps of: having the processor calculate the ratio of the area of ​​a particular hotspot to the area of ​​the corresponding skeletal region of interest, thereby calculating the area percentage of the particular hotspot; scaling (e.g., multiplying) the area percentage by a density coefficient 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 the joints within the corresponding skeletal region of interest (e.g., the density coefficient is the weight ratio of the corresponding skeletal region of interest 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 first subset of hotspots is assigned to a skeletal region of interest, which is a member selected from the group consisting of the pelvic region (e.g., corresponding to the pelvis of a human subject), the lumbar region (e.g., corresponding to the lumbar column of a human subject), and the sacral region (e.g., corresponding to the sacrum of a human subject).

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

[0071] In a particular embodiment, one or more region-dependent correction rates include one or more correction rate pairs, each correction rate pair being associated with a specific skeletal region of interest and including a first member and a second member (of the pair), the first member of the pair being a ventral image correction rate used to adjust the bone infiltration rate calculated for hotspots detected in an annotated ventral bone scan image of an annotated image set, and the second member of the pair being a dorsal image correction rate used to adjust the bone infiltration rate calculated for hotspots detected in an annotated dorsal bone scan image of an annotated image set.

[0072] In a particular embodiment, step (c) includes (e.g., iteratively) the steps of: the processor identifying healthy tissue regions in a set of bone scan images that it has determined do not contain any hotspots (e.g., localized areas of relatively high intensity); the processor calculating a normalization coefficient such that the product of the normalization coefficient and the average intensity of the identified healthy tissue regions is a predetermined intensity level; and the processor normalizing the images of the bone scan image set by the normalization coefficient.

[0073] In a particular embodiment, the method further includes the step of (g) a processor calculating one or more risk index values ​​for a human subject based at least in part on a calculated proportion (e.g., area proportion) of the human subject's skeleton occupied by an initial set of hotspots [e.g., the calculated proportion 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 a particular embodiment, the method includes (h) a step of having a processor select a first subset (e.g., up to all of) an initial set of hotspots based at least in part on the likelihood of displacement values ​​[e.g., deciding whether to include a particular hotspot in the subset from the initial set of hotspots based on whether the likelihood of displacement calculated for that particular hotspot exceeds a threshold], and (i) a step of having the processor render a graphical representation of the first subset [e.g., visual indications of the hotspots (e.g., points, boundaries); e.g., a table listing the identified hotspots along with additional information for each hotspot (e.g., location; e.g., likelihood value)] for display in a graphical user interface (GUI) (e.g., a cloud-based GUI).

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

[0076] In a particular embodiment, the method includes the steps of (k) a processor receiving a user selection of a second subset of an initial set of hotspots via a GUI, and (l) a processor calculating one or more risk index values ​​for a human subject based at least in part on a 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 indicator values ​​indicates the risk that a human subject has and / or develops metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancer).

[0078] In a particular embodiment, metastatic cancer is metastatic prostate cancer.

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

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

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

[0082] In a particular embodiment, the drug (e.g., radiopharmaceutical) is technetium-99m methylenediphosphonate. 99m Includes Tc-MDP).

[0083] In another aspect, the present invention relates to a system for lesion marking and quantitative analysis of nuclear medicine images of human subjects (e.g., sets of bone scan images) (e.g., automated or semi-automated lesion marking and quantitative analysis assisted / checked by a user), the system comprising a processor and a memory having instructions, the instructions, when executed by the processor, to (a) access (e.g., and / or receive) a set of bone scan images of a human subject (e.g., a set of one, two, or more images) (b) The bone scan image set is obtained after administration of a drug (e.g., a radiopharmaceutical) to a human subject (e.g., the bone scan image set includes ventral bone scan images and dorsal bone scan images) (e.g., each image in the bone scan image set comprises multiple pixels, each pixel having a value corresponding to a certain intensity), and (b) each image in the bone scan image set is automatically segmented to show specific anatomical regions of the human subject's skeleton (e.g., specific bones such as the cervical vertebrae, clavicle, ribs, lumbar vertebrae, pelvis, sacrum, scapula, skull, thoracic vertebrae, sternum, femur, humerus, etc.). The method involves identifying one or more skeletal regions of interest for each of one or more sets of bones, thereby obtaining a set of annotated images, wherein the one or more skeletal regions of interest include at least one of: (i) a femoral region corresponding to a portion of the femur of a human subject, wherein the portion of the femur encompasses at least three-quarters of the femur along its length [(e.g., more than about three-quarters (e.g., approximately all)], and (ii) a humeral region corresponding to a portion of the humerus of a human subject, wherein the portion of the humerus encompasses at least three-quarters of the humerus along its length [(e.g., more than about three-quarters (e.g., approximately all)], and (c) automatically detecting an initial set of one or more hotspots, each hotspot corresponding to a high-intensity area in the set of annotated images, wherein the automatic detection includes identifying one or more hotspots using the intensity 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 hotspot in the initial set of hotspots, extract a set of hotspot features associated with the hotspot (e.g., a set of one or more hotspot features) for each hotspot in the initial set of hotspots, based on the set of hotspot features associated with the hotspot. (The intensity of pixels located within a specific identified skeletal region of interest is compared to an associated region-dependent threshold.) One or more region-dependent thresholds include one or more values ​​associated with the femoral region and / or humeral region (e.g., a reduced intensity threshold for the femoral region and / or a reduced intensity threshold for the humeral region), which provide enhanced hotspot detection sensitivity in the femoral region and / or humeral region to compensate for reduced drug uptake within that region. (f) Calculate the likelihood of a hotspot to represent the likelihood of a hotspot to represent a transfer [for example, using one or more machine learning modules (e.g., pre-trained machine learning modules; e.g., artificial neural networks (ANNs)) that take at least a portion of the hotspot features as input for a given hotspot and output the likelihood of a transfer for that hotspot], and (f) render a graphical representation of at least a portion of the initial set of hotspots [for example, visual indications of the hotspots (e.g., points, boundaries) superimposed on one or more members of a bone scan image set and / or annotated image set; for example, a table listing the identified hotspots along with additional information for each hotspot (e.g., location; e.g., likelihood of transfer)] for display in a graphical user interface (GUI) (e.g., cloud-based GUI).

[0084] In a particular embodiment, step (b), the instruction causes the processor to compare each member of the bone scan image set with a corresponding atlas image from the atlas image set, wherein each atlas image includes one or more identifications of one or more skeletal regions of interest (e.g., identifications by graphics superimposed on the atlas image), the skeletal regions of interest including the femoral region and / or the humeral region, and for each image in the bone scan image set, align the corresponding atlas image with the image in the bone scan image set such that the identifications of 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 a particular embodiment, each atlas image includes the identification of (i) a femoral region including at least a portion of the knee region of a human subject, and / or (ii) a humeral region including at least a portion of the elbow region of a human subject, and for each image in the bone scan image set, the instruction causes the processor to align the corresponding atlas image with the bone scan image using the identified knee region and / or identified elbow region as landmarks in the image [for example, 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) to align the corresponding atlas image with the bone scan image].

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

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

[0088] In a particular embodiment, step (c) causes the processor to (e.g., iteratively) identify healthy tissue regions in the bone scan image set that are determined to contain no hotspots (e.g., localized areas of relatively high intensity), calculate a normalization coefficient such that the product of the normalization coefficient 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 coefficient.

[0089] In a particular embodiment, the instruction causes the processor to further calculate one or more risk index values ​​for a human subject based at least in part on a calculated percentage (e.g., area percentage) of the human subject's skeleton occupied by an initial set of hotspots [for example, 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 instruction causes the processor to (h) select a first subset (e.g., up to all of) an initial set of hotspots based at least in part on the likelihood of displacement [e.g., whether to include a particular hotspot in the subset from the initial set of hotspots is determined based on whether the likelihood of displacement calculated for that particular hotspot exceeds a threshold], and (i) render a graphical representation of the first subset [e.g., a visual indication of the hotspots (e.g., points, boundaries); e.g., a table listing the identified hotspots along with additional information for each hotspot (e.g., location; e.g., likelihood)] for display in a graphical user interface (GUI) (e.g., a cloud-based GUI).

[0091] In a particular embodiment, the instruction causes the processor to calculate one or more risk index values ​​for a human subject based at least in part on a calculated proportion (e.g., area proportion) of the human subject's skeleton occupied by a first subset of hotspots [for example, the calculated proportion is the total area of ​​the initial set of hotspots divided by the total area of ​​all identified skeletal regions].

[0092] In a particular embodiment, the instruction causes the processor to (k) receive a user selection of a second subset of an initial set of hotspots via a GUI, and (l) calculate one or more risk index values ​​for a human subject based at least in part on a calculated percentage (e.g., area percentage) of the human subject's skeleton occupied by the second subset of hotspots [for example, 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 indicator values indicates a risk that a human subject has and / or develops 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 indicator values indicates that a human subject has a particular state of metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancers).

[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) (e.g., clinical applications for oncology including lesion marking and quantitative analysis).

[0098] In certain embodiments, a drug (e.g., a radiopharmaceutical) includes technetium 99m methylene diphosphonate ( 99m Tc-MDP).

[0099] In another aspect, the present invention relates to a system for lesion marking and quantitative analysis (e.g., automated or semi-automated lesion marking and quantitative analysis assisted / checked by a user) of nuclear medicine images (e.g., a set of bone scan images) of a human subject, the system comprising a processor and a memory having instructions, the instructions, when executed by the processor, to (a) the processor of the computing device access (e.g., and / or receive) a set of bone scan images of a human subject (e.g., a set of one, two, or more images), the set of bone scan images obtained after administration of a drug (e.g., a radiopharmaceutical) to the human subject (e.g., the set of bone scan images includes ventral bone scan images and dorsal bone scan images) (e.g., each image in the set of bone scan images comprises a plurality of pixels, each pixel having a value corresponding to a certain intensity), and (b) the system automatically segments each image in the set of bone scan images to a specific anatomical region of the skeleton of the human subject (e.g., cervical vertebrae, clavicle, ribs, lumbar vertebrae, pelvis, sacrum, scapula, skull). (i) Identifying one or more skeletal regions of interest, each corresponding to a specific bone and / or set of one or more bones (such as the thoracic vertebrae, sternum, femur, humerus, etc.), thereby obtaining a set of annotated images; (c) Automatically detecting an initial set of one or more hotspots, each hotspot corresponding to a high-intensity area in the set of annotated images, wherein the automatic detection includes detecting a set of potential hotspots using (i) the intensity of pixels in the set of annotated images and (ii) a set of prior thresholds (e.g., the prior thresholds are region-dependent thresholds, depending on the identified skeletal regions of interest where a particular pixel is located); calculating a global threshold scaling rate using the set of potential hotspots; adjusting the prior thresholds using the global threshold scaling rate to obtain a set of adjusted thresholds; and identifying an initial set of hotspots using (i) the intensity of pixels in the set of annotated images and (ii) a set of adjusted thresholds.(d) For each hotspot in the initial set of hotspots, extract a set of hotspot features associated with the hotspot (e.g., one or more sets of hotspot features); (e) For each hotspot in the initial set of hotspots, calculate a transition probability value corresponding to the likelihood that the hotspot will represent a transition, based on the set of hotspot features associated with the hotspot [e.g., using one or more machine learning modules (e.g., pre-trained machine learning modules; e.g., artificial neural networks (ANNs)) that take at least a portion of the hotspot features as input for a particular hotspot and output a transition probability value for that hotspot]; and (f) 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 images of a bone scan image set and / or annotated image set; e.g., a table listing the identified hotspots along with additional information for each hotspot (e.g., location; e.g., transition probability value)] for display in a graphical user interface (GUI) (e.g., a cloud-based GUI).

[0100] In a particular embodiment, the instruction causes the processor to calculate a global threshold scaling rate as a function of a measure of disease burden in a human subject [e.g., the percentage of the subject's skeleton occupied by metastases (e.g., hotspots); e.g., a risk index value], and in step (c), adjust a number of prior thresholds by reducing an adjusted threshold (e.g., with respect to a prior threshold) as the disease burden increases (e.g., as measured by the global threshold scaling rate) to compensate for an underestimation of the hotspot area that occurs with increasing disease burden (e.g., so that the total number and / or size of hotspots increases with the reduction of the adjusted threshold).

[0101] In a particular embodiment, the instruction causes the processor to calculate a global threshold scaling rate as a function (e.g., a nonlinear function) of the proportion (e.g., area proportion) of the identified skeleton region occupied by a 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 skeleton regions).

[0102] In a particular embodiment, the instruction causes the processor to calculate a global threshold scaling rate based on (for example, as a function thereof) a risk index value calculated using a set of potential hotspots.

[0103] In a particular embodiment, step (c) causes the processor to (e.g., iteratively) identify healthy tissue regions in the bone scan image set that are determined to contain no hotspots (e.g., localized areas of relatively high intensity), calculate a normalization coefficient such that the product of the normalization coefficient 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 coefficient.

