Systems and methods for automated and interactive analysis of bone scan images to detect metastases

Abstract

Presented herein are systems and methods that provide improved computer-aided display and analysis of nuclear medicine images. In particular, in certain embodiments, the systems and methods described herein provide improvements to a number of image processing steps for automating analysis of bone scan images to assess a patient's cancer status. For example, improved methods for image segmentation, hot spot detection, automated classification of hot spots as indicative of metastasis, and calculation of risk indices such as bone scan index (BSI) values are provided.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to and the benefit of U.S. Provisional Application No. 62 / 837,955, filed April 24, 2019, the entire contents of which are hereby incorporated by reference. TECHNICAL FIELD

[0003] The present invention relates generally 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

[0004] Nuclear medicine imaging involves the use of radiolabeled compounds, known as radiopharmaceuticals. Radiopharmaceuticals are administered to a patient and accumulate in various regions in the body, the administration and accumulation of which are according to and thus indicative of the biophysical and / or biochemical properties of the tissue in the various regions, e.g., properties affected by the presence of a disease, such as cancer, and / or disease state. For example, certain radiopharmaceuticals will accumulate in regions of abnormal bone formation associated with malignant bone lesions to indicate metastasis after being administered to a patient. Other radiopharmaceuticals can bind to specific receptors, enzymes, and proteins in the body that change during disease evolution. These molecules circulate in the blood after being administered to a patient until they find their intended target. The radiopharmaceutical that has bound stays at the disease site, while the rest of the dose is cleared from the body.

[0005] Nuclear medicine imaging techniques capture images by detecting radiation emitted from the radioactive portion of the radiopharmaceutical. The accumulated radiopharmaceutical acts as a marker so that images can be obtained using common nuclear medicine modalities that depict the location and concentration of disease. Examples of nuclear medicine imaging modalities include bone scan imaging (also known as scintigraphy), single photon emission computed tomography (SPECT), and positron emission tomography (PET). Bone scan, SPECT, and PET imaging systems are found in most hospitals worldwide. The choice of a particular imaging modality depends on and / or determines the particular radiopharmaceutical used. For example, technetium-99m (Tc) labeled compounds are compatible with bone scan and SPECT imaging, while PET imaging typically uses fluoridated compounds labeled with 18F. Compounds 99m Tc methylene diphosphonate (Tc MDP) is a common radiopharmaceutical used for bone scan imaging to detect metastatic cancer. Radiolabeled prostate specific membrane antigen (PSMA) targeting compounds (e.g. 99m Tc 1404 and PyL 99m Tc MDP) is a common radiopharmaceutical used for bone scan imaging to detect metastatic cancer. Radiolabeled prostate specific membrane antigen (PSMA) targeting compounds (e.g. 99m Tc 1404 and PyL TM(Also known as [18F]DCFPyL) can be used in conjunction with SPECT imaging and PET imaging, respectively, and enable high-specificity prostate cancer detection.

[0006] Thus, nuclear medicine imaging is a valuable technique that provides physicians with information that can be used to determine whether a patient has a disease and the extent of the disease in the patient. Physicians can use this information to advise a course of treatment for the patient and track disease progression.

[0007] For example, an oncologist can use nuclear medicine images obtained from a study of a patient as input to assess whether the patient has a particular disease (e.g., prostate cancer), what stage the disease has progressed to, what course of treatment, if any, would be recommended, whether surgical intervention is indicated, and possibly make a prognosis. The oncologist can use a radiologist report in making this assessment. A radiologist report is a technical assessment of the nuclear medicine images prepared by a radiologist for the physician requesting the imaging study and includes, for example, the type of study performed, the medical history, comparisons between images, the technique used to perform the study, the radiologist’s observations and findings, and the radiologist’s overall opinion and recommendations that can be made based on the results of the imaging study. The signed radiologist report is sent to the physician, the study is scheduled for review by the physician, and then the physician discusses the results and treatment recommendations with the patient.

[0008] Thus, the process involves a radiologist performing an imaging study on a patient, analyzing the obtained images, creating a radiologist report, forwarding the report to the requesting physician, the physician making an assessment and treatment recommendations, and the physician communicating the results, recommendations, and risks to the patient. The process can also involve repeating the imaging study due to an inconclusive result, or scheduling further testing based on the initial results. If the imaging study indicates that the patient has a particular disease or condition (e.g., cancer), the physician discusses various treatment options, including surgery and the risks of forgoing surgery, or taking a watchful waiting or active surveillance approach instead of surgery.

[0009] Thus, the process of reviewing and analyzing multiple patient images over time plays a critical role in the diagnosis and treatment of cancer. Therefore, there is a significant need for improved tools that facilitate and improve the accuracy of image review and analysis for cancer diagnosis and treatment. Improving the toolset utilized by physicians, radiologists, and other health care professionals in this way would provide significant improvements in the standard of care and patient experience. SUMMARY

[0010] Presented herein are systems and methods that provide improved computer-aided display and analysis of nuclear medicine images. In particular, in certain embodiments, the systems and methods described herein provide improvements to several image processing steps for automating the analysis of bone scan images to assess a patient's cancer status.

[0011] For example, improved methods for image segmentation, hot spot detection, automated classification of hot spots as indicative of metastasis, and calculation of risk indices such as bone scan index (BSI) values are provided. With the improved image processing techniques, the systems and methods described herein can be used to accurately and reliably detect and quantify lesions based on images to assess 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.

[0012] Bone scan images are widely used to diagnose and assess metastatic cancers. A radiopharmaceutical is injected into a patient, which emits nuclear radiation that can be detected to image the spatial distribution of the radiopharmaceutical in the patient's body. The radiopharmaceutical can be selected to selectively accumulate in various types of tissue associated with cancerous lesions (e.g., abnormal bone formation regions).

[0013] While this approach enables lesions to be visualized as bright spots in bone scan images, it is by no means trivial to accurately identify image regions that are indicative of true metastatic lesions. The radiopharmaceutical can also accumulate in non-cancerous anatomical regions (e.g., in the patient's bladder), and the physician and technician must carefully distinguish hot spots indicative of lesions from these regions as well as noise and artifacts. This work is time-consuming, error-prone, and exhibits significant variability between operators.

[0014] Computer-automated lesion detection and analysis provides a way to address these challenges and can greatly improve the accuracy and repeatability of lesion detection and cancer diagnosis. However, tools for automating lesion detection and analysis rely on a complex combination of imaging processing and artificial intelligence steps. For example, image segmentation to identify bone regions can be used to focus analysis on bone regions. Screening and thresholding steps can be used to automatically detect hot spots, and machine learning methods (e.g., artificial neural networks (ANN)) can be used to quantitatively assess the likelihood that a detected hot spot is indicative of metastasis based on features of the hot spot (e.g., size, shape, and intensity). Finally, in certain embodiments, a total risk index for a patient is calculated using a set of detected hot spots indicative of metastasis, which represents the total likelihood that the patient has and / or will develop metastasis or a particular cancer status. One such risk index is the bone scan index (BSI), which provides an estimated mass fraction of the patient's skeleton that is occupied by metastasis.

[0015] The accuracy of either step can have a significant impact on downstream steps and the overall lesion detection and analysis process. The systems and methods described herein provide several specific improvements to various steps in the automated lesion detection and analysis workflow, improving result accuracy across various patient types and cancer stages.

[0016] First, in certain embodiments, the improved image analysis techniques described herein include an improved bone segmentation method that identifies the entire (e.g., more than three-quarters of the length) humeral region and / or femoral region in bone scan images. Prior methods only identified a limited fraction of the femur and humerus. Here, segmenting a larger portion of the bone allows for the identification of lesions located further up the arms and legs that would have otherwise escaped detection. Furthermore, while reduced radiopharmaceutical uptake in the arms and legs makes it difficult to identify lesions in the arms and legs, the methods described herein overcome this problem using a region-dependent thresholding technique that improves the sensitivity of detection in the femoral region and humeral region.

[0017] Second, the present invention also provides a global thresholding technique that improves hotspot detection accuracy, particularly in high disease burden (e.g., when a patient has multiple lesions). This method detects a preliminary set of potential hotspots, and then adjusts the threshold used for hotspot detection based on a scaling factor computed from this preliminary set. The improvement in hotspot detection provides advantages for downstream computations, improving the linearity of the computed BSI values for patients with high metastasis levels.

[0018] Third, in certain embodiments, the systems and methods described herein improve the accuracy of the automated decision made about whether a hotspot represents a metastasis. Specifically, in certain embodiments, the methods described herein utilize clinical experience that indicates that the selection of hotspots as potential metastases depends not only on the image features of the hotspot itself, but also on information from the entire image. Thus, the methods described herein can also use global features (e.g., the total number of hotspots) as input (e.g., as input to an ANN) in the automated decision-making step for lesion identification.

[0019] Fourth, in certain embodiments, the methods described herein also provide an improvement in the method for computing risk index values based on bone involvement by employing a correction factor that takes into account the potential error in the accuracy of automatically locating hotspots to a particular bone region. This is particularly important for hotspots located in or near the sacral region, a complex three-dimensional structure that can be difficult to identify in a two-dimensional bone scan image. This method improves the accuracy of BSI computation and limits the sensitivity to hotspot localization error.

[0020] Therefore, the systems and methods described herein include several improved image analysis techniques for lesion identification and quantification. These methods improve the accuracy and robustness of bone scan image analysis. As described herein, these methods can be used as part of a cloud-based system to facilitate the review and reporting of patient data and allow for improvements in disease detection, treatment, and monitoring.

[0021] In one aspect, the present invention relates to a method for lesion labeling and quantification analysis (e.g., user-assisted / user-reviewed automated or semi-automated lesion labeling and quantification analysis) of nuclear medicine images (e.g., a set of bone scan images) of a subject, the method comprising: (a) accessing (e.g., and / or receiving) the set of bone scan images of the subject (e.g., a set of one, two, or more images) via a processor of a computing device, the set of bone scan images being obtained after administration of a drug (e.g., a radiopharmaceutical) to the subject (e.g., the set of bone scan images comprising anterior and posterior bone scan images) (e.g., wherein the set of bone scan images contains...). Each image comprises multiple pixels, each pixel having a value corresponding to intensity); (b) each image in the bone scan image set is automatically segmented by the processor to identify one or more bone regions of interest, each of the one or more bone regions of interest corresponding to a specific anatomical region of the subject's skeleton (e.g., a specific bone and / or a set of one or more bones (e.g., cervical vertebrae, clavicle, ribs, lumbar vertebrae, pelvis, sacrum, scapula, skull, thoracic vertebrae, sternum, femur, humerus)), thereby obtaining an annotated image set, wherein the one or more bone regions of interest include at least one of (i) and (ii): (i) a femoral region corresponding to a portion of the subject's femur, the femur (i) a portion of the femur that encompasses at least three-quarters of the femur along its length [(e.g., greater than about three-quarters (e.g., approximately all)]; and (ii) a humeral region corresponding to a portion of the subject's humerus, the humeral portion encompassing at least three-quarters of the humerus along its length [(e.g., greater than about three-quarters (e.g., approximately all)]; (c) an initial set of one or more hotspots automatically detected by the processor, each hotspot corresponding to a high-intensity region in the annotated image set, the automatic detection comprising: identifying the one or more hotspots using the intensity of pixels in the annotated image set and using one or more region-dependent thresholds (e.g., where each...). (d) For each hotspot in the initial hotspot set, the processor extracts a set of hotspot features associated with the hotspot (e.g., a set of one or more of the identified bone regions of interest), and wherein the one or more region dependency thresholds include one or more values ​​associated with the femoral region and / or the humeral region (e.g., a reduced intensity threshold for the femoral region and / or a reduced intensity threshold for the humeral region), the one or more values ​​providing enhanced hotspot detection sensitivity in the femoral region and / or the humeral region to compensate for reduced drug uptake in the femoral region and / or the humeral region;(e) For each hotspot in the initial hotspot set, the processor calculates a transfer probability value corresponding to the likelihood of the hotspot representing a transfer based on the set of hotspot features associated with that hotspot [e.g., using one or more machine learning modules (e.g., pre-trained machine learning modules; e.g., artificial neural networks (ANNs)), which receive at least a portion of the hotspot features as input for a particular hotspot and output the transfer probability value of the hotspot]; and (f) the processor causes a graphical representation of at least a portion of the initial hotspot set to be rendered [e.g., visual indications (e.g., points, boundaries) of hotspots overlaid on one or more members of the bone scan image set and / or annotated image set; e.g., a table listing the identified hotspots and additional information for each hotspot (e.g., location; e.g., transfer probability value)] for display within a graphical user interface (GUI) (e.g., a cloud-based GUI).