[0104] In a particular embodiment, the instruction further causes the processor to calculate one or more risk index values ​​for a human subject based at least in part on a calculated percentage (e.g., area percentage) of the human subject's skeleton occupied by an initial set of hotspots [for example, 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 instruction causes the processor to (h) select a first subset (e.g., up to all of) an initial set of hotspots based at least in part on the likelihood of displacement [e.g., whether to include a particular hotspot in the subset from the initial set of hotspots is determined based on whether the likelihood of displacement calculated for that particular hotspot exceeds a threshold], and (i) render a graphical representation of the first subset [e.g., a visual indication of the hotspots (e.g., points, boundaries); e.g., a table listing the identified hotspots along with additional information for each hotspot (e.g., location; e.g., likelihood)] for display in a graphical user interface (GUI) (e.g., a cloud-based GUI).

[0106] In a particular embodiment, the instruction causes the processor to calculate one or more risk index values ​​for a human subject based at least in part on a calculated proportion (e.g., area proportion) of the human subject's skeleton occupied by a first subset of hotspots [for example, the calculated proportion is the total area of ​​the initial set of hotspots divided by the total area of ​​all identified skeletal regions].

[0107] In a particular embodiment, the instruction causes the processor to (k) receive a user selection of a second subset of an initial set of hotspots via a GUI, and (l) calculate one or more risk index values ​​for a human subject based at least in part on a calculated percentage (e.g., area percentage) of the human subject's skeleton occupied by the second subset of hotspots [for example, 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 indicator values ​​indicates the risk that a human subject has and / or develops metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancer).

[0109] In a particular embodiment, metastatic cancer is metastatic prostate cancer.

[0110] In certain embodiments, at least one of the risk indicator values ​​indicates that the human subject has metastatic cancer of a specific condition (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 the processor of the cloud-based system.

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

[0113] In a particular embodiment, the drug (e.g., radiopharmaceutical) is technetium-99m methylenediphosphonate. 99m Includes Tc-MDP).

[0114] In another aspect, the present invention relates to a system for lesion marking and quantitative analysis of nuclear medicine images of human subjects (e.g., sets of bone scan images) (e.g., automated or semi-automated lesion marking and quantitative analysis assisted / reviewed by a user), the system comprising a processor and a memory having instructions, the instructions, when executed by the processor, to (a) access (e.g., and / or receive) a set of bone scan images of a human subject (e.g., a set of one, two, or more images) (b) The bone scan image set is obtained after administration of a drug (e.g., a radiopharmaceutical) to a human subject (e.g., the bone scan image set includes ventral bone scan images and dorsal bone scan images) (e.g., each image in the bone scan image set comprises multiple pixels, each pixel having a value corresponding to a certain intensity), and (b) each image in the bone scan image set is automatically segmented to show specific anatomical regions of the human subject's skeleton (e.g., cervical vertebrae, clavicle, ribs, lumbar vertebrae, pelvis, sacrum, scapula, skull, thoracic vertebrae, sternum, femur, humerus, etc.). (c) Identifying one or more skeletal regions of interest corresponding to each of a specific bone and / or a set of one or more bones, thereby obtaining a set of annotated images; (d) Automatically detecting an initial set of one or more hotspots, each hotspot corresponding to a high-intensity area in the set of annotated images [e.g., detecting one or more hotspots in the initial set of hotspots includes comparing pixel intensity to one or more thresholds (e.g., these one or more thresholds vary depending on the identified skeletal region of interest where a particular pixel is located)]; (e) 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 transition probability value corresponding to the likelihood that the hotspot represents a transition, based on the set of hotspot features associated with the hotspot [e.g., for a specific hotspot,(f) using one or more machine learning modules (e.g., pre-trained machine learning modules; e.g., artificial neural networks (ANNs)) that take at least a portion of the hotspot features as input and output the transfer likelihood values ​​for those hotspots; and (f) automatically selecting a first subset of an initial set of hotspots (e.g., up to all of them), wherein the selection of a particular hotspot to be included in the first subset is based on (i) the transfer likelihood value calculated for that particular hotspot [e.g., based on a comparison between the likelihood value calculated for that particular hotspot and the likelihood threshold (e.g., if that particular hotspot has a likelihood value greater than the likelihood threshold, it will be included in the first subset)], and (ii) one or more global hotspot features, each being part of the initial set of hotspots (g) at least partially based on global hotspot features (e.g., total number of hotspots in the initial hotspot set, average intensity of hotspots in the initial hotspot set, peak intensity of hotspots in the initial hotspot set, etc.) determined using multiple hotspots within, and rendering a graphical representation of at least a portion of a first subset of hotspots [e.g., visual indications of hotspots (e.g., points, boundaries); e.g., a table listing identified hotspots along with additional information for each hotspot (e.g., location; e.g., estimated value)] for display within a graphical user interface (GUI) (e.g., a cloud-based GUI).

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

[0116] In a particular embodiment, in step (f), the instruction causes the processor to adjust the criteria for selecting which hotspots to include in the first subset based on the total number of hotspots in the initial hotspot set [for example, by relaxing the criteria as the total number of hotspots in the initial hotspot set increases (for example, by reducing the transfer likelihood value compared for each hotspot; for example, by scaling the transfer likelihood value based on the total number of hotspots in the initial hotspot set)].

[0117] In a particular embodiment, in step (f), the instruction causes the processor to use a machine learning module (e.g., an ANN module) to select a first subset [for example, the machine learning module receives, for each hotspot, at least a calculated transfer probability value for that hotspot and one or more global hotspot features, and outputs (i) an adjusted transfer probability value that takes the global hotspot features into account (e.g., a scale value that can be compared to a threshold for selecting hotspots in the first subset), and / or (ii) a binary value (e.g., 0 or 1; e.g., Boolean true or false) indicating whether or not that hotspot should be included in the first subset].

[0118] In a particular embodiment, step (c) causes the processor to (e.g., iteratively) identify healthy tissue regions in the bone scan image set that are determined to contain no hotspots (e.g., localized areas of relatively high intensity), calculate a normalization coefficient such that the product of the normalization coefficient 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 coefficient.

[0119] In a particular embodiment, the instruction further causes the processor to calculate one or more risk index values ​​for a human subject based at least in part on a calculated percentage (e.g., area percentage) of the human subject's skeleton occupied by an initial set of hotspots [for example, 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 instruction causes the processor to (h) select a first subset (e.g., up to all of) an initial set of hotspots based at least in part on the likelihood of displacement [e.g., whether to include a particular hotspot in the subset from the initial set of hotspots is determined based on whether the likelihood of displacement calculated for that particular hotspot exceeds a threshold], and (i) render a graphical representation of the first subset [e.g., a visual indication of the hotspots (e.g., points, boundaries); e.g., a table listing the identified hotspots along with additional information for each hotspot (e.g., location; e.g., likelihood)] for display in a graphical user interface (GUI) (e.g., a cloud-based GUI).

[0121] In a particular embodiment, the instruction causes the processor to calculate one or more risk index values ​​for a human subject based at least in part on a calculated proportion (e.g., area proportion) of the human subject's skeleton occupied by a first subset of hotspots [for example, the calculated proportion is the total area of ​​the initial set of hotspots divided by the total area of ​​all identified skeletal regions].

[0122] In a particular embodiment, the instruction causes the processor to (k) receive a user selection of a second subset of an initial set of hotspots via a GUI, and (l) calculate one or more risk index values ​​for a human subject based at least in part on a calculated percentage (e.g., area percentage) of the human subject's skeleton occupied by the second subset of hotspots [for example, 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 indicator values ​​indicates the risk that a human subject has and / or develops metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancer).

[0124] In a particular embodiment, metastatic cancer is metastatic prostate cancer.

[0125] In certain embodiments, at least one of the risk indicator values ​​indicates that the human subject has metastatic cancer of a specific condition (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 the processor of the cloud-based system.

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

[0128] In a particular embodiment, the drug (e.g., radiopharmaceutical) is technetium-99m methylenediphosphonate. 99m Includes Tc-MDP).

[0129] In another aspect, the present invention relates to a system for lesion marking and quantitative analysis of nuclear medicine images of human subjects (e.g., sets of bone scan images) (e.g., automated or semi-automated lesion marking and quantitative analysis assisted / checked by a user), the system comprising a processor and a memory having instructions, the instructions, when executed by the processor, to (a) access (e.g., and / or receive) a set of bone scan images of a human subject (e.g., a set of one, two, or more images) (b) The bone scan image set includes ventral and dorsal bone scan images (for example, each image in the bone scan image set has multiple pixels, and each pixel has a value corresponding to a certain intensity), and (b) each image in the bone scan image set is automatically segmented to identify one or more skeletal regions of interest, each corresponding to a specific anatomical region of the human skeleton (for example, a specific bone and / or set of bones such as the cervical vertebrae, clavicle, ribs, lumbar vertebrae, pelvis, sacrum, scapula, skull, thoracic vertebrae, sternum, femur, humerus, etc.), thereby annotating (c) obtain a set of annotated images, (d) automatically detect an initial set of one or more hotspots, each hotspot corresponding to a high-intensity area in the set of annotated images [for example, detecting one or more hotspots in the initial set of hotspots involves comparing pixel intensity to one or more thresholds (for example, these one or more thresholds vary depending on a specific skeletal region of interest where a particular pixel is located)], (e) extract for each hotspot in the initial set of hotspots a set of hotspot features associated with the hotspot (for example, one or more sets of hotspot features), and (f) calculate an expected value for each hotspot in the initial set of hotspots, based on the set of hotspot features associated with the hotspot, corresponding to the likelihood that the hotspot represents a transition [for example, one or more machine learning modules that take at least a portion of the hotspot features as input for a particular hotspot and output an expected value for that hotspot (for example,(ii) using a pre-trained machine learning module (e.g., an artificial neural network (ANN)), (f) selecting a first subset of hotspots from the initial set of hotspots (e.g., up to all of them) based at least in part on the estimated value calculated for each hotspot in the initial set of hotspots (e.g., deciding whether to include a particular hotspot from the initial set of hotspots in the pre-selected set of hotspots based on the estimated value calculated for that particular hotspot (e.g., by comparing it to an estimated 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 calculation is performed such that for each particular hotspot in the portion of the first subset, (ii) The process includes: calculating a bone invasion rate based on the ratio of (i) the size (e.g., area) of a specific hotspot to the size (e.g., area) of a specific skeletal region assigned to that hotspot based on its location within a set of annotated images (e.g., by a processor), 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 values ​​of the region-dependent correction factors are selected such that assigning a specific hotspot to a specific skeletal region (e.g., one of several adjacent or nearby skeletal regions such as the sacrum, pelvis, and lumbar region) reduces the degree to which the calculated bone invasion rate varies], 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 a particular embodiment, for each specific hotspot, the calculated bone infiltration rate is estimated as the ratio of the total skeletal mass to the physical volume associated with that particular hotspot.

[0131] In a particular embodiment, the instruction causes the processor to calculate the bone infiltration rate by calculating the ratio of the area of ​​a particular hotspot to the area of ​​the corresponding skeletal region of interest, thereby calculating the area percentage of the particular hotspot; and to scale (e.g., multiply) the area percentage by a density coefficient 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 the joints within the corresponding skeletal region of interest (e.g., the density coefficient is the weight ratio of the corresponding skeletal region of interest to the total skeleton (e.g., of an average human)], thereby calculating the bone infiltration rate for the particular hotspot.

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

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

[0134] In a particular embodiment, one or more region-dependent correction rates include one or more correction rate pairs, each correction rate pair being associated with a specific skeletal region of interest and including a first member and a second member (of the pair), the first member of the pair being a ventral image correction rate used to adjust the bone infiltration rate calculated for hotspots detected in an annotated ventral bone scan image of an annotated image set, and the second member of the pair being a dorsal image correction rate used to adjust the bone infiltration rate calculated for hotspots detected in an annotated dorsal bone scan image of an annotated image set.

[0135] In a particular embodiment, step (c) causes the processor to (e.g., iteratively) identify healthy tissue regions in the bone scan image set that are determined to contain no hotspots (e.g., localized areas of relatively high intensity), calculate a normalization coefficient such that the product of the normalization coefficient 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 coefficient.

[0136] In a particular embodiment, the instruction further causes the processor to calculate one or more risk index values ​​for a human subject based at least in part on a calculated percentage (e.g., area percentage) of the human subject's skeleton occupied by an initial set of hotspots [for example, 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 instruction causes the processor to (h) select a first subset (e.g., up to all of) an initial set of hotspots based at least in part on the likelihood of displacement [e.g., whether to include a particular hotspot in the subset from the initial set of hotspots is determined based on whether the likelihood of displacement calculated for that particular hotspot exceeds a threshold], and (i) render a graphical representation of the first subset [e.g., a visual indication of the hotspots (e.g., points, boundaries); e.g., a table listing the identified hotspots along with additional information for each hotspot (e.g., location; e.g., likelihood)] for display in a graphical user interface (GUI) (e.g., a cloud-based GUI).