[0022] In some embodiments, step (b) includes: comparing each member of the bone scan image set with a corresponding atlas image in the atlas image set, each atlas image including one or more identifications (e.g., graphic identifications superimposed on the atlas image) of the one or more bone regions of interest, the bone regions of interest including the femur region and / or the humerus region; and for each image in the bone scan image set, registering the corresponding atlas image with the image in the bone scan image set such that the identifications of the one or more bone regions of interest in the atlas image are applied (e.g., superimposed on) the image in the bone scan image set.

[0023] In some embodiments, each atlas image includes (i) the identification of the femoral region comprising at least a portion of the subject's knee region and / or (ii) the identification of the humeral region comprising at least a portion of the subject's elbow region, and wherein, for each image in the bone scan image set, the registration of the corresponding atlas image to the bone scan image includes using the knee region and / or the identified elbow region in the image as landmarks [e.g., by identifying the knee region in the bone scan image and matching it with the knee region identified in the corresponding atlas image, and then adjusting the atlas image (e.g., calculating coordinate transformation) to register the corresponding atlas image to the bone scan image].

[0024] In some embodiments, the location of at least one detected hot spot in the initial hot spot set corresponds to a body location in or on the femur that is more than three-quarters of the distance along the femur from the end of the femur oriented toward the subject's hip to the end of the femur oriented toward the subject's knee.

[0025] In some embodiments, the location of at least one detected hot spot in the initial set of hot spots corresponds to a body location in or on the humerus, the body location being located at more than three-quarters of the distance along the humerus from one end of the humerus oriented toward the subject's shoulder to the other end of the humerus oriented toward the subject's elbow.

[0026] In some embodiments, step (c) includes (e.g., iteratively): identifying healthy tissue regions in the images of the bone scan image set by the processor that are determined not to contain any hotspots (e.g., relatively high-intensity local areas); calculating a normalization factor by the processor such that the product of the normalization factor and the average intensity of the identified healthy tissue regions is a predefined intensity level; and normalizing the images in the bone scan image set by the normalization factor.

[0027] In some embodiments, the method further includes: (g) calculating one or more risk index values ​​for the subject based at least in part on a calculated score (e.g., area score) of the subject’s bones being occupied by a second subset of hotspots [e.g., where the calculated score is the ratio of the total area of ​​the initial set of hotspots to the total area of ​​all identified skeletal regions].

[0028] In some embodiments, the method includes: (h) selecting a first subset of the initial hotspot set (e.g., up to the entire subset) by the processor based at least in part on the transfer probability value [e.g., determining whether a particular hotspot is included in the subset based on a transfer probability value calculated for a particular hotspot in the initial hotspot set exceeding a threshold]; and (i) causing the processor to render a graphical representation of the first subset [e.g., visual indications (e.g., points, boundaries) of hotspots overlaid on one or more members of the bone scan image set and / or annotated image set; e.g., a table listing identified hotspots and additional information for each hotspot (e.g., location; e.g., transfer probability value)] for display within a graphical user interface (GUI) (e.g., a cloud-based GUI).

[0029] In some embodiments, the method further includes: (j) calculating one or more risk index values ​​for the subject based at least in part on a calculated score (e.g., area score) of the subject’s bones being occupied by a second subset of hotspots [e.g., where the calculated score is the total area of ​​the initial set of hotspots divided by the total area of ​​all identified skeletal regions].

[0030] In some embodiments, the method includes: (k) receiving, via the processor, a user's selection of a second subset of the initial hotspot set via the GUI; and (l) calculating, via the processor, one or more risk index values ​​for the subject based at least in part on a calculated score (e.g., area score) of the subject's bones occupied by the second subset of hotspots [e.g., where the calculated score is the total area of ​​the second subset of hotspots divided by the total area of ​​all identified skeletal regions].

[0031] In some embodiments, at least one of the risk index values ​​indicates the risk of the subject developing and / or acquiring metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancers).

[0032] In some embodiments, the metastatic cancer is metastatic prostate cancer.

[0033] In some embodiments, at least one of the risk index values ​​indicates that the subject has a specific state of metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancers).

[0034] In some embodiments, the processor is a processor for a cloud-based system.

[0035] In some embodiments, the GUI is part of a general image archiving and communication system (PACS) (e.g., and clinical applications in oncology, including lesion marking and quantitative analysis).

[0036] In some embodiments, the agent (e.g., a radiopharmaceutical) includes technetium-99m methylene diphosphonate ( 99m Tc-MDP).

[0037] In another aspect, the present invention relates to a method for lesion labeling and quantification analysis (e.g., user-assisted / user-reviewed automated or semi-automated lesion labeling and quantification analysis) of nuclear medicine images (e.g., a set of bone scan images) of a subject, the method comprising: (a) accessing (e.g., and / or receiving) the set of bone scan images of the subject (e.g., a set of one, two, or more images) via a processor of a computing device, the set of bone scan images being obtained after administration of a drug (e.g., a radiopharmaceutical) to the subject (e.g., the set of bone scan images comprising anterior and posterior bone scan images) (e.g., wherein each image in the set of bone scan images comprises a plurality of pixels). (i) Each pixel has a value corresponding to its intensity; (b) Each image in the bone scan image set is automatically segmented by the processor to identify one or more skeletal regions of interest (e.g., a specific bone and / or a set of one or more bones (e.g., cervical vertebrae, clavicle, ribs, lumbar vertebrae, pelvis, sacrum, scapula, skull, thoracic vertebrae, sternum, femur, humerus) to obtain an annotated image set, each skeletal region of interest corresponding to a specific anatomical region of the subject's skeleton; (c) An initial set of one or more hotspots is automatically detected by the processor, each hotspot corresponding to a high-intensity region in the annotated image set, the automatic detection including: using (i) the intensity of pixels in the annotated image set and (ii) Multiple preliminary thresholds (e.g., where the multiple preliminary thresholds are region-dependent thresholds, depending on the identified skeletal region of interest where a particular pixel is located) are used to detect a potential hotspot set; a global threshold scaling factor is calculated using the potential hotspot set; the multiple preliminary thresholds are adjusted using the global threshold scaling factor to obtain multiple adjusted thresholds; and the initial hotspot set is identified using (i) the intensity of pixels in the annotated image set and (ii) the multiple adjusted thresholds; (d) for each hotspot in the initial hotspot set, a set of hotspot features (e.g., one or more sets) associated with the hotspot are extracted by the processor; (e) for each hotspot in the initial hotspot set, the processing is performed... The processor calculates a transfer probability value corresponding to the likelihood of the hotspot representation transferring 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., an artificial neural network (ANN) that takes at least a portion of the hotspot features as input for a specific hotspot and outputs the transfer probability value of the hotspot)]; and (f) causes the processor to render a graphical representation of at least a portion of the initial set of hotspots [e.g., visual indications of hotspots (e.g., points, boundaries) overlaid on one or more images in the set of bone scan images and / or the set of annotated images]; e.g., listing the identified hotspots and additional information for each hotspot (e.g., location);For example, a table of transfer probability values ​​can be displayed within a graphical user interface (GUI) (e.g., a cloud-based GUI).

[0038] In some embodiments, the global threshold scaling factor is a function of a measure of the subject's disease burden [e.g., the fraction of the area occupied by the subject's bone transfer (e.g., hotspots); e.g., a risk index value], and wherein the adjustment of the plurality of preliminary thresholds performed at step (c) includes: reducing the adjusted threshold (e.g., relative to the preliminary threshold) as the disease burden increases (e.g., as measured by the global threshold scaling factor) to compensate for the underestimation of hotspot area that occurs with the increased disease burden (e.g., such that the total number and / or size of hotspots increases with the decrease in the adjusted threshold).

[0039] In some embodiments, the global threshold scaling factor is a function (e.g., a non-linear function) of the fraction (e.g., area fraction) of the identified bone regions occupied by the potential hotspot set (e.g., where the global threshold scaling factor is a function of the total area of ​​all hotspots in the initial set divided by the total area of ​​all identified bone regions).

[0040] In some embodiments, the global threshold scaling factor is based on a risk index value calculated using a set of potential hotspots (e.g., a function of the risk index value calculated using a set of potential hotspots).

[0041] In some embodiments, step (c) includes (e.g., iteratively): identifying healthy tissue regions in the images of the bone scan image set by the processor that are determined not to contain any hotspots (e.g., relatively high-intensity local areas); calculating a normalization factor by the processor such that the product of the normalization factor and the average intensity of the identified healthy tissue regions is a predefined intensity level; and normalizing the images in the bone scan image set by the processor based on the normalization factor.

[0042] In some embodiments, the method further includes: (g) calculating one or more risk index values ​​for the subject based at least in part on a calculated score (e.g., area score) of the subject’s bones being occupied by a second subset of hotspots [e.g., where the calculated score is the ratio of the total area of ​​the initial set of hotspots to the total area of ​​all identified skeletal regions].

[0043] In some embodiments, the method includes: (h) selecting a first subset of the initial hotspot set (e.g., up to the entire subset) by the processor based at least in part on the transfer probability value [e.g., determining whether a particular hotspot is included in the subset based on a transfer probability value calculated for a particular hotspot in the initial hotspot set exceeding a threshold]; and (i) causing the processor to render a graphical representation of the first subset [e.g., visual indications (e.g., points, boundaries) of hotspots overlaid on one or more members of the bone scan image set and / or annotated image set; e.g., a table listing identified hotspots and additional information for each hotspot (e.g., location; e.g., transfer probability value)] for display within a graphical user interface (GUI) (e.g., a cloud-based GUI).

[0044] In some embodiments, the method further includes: (j) calculating one or more risk index values ​​for the subject based at least in part on a calculated score (e.g., area score) of the subject’s bones being occupied by a second subset of hotspots [e.g., where the calculated score is the total area of ​​the initial set of hotspots divided by the total area of ​​all identified skeletal regions].

[0045] In some embodiments, the method includes: (k) receiving, via the processor, a user's selection of a second subset of the initial hotspot set via the GUI; and (l) calculating, via the processor, one or more risk index values ​​for the subject based at least in part on a calculated score (e.g., area score) of the subject's bones occupied by the second subset of hotspots [e.g., where the calculated score is the total area of ​​the second subset of hotspots divided by the total area of ​​all identified skeletal regions].

[0046] In some embodiments, at least one of the risk index values ​​indicates the risk of the subject developing and / or acquiring metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancers).

[0047] In some embodiments, the metastatic cancer is metastatic prostate cancer.

[0048] In some embodiments, at least one of the risk index values ​​indicates that the subject has a specific state of metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancers).

[0049] In some embodiments, the processor is a processor for a cloud-based system.

[0050] In some embodiments, the GUI is part of a general image archiving and communication system (PACS) (e.g., and clinical applications in oncology, including lesion marking and quantitative analysis).

[0051] In some embodiments, the agent (e.g., a radiopharmaceutical) includes technetium-99m methylene diphosphonate ( 99m Tc-MDP).