[0138] In a particular embodiment, the instruction causes the processor to calculate one or more risk index values ​​for a human subject based at least in part on a calculated proportion (e.g., area proportion) of the human subject's skeleton occupied by a first subset of hotspots [for example, the calculated proportion is the total area of ​​the initial set of hotspots divided by the total area of ​​all identified skeletal regions].

[0139] In a particular embodiment, the instruction causes the processor to (k) receive a user selection of a second subset of an initial set of hotspots via a GUI, and (l) calculate one or more risk index values ​​for a human subject based at least in part on a calculated percentage (e.g., area percentage) of the human subject's skeleton occupied by the second subset of hotspots [for example, 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 indicator values ​​indicates the risk that a human subject has and / or develops metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancer).

[0141] In a particular embodiment, metastatic cancer is metastatic prostate cancer.

[0142] In certain embodiments, at least one of the risk indicator values ​​indicates that the human subject has metastatic cancer of a specific condition (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 the processor of the cloud-based system.

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

[0145] In a particular embodiment, the drug (e.g., radiopharmaceutical) is technetium-99m methylenediphosphonate. 99m Includes Tc-MDP).

[0146] In another aspect, the present invention relates to a computer-aided image analysis device [e.g., a computer-aided detection (CADe) device; e.g., a computer-aided diagnostic (CADx) device] comprising any one of the embodiments and models 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 [for example, for receiving, transferring, storing, displaying, manipulating, quantifying, and reporting digital medical images acquired using nuclear medicine imaging; for example, the device provides clinical applications for oncology, including general picture archiving and communication system (PACS) tools, as well as lesion marking and quantitative analysis].

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

[0149] In a particular embodiment, the device is programmed to be used for analyzing bone scan images to evaluate and / or detect prostate cancer.

[0150] In certain embodiments, the device includes a label indicating that the device is intended for use by trained healthcare professionals and / or researchers [e.g., for receiving, transferring, storing, displaying, manipulating, quantifying, and reporting digital medical images acquired using nuclear medicine imaging; for example, the device provides a general picture archiving and communication system (PACS) tool as well as / or clinical applications for oncology, including lesion marking and quantitative analysis].

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

[0152] In a particular embodiment, the label further specifies that the device is used to analyze bone scan images for the evaluation and / or detection of prostate cancer.

[0153] Embodiments described in relation to one aspect of the present invention may be applied to other aspects of the present invention (for example, features of an embodiment described in relation to one independent claim, e.g., a method claim, are intended to be applicable to other embodiments of another independent claim, e.g., an apparatus claim, and vice versa).

[0154] The aforementioned and other purposes, aspects, features, and advantages of this disclosure will be made clearer and better understood by referring to the following description in conjunction with the accompanying drawings. [Brief explanation of the drawing]

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

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

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

[0158] [Figure 4] Figure 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 Figure 1, according to an explanatory embodiment.

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

[0160] [Figure 6] Figure 6 is a block flowchart of a process for processing whole-body bone scan images and determining bone scan index (BSI) values, according to an explanatory embodiment.

[0161] [Figure 7] Figure 7 is a diagram of a whole-body bone scan image set showing a skeletal atlas superimposed on ventral and dorsal bone scan images for skeletal segmentation, according to an embodiment for illustrative purposes.

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

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

[0164] [Figure 9-2] Figure 9B is a screenshot of a GUI window displaying a list of patients presented to the proposed new device, according to an explanatory embodiment.

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

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

[0167] [Figure 10CD] Figure 10C is a screenshot of a portion of a GUI window showing options for a color map for displaying bone scan images, according to an explanatory embodiment.

[0168] Figure 10D is a screenshot of a portion of a GUI window showing options for a color map for displaying bone scan images, according to an explanatory embodiment.

[0169] [Figure 11-1] Figure 11A is a screenshot of a portion of the GUI window for viewing bone scan images according to an explanatory embodiment.

[0170] [Figure 11-2] Figure 11B is a screenshot of a portion of the GUI window for viewing bone scan images according to an explanatory embodiment.

[0171] Figure 11C is a screenshot of a portion of the GUI window for viewing bone scan images, illustrating the zoom function according to an explanatory embodiment.

[0172] Figure 11D is a screenshot of a portion of the GUI window for viewing bone scan images, illustrating the zoom function according to an explanatory embodiment.

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

[0174] Figure 12B is a screenshot of a GUI window displaying a bone scan image using an intensity window that ranges up to the maximum intensity value, according to an explanatory embodiment.

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

[0176] Figure 12D is a screenshot of a portion of the GUI in Figure 12B, showing another graphical control for adjusting the intensity window threshold.

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

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

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

[0180] Figure 13B is a screenshot of a GUI that shows the local intensity value at the location of the mouse pointer displayed in the corner (lower left) of the GUI, according to an embodiment for illustrative purposes.

[0181] [Figure 14-1] Figure 14A is a screenshot of the GUI showing bone scan images from different selectable surveys, displayed via different GUI tabs, according to an embodiment for illustrative purposes.

[0182] Figure 14B is a screenshot of the GUI showing bone scan images from different studies displayed side-by-side simultaneously, according to an embodiment for illustrative purposes.

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

[0184] Figure 14D is a screenshot of a GUI showing ventral and dorsal images displayed side-by-side, according to an explanatory embodiment, where the visibility of various images can be selected via checkboxes within the GUI.

[0185] [Figure 15] Figure 15A is a screenshot of a GUI showing the display of total image intensity and total skeletal intensity according to an explanatory embodiment.

[0186] Figure 15B is a screenshot of the GUI showing the display of total image intensity according to an explanatory embodiment.

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

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

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

[0190] [Figure 17-2] Figure 17B is a screenshot of the GUI showing the identified hotspots according to an explanatory embodiment, and the updated BSI values ​​after excluding hotspots that were automatically detected based on user input via the GUI.

[0191] Figure 17C is a screenshot of the GUI showing the identified hotspots according to an explanatory embodiment, and the updated BSI values ​​after including previously excluded automatically detected hotspots based on user input via the GUI.

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

[0193] Figure 18B is a screenshot of a GUI showing a toggle graphical control for enabling user selection of hotspots to be included and / or excluded, according to an embodiment for illustrative purposes.

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

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

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

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

[0198] Figure 20B is a screenshot of a GUI displaying a graph showing the change in calculated BSI values ​​over time, according to an explanatory embodiment.

[0199] [Figure 21] Figure 21 is a screenshot of a GUI that provides a table listing the number of hotspots in a specific anatomical region used to calculate BSI values ​​in various studies, according to an embodiment for illustrative purposes.

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

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

[0202] [Figure 23] Figure 23 shows a set of bone scan images with atlas-based skeletal segmentation superimposed, comparing the use of a limited atlas (left) with a full-length atlas (right), according to an embodiment for illustrative purposes.

[0203] [Figure 24] Figure 24 is a block flowchart illustrating an exemplary process for improved image analysis via segmentation of the entire humeral and / or femoral region, and enhanced hotspot detection therein, according to an embodiment for illustrative purposes.

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

[0205] [Figure 25-2] Figure 25B is a block flowchart of an exemplary process for improved image analysis using a global threshold scaling technique, according to an embodiment for illustrative purposes.

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

[0207] [Figure 27] Figure 27 is a diagram of two sets of bone scan images showing portions around the sacrum, pelvis, and lumbar spine region, according to an embodiment for illustrative purposes.

[0208] [Figure 28] Figure 28 is a block flowchart of an exemplary process for calculating risk index values ​​based on bone infiltration rate and region-dependent correction rates, taking into account potential errors in hotspot localization, according to an embodiment for illustrative purposes.

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

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

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

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

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

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

[0215] [Figure 33-2] Figure 33B shows an exemplary microservices architecture.

[0216] [Figure 34] Figure 34 is a block diagram of an exemplary cloud computing environment used in a particular embodiment.

[0217] [Figure 35] Figure 35 is a block diagram of an exemplary computing device and an exemplary mobile computing device used in a particular embodiment. [Modes for carrying out the invention]

[0218] The features and advantages of this disclosure will become more apparent from the detailed description below, which is to be interpreted in conjunction with the drawings. In the drawings, similar reference numerals throughout identify corresponding elements. In the drawings, similar reference numerals generally indicate identical, functionally similar, and / or structurally similar elements.

[0219] Detailed explanation The systems, devices, methods, and processes of the present invention for which patent rights are claimed are intended to include modifications and adaptations developed using information from the embodiments described herein. Adaptations and / or modifications of the systems, devices, methods, and processes described herein can be made by those skilled in the art.

[0220] Throughout this description, where articles, devices, and systems are described as having, including, or comprising certain components, or processes and methods are described as having, including, or comprising certain steps, it is intended that, in addition, there exist articles, devices, and systems of the present invention that are essentially composed of or comprise the described components, and processes and methods of the present invention that are essentially composed of or comprise the described processing steps.

[0221] It should be understood that the order of the steps or the order in which certain operations are performed is not important, as long as the invention is operable. Furthermore, two or more steps or operations may be performed simultaneously.

[0222] Any reference to any document in this specification, for example in the background art section, does not constitute an endorsement that such document constitutes prior art with respect to any of the claims presented herein. The background art section is provided for clarity purposes and is not intended to be 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. Where used in this application, the term “comprise” and its variations such as “comprising” and “comprises” are not intended to exclude other additions, components, integers, or steps. Where used in this application, the terms “about” and “approximately” are used as synonyms. Any numbers used in this application, with or without “about / approximately,” are intended to cover ordinary variations as understood by those skilled in the art. In certain embodiments, the terms “approximately” or “about” mean, unless otherwise specified or evident from the context, a range of values ​​that fall within 25%, 20%, 19%, 18%, 17%, 16%, 15%, 14%, 13%, 12%, 11%, 10%, 9%, 8%, 7%, 6%, 5%, 4%, 3%, 2%, 1%, or less than 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 than or less than the stated reference value) (except when such a number exceeds 100% of the possible value).

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

[0226] The systems and methods described herein relate to improved computer-aided visualization and analysis of nuclear medicine images. Specifically, in certain embodiments, the systems and methods described herein result in improvements to several image processing steps used in the automated analysis of bone scan images for assessing a patient's cancer status. For example, improved techniques are provided for image segmentation, hotspot detection, automated classification of hotspots as representing metastases, and calculation of risk indicators such as bone scan index (BSI) values. The automated BSI calculation technique is 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 (with U.S. Patent Application No. 15 / 282,422 being a reissue), and PCT Application No. PCT / US17 / 58418 filed October 26, 2017, the contents of which are incorporated in whole by reference. PCT application PCT / US2017 / 058418, filed on 26 October 2017, which is also incorporated herein in its entirety, describes a cloud-based platform capable of functioning as a platform for providing image analysis and BSI calculation tools in accordance with the methods described herein.

[0227] Specifically, in certain embodiments, bone scan images are acquired 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., enlarged vascular system, abnormal bone formation) or due to the (drug-mediated) recognition of specific biomolecules that are 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 used to image the spatial distribution of the drug within the subject.

[0228] For example, in a particular embodiment, a gamma camera is used to acquire bone scan images as a two-dimensional scan. For instance, two images, a ventral image and a dorsal image, are acquired to form a bone scan image set. Areas of the body where the drug accumulates at high concentrations appear in the bone scan images as areas of high intensity, i.e., bright spots. The drug may accumulate in cancerous bone lesions, as described above, as well as in other areas such as the bladder of the target.

[0229] A series of image processing steps are performed to accurately identify regions of bone scan images representing lesions and to generate a quantitative estimate of tumor burden. Specifically, the bone scan images are segmented to identify regions corresponding to the bones of the target skeleton and to form a set of annotated images. Regions with high intensity relative to the surrounding area are identified within the skeletal region 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 metrics such as radius, eccentricity), and / or measures of hotspot intensity (e.g., peak intensity, mean intensity, integrated intensity), are extracted and used to determine a metastasis probability value for each hotspot, representing the likelihood that the hotspot represents metastasis. For example, in a particular embodiment, the metastasis probability value is calculated using an artificial neural network (ANN) that takes a set of hotspot features as input for each hotspot and outputs a metastasis probability value.

[0230] The likelihood of metastasis can be used to automatically filter an initial set of hotspots to determine a subset used in calculating a risk index indicating the risk of a subject having and / or developing metastatic cancer. By filtering hotspots in this way, only those determined to have a high likelihood of metastasis are included in the risk index calculation. In certain embodiments, a graphical representation of the hotspots and / or their likelihood values ​​may be rendered for the user to see, for example, as a table of markings and / or information overlaid on an annotated image, allowing the user to select a subset of hotspots to use in the risk index calculation. This allows the user to enhance the automatic selection of hotspots 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 in lesion detection and risk index calculation. For example, this disclosure includes an improved segmentation technique that identifies the entire humeral region (e.g., more than three-quarters of the length) and / or the entire femoral region (e.g., more than three-quarters of the length). Previously, only limited proportions of the femur and humerus were identified. By segmenting larger (e.g., entire) portions of these bones, it becomes possible to identify lesions located towards the extremities of the subject's arms and legs. To account for reduced drug uptake at such extremities, 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 femoral and humeral regions to increase detection sensitivity in those regions.

[0232] In another improved approach, the systems and methods described herein may utilize a global threshold scaling technique to detect hotspots. This approach first detects an initial set of hotspots by identifying a set of potential hotspots using multiple prior region-dependent thresholds. Using this set of potential hotspots, a global threshold scaling rate is calculated based on the area percentage of the subject's skeleton occupied by the set of potential hotspots. The prior thresholds are then adjusted using the global threshold scaling rate, and the adjusted thresholds are used to detect the initial set of hotspots. This approach has been found to ultimately increase the linearity of the risk index calculated using the initial set of hotspots, particularly in cases of high disease burden, such as when the subject suffers from numerous lesions and a large proportion of the skeleton is occupied by hotspots.

[0233] This disclosure also includes improvements to hotspot selection and risk index calculation. For example, the methods described herein may use a hotspot pre-selection technique that filters hotspots not only based on the calculated metastasis probability but also on global hotspot features that measure the overall nature 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 the total number of hotspots, such as the average number of hotspots per region; measures of overall hotspot intensity, such as peak or average hotspot intensity; and measures of total hotspot size, such as the total area of ​​hotspots or average hotspot size. This allows the processing method to leverage clinical experience that shows hotspot selection depends on the rest of the image. Specifically, the probability of a hotspot being selected is higher when there are many other hotspots and lower when it is the only hotspot. Therefore, selecting or filtering hotspots based solely on their individual metastasis probability may result in an underestimation of the calculated risk index value in subjects with many hotspots. Incorporating global features as described herein can improve outcomes in patients with many hotspots.

[0234] Finally, the systems and methods described herein also provide improvements to techniques for calculating risk index values ​​based on bone invasion, such as the Bone Scan Index (BSI). For example, BSI is a risk index value that provides an estimate of the proportion of a subject's total skeletal mass that is occupied by cancerous lesions. Calculating BSI involves calculating the bone invasion rate for each specific 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 subject's BSI value is calculated by summing up scaled versions (e.g., to convert the area ratio to relative mass). However, the difficulty in accurately pinpointing the specific skeletal region in which a particular hotspot is located can lead to errors in the BSI value. Bone scan images are two-dimensional images, while the underlying skeleton of the subject is a three-dimensional structure. Therefore, a hotspot may be misidentified as being located in a certain region when it actually represents a lesion located in a different bone. This is particularly challenging in the case of the sacrum and the adjacent pelvic and lumbar regions. To address this challenge, this disclosure includes a modified risk index calculation method that scales the bone invasiveness rate in a manner that accounts for potential errors in localizing hotspots using a region-dependent correction factor. This method improves the accuracy of the BSI calculation and limits its sensitivity to hotspot localization.

[0235] Therefore, the systems and methods described herein include several improved image analysis techniques for the identification and quantification of lesions. These techniques 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 the 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, proton emission tomography (PET) imaging, and single-photon emission computed tomography (SPECT) imaging.

[0237] As used herein, an "image", such as a 3D image of a mammal, includes any visual representation, such as a photograph, video frame, streaming video, and any electronic, digital, or mathematical analog of a photograph, video frame, or streaming video. Any device described herein includes, in certain embodiments, a display for displaying an image or any other result produced by a processor. Any method described herein includes, in certain embodiments, the 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 regarding 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 administering a 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" a drug 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 cerebrospinal fluid, or injection into a body compartment.

[0240] As used herein, “radionuclide” means a component comprising a radioisotope consisting of at least one element. Suitable exemplary radionuclides include, but are not limited to, those described herein. In some embodiments, the radionuclide is a radionuclide used in proton emission tomography (PET). In some embodiments, the radionuclide is a radionuclide used in single-photon emission computed tomography (SPECT). In some embodiments, the non-restrictive 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 192It includes Ir.

[0241] As used herein, the term “radiopharmaceutical” means a compound comprising a radionuclide. In certain embodiments, the radiopharmaceutical is used for diagnostic and / or therapeutic purposes. In certain embodiments, the radiopharmaceutical comprises a small molecule labeled with one or more radionuclides, an antibody labeled with one or more radionuclides, and an antigen-binding moiety of an antibody labeled with one or more radionuclides.

[0242] Nuclear medicine images (e.g., PET scans; SPECT scans; whole-body bone scans; synthetic PET-CT images; synthetic SPECT-CT images) form images by detecting radiation emitted from the radionuclides of radiopharmaceuticals. 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 leverage different biological mechanisms and / or specific, unique enzyme or receptor binding interactions, and thus, when administered to a patient, can selectively concentrate in specific types of tissues and / or regions within the patient's body. Regions within the patient's body with higher concentrations of radiopharmaceutical emit more radiation than other regions, and these regions appear brighter in nuclear medicine images. Therefore, the variation in intensity within nuclear medicine images can be used to map the distribution of radiopharmaceuticals within a patient's body. This mapped distribution of radiopharmaceuticals within the patient's body can be used, for example, to infer the presence of cancerous tissue in various regions of the patient's body.

[0243] For example, when administered to a patient, methylenediphosphonate technetium-99m( 99m Tc MDP selectively accumulates within the patient's skeletal region, particularly in areas with abnormal bone formation associated with malignant bone lesions. This selective concentration of the radiopharmaceutical in such areas creates identifiable hotspots, i.e., localized areas of high intensity within nuclear medicine images. Therefore, by identifying such hotspots within a whole-body scan of a patient, the presence of malignant bone lesions associated with metastatic prostate cancer can be inferred. As described below, risk indicators correlate with the patient's overall survival and other prognostic indicators, such as disease status, progression, and treatment response, in the patient... 99m It can be calculated based on an automated analysis of the intensity variability in whole-body scans obtained after administration of Tc MDP. In certain embodiments, other radiopharmaceuticals may also be used. 99m It can be used in the same way as Tc MDP.

[0244] In certain embodiments, the specific radiopharmaceutical used depends on the specific nuclear medicine imaging modality used. For example, sodium 18F fluoride (NaF) is also used. 99m Like Tc MDP, it accumulates in bone lesions but can be used in PET imaging. In certain embodiments, PET imaging may use a radioactive form of vitamin choline that is readily absorbed by prostate cancer cells.

[0245] In certain embodiments, radiopharmaceuticals that selectively bind to specific target proteins or receptors, particularly those whose expression is increased in cancer tissue, may be used. Such target proteins or receptors include, but are not limited to, tumor antigens such as CEA expressed in colorectal carcinoma, Her2 / neu expressed in multiple 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 in virtually all prostate cancers. Furthermore, its expression is further increased in poorly differentiated metastatic carcinomas and hormone-refractory carcinomas. Therefore, a radiopharmaceutical corresponding to a PSMA conjugate labeled with one or more radionuclides (e.g., a compound with high affinity for PSMA) 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 conjugate can be used to identify the presence of cancerous tissue within the prostate. In certain embodiments, as applicable when the disease is metastatic, nuclear medicine images obtained using a radiopharmaceutical containing a PSMA conjugate can be used to identify the presence of cancerous tissue in various regions, including not only the prostate but also other organs and tissue regions such as the lungs, lymph nodes, and bones.

[0247] Specifically, when administered to a patient, the radionuclide-labeled PSMA conjugate selectively accumulates in cancer tissue based on its affinity for PSMA. 99m Similar to the explanation above regarding Tc MDP, selective concentration of the radionuclide-labeled PSMA conjugate to specific sites within the patient's body creates detectable hotspots in nuclear medicine imaging. As the PSMA conjugate concentrates within various cancerous tissues and areas of the body expressing PSMA, localized cancers within the patient's prostate and / or metastatic cancers 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 indicators, such as disease status, progression, and treatment response, can be calculated based on automated analysis of the intensity variability in nuclear medicine imaging obtained after the patient has been administered the PSMA conjugate radiopharmaceutical.

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

[0249] Various PSMA binders and their versions labeled with radionuclides are described in U.S. Patents 8,778,305, 8,211,401, and 8,962,799, each of which is incorporated herein by reference in its entirety. Several PSMA binders and their versions labeled with radionuclides are also described in PCT application PCT / US2017 / 058418, filed on 26 October 2017, which is incorporated herein by reference in its entirety. B. Bone scanning imaging devices for image analysis

[0250] In certain embodiments, the computer-aided image analysis devices described herein are intended to be used by trained healthcare professionals and researchers for the reception, transmission, storage, 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 known as bone scan imaging) is a widely used imaging modality for assessing the skeletal disease burden. The current standard for assessing disease progression based on bone scan images is based on the quasi-quantitative modified Prostate Cancer Working Group 2 and Prostate Cancer Working Group 3 (PCWG) criteria. The definition of these criteria relies on the appearance of new lesions, as interpreted by a trained interpreter, either (i) two new lesions at the first and second post-hoc scans compared to the pre-treatment scan, followed by two additional lesions (known as the 2+2 rule), or (ii) two new confirmed lesions for the subsequent first post-hoc scan. However, this quasi-quantitative assessment method for counting lesions has several limitations. Specifically, the assessment is affected by manual variability, is limited in its assessment of disease progression, and cannot accurately assess the fusional / spread disease burden, such as in metastatic prostate cancer.

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

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

[0254] In certain embodiments, the systems and methods described herein may be implemented as cloud-based platforms for automated and semi-automated image analysis for the detection and assessment of cancer status in patients. 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 specific image files, such as DICOM files. The software conforms to the Digital Imaging and Communications in Medicine (DICOM 3) standard.

[0255] In certain embodiments, a device (e.g., a computer-aided image analysis tool) following the system and methods described in the present invention is programmed to suit an intended user, typically a healthcare professional who uses the software to view patient images and analyze the results. The user interacts with this 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 interact with 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 analysis is reviewed by a physician, who may be guided through the quality control and reporting workflow. Once the quality control of the automated assessment is certified, a report can be created and signed. The service can be configured to comply with HIPAA and 21 CFR part 11. i. Access to the Service

[0257] In certain embodiments, access to the software tools according to 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 verification code sent as a text message to the phone number associated with the username, password, and account. ii. System Requirements

[0258] In certain embodiments, the requirements implemented by software include one or more of the following. · A computer with Windows or OS X with Internet access, · Chrome browser, · An available personal mobile phone (for multi-factor authentication only).

[0259] In certain embodiments, the requirements implemented by the user include one or more of the following. · Chrome browser a. At least version 54 b. JavaScript® must be permitted c. HTML5 is required d. Writing to local storage and session storage is required · Display resolution of at least 1280 × 960 iii. Image Requirements

[0260] In a 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" The image type (0008, 0008) must be "ORIGINAL\PRIMARY\WHOLE BODY\EMISSION". • The survey date (0008, 0020) must include a valid date. • The number of frames (0028, 0008) must be 1 or 2. • The number of slices (0054, 0081) must be 1 or 2. The 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. • The patient's 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 a single multi-frame series (one file) containing both frames. Image pixel data should be within the 16-bit range. Images with pixels in the range of 0 to 255 (8 bits) are not sufficient.

[0261] In a 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 upper part of the tibia and the upper part of the forearm of each arm. • No filtering or other post-processing techniques should be applied to the images.

[0262] In a particular embodiment, whole-body bone scintigraphy images are acquired in accordance with relevant guidelines such as "EANM Bone Scintigraphy: Tumor Imaging Procedure Guidelines" and "ACR-SPR Practice Parameters for Performing Bone Scintigraphy (Bone Scan)". iv. Workflow

[0263] Figure 1 is a block flowchart showing a quality control and reporting workflow 100 for generating a BSI report according to an embodiment for illustrative purposes. In a particular embodiment, the systems and methods described herein include a GUI for guiding a user (e.g., a healthcare professional) through the review and analysis of patient image data to calculate automated BSI values ​​and generate a report. The workflow 100 allows the user to select 102 and upload 104 image files of patient data, which are 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, from which the user can select a specific patient from a list for inspection / analysis. When a column corresponding to a specific patient is selected in the list in Figure 2, a new and / or updated window 300, showing the patient information, is displayed, as shown in Figure 3.

[0265] Moving to Figure 4, the user may then access the guided analysis software inspection page 400. The inspection page 400 provides the user (e.g., a healthcare professional such as an internist) with a GUI that allows them to examine the automated identification of hotspots performed by the image data and software backend in order to calculate the automated BSI index of a patient. The automated BSI calculation technique is described in more detail herein, as well as in previous versions (without improvements of this disclosure) which 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 (with U.S. Patent Application No. 15 / 282,422 being a reissue), and PCT Application No. PCT / US17 / 58418 filed October 26, 2017, the contents of which are incorporated herein by reference in their entirety.