[0052] In another aspect, the present invention relates to a method for lesion labeling and quantification analysis (e.g., user-assisted / user-reviewed automated or semi-automated lesion labeling and quantification analysis) of nuclear medicine images (e.g., a set of bone scan images) of a subject, the method comprising: (a) accessing (e.g., and / or receiving) the set of bone scan images of the subject (e.g., a set of one, two, or more images) via a processor of a computing device, the set of bone scan images being obtained after administration of a drug (e.g., a radiopharmaceutical) to the subject (e.g., the set of bone scan images comprising anterior and posterior bone scan images) (e.g., wherein each image in the set of bone scan images comprises a plurality of pixels, each pixel...). (a) The processor automatically segments each image in the bone scan image set to identify one or more skeletal regions of interest (e.g., a specific bone and / or a set of one or more bones (e.g., cervical vertebrae, clavicle, ribs, lumbar vertebrae, pelvis, sacrum, scapula, skull, thoracic vertebrae, sternum, femur, humerus) to obtain an annotated image set, each skeletal region of interest corresponding to a specific anatomical region of the subject's skeleton; (c) The processor automatically detects an initial set of one or more hotspots, each hotspot corresponding to a high-intensity region in the annotated image set [e.g., wherein detecting the one or more hotspots in the initial hotspot set includes comparing pixel intensity with one or more thresholds (e.g., wherein...]. (d) For each hotspot in the initial hotspot set, the processor extracts a set of hotspot features associated with the hotspot (e.g., a set of one or more); (e) For each hotspot in the initial hotspot set, the processor calculates a transfer probability value corresponding to the probability of the hotspot representing 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)), which receive at least a portion of the hotspot features as input for a specific hotspot and output the hotspot features]. (f) Selecting a first subset of the initial hotspot set by the processor (e.g., up to all subsets), wherein the selection of a specific hotspot to be included in the first subset is based at least in part on: (i) a transfer probability value calculated for the specific hotspot [e.g., based on a comparison of the probability value calculated for the specific hotspot with a probability threshold (e.g., if the probability value is greater than the probability threshold, then the specific hotspot is included in the first subset)]; and (ii) one or more global hotspot features, each global hotspot feature being determined using multiple hotspots in the initial hotspot set (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.).(g) The processor causes the rendering of a graphical representation of at least a portion of the first subset of hotspots [e.g., visual indications (e.g., points, boundaries) of hotspots overlaid on one or more images in the bone scan image set and / or annotated image set; e.g., a table listing identified hotspots and additional information (e.g., location; e.g., probability value) for each hotspot] for display within a graphical user interface (GUI) (e.g., a cloud-based GUI).

[0053] In some embodiments, the one or more global hotspot features include the total number of hotspots in the initial hotspot set.

[0054] In some embodiments, step (f) includes adjusting the criteria for selecting hotspots to be included in the first subset based on the total number of hotspots in the initial hotspot set [e.g., by relaxing the criteria as the total number of hotspots in the initial hotspot set increases (e.g., by reducing the transfer probability threshold compared with each hotspot transfer probability value; e.g., by scaling the transfer probability value based on the total number of hotspots in the initial hotspot set)].

[0055] In some embodiments, step (f) includes using a machine learning module to select a first subset (e.g., an ANN module) [e.g., where the machine learning module receives at least a transition probability value computed for each hotspot and one or more global hotspot features for each hotspot, and outputs (i) an adjusted transition probability value that takes the global hotspot features into account (e.g., a value on a scale that can be compared with a threshold used to select hotspots in the first subset) and / or (ii) a binary (e.g., 0 or 1; e.g., Boolean true or false) value indicating whether a hotspot should or should not be included in the first subset].

[0056] In some embodiments, step (c) includes (e.g., iteratively): identifying healthy tissue regions in the images of the bone scan image set by the processor that are determined not to contain any hotspots (e.g., relatively high-intensity local areas); calculating a normalization factor by the processor such that the product of the normalization factor and the average intensity of the identified healthy tissue regions is a predefined intensity level; and normalizing the images in the bone scan image set by the processor based on the normalization factor.

[0057] In some embodiments, the method further includes: (g) calculating one or more risk index values ​​for the subject based at least in part on a calculated score (e.g., area score) of the subject’s bones being occupied by a second subset of hotspots [e.g., where the calculated score is the ratio of the total area of ​​the initial set of hotspots to the total area of ​​all identified skeletal regions].

[0058] In some embodiments, the method includes: (h) selecting a first subset of the initial hotspot set (e.g., up to the entire subset) by the processor based at least in part on the transfer probability value [e.g., determining whether a particular hotspot is included in the subset based on a transfer probability value calculated for a particular hotspot in the initial hotspot set exceeding a threshold]; and (i) causing the processor to render a graphical representation of the first subset [e.g., visual indications (e.g., points, boundaries) of hotspots overlaid on one or more members of the bone scan image set and / or annotated image set; e.g., a table listing identified hotspots and additional information for each hotspot (e.g., location; e.g., transfer probability value)] for display within a graphical user interface (GUI) (e.g., a cloud-based GUI).

[0059] In some embodiments, the method further includes: (j) calculating one or more risk index values ​​for the subject based at least in part on a calculated score (e.g., area score) of the subject’s bones being occupied by a second subset of hotspots [e.g., where the calculated score is the total area of ​​the initial set of hotspots divided by the total area of ​​all identified skeletal regions].

[0060] In some embodiments, the method includes: (k) receiving, via the processor, a user's selection of a second subset of the initial hotspot set via the GUI; and (l) calculating, via the processor, one or more risk index values ​​for the subject based at least in part on a calculated score (e.g., area score) of the subject's bones occupied by the second subset of hotspots [e.g., where the calculated score is the total area of ​​the second subset of hotspots divided by the total area of ​​all identified skeletal regions].

[0061] In some embodiments, at least one of the risk index values ​​indicates the risk of the subject developing and / or acquiring metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancers).

[0062] In some embodiments, the metastatic cancer is metastatic prostate cancer.

[0063] In some embodiments, at least one of the risk index values ​​indicates that the subject has a specific state of metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancers).

[0064] In some embodiments, the processor is a processor for a cloud-based system.

[0065] In some embodiments, the GUI is part of a general image archiving and communication system (PACS) (e.g., and clinical applications in oncology, including lesion marking and quantitative analysis).

[0066] In some embodiments, the agent (e.g., a radiopharmaceutical) includes technetium-99m methylene diphosphonate ( 99m Tc-MDP).

[0067] In another aspect, the present invention relates to a method for lesion labeling and quantification analysis (e.g., user-assisted / user-reviewed automated or semi-automated lesion labeling and quantification analysis) of nuclear medicine images (e.g., a set of bone scan images) of a subject, the method comprising: (a) accessing (e.g., and / or receiving) the set of bone scan images of the subject (e.g., a set of one, two, or more images) via a processor of a computing device (e.g., the set of bone scan images includes anterior bone scan images and posterior bone scan images) (e.g., wherein each image in the set of bone scan images comprises a plurality of pixels, each pixel...). (a) The processor automatically segments each image in the bone scan image set to identify one or more skeletal regions of interest (e.g., a specific bone and / or a set of one or more bones (e.g., cervical vertebrae, clavicle, ribs, lumbar vertebrae, pelvis, sacrum, scapula, skull, thoracic vertebrae, sternum, femur, humerus) to obtain an annotated image set, each skeletal region of interest corresponding to a specific anatomical region of the subject's skeleton; (c) The processor automatically detects an initial set of one or more hotspots, each hotspot corresponding to a high-intensity region in the annotated image set [e.g., where the detected hotspots have values ​​corresponding to intensity]. The one or more hotspots in the initial hotspot set include comparing pixel intensity with one or more thresholds (e.g., where the one or more thresholds vary according to the identified skeletal region of interest where the particular pixel is located)]; (d) for each hotspot in the initial hotspot set, the processor extracts a set of hotspot features associated with the hotspot (e.g., a set of one or more); (e) for each hotspot in the initial hotspot set, the processor calculates a probability value corresponding to the likelihood of the hotspot representing a shift 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., an artificial neural network (ANN) that receives at least a portion of the hotspot features as input for a particular hotspot and outputs the probability value of the hotspot)]; (f) the processor selects a first subset of hotspots in the initial hotspot set (e.g., up to all hotspots) based at least in part on the probability values ​​calculated for each hotspot in the initial hotspot set [e.g., by determining whether to include the particular hotspot in the pre-selected hotspot set based on the probability values ​​calculated for the particular hotspot in the initial hotspot set (e.g., by comparing the probability value with a probability threshold)];(g) Calculating one or more risk index values ​​(e.g., bone scan index values) using at least a portion (e.g., up to the entire first subset) of hotspots via the processor, the calculation comprising: determining one or more bone accumulation factors for each specific hotspot of the first subset based on (i) the ratio of the size (e.g., area) of the specific hotspot to (ii) the size (e.g., area) of a specific bone region to which the specific hotspot was assigned (e.g., via the processor), the assignment being based on the specific hotspot in the banded area. Record the location within the image set; adjust the bone involvement factor using one or more region-dependent correction factors [e.g., each region-dependent correction factor is associated with one or more bone regions; for example, where the region-dependent correction factor has a value selected to reduce the degree to which a particular hotspot is assigned to a particular bone region (e.g., multiple adjacent or nearby bone regions (e.g., sacral region, pelvic region, and lumbar region)) results in a fluctuation in the calculated bone involvement factor], thereby obtaining one or more adjusted bone involvement factors; and sum the adjusted bone involvement factors to determine the one or more risk index values.

[0068] In some embodiments, for each specific hotspot, the calculated bone accumulation factor estimates the proportion of total bone mass occupied by the body volume associated with that specific hotspot.

[0069] In some embodiments, calculating the bone accumulation factor includes: calculating an area fraction of the specific hotspot by the processor, which is the ratio of the area of ​​the specific hotspot to the area of ​​the corresponding bone region of interest; and scaling (e.g., multiplying) the area fraction by a density coefficient associated with the bone region of interest to which the specific hotspot was assigned [e.g., taking into account the weight and / or density of bone in the corresponding bone region of interest (e.g., where the density coefficient is the weight fraction of the corresponding bone region of interest relative to the total bone (e.g., in a typical person)] to calculate the bone accumulation factor of the specific hotspot.

[0070] In some embodiments, at least a portion of the hotspots in the first subset are assigned to the skeletal region of interest, which is a member of a group selected from the pelvic region (e.g., corresponding to the subject's pelvis), the lumbar region (e.g., corresponding to the subject's lumbar spine), and the sacral region (e.g., corresponding to the subject's sacrum).

[0071] In some embodiments, the one or more zone-dependent correction factors include a sacral zone correction factor, which is associated with the sacral zone and used to adjust the bone involvement factor identified (e.g., processed) as a hotspot located in the sacral zone, and wherein the sacral zone correction factor has a value less than 1 (e.g., less than 0.5).

[0072] In some embodiments, the one or more region-dependent correction factors include one or more pairs of correction factors, each pair of correction factors being associated with a specific bone region of interest and including a first member and a second member (in the pair), wherein: the first member of the pair is an anterior image correction factor and is used to adjust the bone accumulation factor calculated for hotspots detected in the annotated anterior bone scan images in the annotated image set, and the second member of the pair is a posterior image correction factor and is used to adjust the bone accumulation factor calculated for hotspots detected in the annotated posterior bone scan images in the annotated image set.

[0073] In some embodiments, step (c) includes (e.g., iteratively): identifying healthy tissue regions in the images of the bone scan image set by the processor that are determined not to contain any hotspots (e.g., relatively high-intensity local areas); calculating a normalization factor by the processor such that the product of the normalization factor and the average intensity of the identified healthy tissue regions is a predefined intensity level; and normalizing the images in the bone scan image set by the processor based on the normalization factor.

[0074] In some embodiments, the method further includes: (g) calculating one or more risk index values ​​for the subject based at least in part on a calculated score (e.g., area score) of the subject’s bones being occupied by a second subset of hotspots [e.g., where the calculated score is the ratio of the total area of ​​the initial set of hotspots to the total area of ​​all identified skeletal regions].

[0075] In some embodiments, the method includes: (h) selecting a first subset of the initial hotspot set (e.g., up to the entire subset) by the processor based at least in part on the transfer probability value [e.g., determining whether a particular hotspot is included in the subset based on a transfer probability value calculated for a particular hotspot in the initial hotspot set exceeding a threshold]; and (i) causing the processor to render a graphical representation of the first subset [e.g., visual indications (e.g., points, boundaries) of hotspots overlaid on one or more members of the bone scan image set and / or annotated image set; e.g., a table listing identified hotspots and additional information for each hotspot (e.g., location; e.g., transfer probability value)] for display within a graphical user interface (GUI) (e.g., a cloud-based GUI).

[0076] In some embodiments, the method further includes: (j) calculating one or more risk index values ​​for the subject based at least in part on a calculated score (e.g., area score) of the subject’s bones being occupied by a second subset of hotspots [e.g., where the calculated score is the total area of ​​the initial set of hotspots divided by the total area of ​​all identified skeletal regions].