[0266] The inspection page allows the user to inspect hotspots representing cancerous lesions automatically identified in the image by the software.106 The user must use the inspection page GUI to edit the set of areas identified as hotspots,108 and confirm that the image quality, skeletal segmentation (drawn by contour lines in the screenshot), and set of identified hotspots have been inspected and accepted before proceeding to report generation.110 Once the user's inspection and quality control are confirmed,112 a report like the report shown in Figure 5 may be generated,114 including the final BSI calculation. v. Image Processing

[0267] As shown in the block flowchart of Figure 6, the systems and methods described herein may be used in process 600 for automated detection and pre-selection of hotspots, receipt of user validation of a pre-selected set of hotspots, and in particular for calculation of risk indicator values ​​such as bone scan index (BSI) values.

[0268] In particular, in certain embodiments, as shown in Figure 6, in segmentation step 602, the software tool described herein places patient images into a reference coordinate frame using an image alignment algorithm. This is done by non-rigidly adjusting the atlas image to each patient image. By fitting the atlas image to the patient, the patient image can be segmented into skeletal and background regions. The skeletal region can be further divided into smaller skeletal regions of interest, which is also referred to herein as localization. In 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. Hotspots are detected in hotspot detection step 606. In certain embodiments, hotspots are detected using thresholding rules. In another step, hotspot pre-selection 608 is performed. In certain embodiments, hotspot pre-selection is based on image analysis and machine learning techniques, aiming to pre-select important hotspots to be included in a set of pre-selected hotspots, which can be checked by the user in hotspot validation step 610. This pre-selection is based on a range of characteristics of the hotspot, such as its 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 the user must make (for example, to select hotspots to use for calculating risk indicators such as BSI values). 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 pre-selection of hotspots, or manually include and / or exclude hotspots if necessary, so that a report can be generated.

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

number

[0270] In certain embodiments, skeletal segmentation is performed using a registration-based segmentation method in which an atlas image is aligned with the bone scan image to be segmented. In this method, the device automatically contours the skeleton into separate skeletal regions by elastically fitting a manually contoured, annotated template image set to each image set being analyzed. This annotated template image set is known as a skeletal atlas image set. This atlas image set is structured in the same way 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 point when analyzing the scan. The atlas is manually annotated with regions of interest (skeletal regions) that can be transferred to a new scan to accurately calculate the BSI. An exemplary atlas image set 700 with 31 manually drawn skeletal regions is shown in Figure 7. As shown in Figure 7, like the bone scan image set, the atlas image set 700 may include a ventral image 702 and a dorsal image 704.

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

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

[0273] Construction of a skeletal atlas. The exemplary aBSI devices described herein rely 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 so that the distance between similar structures in the two images is minimized. Structure is defined by edges and ridges (lines) in each image. Therefore, global intensity differences and texture patterns are ignored by the algorithm. As a result, a suitable atlas image exhibits the following two important properties: • Display the same pattern of edges and ridges, and The (elastic) transformations required to align the images are generally minimized across the expected set of anatomical variations between the images being analyzed.

[0274] In certain embodiments, to satisfy these requirements, an atlas image based on a database of actual, normal (without metastasis 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. Then, an average transformation is calculated from all the resulting transformations. Subsequently, all images are transformed to match this anatomical average to represent the average anatomical structure in the database. In this process, the intensity is also normalized to create a typical bone scan image suitable as an anatomical reference. The schematic diagram shown in Figure 8 conceptually illustrates this idea, namely how the average anatomical structure and intensity (central image 802) can be inferred from multiple bone scan images 804a, 804b, 804c, and 804d in the database.

[0275] In certain embodiments, the average anatomical structure used in the atlas image set rapidly converges to a stable estimate as the number of scans included increases. A relatively small number of scans (e.g., 30) may be sufficient to create a representative reference image. Furthermore, since this algorithm is driven by the major structures in the image and does not respond to differences in shape and / or size, a single atlas can be applied to any bone scan image for skeletal segmentation. Intensity normalization and hotspot detection

[0276] One challenge in reading scintigraphy images, such as bone scans, is that intensity levels can vary between scans due to various parameters such as injection dose, time from injection to scan, scan time, body type, camera hardware, and settings. In certain embodiments, to facilitate reading for the user and as part of a quantification pipeline (as shown in Figure 6, for example), the input bone scan is normalized so that the average intensity of healthy bone tissue is scaled to a predetermined baseline level. At this stage of the quantification pipeline, skeletal segmentation has already been performed, and the pixels belonging to the skeleton are known. However, in order to measure the average intensity of healthy tissue, and thus to normalize the scan, areas of high intensity must be identified and excluded. Detecting these hotspots within the skeleton is straightforward if the image has already been normalized. This is a chicken-or-the-egg kind of problem: here, hotspot detection is needed to normalize the image, and hotspot detection depends on the normalized image. Therefore, this challenge can be addressed using an iterative approach such as the steps listed below. (1) Estimate normalization by assuming that all bone tissue is healthy (no hotspots). (2) Detect hotspots in light of the current normalization. (3) Estimate the normalization in light of the current set of hotspots. (4) Repeat steps (2) and (3) until convergence occurs.

[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 Gaussian difference-of-Gaussians band-pass filter that emphasizes high-intensity small regions relative to their surroundings. This filtered image is then thresholded to a certain level based on region.

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

[0279] The output of hotspot detection is a set of ROIs (Regions of Interest) representing hotspots in the image, as well as normalization coefficients used to set initial maximum and minimum thresholds for image windowing. Pre-selection of hotspots

[0280] Hotspots may be classified as either to be included or excluded for pre-selection using a data-driven learning method based on an artificial neural network (ANN). The ANN may be adapted / trained on a training database that includes patients ranging from normal bone scans to bones with numerous widespread hotspots.

[0281] Each hotspot in the training database is characterized using a set of features (measurements) regarding its size, location, orientation, shape, and texture. These features are fed to the ANN first during the training phase, when the ANN's parameters are set to maximize classification performance in the cross-validation study, and then within the actual software used to classify whether to include or exclude the hotspots.

[0282] Pre-selection training analyzes hotspots and their immediate neighbors within the image to make the classifier robust to large-scale differences in the training material. Therefore, the classifier becomes 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 ANN parameters, separate networks can be constructed for each different bone localization. Then, the set of input features for each of those ANNs will differ somewhat. For example, symmetry features are only applicable to localizations that have a natural symmetrical counterpart.

[0283] In addition, hotspots in the training set are typically manually labeled as to be included or excluded by a healthcare professional trained in reading bone scans. Target labels may be validated by a second healthcare professional. In one exemplary method, US and European procedural guidelines for bone scintigraphy are combined, and the equipment used to obtain bone scans is the same in both the US and Europe. Furthermore, criteria used to interpret bone scans for clinical trials, such as the Prostate Working Group 2 criteria, are used globally. This is based on the common knowledge in the nuclear medicine community that variability in the appearance of bone scans due to cancerous disease is far more pronounced than small variability that may be measured within normal bone density, for example, between different races. The parameters of the ANN are optimized so that the resulting classifier mimics the choices of healthcare professionals. Cross-validation techniques may be used to avoid bias toward the training set.

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

[0285] The ANNs described herein may be implemented via one or more machine learning modules. As used herein, the term “machine learning module” means 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 (e.g., an image (e.g., a 2D image; e.g., a 3D image), a dataset, etc.). For example, a machine learning module may take a 3D image of a subject (e.g., a CT image; e.g., an MRI image) as input and, for each voxel in the image, determine a value that represents the likelihood that the voxel lies in a region of the 3D image corresponding 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, for example, as separate software applications. Machine learning modules may be software and / or hardware. For example, a machine learning module may be implemented entirely in software, or certain functions of a CNN module may be performed via specialized hardware (e.g., an application-specific integrated circuit (ASIC)). E. Graphical User Interface and Image Display

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

[0287] In certain embodiments, a GUI system using the techniques described herein facilitates image viewing. For example, automatic image resizing to fit the screen size may be provided. Figure 11A shows a GUI embodiment without this feature, where the user changes the image size by clicking a zoom icon. Figure 11B shows a screenshot of a GUI where the image is automatically resized to fill the screen vertically. Figures 11C and 11D show two techniques for providing zoom functionality for viewing images in detail. In the embodiment shown in Figure 11C, zooming occurs around the mouse pointer when the scroll wheel button is clicked and held. In the embodiment shown in Figure 11D, zoom and pan functionality is provided using the mouse scroll wheel as well as click and drag actions.

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

[0289] Figures 13A and 13B show screenshots of exemplary GUI windows where local intensity is displayed. Specifically, in certain embodiments, local intensity is displayed when the user hovers the mouse pointer over an image. In the embodiment shown in Figure 13A, the local intensity value is displayed at the location of the mouse pointer. In the embodiment shown in Figure 13B, the intensity is displayed in the lower left corner of the image instead of next to the mouse pointer and obscuring part of the image (for example, providing improved safety accordingly).

[0290] Figures 14A–14D show exemplary GUI embodiments for displaying multiple bone scan images. In the embodiments shown in Figures 14A and 14C, a tabular method is used to allow the user to switch between different sets of bone scan images for different studies, or to switch between ventral and dorsal images. In the embodiments shown in Figures 14B and 14C, checkboxes are used to switch the visibility of different images and to display multiple images from different studies side by side.

[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 skeletal intensity according to an explanatory embodiment. In certain embodiments, such as the GUI shown in Figure 15B, only the total image intensity is displayed (total skeletal intensity is not displayed) to provide a simpler user interface (for example, to avoid confusion).

[0292] A GUI tool for inspecting bone scan images may display iconographic indications of detected (e.g., pre-selected) hotspots overlaid on the bone scan image, as shown, for example, in Figures 16A and 16B. In certain embodiments, a table of hotspots, as shown in Figure 17A, is also displayed. In certain embodiments, the GUI tool described herein allows the user to select or deselect pre-selected hotspots to include 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 Figures 17B and 17C, allowing the user to observe its changes (comparing 1704a to 1704b) along with the selection and / or deselection of various hotspots (1702a and 1702b). Various graphical control techniques may be used to allow the user to select to include or exclude pre-selected hotspots. For example, Figure 18A shows a popup control that appears after a user right-click action. Figure 18B shows a toggle switch 1802 ("Edit Hotspots") that can be turned on or off. When the toggle switch is on, the user can click on the various hotspots to select or deselect them. In certain embodiments, the GUI tool may include safety / quality control features, such as a pop-up window shown in Figure 18C, which prompts the user to verify quality control requirements before a report is generated.

[0293] In certain embodiments, once BSI values ​​for various studies are calculated, they are displayed for user review. For example, the calculated BSI values ​​may be displayed in a table, as shown in Figure 19A. In certain embodiments, the calculated BSI values ​​for different studies are displayed above the displayed bone scan images for the corresponding studies, as shown in Figure 19B. A chart showing the time evolution of the calculated BSI values ​​may also be displayed, as shown in Figures 20A and 20B. Additional information regarding the calculated BSI values ​​may be displayed. Figure 21 shows a screenshot of a GUI that provides a table listing the number of hotspots in a particular anatomical region used to calculate BSI values ​​for different studies, according to an embodiment for illustrative purposes. In certain embodiments, the systems and methods described herein enable automated report generation. Figures 22A and 22B show examples of automatically generated reports. F. Improved image processing techniques

[0294] In certain embodiments, the systems and methods described herein include modifications to one or more of the image processing steps shown in Figure 6. i. Skeletal segmentation

[0295] As described herein, a set of skeletal atlas images may be used for image segmentation. The set of skeletal atlas images includes a pair of template bone scan images (ventral and dorsal) representing a typical normal bone scan, as well as manual contouring of 31 skeletal regions. These regions are distorted to match the current patient image to be analyzed. In certain embodiments, a limited atlas is used with skeletal identification covering only three-quarters of the femur and humerus. In certain embodiments, an improved full-length atlas is used, which includes identification of skeletal regions covering the entire femur and humerus. Figure 23 shows a set of bone scan images with skeletal segmentation based on the atlas, comparing the use of the limited atlas 2300a (left) and the full-length atlas 2300b (right). As shown in Figure 23, the restricted atlas 2300a includes only three-quarters or less of the humerus 2302a and femur 2304a, while 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 enhance the stability of skeletal segmentation. Specifically, the enlarged atlas allows the knee and elbow to be used as references during alignment. Such reference points are beneficial for image analysis due to their clear contrast and provide improved safety. The enlarged atlas also allows for the detection of hotspots in the extremities. ii. Hotspot detection threshold

[0296] As described herein, an initial set of candidate hotspots is found using image intensity thresholding. In certain embodiments, a global, fixed threshold is used so that the same value is used across all skeletal regions of interest and all images. Another improved technique involves setting region thresholds that vary across different skeletal regions of interest. For example, this technique has increased detection sensitivity by allowing a reduced threshold to be used for the femoral and humeral regions (e.g., lowered from 650 to 500). The femoral and humeral regions show less uptake than other skeletal regions in the bone scan image. Therefore, a lower threshold can be used for those regions to achieve a similar level of sensitivity as the rest of the body. This functionality is enabled by setting individual thresholds for different skeletal regions, which leads to increased detection of lower intensity hotspots within those skeletal regions.