[0077] In some embodiments, the method includes: (k) receiving, via the processor, a user's selection of a second subset of the initial hotspot set via the GUI; and (l) calculating, via the processor, one or more risk index values ​​for the subject based at least in part on a calculated score (e.g., area score) of the subject's bones occupied by the second subset of hotspots [e.g., where the calculated score is the total area of ​​the second subset of hotspots divided by the total area of ​​all identified skeletal regions].

[0078] In some embodiments, at least one of the risk index values ​​indicates the risk of the subject developing and / or acquiring metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancers).

[0079] In some embodiments, the metastatic cancer is metastatic prostate cancer.

[0080] In some embodiments, at least one of the risk index values ​​indicates that the subject has a specific state of metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancers).

[0081] In some embodiments, the processor is a processor for a cloud-based system.

[0082] In some embodiments, the GUI is part of a general image archiving and communication system (PACS) (e.g., and clinical applications in oncology, including lesion marking and quantitative analysis).

[0083] In some embodiments, the agent (e.g., a radiopharmaceutical) includes technetium-99m methylene diphosphonate ( 99m Tc-MDP).

[0084] In another aspect, the present invention relates to a system for lesion labeling and quantification analysis (e.g., user-assisted / user-reviewed automated or semi-automated lesion labeling and quantification analysis) of nuclear medicine images (e.g., a set of bone scan images) of a subject, the system comprising: a processor; and a memory having instructions thereon, wherein the instructions, when executed by the processor, cause the processor to: (a) access (e.g., and / or receive) the set of bone scan images of the subject (e.g., a set of one, two, or more images), the set of bone scan images being obtained after administration of a drug (e.g., a radiopharmaceutical) to the subject (e.g., the set of bone scan images includes pre-existing conditions). (a) Lateral and posterior bone scan images (e.g., each image in the set of bone scan images comprises multiple pixels, each pixel having a value corresponding to intensity); (b) each image in the set of bone scan images is automatically segmented to identify one or more skeletal regions of interest (e.g., a specific bone and / or a set of one or more bones (e.g., cervical vertebrae, clavicle, ribs, lumbar vertebrae, pelvis, sacrum, scapula, skull, thoracic vertebrae, sternum, femur, humerus)) to obtain an annotated image set, wherein each of the one or more skeletal regions of interest corresponds to a specific anatomical region of the subject's skeleton, wherein the one or more skeletal regions of interest include at least one of (i) and (ii): (i) the femur region, (i) a portion of the femur of the subject, the femoral portion encompassing at least three-quarters of the femur along its length [(e.g., greater than about three-quarters (e.g., approximately all)]; and (ii) a humeral region, corresponding to a portion of the humerus of the subject, the humeral portion encompassing at least three-quarters of the humerus along its length [(e.g., greater than about three-quarters (e.g., approximately all)]; (c) an initial set of automatically detected hotspots, each hotspot corresponding to a high-intensity region in the annotated image set, the automatic detection comprising: identifying the one or more hotspots using pixel intensity in the annotated image set and using one or more region-dependent thresholds. Hotspots (e.g., where each region dependency threshold is associated with an identified bone region of interest, such that the intensity of a pixel located within a particular identified bone region is compared to the associated region dependency threshold), and wherein the one or more region dependency thresholds include one or more values ​​associated with the femoral region and / or the humeral region (e.g., a reduced intensity threshold for the femoral region and / or a reduced intensity threshold for the humeral region), the one or more values ​​providing enhanced hotspot detection sensitivity for the femoral region and / or the humeral region to compensate for reduced drug uptake in the femoral region and / or the humeral region; (d) for each hotspot in the initial hotspot set, extracting a set of hotspot features associated with the hotspot (e.g., a set of one or more);(e) For each hotspot in the initial hotspot set, calculate a transfer probability value corresponding to the likelihood of the hotspot representing a transfer based on the set of hotspot features associated with that hotspot [e.g., using one or more machine learning modules (e.g., pre-trained machine learning modules; e.g., artificial neural networks (ANNs)), which receive at least a portion of the hotspot features as input for a particular hotspot and output the transfer probability value of the hotspot]; and (f) cause a graphical representation of at least a portion of the initial hotspot set to be rendered [e.g., visual indications (e.g., points, boundaries) of hotspots overlaid on one or more members of the bone scan image set and / or annotated image set; e.g., a table listing the identified hotspots and additional information for each hotspot (e.g., location; e.g., transfer probability value)] for display within a graphical user interface (GUI) (e.g., a cloud-based GUI).

[0085] In some embodiments, the instruction at step (b) causes the processor to: compare each member of the bone scan image set with a corresponding atlas image in the atlas image set, each atlas image including one or more identifications (e.g., graphic identifications superimposed on the atlas image) of the one or more bone regions of interest, the bone regions of interest including the femur and / or the humerus; and for each image in the bone scan image set, register the corresponding atlas image with the image in the bone scan image set such that the identifications of the one or more bone regions of interest in the atlas image are applied (e.g., superimposed on) the image in the bone scan image set.

[0086] In some embodiments, each atlas image includes the identification of (i) the femoral region comprising at least a portion of the subject's knee region and / or (ii) the humeral region comprising at least a portion of the subject's elbow region, and wherein, for each image in the set of bone scan images, the instructions cause the processor to use the identified knee region and / or the identified elbow region in the image as landmarks to register the corresponding atlas image to the bone scan image [e.g., by identifying the knee region in the bone scan image and matching it with the knee region identified in the corresponding atlas image, and then adjusting the atlas image (e.g., calculating coordinate transformation) to register the corresponding atlas image to the bone scan image].

[0087] In some embodiments, the location of at least one detected hot spot in the initial hot spot set corresponds to a body location in or on the femur that is more than three-quarters of the distance along the femur from the end of the femur oriented toward the subject's hip to the end of the femur oriented toward the subject's knee.

[0088] In some embodiments, the location of at least one detected hot spot in the initial set of hot spots corresponds to a body location in or on the humerus, the body location being located at more than three-quarters of the distance along the humerus from one end of the humerus oriented toward the subject's shoulder to the other end of the humerus oriented toward the subject's elbow.

[0089] In some embodiments, the instructions at step (c) cause the processor (e.g., iteratively) to: identify healthy tissue regions in the images of the bone scan image set that are determined not to contain any hot spots (e.g., local areas with relatively high intensity); calculate a normalization factor such that the product of the normalization factor and the average intensity of the identified healthy tissue regions is a predefined intensity level; and normalize the images in the bone scan image set by the normalization factor.

[0090] In some embodiments, the instructions further cause the processor to: (g) calculate one or more risk index values ​​for the subject based at least in part on a calculated score (e.g., area score) of the subject’s bones being occupied by the initial hotspot set [e.g., where the calculated score is the ratio of the total area of ​​the initial hotspot set to the total area of ​​all identified skeletal regions].

[0091] In some embodiments, the instructions cause the processor to: (h) select a first subset (e.g., up to all) of the initial hotspot set based at least in part on a transfer probability value [e.g., determine whether a particular hotspot is included in the subset based on a transfer probability value calculated for a particular hotspot in the initial hotspot set exceeding a threshold]; and (i) cause to render a graphical representation of the first subset [e.g., visual indications (e.g., points, boundaries) of hotspots overlaid on one or more members of the bone scan image set and / or annotated image set; e.g., a table listing the identified hotspots and additional information (e.g., location; e.g., probability value) for each hotspot] for display within a graphical user interface (GUI) (e.g., a cloud-based GUI).

[0092] In some embodiments, the instructions cause the processor to: (j) calculate one or more risk index values ​​for the subject based at least in part on a calculated score (e.g., area score) of the subject’s bones being occupied by a first subset of hotspots [e.g., where the calculated score is the total area of ​​the initial set of hotspots divided by the total area of ​​all identified skeletal regions].

[0093] In some embodiments, the instructions cause the processor to: (k) receive a user's selection of a second subset of the initial hotspot set via the GUI; and (l) calculate one or more risk index values ​​for the subject based at least in part on a calculated score (e.g., area score) of the subject's bones occupied by the second subset of hotspots [e.g., where the calculated score is the total area of ​​the second subset of hotspots divided by the total area of ​​all identified skeletal regions].

[0094] In some embodiments, at least one of the risk index values ​​indicates the risk of the subject developing and / or acquiring metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancers).

[0095] In some embodiments, the metastatic cancer is metastatic prostate cancer.

[0096] In some embodiments, at least one of the risk index values ​​indicates that the subject has a specific state of metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancers).

[0097] In some embodiments, the system is a cloud-based system. In some embodiments, the processor is a cloud-based system processor.

[0098] In some embodiments, the GUI is part of a general image archiving and communication system (PACS) (e.g., and clinical applications in oncology, including lesion marking and quantitative analysis).

[0099] In some embodiments, the agent (e.g., a radiopharmaceutical) includes technetium-99m methylene diphosphonate ( 99m Tc-MDP).

[0100] In another aspect, the present invention relates to a system for lesion labeling and quantification analysis (e.g., user-assisted / user-reviewed automated or semi-automated lesion labeling and quantification analysis) of nuclear medicine images (e.g., a set of bone scan images) of a subject, the system comprising: a processor; and a memory having instructions thereon, wherein the instructions, when executed by the processor, cause the processor to: (a) access (e.g., and / or receive) the set of bone scan images of the subject (e.g., one, two, or more images) via a processor of a computing device, the set of bone scan images being obtained after administration of a drug (e.g., a radiopharmaceutical) to the subject (e.g., the set of bone scan images includes anterior bone scans). (a) tracing images and posterior bone scan images (e.g., each image in the bone scan image set includes multiple pixels, each pixel having a value corresponding to intensity); (b) automatically segmenting each image in the bone scan image set to identify one or more skeletal regions of interest (e.g., a specific bone and / or a set of one or more bones (e.g., cervical vertebrae, clavicle, ribs, lumbar vertebrae, pelvis, sacrum, scapula, skull, thoracic vertebrae, sternum, femur, humerus) to obtain an annotated image set, each skeletal region of interest corresponding to a specific anatomical region of the subject's skeleton; (c) automatically detecting an initial set of one or more hotspots, each hotspot corresponding to a high-intensity region in the annotated image set, the automatic detection including: using (i) Detecting a potential hotspot set using the intensity of pixels in the annotated image set and (ii) multiple preliminary thresholds (e.g., where the multiple preliminary thresholds are region-dependent thresholds that depend on the identified skeletal region of interest where a particular pixel is located); calculating a global threshold scaling factor using the potential hotspot set; adjusting the multiple preliminary thresholds using the global threshold scaling factor to obtain multiple adjusted thresholds; and identifying the initial hotspot set using (i) the intensity of pixels in the annotated image set and (ii) the multiple adjusted thresholds; (d) extracting a set of hotspot features (e.g., one or more sets) associated with each hotspot in the initial hotspot set; (e) for the initial hotspots For each hotspot in the set, a transfer probability value corresponding to the likelihood of the hotspot representing a transfer is calculated 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., an artificial neural network (ANN) that takes at least a portion of the hotspot features as input for a particular hotspot and outputs the transfer probability value of the hotspot)]; and (f) causing a graphical representation of at least a portion of the initial set of hotspots to be rendered [e.g., visual indications of hotspots (e.g., points, boundaries) overlaid on one or more images in the set of bone scan images and / or the set of annotated images]; e.g., listing the identified hotspots and additional information for each hotspot (e.g., location);For example, a table of transfer probability values ​​can be displayed within a graphical user interface (GUI) (e.g., a cloud-based GUI).

[0101] In some embodiments, the instructions cause the processor to: calculate a global threshold scaling factor, which is a measure of the subject's disease burden [e.g., the fraction of the area occupied by the subject's bone transfer (e.g., hotspots); e.g., a risk index value]; and at step (c) adjust the plurality of initial thresholds by decreasing an adjusted threshold (e.g., relative to an initial threshold) as the disease burden increases (e.g., as measured by the global threshold scaling factor) to compensate for the underestimation of hotspot area that occurs with the increase in disease burden (e.g., such that the total number and / or size of hotspots increases as the adjusted threshold decreases).

[0102] In some embodiments, the instructions cause the processor to calculate the global threshold scaling factor as a function (e.g., a non-linear function) of the fraction (e.g., area fraction) occupied by the potential hotspot set of the identified skeletal regions (where the global threshold scaling factor is a function of the total area of ​​all hotspots in the initial set divided by the total area of ​​all identified skeletal regions).