[0297] Figure 24 shows an exemplary process 2400 for lesion marking and analysis using the improved segmentation and region-dependent thresholding techniques 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 a skeletal region of interest, which includes a femoral region 2422 corresponding to more than three-quarters of the length of the subject's femur, and / or a humeral region 2424 corresponding to more than three-quarters of the length of the subject's humerus. The identified skeletal region of interest is then analyzed to automatically detect an initial set of hotspots 2430. As described herein, this step of automatically detecting hotspots may include applying a thresholding operation to each skeletal region of interest using region-dependent thresholds, i.e., thresholds whose values ​​are not uniform and differ for each skeletal region. Specifically, lower threshold values ​​(2432 and 2434, respectively) may be used for the femoral and / or humeral regions to increase detection sensitivity in these regions, thereby taking into account reduced uptake of drugs (e.g., radiopharmaceuticals) in them.

[0298] Once an initial set of hotspots is detected, a set of hotspot features is extracted for each hotspot,2440 which is used to calculate a transfer probability value for each hotspot.2450 In certain embodiments, the detected hotspots may be rendered for graphical display to the user, along with additional information such as the calculated probability value.2460 The hotspots may be filtered (pre-selected) based on the calculated probability value to be included in a first subset, either for presentation to the user and / or for use in calculating risk indicators such as BSI values. The user may review the detected hotspots, e.g., the initial set of hotspots, or the first subset containing filtered hotspots, via the graphical display and confirm or reject hotspots to include in a second final subset. This final subset may then be used to calculate risk indicator values, thereby incorporating the user's expertise into the decision-making process.

[0299] In certain embodiments, global dynamic threshold adjustment is used. This method involves reviewing the resulting BSI value and fine-tuning the global threshold scaling rate to adapt to high-burden diseases. The scaling is calculated according to the following formula:

number

[0300] Specifically, in certain embodiments, the global scaling rate method is a data-driven method that takes into account errors that may cause BSI values ​​to be underestimated 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 produces realistic bone scan images and also takes into account specific camera and examination parameters. Thus, for certain known input parameters, realistic bone scan images can be produced so that the volume of the lesion, the volume of the skeleton, and therefore the ground truth as seen from the BSI value are known. Thus, this method allows the BSI value calculated via the image analysis method described herein to be verified and compared against the BSI value of the known ground truth determined from the input parameters of the image simulation. By performing numerous simulations with varying degrees of disease burden, it was demonstrated that previous systems that did not use the global threshold scaling method described herein non-linearly underestimated BSI values ​​for higher disease burdens. The form of the nonlinear function used for the global threshold scaling rate in Equation 1 is based on error patterns observed in simulation studies and is intended to correct for nonlinear underestimations of observed and calculated BSI values.

[0301] Figure 25A is a graph of the scaling rate as a function of b. As shown in the graph, multiplying the pre-threshold by the global threshold scaling rate reduces the threshold as the disease burden, measured by the percentage of hotspot area b, increases. As a result of the reduced (adjusted) threshold, larger areas are 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 rate is used to adjust the threshold used for hotspot detection, correcting the observed underestimation of the calculated BSI value at high levels of disease.

[0302] Figure 25B illustrates an exemplary process 2500 for using a global threshold scaling technique to detect hotspots. In the 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 the identification of skeletal regions of interest. In certain embodiments, the segmentation may include the identification of the full-length femur and / or humerus, as described with respect to Figure 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 using a pre-threshold 2532. The set of potential hotspots is then used to calculate a global threshold scaling rate 2534, as described herein, for example, and then used to adjust the pre-threshold 2536. The adjusted thresholds are then used to automatically detect hotspots for inclusion in the initial set of hotspots 2538. Similar to process 2400, once an initial set of hotspots is detected, a set of hotspot features is extracted for each hotspot, and for each hotspot, this is used to calculate the transfer probability value. The detected hotspots can then be rendered for graphical display to the user. iii. Pre-selection of hotspots

[0303] As described herein, hotspots are classified to determine whether or not they should be pre-selected. In certain embodiments, the classification of hotspots is performed via a two-step process. In certain embodiments, the first step is to classify a particular hotspot using its local characteristics, but not classifying the rest of the image if, for example, the patient has no 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 rest of the image.

[0304] Clinical experience indicates that hotspot selection depends on the rest of the image. The probability of a hotspot being selected is higher when there are many other hotspots and lower when it is the only hotspot. Therefore, using a single-step process may result in an underestimation of BSI values ​​in patients with many hotspots. Using a two-step process can improve outcomes in patients with many hotspots and high BSI. Hotspot selection using global hotspot features may be performed using machine learning modules. For example, in a particular embodiment, a first machine learning module may be used to calculate the likelihood of metastasis for each hotspot, while a second machine learning module (e.g., implementing a different ANN) can receive the calculated likelihood along with the global hotspot features and determine whether the hotspot should be included in a pre-selected subset of hotspots.

[0305] Figure 26 shows an exemplary process 2600 for pre-selecting hotspots using global hotspot features. Process 2600 begins by accessing a set of bone scan images 2610, automatically segmenting the images (members) of that bone scan image set to identify skeletal regions 2620, and automatically detecting an initial set of hotspots 2630. The segmentation step 2620 and the hotspot detection step 2630 may utilize the improved segmentation method of process 2400 and / or the global threshold scaling method of process 2500. Hotspot features are extracted for each hotspot 2640, and for each hotspot, they are used to calculate a metastasis probability value 2650. Process 2600 uses the metastasis probability value in conjunction with the global hotspot features to pre-select a first subset of initial hotspots 2662. This first subset allows the hotspots to be filtered, so that a smaller, more targeted set of hotspots that have been automatically determined to be more likely cancerous lesions are displayed to the user 2664. iv. Atlas weight

[0306] In certain embodiments, the correction rates for the sacral, pelvic, and lumbar regions are adjusted so that hotspots of equal area correspond to more uniform measurements of BSI-associated lesions. In certain embodiments, no such adjustment is made, and the sacral region differs significantly from the adjacent pelvic and lumbar regions. Figure 27 shows those regions as outlined in the skeletal atlas, namely the sacral region 2706, the pelvic region 2702, and the lumbar region 2704.

[0307] To calculate the BSI value, the proportion of each selected hotspot to the total skeleton is calculated, and the BSI value is calculated as the sum of all such proportions. For each hotspot, the proportion is calculated as follows: Divide the hotspot size by the size of the corresponding skeletal region (e.g., skull, ribs, lumbar vertebrae, pelvis) obtained from the segmentation of the skeleton, and multiply by the constant for the weight ratio of the current skeletal region to the weight of the total skeleton. There is one of these constants for each skeletal region, which can be determined based on International Commission on Radiological Protection (ICRP) publication 23.

[0308] Co-occurrence is, formula,

number

[0309] In certain embodiments, this basic method works well in most skeletal regions but not in the sacral region, which is surrounded by the pelvic 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 them to the correct regions. Depending on the localization of the assigned hotspots (e.g., to the pelvis, lumbar column, or sacrum), hotspots of similar size may have large differences in their respective contributions to the calculated BSI ratio. To reduce such differences, the coefficient c in the above formula is adjusted with respect to the sacrum so that the difference from the pelvic region to the lumbar region is smoother. Specifically, the correction rate is the ratio

number

[0310] Figure 28 shows an exemplary process 2800 for adjusting for skeletal complications and utilizing the correction rates described herein to calculate risk index values ​​therefrom. Process 2800 includes the steps of accessing a bone scan image 2810, automatically segmenting the bone scan image 2820 to identify skeletal regions, automatically detecting an initial set of hotspots 2830, and extracting hotspot features 2840 and calculating a likely metastasis value for each hotspot 2850. The segmentation step 2820 and the hotspot detection step 2830 may utilize the improved segmentation method of process 2400 and / or the global threshold scaling method of process 2500. A first subset of the initial set of hotspots is automatically selected 2860, at least based on the likely metastasis values ​​calculated in step 2850. Global hotspot features, such as those described with respect to process 2600, may be used, or user input received, for example, through interaction with a GUI, may be used to select the first subset. The first subset of hotspots is then used to calculate the risk index values ​​for the subject.2870 As described herein, the calculation of the risk index may include calculating the bone invasiveness rate2872, adjusting the bone invasiveness rate using a region-dependent adjustment rate2874, and summing the adjusted bone invasiveness rates2876. [Examples]

[0311] G. Examples: BSI calculation performance This embodiment demonstrates the linearity, precision, and reproducibility of the calculated BSI values. i. Linearity and precision

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

[0313] The coefficients of variability and standard deviation of automated BSI values ​​for each of five predefined tumor loads associated with various localizations were determined for a second set of 50 simulated bone scans. The coefficient of variability for each of the five predefined phantom BSIs was either less than 30% or equal to 30%. The results are shown in Table 2 below. Table 2. Automated BSI value variability and target values ​​for each of the five predefined tumor burdens. Coefficient of the septum deviation [Table 2] iii. Reproducibility using different cameras

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

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

[0316] Figure 27 shows a Bland-Altman figure evaluating the reproducibility of automated BSI readings from repeated bone scans in 35 metastatic patients. The coefficient of repeatability (2xSD) was 0.36 and -0.36 (horizontal dotted line), the mean and median BSI difference was 0.0 (solid horizontal line), and the standard deviation was 0.18. 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 two aBSI values ​​obtained from repeated scans. Table 5. Descriptive statistics showing the reproducibility of automated BSI calculations from repeated bone scans in 35 metastatic patients. [Table 5-1] [Table 5-2] vi. Comparison with previously approved devices

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

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

[0319] Figure 34 shows an illustrative network environment 3400 for use in the methods and systems described herein. For a brief overview, a block diagram of an exemplary cloud computing environment 3400 is shown and described with reference to Figure 34. 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, computing resources may include any hardware and / or software used to process data. For example, computing resources may include hardware and / or software capable of running 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 within the cloud computing environment 3400. In some implementations, resource providers 3402 may be connected through a computer network 3408. Each resource provider 3402 may be connected to one or more computing devices 3404a, 3404b, 3404c (collectively 3404) via a computer network 3408.

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

[0321] Figure 35 shows examples of computing device 3500 and mobile computing device 3550 that may be used in the methods and systems described herein. 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 appropriate 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 shown herein, their connections and relationships, and their functions are intended to be examples only and not limiting.

[0322] The computing device 3500 includes a processor 3502, memory 3504, storage device 3506, a high-speed interface 3508 connecting to memory 3504 and multiple high-speed expansion ports 3510, and a low-speed interface 3512 connecting to low-speed expansion port 3514 and storage device 3506. Each of the processor 3502, memory 3504, storage device 3506, high-speed interface 3508, high-speed expansion ports 3510, and low-speed interface 3512 is interconnected using various buses and may be mounted on a common motherboard or in other manner as appropriate. The processor 3502 can process instructions to be executed within the computing device 3500, including instructions stored in memory 3504 or storage device 3506 for displaying GUI graphic information on external input / output devices 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 multiple types of memory. Furthermore, multiple computing devices may be connected, each providing a portion of the required operation (e.g., as a server bank, a group of blade servers, or a multiprocessor system). Therefore, when this term is used herein and it is stated that multiple functions are performed by a “processor,” this includes embodiments in which multiple functions are performed by any number of processors (one or more) of any number of computing devices. Moreover, when it is stated that a particular function is performed by a “processor,” this includes embodiments in which that function is performed by any number of processors (one or more) of any number of computing devices (e.g., in a distributed computing system).

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

[0324] The storage device 3506 can provide high-capacity storage to the computing device 3500. In some implementations, the storage device 3506 may be an array of devices including computer-readable media such as floppy disk devices, hard disk devices, optical disk devices, or tape devices, flash memory, or other similar solid-state memory devices, or devices in a storage area network or other configuration. Instructions may be stored in an information carrier. When an instruction is executed by one or more processing devices (e.g., processor 3502), it performs one or more methods, such as those described above. Instructions may also be stored in one or more storage devices, such as a computer or machine-readable media (e.g., memory 3504, storage device 3506, or memory on processor 3502).

[0325] The high-speed interface 3508 manages bandwidth-intensive operations of the computing device 3500, while the low-speed interface 3512 manages less bandwidth-intensive operations. Such function assignments are merely examples. In some implementations, the high-speed interface 3508 is coupled to memory 3504, display 3516 (e.g., through a graphics processor or accelerator), and high-speed expansion port 3510, which may accept various expansion cards (not shown). In this implementation, the low-speed interface 3512 is coupled to storage device 3506 and 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, pointing device, scanner, or networking devices such as switches or routers, for example, through a network adapter.

[0326] The computing device 3500 can 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 within 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 of a mobile device (not shown), such as a mobile computing device 3550. Each of such devices may include one or more computing devices 3500 and mobile computing devices 3550, and the entire system may consist of multiple computing devices communicating with each other.

[0327] The mobile computing device 3550 includes, among other components, an input / output device such as a processor 3552, memory 3564, and 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 is interconnected using various buses, and some of these components may be mounted on a common motherboard or in other ways as appropriate.