[0103] In some embodiments, the instructions cause the processor to calculate a global threshold scaling factor based on (e.g., according to) a risk index value calculated using a set of potential hotspots.

[0104] In some embodiments, the instructions at step (c) cause the processor (e.g., iteratively) to: identify healthy tissue regions in the images of the bone scan image set that are determined not to contain any hot spots (e.g., local areas with relatively high intensity); calculate a normalization factor such that the product of the normalization factor and the average intensity of the identified healthy tissue regions is a predefined intensity level; and normalize the images in the bone scan image set by the normalization factor.

[0105] In some embodiments, the instructions further cause the processor to: (g) calculate one or more risk index values ​​for the subject based at least in part on a calculated score (e.g., area score) of the subject’s bones being occupied by the initial hotspot set [e.g., where the calculated score is the ratio of the total area of ​​the initial hotspot set to the total area of ​​all identified skeletal regions].

[0106] In some embodiments, the instructions cause the processor to: (h) select a first subset (e.g., up to all) of the initial hotspot set based at least in part on a transfer probability value [e.g., determine whether a particular hotspot is included in the subset based on a transfer probability value calculated for a particular hotspot in the initial hotspot set exceeding a threshold]; and (i) cause to render a graphical representation of the first subset [e.g., visual indications (e.g., points, boundaries) of hotspots overlaid on one or more members of the bone scan image set and / or annotated image set; e.g., a table listing the identified hotspots and additional information (e.g., location; e.g., probability value) for each hotspot] for display within a graphical user interface (GUI) (e.g., a cloud-based GUI).

[0107] In some embodiments, the instructions cause the processor to: (j) calculate one or more risk index values ​​for the subject based at least in part on a calculated score (e.g., area score) of the subject’s bones being occupied by a first subset of hotspots [e.g., where the calculated score is the total area of ​​the initial set of hotspots divided by the total area of ​​all identified skeletal regions].

[0108] In some embodiments, the instructions cause the processor to: (k) receive a user's selection of a second subset of the initial hotspot set via the GUI; and (l) calculate one or more risk index values ​​for the subject based at least in part on a calculated score (e.g., area score) of the subject's bones occupied by the second subset of hotspots [e.g., where the calculated score is the total area of ​​the second subset of hotspots divided by the total area of ​​all identified skeletal regions].

[0109] In some embodiments, at least one of the risk index values ​​indicates the risk of the subject developing and / or acquiring metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancers).

[0110] In some embodiments, the metastatic cancer is metastatic prostate cancer.

[0111] In some embodiments, at least one of the risk index values ​​indicates that the subject has a specific state of metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancers).

[0112] In some embodiments, the system is a cloud-based system. In some embodiments, the processor is a cloud-based system processor.

[0113] In some embodiments, the GUI is part of a general image archiving and communication system (PACS) (e.g., and clinical applications in oncology, including lesion marking and quantitative analysis).

[0114] In some embodiments, the agent (e.g., a radiopharmaceutical) includes technetium-99m methylene diphosphonate ( 99m Tc-MDP).

[0115] In another aspect, the present invention relates to a system for lesion labeling and quantification analysis (e.g., user-assisted / user-reviewed automated or semi-automated lesion labeling and quantification analysis) of nuclear medicine images (e.g., a set of bone scan images) of a subject, the system comprising: a processor; and a memory having instructions thereon, wherein the instructions, when executed by the processor, cause the processor to: (a) access (e.g., and / or receive) the set of bone scan images of the subject (e.g., a set of one, two, or more images), the set of bone scan images being obtained after administration of a drug (e.g., a radiopharmaceutical) to the subject (e.g., the set of bone scan images comprising anterior and posterior bone scan images). (a) (e.g., each image in the bone scan image set includes multiple pixels, each pixel having a value corresponding to intensity); (b) automatically segmenting each image in the bone scan image set to identify one or more skeletal regions of interest (e.g., a specific bone and / or a set of one or more bones (e.g., cervical vertebrae, clavicle, ribs, lumbar vertebrae, pelvis, sacrum, scapula, skull, thoracic vertebrae, sternum, femur, humerus) to obtain an annotated image set, each skeletal region of interest corresponding to a specific anatomical region of the subject's skeleton; (c) automatically detecting an initial set of one or more hotspots, each hotspot corresponding to a high-intensity region in the annotated image set [e.g., wherein the one or more hotspots detected in the initial set of hotspots include...]. (d) Compare pixel intensity with one or more thresholds (e.g., where the one or more thresholds vary depending on the identified skeletal region of interest where the particular pixel is located); (e) For each hotspot in the initial hotspot set, extract a set of hotspot features associated with the hotspot (e.g., one or more sets); (e) For each hotspot in the initial hotspot set, calculate a transition probability value corresponding to the probability of the hotspot representing 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)), which receive at least a portion of the hotspot features as input for a particular hotspot and...] (f) Automatically select a first subset of the initial hotspot set (e.g., up to all hotspots), wherein the selection of a specific hotspot to be included in the first subset is based at least in part on: (i) a transfer probability value calculated for the specific hotspot [e.g., based on comparing the probability value calculated for the specific hotspot with a probability threshold (e.g., if the probability value is greater than the probability threshold, then the specific hotspot is included in the first subset)]; and (ii) one or more global hotspot features, each global hotspot feature being determined using multiple hotspots in the initial hotspot set (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.).(g) causing a graphical representation of at least a portion of the first subset of hotspots to be rendered [e.g., visual indications (e.g., points, boundaries) of hotspots overlaid on one or more images in the bone scan image set and / or annotated image set; e.g., a table listing identified hotspots and additional information (e.g., location; e.g., probability value) for each hotspot] for display within a graphical user interface (GUI) (e.g., a cloud-based GUI).

[0116] In some embodiments, the one or more global hotspot features include the total number of hotspots in the initial hotspot set.

[0117] In some embodiments, the instruction at step (f) causes the processor to adjust the criteria for selecting hotspots to be included in the first subset based on the total number of hotspots in the initial hotspot set [e.g., by relaxing the criteria as the total number of hotspots in the initial hotspot set increases (e.g., by reducing the transfer probability threshold compared with each hotspot transfer probability value; e.g., by scaling the transfer probability value based on the total number of hotspots in the initial hotspot set)].

[0118] In some embodiments, the instruction at step (f) causes the processor to use a machine learning module to select a first subset (e.g., an ANN module) [e.g., where the machine learning module receives at least a transition probability value computed for each hotspot and one or more global hotspot features, and outputs (i) an adjusted transition probability value that takes the global hotspot features into account (e.g., a value on a scale that can be compared with a threshold used to select hotspots in the first subset) and / or (ii) a binary (e.g., 0 or 1; e.g., Boolean true or false) value indicating whether a hotspot should or should not be included in the first subset].

[0119] In some embodiments, the instructions at step (c) cause the processor (e.g., iteratively) to: identify healthy tissue regions in the images of the bone scan image set that are determined not to contain any hot spots (e.g., local areas with relatively high intensity); calculate a normalization factor such that the product of the normalization factor and the average intensity of the identified healthy tissue regions is a predefined intensity level; and normalize the images in the bone scan image set by the normalization factor.

[0120] In some embodiments, the instructions further cause the processor to: (g) calculate one or more risk index values ​​for the subject based at least in part on a calculated score (e.g., area score) of the subject’s bones being occupied by the initial hotspot set [e.g., where the calculated score is the ratio of the total area of ​​the initial hotspot set to the total area of ​​all identified skeletal regions].

[0121] In some embodiments, the instructions cause the processor to: (h) select a first subset (e.g., up to all) of the initial hotspot set based at least in part on a transfer probability value [e.g., determine whether a particular hotspot is included in the subset based on a transfer probability value calculated for a particular hotspot in the initial hotspot set exceeding a threshold]; and (i) cause to render a graphical representation of the first subset [e.g., visual indications (e.g., points, boundaries) of hotspots overlaid on one or more members of the bone scan image set and / or annotated image set; e.g., a table listing the identified hotspots and additional information (e.g., location; e.g., probability value) for each hotspot] for display within a graphical user interface (GUI) (e.g., a cloud-based GUI).

[0122] In some embodiments, the instructions cause the processor to: (j) calculate one or more risk index values ​​for the subject based at least in part on a calculated score (e.g., area score) of the subject’s bones being occupied by a first subset of hotspots [e.g., where the calculated score is the total area of ​​the initial set of hotspots divided by the total area of ​​all identified skeletal regions].

[0123] In some embodiments, the instructions cause the processor to: (k) receive a user's selection of a second subset of the initial hotspot set via the GUI; and (l) calculate one or more risk index values ​​for the subject based at least in part on a calculated score (e.g., area score) of the subject's bones occupied by the second subset of hotspots [e.g., where the calculated score is the total area of ​​the second subset of hotspots divided by the total area of ​​all identified skeletal regions].

[0124] In some embodiments, at least one of the risk index values ​​indicates the risk of the subject developing and / or acquiring metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancers).

[0125] In some embodiments, the metastatic cancer is metastatic prostate cancer.

[0126] In some embodiments, at least one of the risk index values ​​indicates that the subject has a specific state of metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancers).

[0127] In some embodiments, the system is a cloud-based system. In some embodiments, the processor is a cloud-based system processor.

[0128] In some embodiments, the GUI is part of a general image archiving and communication system (PACS) (e.g., and clinical applications in oncology, including lesion marking and quantitative analysis).

[0129] In some embodiments, the agent (e.g., a radiopharmaceutical) includes technetium-99m methylene diphosphonate ( 99m Tc-MDP).

[0130] In another aspect, the present invention relates to a system for lesion labeling and quantification analysis (e.g., user-assisted / user-reviewed automated or semi-automated lesion labeling and quantification analysis) of nuclear medicine images (e.g., a set of bone scan images) of a subject, the system comprising: a processor; and a memory having instructions thereon, wherein the instructions, when executed by the processor, cause the processor to: (a) access (e.g., and / or receive) the set of bone scan images of the subject (e.g., a set of one, two, or more images) (e.g., the set of bone scan images includes anterior bone scan images and posterior bone scan images). (a) Automatically segmenting each image in the bone scan image set to identify one or more skeletal regions of interest (e.g., a specific bone and / or a set of one or more bones (e.g., cervical vertebrae, clavicle, ribs, lumbar vertebrae, pelvis, sacrum, scapula, skull, thoracic vertebrae, sternum, femur, humerus) to obtain an annotated image set, each skeletal region of interest corresponding to a specific anatomical region of the subject's skeleton; (c) Automatically detecting an initial set of one or more hotspots, each hotspot corresponding to a specific anatomical region of the subject's skeleton. (d) High-intensity regions in the annotated image set [e.g., where detecting one or more hotspots in the initial hotspot set includes comparing pixel intensity with one or more thresholds (e.g., where the one or more thresholds vary according to the identified skeletal region of interest where the particular pixel is located)]; (e) For each hotspot in the initial hotspot set, extract a set of hotspot features associated with the hotspot (e.g., one or more sets); (f) For each hotspot in the initial hotspot set, compute a probability value corresponding to the probability of hotspot representation shift 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., an artificial neural network (ANN) that receives at least a portion of the hotspot features as input for a particular hotspot and outputs the probability value of the hotspot)]; (g) Select a first subset of hotspots in the initial hotspot set (e.g., up to all hotspots) at least partially based on the probability value computed for each hotspot in the initial hotspot set [e.g., by determining whether to include the particular hotspot in the pre-selected hotspot set based on the probability value computed for the particular hotspot in the initial hotspot set (e.g., by comparing the probability value with a probability threshold)];(g) Calculating one or more risk index values ​​(e.g., bone scan index values) using at least a portion (e.g., up to the entire first subset) of hotspots via the processor, the calculation comprising: for each specific hotspot in the first subset, calculating a bone accumulation factor based on (i) the ratio of the size (e.g., area) of the specific hotspot to (ii) the size (e.g., area) of a specific bone region to which the specific hotspot was assigned (e.g., via the processor), thereby determining one or more bone accumulation factors, the assignment being based on the specific hotspot in the band Record the location within the image set; adjust the bone involvement factor using one or more region-dependent correction factors [e.g., each region-dependent correction factor is associated with one or more bone regions; for example, where the region-dependent correction factor has a value selected to reduce the degree to which a particular hotspot is assigned to a particular bone region (e.g., multiple adjacent or nearby bone regions (e.g., sacral region, pelvic region, and lumbar region)) results in a fluctuation in the calculated bone involvement factor], thereby obtaining one or more adjusted bone involvement factors; and sum the adjusted bone involvement factors to determine the one or more risk index values.