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

[0329] The processor 3552 can communicate with the 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, an OLED (organic light-emitting diode) display, or other suitable display technology. The display interface 3556 may include suitable circuitry for driving the display 3554 to present graphic information and other information to the user. The control interface 3558 may receive commands from the user and translate them for transmission to the processor 3552. Additionally, an external interface 3562 may 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 enable wired communication in some implementations and wireless communication in others, and multiple interfaces may be used.

[0330] Memory 3564 stores information within the mobile computing device 3550. Memory 3564 may be implemented as 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 via an expansion interface 3572, which may include, for example, a SIMM (Single Inline Memory Module) card interface. The expansion memory 3574 can provide additional storage space for the mobile computing device 3550, or it may store applications or other information for the mobile computing device 3550. Specifically, the expansion memory 3574 may contain instructions for executing or supplementing the processes described above, and may also contain secure information. 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 the secure use of the mobile computing device 3550. Furthermore, secure applications, such as those that store identification information in a hack-proof format on the SIMM card, can be provided via the SIMM card along with additional information.

[0331] The memory may include, for example, flash memory and / or NVRAM memory (non-volatile random access memory), as discussed below. In some implementations, instructions are stored in an information carrier and, when executed by one or more processing devices (e.g., processor 3552), perform one or more actions, such as those described above. Instructions may also be stored in one or more storage devices, such as one or more computers or machine-readable media (e.g., memory 3564, extended memory 3574, or memory on processor 3552). In some implementations, instructions may be received in the propagating 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 enables communication under various modes or protocols, including, among others, GSM® voice call (Global System for Mobile Communication), SMS (Short Message Service), EMS (Enhanced Messaging Service), or MMS messaging (Multimedia Messaging Service), CDMA (Code Division Multiple Access), TDMA (Time Division Multiple Access), PDC (Personal Digital Cellular), WCDMA® (Wideband Code Division Multiple Access), CDMA2000, or GPRS (General Packet Radio Service). Such communication may be performed, for example, through a transceiver 3568 using radio frequencies. Alternatively, short-range communication may be performed using Bluetooth®, Wi-Fi®, or other such transceivers (not shown). Furthermore, the GPS (Global Positioning System) receiver module 3570 may 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 by voice using the audio codec 3560, which can receive speech information from the user and convert it into usable digital information. The audio codec 3560 can also generate audible sound for the user, for example, through a speaker in the handset of the mobile computing device 3550. Such sound may include sounds from voice calls, recorded sounds (e.g., voice messages, music files, etc.), and sounds generated by applications running on the mobile computing device 3550.

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

[0335] The various implementations of the systems and technologies described herein can be realized as digital electronic circuits, integrated circuits, specially designed ASICs (application-specific integrated circuits), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include implementations as one or more computer programs executable and / or interpretable on a programmable system including at least one programmable processor, which may be specialized or general-purpose, and which are coupled to receive and transmit data and instructions from 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 code) include machine instructions for a programmable processor and can be implemented in high-level procedural and / or object-oriented programming languages, as well as / or assembly language and / or machine language. As used herein, the terms machine-readable medium and computer-readable medium refer to any computer program product, apparatus and / or device (e.g., magnetic disks, optical disks, memory, programmable logic devices (PLDs)) used to provide machine instructions and / or data to a programmable processor, including machine-readable medium that receives 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 user interaction, the systems and technologies described herein can be implemented on a computer having a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal disk) monitor), as well as a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. User interaction can also be enabled using other types of devices; for example, the feedback provided to the user may be any form of sensory feedback (e.g., visual feedback, audible feedback, or haptic feedback), and user input may be received in any form, including acoustic, speech, or haptic input.

[0338] The systems and technologies described herein may be implemented within a computing system that includes backend components (e.g., as a data server), middleware components (e.g., an application server), or frontend components (e.g., a client computer having a graphical user interface or web browser that allows a user to interact with an implementation of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system may be interconnected by digital data communications (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the internet.

[0339] Computer systems may include clients and servers. Clients and servers are generally remote from each other and typically interact through a communication network. The client-server relationship arises from computer programs running on each computer that have a client-server relationship with each other. In some implementations, the modules and / or services described herein may be separate, combined, or integrated into a single or combined module and / or service. The modules and / or services depicted in the diagrams are not intended to limit the systems described herein to the software architecture shown therein.