[0131] In some embodiments, for each specific hotspot, the calculated bone accumulation factor estimates the proportion of total bone mass occupied by the body volume associated with that specific hotspot.

[0132] In some embodiments, the instructions cause the processor to calculate the bone accumulation factor by: calculating the area fraction of the specific hotspot to the area of ​​the corresponding bone region of interest, thereby calculating the area fraction of the specific hotspot; and scaling (e.g., multiplying) the area fraction by a density coefficient associated with the bone region of interest to which the specific hotspot was assigned [e.g., taking into account the weight and / or density of the bone in the corresponding bone region of interest (e.g., where the density coefficient is the weight fraction of the corresponding bone region of interest relative to the total bone (e.g., in a normal person)], thereby calculating the bone accumulation factor of the specific hotspot.

[0133] In some embodiments, at least a portion of the hotspots in the first subset are assigned to the skeletal region of interest, which is a member of a group selected from the pelvic region (e.g., corresponding to the subject's pelvis), the lumbar region (e.g., corresponding to the subject's lumbar spine), and the sacral region (e.g., corresponding to the subject's sacrum).

[0134] In some embodiments, the one or more zone-dependent correction factors include a sacral zone correction factor, which is associated with the sacral zone and used to adjust the bone involvement factor identified (e.g., processed) as a hotspot located in the sacral zone, and wherein the sacral zone correction factor has a value less than 1 (e.g., less than 0.5).

[0135] In some embodiments, the one or more region-dependent correction factors include one or more pairs of correction factors, each pair of correction factors being associated with a specific bone region of interest and including a first member and a second member (in the pair), wherein: the first member of the pair is an anterior image correction factor and is used to adjust the bone accumulation factor calculated for hotspots detected in the annotated anterior bone scan images in the annotated image set, and the second member of the pair is a posterior image correction factor and is used to adjust the bone accumulation factor calculated for hotspots detected in the annotated posterior bone scan images in the annotated image set.

[0136] In some embodiments, the instructions at step (c) cause the processor (e.g., iteratively) to: identify healthy tissue regions in the images of the bone scan image set that are determined not to contain any hot spots (e.g., local areas with relatively high intensity); calculate a normalization factor such that the product of the normalization factor and the average intensity of the identified healthy tissue regions is a predefined intensity level; and normalize the images in the bone scan image set by the normalization factor.

[0137] In some embodiments, the instructions further cause the processor to: (g) calculate one or more risk index values ​​for the subject based at least in part on a calculated fraction (e.g., area fraction) of the subject’s bones being occupied by the initial hotspot set [e.g., where the calculated fraction is the ratio of the total area of ​​the initial hotspot set to the total area of ​​all identified skeletal regions].

[0138] In some embodiments, the instructions cause the processor to: (h) select a first subset (e.g., up to all) of the initial hotspot set based at least in part on a transfer probability value [e.g., determine whether a particular hotspot is included in the subset based on a transfer probability value calculated for a particular hotspot in the initial hotspot set exceeding a threshold]; and (i) cause to render a graphical representation of the first subset [e.g., visual indications (e.g., points, boundaries) of hotspots overlaid on one or more members of the bone scan image set and / or annotated image set; e.g., a table listing the identified hotspots and additional information (e.g., location; e.g., probability value) for each hotspot] for display within a graphical user interface (GUI) (e.g., a cloud-based GUI).

[0139] In some embodiments, the instructions cause the processor to: (j) calculate one or more risk index values ​​for the subject based at least in part on a calculated score (e.g., area score) of the subject’s bones being occupied by a first subset of hotspots [e.g., where the calculated score is the total area of ​​the initial set of hotspots divided by the total area of ​​all identified skeletal regions].

[0140] In some embodiments, the instructions cause the processor to: (k) receive a user's selection of a second subset of the initial hotspot set via the GUI; and (l) calculate one or more risk index values ​​for the subject based at least in part on a calculated score (e.g., area score) of the subject's bones occupied by the second subset of hotspots [e.g., where the calculated score is the total area of ​​the second subset of hotspots divided by the total area of ​​all identified skeletal regions].

[0141] In some embodiments, at least one of the risk index values ​​indicates the risk of the subject developing and / or acquiring metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancers).

[0142] In some embodiments, the metastatic cancer is metastatic prostate cancer.

[0143] In some embodiments, at least one of the risk index values ​​indicates that the subject has a specific state of metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancers).

[0144] In some embodiments, the system is a cloud-based system. In some embodiments, the processor is a cloud-based system processor.

[0145] In some embodiments, the GUI is part of a general image archiving and communication system (PACS) (e.g., and clinical applications in oncology, including lesion marking and quantitative analysis).

[0146] In some embodiments, the agent (e.g., a radiopharmaceutical) includes technetium-99m methylene diphosphonate ( 99m Tc-MDP).

[0147] In another aspect, the present invention relates to a computer-aided image analysis apparatus [e.g., a computer-aided detection (CADe) apparatus; e.g., a computer-aided diagnostic (CADx) apparatus] that includes any of the aspects and embodiments described herein (for example, in paragraphs

[0083] to

[00145] ).

[0148] In some embodiments, the device is programmed for use by trained healthcare professionals and / or researchers [e.g., to receive, transmit, store, display, manipulate, quantify, and report digital medical images acquired using nuclear medicine imaging; for example, where the device provides a Picture Archiving and Communication System (PACS) tool for oncology and / or clinical applications, including lesion marking and quantitative analysis].

[0149] In some embodiments, the device is programmed to analyze bone scan images to assess and / or detect metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancers).

[0150] In some embodiments, the device is programmed to analyze bone scan images to assess and / or detect prostate cancer.

[0151] In some embodiments, the device includes a label indicating that the device is intended for use by trained healthcare professionals and / or researchers [e.g., to receive, transmit, store, display, manipulate, quantify, and report digital medical images acquired using nuclear medicine imaging; e.g., where the device provides a Picture Archiving and Communication System (PACS) tool for oncology and / or clinical applications, including lesion marking and quantitative analysis].

[0152] In some embodiments, the markings further specify that the device is intended for analyzing bone scan images to assess and / or detect metastatic cancer (e.g., metastatic prostate cancer, metastatic breast cancer, metastatic lung cancer, and other metastatic bone cancers).

[0153] In some embodiments, the markings further specify that the device is intended for use in analyzing bone scan images to assess and / or detect prostate cancer.

[0154] The embodiments described with respect to one aspect of the invention may be applied to another aspect of the invention (for example, features of an embodiment described with respect to one independent claim (e.g., a method claim) are considered to be applicable to other embodiments of other independent claims, such as system claims, and vice versa). Attached Figure Description

[0155] The foregoing and other objects, aspects, features, and advantages of the invention will become clearer and better understood from the following description taken in conjunction with the accompanying drawings, in which:

[0156] Figure 1 This is a block flowchart illustrating the quality control and reporting workflow for generating BSI reports according to an illustrative embodiment.

[0157] Figure 2 This is a screenshot of a graphical user interface (GUI) for selecting patient data for review, according to an illustrative embodiment. Figure 1 The software-based implementation of the quality control and reporting workflows shown in the document is used in conjunction with this implementation.

[0158] Figure 3 These are screenshots of a graphical user interface (GUI) for reviewing patient information according to an illustrative embodiment, the graphical user interface being...Figure 1 The software-based implementation of the quality control and reporting workflows shown in the document is used in conjunction with this implementation.

[0159] Figure 4 These are screenshots of a graphical user interface (GUI) for reviewing patient image data and editing hotspot selection according to an illustrative embodiment, the graphical user interface being... Figure 1 The software-based implementation of the quality control and reporting workflows shown in the document is used in conjunction with this implementation.

[0160] Figure 5 According to the illustrative embodiments Figure 1 The screenshot shows an automated report generated by the user following a software-based implementation of the quality control and reporting workflow.

[0161] Figure 6 This is a block flowchart illustrating the process of processing whole-body bone scan images and determining bone scan index (BSI) values ​​according to an illustrative embodiment.

[0162] Figure 7 It is a set of whole-body bone scan images illustrating a skeletal atlas according to an illustrative embodiment, the set of whole-body bone scan images being overlaid on anterior and posterior bone scan images for bone segmentation.

[0163] Figure 8 This is a schematic diagram illustrating the construction of a skeletal atlas from multiple patient images according to an illustrative embodiment.

[0164] Figure 9A This is a screenshot of a GUI window displaying a list of patients presented in an approval device according to an illustrative embodiment.

[0165] Figure 9B This is a screenshot of a GUI window displaying a list of patients presented in the novel device according to an illustrative embodiment.

[0166] Figure 10A This is a screenshot of a GUI window used to display and review bone scan images and calculated BSI values, according to an illustrative embodiment.

[0167] Figure 10B This is a screenshot of a GUI window used to display and review bone scan images and calculated BSI values, according to an illustrative embodiment.

[0168] Figure 10C This is a screenshot showing a portion of the color map options for displaying a bone scan image in a GUI window according to an illustrative embodiment.

[0169] Figure 10DThis is a screenshot showing a portion of the color map options for displaying a bone scan image in a GUI window according to an illustrative embodiment.

[0170] Figure 11A This is a screenshot of a portion of a GUI window used to view bone scan images according to an illustrative embodiment.

[0171] Figure 11B This is a screenshot of a portion of a GUI window used to view bone scan images according to an illustrative embodiment.

[0172] Figure 11C This is a screenshot illustrating a portion of the zoom feature in a GUI window used to view bone scan images according to an illustrative embodiment.

[0173] Figure 11D This is a screenshot illustrating a portion of the zoom feature in a GUI window used to view bone scan images according to an illustrative embodiment.

[0174] Figure 12A This is a screenshot of a GUI window that displays bone scan images using an intensity window, which only covers a limited range of intensity values.

[0175] Figure 12B This is a screenshot of a GUI window displaying a bone scan image using an intensity window, according to an illustrative embodiment, the intensity window ranging up to the maximum intensity value.

[0176] Figure 12C yes Figure 12A A screenshot of a portion of the GUI, showing graphical controls for adjusting the intensity window threshold.

[0177] Figure 12D yes Figure 12B A screenshot of a portion of the GUI, showing another graphical control used to adjust the intensity window threshold.

[0178] Figure 12E This is a screenshot of the GUI that displays anterior and posterior images of a collection of bone scan images, with each image displayed using a separate intensity window.

[0179] Figure 12F This is a screenshot of the GUI that displays the anterior and posterior images of a collection of bone scan images, with the same intensity window used for both images.

[0180] Figure 13A This is a screenshot of a portion of a GUI that illustrates local intensity values, displayed at the location of the mouse pointer, according to an illustrative embodiment.

[0181] Figure 13BThis is a screenshot of a GUI showing local intensity values ​​according to an illustrative embodiment, the local intensity values ​​being located at the position of the mouse pointer displayed in the corner (lower left corner) of the GUI.

[0182] Figure 14A These are screenshots of GUIs showing bone scan images from different studies, displayed according to illustrative embodiments and selectable via different GUI tabs.

[0183] Figure 14B These are screenshots of a GUI showing bone scan images from different studies, displayed side-by-side simultaneously, according to an illustrative embodiment.

[0184] Figure 14C The screenshots are of a GUI showing a front and rear image according to an illustrative embodiment, the front and rear images being displayed and selectable via different GUI tabs.

[0185] Figure 14D The screenshot shows a GUI of a front and rear image according to an illustrative embodiment, which is displayed side by side and simultaneously, wherein the visibility of the various images can be selected by checkboxes in the GUI.

[0186] Figure 15A This is a screenshot of a GUI showing the display of total image intensity and total bone intensity according to an illustrative embodiment.

[0187] Figure 15B This is a screenshot of the GUI showing the display of total image intensity according to an illustrative embodiment.

[0188] Figure 16A The screenshot shows a GUI of an identified hotspot, as shown in the illustrative embodiment, which is displayed as a highlighted area overlaid on a bone scan image.

[0189] Figure 16B The screenshot shows a GUI of an identified hotspot, as shown in the illustrative embodiment, which is displayed as a highlighted area overlaid on a bone scan image.

[0190] Figure 17A This is a screenshot of a GUI showing a hotspot table according to an illustrative embodiment, the hotspot table listing hotspots identified within a set of bone scan images.

[0191] Figure 17B The screenshot is a screenshot of the GUI according to an illustrative embodiment, showing the identified hotspots and illustrating the updated BSI value after automatically detected hotspots are excluded based on user input via the GUI.

[0192] Figure 17C The screenshot is a screenshot of the GUI according to an illustrative embodiment, showing the identified hotspots and illustrating the updated BSI value after the user inputs via the GUI to include previously excluded automatically detected hotspots.

[0193] Figure 18A These are screenshots illustrating a GUI for a pop-up graphical control used to include and / or exclude automatically identified hotspots, according to an illustrative embodiment.

[0194] Figure 18B The screenshots illustrate a GUI that allows users to select which hotspots to include and / or exclude, based on an illustrative embodiment.

[0195] Figure 18C This is a screenshot of the GUI of a window that prompts the user to follow a quality control workflow before generating a report, according to an illustrative embodiment.

[0196] Figure 19A These are screenshots of a GUI showing BSI values ​​calculated for different studies and hotspots contained in a table, according to an illustrative embodiment.

[0197] Figure 19B The screenshot is a GUI according to an illustrative embodiment, which displays calculated BSI values ​​as a title above the window, thereby showing bone scan images from different studies.

[0198] Figure 20A This is a screenshot of a GUI according to an illustrative embodiment, showing a graph of calculated BSI values ​​changing over time.

[0199] Figure 20B This is a screenshot of a GUI according to an illustrative embodiment, showing a graph of calculated BSI values ​​changing over time.

[0200] Figure 21 The image is a screenshot of a GUI according to an illustrative embodiment, which provides a table listing the number of hotspots in a particular anatomical region, the table being used to calculate BSI values ​​in different studies.

[0201] Figure 22A The image is a screenshot of an automatically generated report based on an illustrative embodiment, the report being based on a method for automating the analysis of bone scan images and calculating BSI values.

[0202] Figure 22B The image is a screenshot of an automatically generated report based on an illustrative embodiment, the report being based on a method for automating the analysis of bone scan images and calculating BSI values.

[0203] Figure 23 It is a collection of bone scan images overlaid with atlas-based skeletal segments according to an illustrative embodiment, which compares the use of a finite atlas (left) with a full-length atlas (right).

[0204] Figure 24 This is a block flowchart of an exemplary process for improving image analysis by segmenting and enhancing hotspot detection in the humeral and / or femoral regions according to an illustrative embodiment.

[0205] Figure 25A This is a graph illustrating how the global threshold scaling factor changes as disease burden measures, based on an illustrative embodiment.

[0206] Figure 25B This is a block flowchart illustrating an exemplary process of improving image analysis using a global threshold scaling method according to an illustrative embodiment.

[0207] Figure 26 This is a block flowchart illustrating an instance of a process for pre-selecting a first subset of hotspots using global hotspot features, according to an illustrative embodiment.

[0208] Figure 27 It is a set of two bone scan images according to an illustrative embodiment, the bone scan images showing portions of the sacrum, pelvis and lumbar region.

[0209] Figure 28 This is a block flowchart illustrating an exemplary process for calculating a risk index value based on a skeletal involvement factor and a region dependency correction factor, according to an illustrative embodiment, wherein the region dependency correction factor takes into account potential errors in hotspot localization.

[0210] Figure 29 It is a graph that compares BSI values ​​calculated by an automated, software-based method according to the aspects and embodiments described herein with known analytical standards.

[0211] Figure 30 It is a chart that shows the reproducibility of automated BSI values ​​obtained from multiple scans of 50 simulated prostheses.

[0212] Figure 31 It is a chart that shows the reproducibility of automated BSI values ​​obtained from multiple scans of 35 metastatic patients.

[0213] Figure 32A This is a graph illustrating the performance of a current version of the image analysis software employing an improved embodiment as described herein.

[0214] Figure 32B This is a graph illustrating the performance of a previous version of the image analysis software according to an illustrative embodiment.

[0215] Figure 33A It is a diagram illustrating an exemplary cloud platform architecture.

[0216] Figure 33B It is a diagram illustrating an instanced microservice architecture.

[0217] Figure 34 This is a block diagram of an exemplary cloud computing environment used in some embodiments.

[0218] Figure 35 This is a block diagram of an instance computing device and an instance mobile computing device used in some embodiments.

[0219] The features and advantages of the invention will become clearer when read in conjunction with the detailed description set forth below, in conjunction with the accompanying drawings, throughout which similar reference characters identify corresponding elements. In the drawings, similar element symbols generally indicate identical, functionally similar, and / or structurally similar elements. Detailed Implementation

[0220] It is anticipated that the systems, apparatus, methods, and processes of the claimed invention encompass variations and modifications made using information from the embodiments described herein. Changes and / or modifications to the systems, apparatus, methods, and processes described herein can be made by those skilled in the art.

[0221] Throughout this description, where articles, apparatuses, and systems are described as having, containing, or including specific components, or processes and methods are described as having, containing, or including specific steps, it is anticipated that there are additional articles, apparatuses, and systems of the invention that are substantially composed of or comprised of the described components, and processes and methods of the invention that are substantially composed of or comprised of the described processing steps.

[0222] It should be understood that the order of the steps or the order in which specific actions are performed is not important, as long as the invention remains operable. Furthermore, two or more steps or actions can be performed simultaneously.

[0223] For example, any disclosure mentioned in the Background section herein is not an admission that such disclosure constitutes prior art with respect to any of the claims presented herein. The Background section is presented for clarity and is not intended to be a description of prior art with respect to any claim.

[0224] Headings are provided for the convenience of the reader – the presence and / or placement of headings are not intended to limit the scope of the subject matter described herein.

[0225] In this application, unless otherwise stated, the use of "or" means "and / or". As used in this application, the term "comprise" and variations thereof (e.g., comprising / comprises) are not intended to exclude other additives, components, integers, or steps. As used in this application, the terms "about" and "approximately" are used equivalently. Any numerical value used in this application, with or without the "about / approximately" modifier, is intended to cover any normal fluctuation known to a person of ordinary skill in the relevant art. In some embodiments, the terms "about" or "about" refer to a range of values ​​(greater or less than) in any direction within 25%, 20%, 19%, 18%, 17%, 16%, 15%, 14%, 13%, 12%, 11%, 10%, 9%, 8%, 7%, 6%, 5%, 4%, 3%, 2%, 1%, or less of the reference value, unless otherwise stated or otherwise understood from the context (except where this number would exceed 100% of the possible value).

[0226] The articles “a (a and an)” are used herein to refer to one or more (i.e., at least one) grammatical objects of the article. For example, “an element” means one or more elements. Therefore, in this specification and the appended claims, the singular forms “a (a / an)” and “described” include multiple indicators unless the context clearly indicates otherwise. Thus, for example, referring to a pharmaceutical composition including “a pharmaceutical agent” includes referring to two or more pharmaceutical agents.

[0227] The systems and methods described herein relate to improved computer-aided display and analysis of nuclear medicine images. Specifically, in some embodiments, the systems and methods described herein provide improvements to several image processing steps for automating the analysis of bone scan images to assess a patient's cancer status. For example, improved methods are provided for image segmentation, hotspot detection, automated classification of hotspots to represent metastasis, and calculation of risk indices such as Bone Scanning Index (BSI) values. Automated BSI calculation techniques are described in detail in U.S. Patent Application No. 15 / 282,422, filed September 30, 2016; U.S. Patent Application No. 8,855,387, filed October 7, 2014 (which is refiled under U.S. Patent Application No. 15 / 282,422), filed October 26, 2017; and PCT Application No. PCT / US17 / 58418, filed October 26, 2017, the contents of each of these applications are hereby incorporated by reference in their entirety. Also incorporated in this case is PCT application PCT / US2017 / 058418, filed on October 26, 2017, which describes a cloud-based platform that can be used as a platform to provide image analysis and BSI calculation tools according to the methods described herein.

[0228] Specifically, in some embodiments, bone scan images are acquired after a drug (e.g., a radiopharmaceutical) has been administered to the subject. The administered drug accumulates in cancerous bone lesions due to the physical properties of the underlying tissue (e.g., increased vascularity, abnormal bone formation) or because of the identification (by the drug) of specific biomolecules selectively expressed or overexpressed in tumor form, such as prostate-specific membrane antigen (PSMA). The drug includes a radionuclide that emits nuclear radiation, which can be detected and thus used to image the spatial distribution of the drug within the subject.

[0229] For example, in some embodiments, a gamma camera is used to acquire bone scan images in a two-dimensional scan format. For instance, two images – an anterior image and a posterior image – are acquired to form a set of bone scan images. Body areas with accumulated high concentrations of the drug are represented as high-intensity areas (i.e., bright spots) in the bone scan images. The drug can accumulate in, for example, cancerous bone lesions as described above, as well as in other areas (e.g., the subject's bladder).

[0230] To accurately identify regions representing lesions in bone scan images and generate a quantitative estimate of tumor burden, a series of image processing steps are performed. Specifically, the bone scan images are segmented to identify regions corresponding to individual bones in the subject's skeleton, thereby forming an annotated image set. High-intensity regions are identified within the bone regions relative to their surroundings, and these high-intensity regions are compared to a threshold to detect an initial set of hotspots. Features of the initial hotspots (e.g., hotspot size (e.g., area), hotspot shape (e.g., described by various metrics such as radius, eccentricity, etc.), and / or measures of hotspot intensity (e.g., peak intensity, average intensity, integral intensity, etc.) are extracted and used to determine a metastasis probability value for each hotspot, the metastasis probability value representing the likelihood that the hotspot represents metastasis. For example, in some embodiments, an artificial neural network (ANN) that receives a set of hotspot features as input and outputs a metastasis probability value for each hotspot is used to compute the metastasis probability value.

[0231] Metastasis probability values ​​can be used to automatically filter the initial set of hotspots to determine a subset for calculating a risk index indicating the risk of a subject developing and / or acquiring metastatic cancer. By filtering hotspots in this way, the risk index calculation includes only those hotspots determined to have a high probability of metastasis. In some embodiments, a graphical representation of the hotspots and / or their probability values ​​is rendered and displayed to the user (e.g., as overlay markers and / or an information table on an annotated image), allowing the user to select a subset of hotspots for use in calculating the risk index. This allows the user to enhance the automated selection of hotspots to utilize their input in calculating the risk index.

[0232] The method described herein incorporates several improvements to the aforementioned image processing steps, thereby providing improved accuracy in lesion detection and risk index calculation. For example, the invention includes an improved segmentation method for identifying the entire (e.g., more than three-quarters of the length) humerus and / or the entire (e.g., more than three-quarters of the length) femur region. Previously, only limited fractions of the femur and humerus could be identified. Segmenting larger (e.g., entire) portions of these bones allows for the identification of lesions located further away in the extremities of the subject's arms and legs. To account for reduced drug uptake in the extremities, the method described herein also utilizes a region-dependent threshold in the hotspot detection step. Different bone regions have different region-dependent thresholds, and lower values ​​are found in the femur and humerus regions to improve detection sensitivity in these regions.

[0233] In another improved approach, the system and method described herein may use a global threshold scaling technique to detect hotspots. This method first identifies an initial set of potential hotspots by using multiple preliminary region-dependent thresholds. A global threshold scaling factor is then calculated using the potential hotspot set based on the area fraction of the subject's bone occupied by the potential hotspot set. The preliminary thresholds are then adjusted using the global threshold scaling factor, and the adjusted thresholds are used to detect the initial set of hotspots. This method is found to ultimately improve the linearity of the risk index calculated using the initial set of hotspots, particularly for high levels of disease burden—for example, in cases where the subject has numerous lesions and a large portion of their bone is occupied by hotspots.