[0340] While the present invention has been illustrated and described in detail with respect to certain preferred embodiments, those skilled in the art will understand that various forms and details can be modified 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) A method for lesion marking and quantitative analysis of nuclear medicine images in human subjects, (a) A step of accessing a set of bone scan images of a human subject using the processor of a computing device, wherein the set of bone scan images is obtained after administration of a drug to the human subject; (b) The processor automatically segments each image in the bone scan image set to identify one or more skeletal regions of interest, each corresponding to a specific anatomical region of the human skeleton, thereby obtaining a set of annotated images, wherein the one or more skeletal regions of interest are (i) A femoral region corresponding to a portion of the femur of the human subject, wherein the portion of the femur is a femoral region encompassing at least three-quarters of the femur along the length of the femur, and (ii) A humeral region corresponding to a portion of the humerus of the human subject, wherein the portion of the humerus is a humeral region encompassing at least three-quarters of the humerus along its length, A step including at least one of the following, (c) A step of the processor automatically detecting an initial set of one or more hotspots, each hotspot corresponding to a high-intensity area in the set of annotated images, and the automatic detection includes identifying the one or more hotspots using the intensity 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 femoral region and / or the humeral region to provide enhanced hotspot detection sensitivity in the femoral region and / or the humeral region to compensate for reduced drug uptake in the region, (d) For each hotspot in the initial set of hotspots, the processor extracts a set of hotspot features associated with the hotspot; (e) For each hotspot in the initial set of hotspots, the processor calculates a transition probability value corresponding to the likelihood that the hotspot will represent a transition, based on the set of hotspot features associated with the hotspot; (f) A method comprising the step of causing the processor to render a graphic representation of at least a portion of the initial set of hotspots for display in a graphical user interface (GUI). (Item 2) Step (b) is, A step of comparing each member of the bone scan image set with a corresponding atlas image from the atlas image set, wherein each atlas image includes one or more identifications of the one or more skeletal regions of interest, and the skeletal regions of interest include the femoral region and / or the humeral region, The method according to item 1, comprising the step of aligning the corresponding atlas image with the image in the bone scan image set, such that the identification of one or more skeletal regions of interest in the atlas image is applied to the image in the bone scan image set. (Item 3) The method of item 2, wherein each atlas image includes the identification of (i) the femoral region including at least a portion of the knee region of the human subject, and / or (ii) the humeral region including at least a portion of the elbow region of the human subject, and for each image in the set of bone scan images, the alignment of the corresponding atlas image to the bone scan image includes using the identified knee region and / or the identified elbow region as landmarks in the image. (Item 4) The method according to any one of the above items, wherein the location of at least one detected hotspot from the initial set of hotspots corresponds to a physical location in or on the femur that is located more than three-quarters of the distance along the femur from the end of the femur facing the buttocks of the human subject to the end of the femur facing the knee of the human subject. (Item 5) The method according to any one of the above items, wherein the location of at least one detected hotspot from the initial set of hotspots corresponds to a physical location in or on the humerus, which is located more than three-quarters of the distance along the humerus from the shoulder end of the humerus facing the elbow end of the human subject. (Item 6) Step (c) is The processor performs the steps of identifying healthy tissue regions within the bone scan image set in which it is determined that no hotspots are present, The processor performs the steps of calculating a normalization coefficient such that the product of the normalization coefficient and the average intensity of the identified healthy tissue area is a predetermined intensity level, The method according to any one of the items, comprising the step of normalizing the images of the bone scan image set by the normalization coefficient using the processor. (Item 7) (g) The method of any one of the items, further comprising the step of the processor calculating one or more risk index values ​​for the human subject based at least in part on the calculated proportion of the human subject's skeleton occupied by the initial set of hotspots. (Item 8) (h) The step of the processor selecting a first subset of the initial set of hotspots based at least in part on the expected transfer value, (i) The step of causing the processor to render the graphic representation of the first subset for display in a graphical user interface (GUI), The method described in any one of the above items, including: (Item 9) (j) The method of item 8, further comprising the step of the processor calculating one or more risk index values ​​for the human subject based at least in part on the calculated proportion of the human subject's skeleton occupied by the first subset of hotspots. (Item 10) (k) The processor receives a user selection of a second subset of the initial set of hotspots via the GUI, (l) The processor calculates one or more risk index values ​​for the human subject based at least in part on the calculated proportion of the human subject's skeleton occupied by the second subset of hotspots. The method described in any one of the above items, including: (Item 11) The method according to any one of items 7 to 10, wherein at least one of the risk indicator values ​​indicates the risk that the human subject has and / or develops metastatic cancer. (Item 12) The method according to item 11, wherein the metastatic cancer is metastatic prostate cancer. (Item 13) The method according to any one of items 7 to 10, wherein at least one of the risk indicator values ​​indicates that the human subject has metastatic cancer of a specific condition. (Item 14) The method according to any one of the above items, wherein the processor is a processor for a cloud-based system. (Item 15) The method according to any one of the above items, wherein the GUI is part of a general photo archiving and communication system (PACS). (Item 16) The aforementioned drug is technetium-99m methylenediphosphonate ( 99m The method according to any one of the above items, including Tc-MDP. (Item 17) A method for lesion marking and quantitative analysis of nuclear medicine images in human subjects, (a) A step of accessing a set of bone scan images of a human subject using the processor of a computing device, wherein the set of bone scan images is obtained after administration of a drug to the human subject; (b) The processor automatically segments each image in the bone scan image set to identify one or more skeletal regions of interest, each corresponding to a specific anatomical region of the human skeleton, thereby obtaining a set of annotated images; (c) The processor automatically detects an initial set of one or more hotspots, wherein each hotspot corresponds to a high-intensity area in the set of annotated images, and the automatic detection is (i) the intensity of pixels in the annotated set of images, and (ii) a set of prior thresholds, to detect a set of potential hotspots. Calculate the global threshold scaling rate using the set of potential hotspots. Adjust the multiple pre-thresholds using the global threshold scaling rate, thereby obtaining multiple adjusted thresholds, and The steps include: (i) identifying an initial set of hotspots using the intensity 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, the processor extracts a set of hotspot features associated with the hotspot; (e) For each hotspot in the initial set of hotspots, the processor calculates a transition probability value corresponding to the likelihood that the hotspot will represent a transition, based on the set of hotspot features associated with the hotspot; (f) A method comprising the step of causing the processor to render a graphic representation of at least a portion of the initial set of hotspots for display in a graphical user interface (GUI). (Item 18) The method of item 17, wherein the global threshold scaling rate is a function of a measure of the disease burden of the human subject, and the adjustment of the plurality of prior thresholds performed in step (c) is performed by reducing the adjusted thresholds as the disease burden increases in order to compensate for underestimation of hotspot areas that occur with increasing disease burden. (Item 19) The method according to item 17 or 18, wherein the global threshold scaling rate is a function of the proportion of the identified skeleton region occupied by the set of potential hotspots. (Item 20) The method according to 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) A method for lesion marking and quantitative analysis of nuclear medicine images in human subjects, (a) A step of accessing a set of bone scan images of a human subject using the processor of a computing device, wherein the set of bone scan images is obtained after administration of a drug to the human subject; (b) The processor automatically segments each image in the bone scan image set to identify one or more skeletal regions of interest, each corresponding to a specific anatomical region of the human skeleton, thereby obtaining a set of annotated images; (c) The step of the processor automatically detecting one or more initial sets of hotspots, wherein each hotspot corresponds to a high-intensity area in the set of annotated images, (d) For each hotspot in the initial set of hotspots, the processor extracts a set of hotspot features associated with the hotspot; (e) For each hotspot in the initial set of hotspots, the processor calculates a transition probability value corresponding to the likelihood that the hotspot will represent a transition, based on the set of hotspot features associated with the hotspot; (f) The processor selects a first subset of the initial set of hotspots, wherein the selection of specific hotspots to be included in the first subset is (i) The expected transfer value calculated for the specific hotspot, and (ii) One or more global hotspot features, each determined using multiple hotspots in an initial set of hotspots, Steps based at least partially on, (g) A method comprising the step of causing the processor to render a graphic representation of at least a portion of the first subset of hotspots for display in a graphical user interface (GUI). (Item 22) The method according to item 21, wherein the one or more global hotspot features include the total number of hotspots in the initial hotspot set. (Item 23) The method according to item 22, wherein step (f) includes adjusting the criteria for selecting which hotspots to include in the first subset based on the total number of hotspots in the initial set of hotspots. (Item 24) The method according to any one of items 21 to 23, wherein step (f) includes selecting the first subset using a machine learning module. (Item 25) A method for lesion marking and quantitative analysis of nuclear medicine images in human subjects, (a) The computing device's processor accesses the set of bone scan images of the human subject, (b) The processor automatically segments each image in the bone scan image set to identify one or more skeletal regions of interest, each corresponding to a specific anatomical region of the human skeleton, thereby obtaining a set of annotated images; (c) The step of the processor automatically detecting one or more initial sets of hotspots, wherein each hotspot corresponds to a high-intensity area in the set of annotated images, (d) For each hotspot in the initial set of hotspots, the processor extracts a set of hotspot features associated with the hotspot; (e) For each hotspot in the initial set of hotspots, the processor calculates an expected value corresponding to the likelihood that the hotspot will represent a transition, based on the set of hotspot features associated with the hotspot; (f) The step of selecting a first subset of the hotspots from the initial set of hotspots based at least in part on the estimated value calculated by the processor for each hotspot in the initial set of hotspots, (g) The processor calculates one or more risk indicator values ​​using at least a portion of the first subset of hotspots, the calculation of For each specific hotspot of the portion of the first subset, the steps of: (ii) calculating a bone infiltration rate based on the ratio of (i) the size (e.g., area) of the specific hotspot to the size of a specific skeletal region to which the specific hotspot is assigned based on its location within the set of annotated images, thereby determining one or more bone infiltration rates; The steps of adjusting the bone infiltration rate using one or more region-dependent correction rates to obtain one or more adjusted bone infiltration rates, and A method comprising the step of determining one or more risk index values ​​by summing the adjusted bone infiltration rates. (Item 26) The method according to item 25, wherein for each specific hotspot, the calculated bone infiltration rate is estimated as the ratio of the total skeletal mass to the physical volume associated with the specific hotspot. (Item 27) The step of calculating the bone infiltration rate is The processor calculates the ratio of the area of ​​the specific hotspot to the area of ​​the corresponding skeletal region of interest, thereby calculating the area ratio of the specific hotspot. The method of item 26, comprising scaling the area percentage by a density coefficient associated with the skeletal region of interest to which the particular hotspot is assigned, thereby calculating the bone infiltration rate of the particular hotspot. (Item 28) The method according to any one of items 25 to 27, wherein at least a portion of the hotspots of the first subset is assigned to a skeletal region of interest, which is a member selected from the group consisting of the pelvic region, the lumbar region, and the sacral region. (Item 29) The method according to any one of items 25 to 28, wherein the one or more region-dependent correction rates include a sacral region correction rate associated with the sacral region, which is used to adjust the rate of bone infiltration of a hotspot identified as located within the sacral region, and the sacral region correction rate 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 specific 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 ratio, which is used to adjust the bone infiltration rate calculated for hotspots detected in the annotated ventral bone scan images of the annotated image set. The method according to any one of items 25 to 29, wherein the second member of the pair is a dorsal image correction rate, which is used to adjust the bone infiltration rate calculated for hotspots detected in the annotated dorsal bone scan images of the annotated set of images. (Item 31) A system for lesion marking and quantitative analysis of nuclear medicine images in human subjects, Processor and A memory having instructions, wherein when an instruction is executed by the processor, the processor has (a) Accessing the set of bone scan images of the human subject, wherein the set of bone scan images was obtained after the administration of the drug to 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 specific anatomical region of the human skeleton, thereby obtaining a set of annotated images, wherein the one or more skeletal regions of interest are (i) A femoral region corresponding to a portion of the femur of the human subject, wherein the portion of the femur is a femoral region encompassing at least three-quarters of the femur along the length of the femur, and (ii) A humeral region corresponding to a portion of the humerus of the human subject, wherein the portion of the humerus is a humeral region encompassing at least three-quarters of the humerus along its length, Including at least one of the following, (c) Automatically detecting an initial set of one or more hotspots, each hotspot corresponding to a high-intensity area in the set of annotated images, and the automatic detection includes identifying the one or more hotspots using the intensity 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 femoral region and / or the humeral region to provide enhanced hotspot detection sensitivity in the femoral region and / or the humeral region to compensate for reduced drug uptake in the region, (d) For each hotspot in the initial set of hotspots, extract a set of hotspot features associated with that hotspot, (e) For each hotspot in the initial set of hotspots, calculate a transition probability value corresponding to the likelihood that the hotspot will represent a transition, based on the set of hotspot features associated with the hotspot; (f) A system that causes a graphical representation of at least a portion of the initial set of hotspots to be rendered for display within a graphical user interface (GUI). (Item 32) A system for lesion marking and quantitative analysis of nuclear medicine images in human subjects, Processor and A memory having instructions, wherein when an instruction is executed by the processor, the processor has (a) Accessing the set of bone scan images of the human subject using the processor of a computing device, wherein the set of bone scan images was obtained after the administration of the drug to 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 specific anatomical region of the human skeleton, thereby obtaining a set of annotated images, (c) Automatically detecting an initial set of one or more hotspots, each hotspot corresponding to a high-intensity area in the set of annotated images, and the automatic detection is (i) the intensity of pixels in the annotated set of images, and (ii) a set of prior thresholds, to detect a set of potential hotspots. Calculate the global threshold scaling rate using the set of potential hotspots. Adjust the multiple pre-thresholds using the global threshold scaling rate, thereby obtaining multiple adjusted thresholds, and (i) the intensity of pixels in the annotated image set, and (ii) the plurality of adjusted thresholds, to identify an initial set of hotspots, (d) For each hotspot in the initial set of hotspots, extract a set of hotspot features associated with that hotspot, (e) For each hotspot in the initial set of hotspots, calculate a transition probability value corresponding to the likelihood that the hotspot will represent a transition, based on the set of hotspot features associated with the hotspot; (f) A system that causes a graphical representation of at least a portion of the initial set of hotspots to be rendered for display within a graphical user interface (GUI). (Item 33) A system for lesion marking and quantitative analysis of nuclear medicine images in human subjects, Processor and A memory having instructions, wherein when an instruction is executed by the processor, the processor has (a) Accessing the set of bone scan images of the human subject, wherein the set of bone scan images was obtained after the administration of the drug to 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 specific anatomical region of the human skeleton, thereby obtaining a set of annotated images, (c) Automatically detecting an initial set of one or more hotspots, each hotspot corresponding to a high-intensity area in the set of annotated images, (d) For each hotspot in the initial set of hotspots, extract a set of hotspot features associated with that hotspot, (e) For each hotspot in the initial set of hotspots, calculate a transition probability value corresponding to the likelihood that the hotspot will represent 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 the selection of specific hotspots to be included in the first subset is (i) The expected transfer value calculated for the specific hotspot, and (ii) One or more global hotspot features, each determined using multiple hotspots in an initial set of hotspots, Based at least partially on, (g) A system that causes a graphical representation of at least a portion of the first subset of hotspots to be rendered for display within a graphical user interface (GUI). (Item 34) A system for lesion marking and quantitative analysis of nuclear medicine images in human subjects, Processor and A memory having instructions, wherein when an instruction is executed by the processor, the processor has (a) Accessing the set of bone scan images 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 specific anatomical region of the human skeleton, thereby obtaining a set of annotated images, (c) Automatically detecting an initial set of one or more hotspots, each hotspot corresponding to a high-intensity area in the set of annotated images, (d) For each hotspot in the initial set of hotspots, extract a set of hotspot features associated with that hotspot, (e) For each hotspot in the initial set of hotspots, calculate an expected value corresponding to the likelihood that the hotspot will represent a transition, based on the set of hotspot features associated with the hotspot; (f) Selecting a first subset of the hotspots from the initial set of hotspots, based at least in part on the estimated value calculated for each hotspot in the initial set of hotspots, (g) to calculate one or more risk indicator values ​​using at least a portion of the first subset of hotspots, and to perform the calculation, For each specific hotspot of the portion of the first subset, calculate the bone infiltration rate based on the ratio of (i) the size of the specific hotspot to (ii) the size of a specific skeletal region to which the specific hotspot is assigned based on its location within the set of annotated images, thereby determining one or more bone infiltration rates. Adjusting the bone infiltration rate using one or more region-dependent correction rates to obtain one or more adjusted bone infiltration rates, and A system comprising determining one or more risk index values ​​by summing the adjusted bone infiltration rates. (Item 35) A computer-aided image analysis device comprising a system according to any one of items 31 to 34. (Item 36) The device described in item 35, which is programmed for use by trained healthcare professionals and / or researchers. (Item 37) The device described in item 36, which is programmed for use in the analysis of bone scan images for the evaluation and / or detection of metastatic cancer. (Item 38) The device described in item 36 or 37, which is programmed for use in analyzing bone scan images for evaluating and / or detecting prostate cancer. (Item 39) The device described in any one of paragraphs 35 to 38, which has a label indicating that the device is intended to be used by trained healthcare professionals and / or researchers. (Item 40) The device as described in item 39, wherein the label further specifies that the device is intended to be used for the analysis of bone scan images for the evaluation and / or detection of metastatic cancer. (Item 41) The device as described in item 39 or 40, further specifying that the label indicates the device is intended for use in the analysis of bone scan images for the evaluation and / or detection of prostate cancer.

Claims

1. A method for lesion marking and quantitative analysis of nuclear medicine images in human subjects, (a) A step of accessing a set of bone scan images of a human subject using the processor of a computing device, wherein the set of bone scan images is obtained after administration of a drug to the human subject; (b) The processor automatically segments each image in the bone scan image set to identify one or more skeletal regions of interest, each corresponding to a specific anatomical region of the human skeleton, thereby obtaining a set of annotated images; (c) The step of the processor automatically detecting an initial set of one or more hotspots, each hotspot corresponding to a high-intensity area in the set of annotated images, and the step of automatic detection is (i) the intensity of pixels in the annotated set of images, and (ii) a set of prior thresholds, to detect a set of potential hotspots. Using the aforementioned set of potential hotspots, calculate the global threshold scaling rate. Adjust the multiple pre-thresholds using the global threshold scaling rate, thereby obtaining multiple adjusted thresholds, and (i) the intensity of pixels in the annotated set of images, and (ii) the multiple adjusted thresholds, to identify the initial set of hotspots. Steps including, (d) For each hotspot in the initial set of hotspots, the processor extracts a set of hotspot features associated with the hotspot; (e) For each hotspot in the initial set of hotspots, the processor calculates a transition probability value corresponding to the likelihood that the hotspot will represent a transition, based on the set of hotspot features associated with the hotspot; (f) The step of causing the processor to render a graphic representation of at least a portion of the initial set of hotspots for display in a graphical user interface (GUI), Methods that include...

2. The method according to claim 1, wherein the global threshold scaling rate is a function of a measure of the disease burden in the human subject, and adjusting the plurality of prior thresholds performed in step (c) includes reducing the adjusted thresholds as the disease burden increases in order to compensate for underestimation of hotspot areas that occur with increasing disease burden.

3. The method according to claim 1 or 2, wherein the global threshold scaling rate is a function of the proportion of the identified skeleton region occupied by the set of potential hotspots.

4. The 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 system for lesion marking and quantitative analysis of nuclear medicine images in human subjects, Processor and A memory having instructions, wherein when an instruction is executed by the processor, the processor has (a) Accessing the bone scan image set of the human subject by the processor of a computing device, wherein the bone scan image set was obtained after the administration of the drug to 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 specific anatomical region of the human skeleton, thereby obtaining a set of annotated images, (c) Automatically detecting an initial set of one or more hotspots, each hotspot corresponding to a high-intensity area in the set of annotated images, and the automatic detection is (i) the intensity of pixels in the annotated set of images, and (ii) a set of prior thresholds, to detect a set of potential hotspots. Using the aforementioned set of potential hotspots, calculate the global threshold scaling rate. Adjust the multiple pre-thresholds using the global threshold scaling rate, thereby obtaining multiple adjusted thresholds, and (i) the intensity of pixels in the annotated set of images, and (ii) the multiple adjusted thresholds, to identify the initial set of hotspots. This includes, (d) For each hotspot in the initial set of hotspots, extract a set of hotspot features associated with that hotspot, (e) For each hotspot in the initial set of hotspots, calculate a transition probability value corresponding to the likelihood that the hotspot will represent a transition, based on the set of hotspot features associated with the hotspot. (f) A system that causes a graphical representation of at least a portion of the initial set of hotspots to be rendered for display within a graphical user interface (GUI).

6. A computer-aided image analysis device comprising the system described in claim 5.

7. The device according to claim 6, wherein the device is programmed for use by trained healthcare professionals and / or researchers.

8. The device according to claim 7, wherein the device is programmed for use in analyzing bone scan images for evaluating and / or detecting metastatic cancer.

9. The device according to claim 7 or 8, wherein the device is programmed to be used for analyzing bone scan images for evaluating and / or detecting prostate cancer.

10. The device according to any one of claims 6 to 9, comprising a label indicating that the device is intended to be used by trained healthcare professionals and / or researchers.

11. The device according to claim 10, wherein the label further specifies that the device is intended to be used for analyzing bone scan images for the evaluation and / or detection of metastatic cancer.

12. The device according to either claim 10 or 11, wherein the label further specifies that the device is intended to be used for analyzing bone scan images for the evaluation and / or detection of prostate cancer.

Images