[0234] This invention also includes improvements to hotspot selection and risk index calculation. For example, the method described herein may use a hotspot pre-selection technique that filters hotspots not only based on a calculated transfer probability value but also on global hotspot features that measure the properties of the entire initial set of hotspots (e.g., the total number of hotspots in the set). Other examples of global hotspot features include: other measurements of the total number of hotspots, such as the average number of hotspots per area; measurements of the total hotspot intensity, such as peak or average hotspot intensity; and measurements of the total hotspot size, such as the total area of ​​hotspots, average hotspot size, etc. This allows the processing method to utilize clinical experience that suggests hotspot selection depends on the rest of the image. Specifically, if many other hotspots are present, the probability of a hotspot being selected is higher, and if only one hotspot is present, the probability of it being selected is lower. Selecting or filtering hotspots based solely on individual transfer probability values ​​can therefore lead to an underestimation of the risk index value calculated for subjects with many hotspots. Incorporating the global features described herein can improve performance for patients with many hotspots.

[0235] Finally, the systems and methods described herein also provide improvements to methods for calculating risk index values ​​(e.g., bone scan index (BSI)) based on bone involvement. For example, BSI is a risk index value that provides an estimate of the fraction of a subject's total bone mass occupied by cancerous lesions. Calculating BSI involves calculating a bone involvement factor for each specific hot spot based on the ratio of the area of ​​the specific hot spot to the area of ​​the bone region in which the specific hot spot is located. A scaled version (e.g., to convert the area ratio into relative mass) is summed to calculate the subject's BSI value. However, difficulties in correctly locating the specific bone region in which a specific hot spot is located can lead to errors in the BSI value. Bone scan images are two-dimensional images, but the subject's underlying skeleton is a three-dimensional structure. Therefore, hot spots may be incorrectly identified as being located in one region when, in fact, the hot spot represents a lesion located in a different bone. This is a particular challenge for the sacrum and adjacent pelvic and lumbar regions. To address this challenge, the present invention includes a modified risk index calculation method that uses a region-dependent correction factor to scale the bone involvement factor to account for potential errors in hot spot localization. This method improves the accuracy of BSI calculation and limits the sensitivity to hotspot location.

[0236] Therefore, the systems and methods described herein include several improved image analysis techniques for lesion identification and quantification. These methods improve the accuracy and robustness of bone scan image analysis. As described herein, the methods can be used as part of a cloud-based system that facilitates the review and reporting of patient data and allows for improved disease detection, treatment, and monitoring.

[0237] A. Nuclear medicine images

[0238] Nuclear imaging modalities (e.g., bone scan imaging, positron emission tomography (PET) imaging, and single-photon emission tomography (SPECT) imaging) are used to obtain nuclear medicine images.

[0239] As used herein, “image” (for example, a 3D image of a mammal) includes any visual representation, such as a photograph, video frame, streaming video, and any electronic, digital, or mathematical analogue of a photograph, video frame, or streaming video. In some embodiments, any device described herein includes a display for displaying an image or any other result produced by a processor. In some embodiments, any method described herein includes the step of displaying an image or any other result produced by the method. As used herein, “3D” or “three-dimensional” when referring to “image” means conveying information about three dimensions. A 3D image may be rendered as a dataset in three-dimensional form and / or may be displayed as a two-dimensional representation of a collection, or displayed as a three-dimensional representation.

[0240] In some embodiments, nuclear medicine images use imaging agents comprising radiopharmaceuticals. Nuclear medicine images are obtained after administration of radiopharmaceuticals to a patient (e.g., a test subject) and provide information about the distribution of the radiopharmaceuticals within the patient's body. Radiopharmaceuticals are compounds comprising radionuclides.

[0241] As used herein, “administered” means the introduction of a substance (e.g., an imaging agent) into a subject’s body. Generally, any route of administration may be used, including, for example, non-gut (e.g., intravenous), oral, topical, subcutaneous, peritoneal, intra-arterial, inhalation, vaginal, rectal, nasal, into the cerebrospinal fluid, or instillation into a body compartment.

[0242] As used herein, “radionoid” means a portion of a radioactive isotope comprising at least one element. Illustratively suitable radionoids include (but are not limited to) the radionoids described herein. In some embodiments, the radionoid is a radionoid used in positron emission tomography (PET). In some embodiments, the radionoid is a radionoid used in single-photon emission computational tomography (SPECT). In some embodiments, a non-limiting list of radionoids 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、 105Rh、 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 192 Ir.

[0243] As used herein, the term "radiomedicine" refers to a compound comprising a radionuclide. In some embodiments, radiomedicines are used for diagnostic and / or therapeutic purposes. In some embodiments, a radiomedicine comprises: a small molecule labeled with one or more radionuclides, an antibody labeled with one or more radionuclides, and an antigen-binding portion of the antibody labeled with one or more radionuclides.

[0244] Nuclear medicine imaging (e.g., PET scans; SPECT scans; whole-body bone scans; synthetic PET-CT images; synthetic SPECT-CT images) detects radiation emitted from radionuclides of radiopharmaceuticals to form images. The distribution of specific radiopharmaceuticals within a patient can be determined through biological mechanisms (e.g., blood flow or perfusion) and through specific enzyme or receptor binding reactions. Different radiopharmaceuticals can be designed to utilize different biological mechanisms and / or specific enzyme or receptor binding reactions, and thus, different radiopharmaceuticals, when administered to a patient, will selectively concentrate in specific types of tissues and / or regions within the patient's body. Regions within the patient with higher concentrations of radiopharmaceuticals than other regions will emit greater amounts of radiation, causing these regions to appear brighter in nuclear medicine images. Therefore, intensity variations within nuclear medicine images can be used to map the distribution of radiopharmaceuticals within the 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.

[0245] For example, after being administered to patients, technetium 99m methylene diphosphonate ( 99m Tc MDP selectively accumulates in the skeletal region of the patient, specifically in sites of abnormal bone formation associated with malignant bone lesions. The selective concentration of radiopharmaceuticals at these sites creates identifiable hotspots—high-intensity regionalized areas in nuclear medicine imaging. Therefore, the presence of malignant bone lesions associated with metastatic prostate cancer can be inferred by identifying these hotspots within a patient's whole-body scan. As described below, this can be based on the administration of radiopharmaceuticals to the patient... 99mAutomated analysis of intensity changes in whole-body scans obtained after Tc MDP is used to calculate risk indices associated with overall patient survival and other prognostic measures such as disease status, progression, and treatment effectiveness. In some embodiments, it can also be combined with... 99m Other radiopharmaceuticals are used in a similar manner to Tc MDP.

[0246] In some embodiments, the specific radiopharmaceuticals used depend on the specific nuclear medicine imaging modality used. For example, sodium 18F fluoride (NaF) also accumulates in bone lesions, similar to... 99m Tc MDP, but can be used in conjunction with PET imaging. In some embodiments, PET imaging may also utilize the radioactive form of vitamin choline, which is readily absorbed by prostate cancer cells.

[0247] In some embodiments, radiopharmaceuticals that selectively bind to specific proteins or receptors of interest—specifically, radiopharmaceuticals whose expression is increased in cancerous tissue—may be used. Such proteins or receptors of interest include, but are not limited to, tumor antigens, such as CEA, which is expressed in rectal cancer; Her2 / neu, which is expressed in a variety of cancers; BRCA1 and BRCA2, which are expressed in breast and ovarian cancers; and TRP-1 and -2, which are expressed in melanoma.

[0248] For example, prostate-specific membrane antigen (PSMA) is upregulated in prostate cancer (including metastatic disease). Almost all prostate cancers express PSMA, and PSMA expression is further increased in poorly differentiated metastatic and hormone-resistant cancers. Therefore, radiopharmaceuticals corresponding to PSMA-binding agents labeled with one or more radionuclides (e.g., compounds with high affinity for PSMA) can be used to obtain nuclear medicine images of a patient from which the presence and / or status of prostate cancer in various regions of the patient (e.g., including but not limited to skeletal regions) can be assessed. In some embodiments, nuclear medicine images obtained using PSMA-binding agents are used to identify the presence of cancerous tissue within the prostate when the disease is in a localized state. In some embodiments, nuclear medicine images obtained using radiopharmaceuticals including PSMA-binding agents are used to identify the presence of cancerous tissue in various regions that include not only the prostate but also other relevant organ and tissue regions, such as the lungs, lymph nodes, and bone, when the disease is metastatic.

[0249] Specifically, when administered to patients, the radiolabeled PSMA-binding agent selectively accumulates within cancerous tissue based on its affinity for PSMA. This is consistent with the above discussion regarding... 99mSimilar to the approach described in Tc MDP, the selective concentration of radiolabeled PSMA-binding agents at specific sites within a patient's body creates detectable hotspots in nuclear medicine imaging. When PSMA-binding agents concentrate in various cancerous tissues and regions expressing PSMA, localized cancers within the patient's prostate and / or metastatic cancers in various regions of the patient's body can be detected and evaluated. Following administration of PSMA-binding radiopharmaceuticals to patients, risk indices related to overall survival and other prognostic measures such as disease status, progression, and treatment efficacy can be calculated based on automated analysis of intensity changes obtained from nuclear medicine imaging.

[0250] Various radionuclide-labeled PSMA binders can be used as radiopharmaceutical imaging agents in nuclear medicine imaging to detect and evaluate prostate cancer. In some embodiments, the specific radionuclide-labeled PSMA binder used depends on several factors, such as the specific imaging modality (e.g., PET; e.g., SPECT) and the specific area (e.g., organ) of the patient to be imaged. For example, some radionuclide-labeled PSMA binders are suitable for PET imaging, while others are suitable for SPECT imaging. For example, some radionuclide-labeled PSMA binders facilitate imaging of the patient's prostate and are primarily used when the disease is localized, while others facilitate imaging of organs and areas throughout the patient's body and are used to evaluate metastatic prostate cancer.

[0251] Various versions of PSMA binding agents and their radiolabeled forms are described in U.S. Patent Nos. 8,778,305, 8,211,401, and 8,962,799, the entire contents of each of which are incorporated herein by reference. Several versions of PSMA binding agents and their radiolabeled forms are also described in PCT application PCT / US2017 / 058418, filed October 26, 2017, the entire contents of which are incorporated herein by reference.

[0252] B. Bone scanning imaging device for image analysis

[0253] In some embodiments, the computer-aided image analysis apparatus described herein is intended for use by trained healthcare professionals and researchers to receive, transmit, store, display, manipulate, quantify, and report digital medical images acquired using nuclear medicine (NM) imaging. In some embodiments, such apparatus provides universal picture archiving and communication system (PACS) tools and clinical applications in oncology, including lesion marking and quantitative analysis.

[0254] C. Bone scan image analysis and bone scan index calculation

[0255] Bone scintillation (also known as bone scan imaging) is an imaging modality widely used to assess the burden of skeletal disease. Current standards for assessing disease progression based on bone scan images are based on semi-quantitative modified prostate cancer working group 2 and prostate cancer working group 3 (PCWG) guidelines. These guidelines are defined based on the presence of new lesions as interpreted by a trained reader: (i) the discovery of two new lesions followed by two additional lesions at the first and second follow-up scans compared to the pretreatment scan (known as the 2+2 rule); or (ii) the discovery of two new, newly confirmed lesions relative to the subsequent first follow-up scan. However, this semi-quantitative assessment method, which counts the number of lesions, has several limitations. Specifically, the assessment is subject to human variability, is constrained to assessing disease progression, and cannot accurately assess the burden of fusion / metastatic disease, for example, associated with metastatic prostate cancer.

[0256] Therefore, the need for automated and quantitative assessment of bone scans remains far from being met. The Automated Bone Scan Index (BSI), developed by EXINIDiagnostics AB and Lund Sweden, is a fully quantitative assessment of skeletal disease in patients based on bone scans, evaluated as a fraction of total bone mass. The BSI has undergone rigorous pre-analytical and analytical validation as an objective measure of quantitative changes in bone scans related to disease burden. In a recent phase 3 prospective study, BSI assessment has been shown to pose a risk of stratification in patients with metastatic prostate cancer.

[0257] The systems and methods described herein relate to improved computer-aided approaches for analyzing bone scan images and calculating BSI in an automated and semi-automated user-guided manner. A GUI tool is also described herein that facilitates the review and automated analysis of bone scan images for determining a user's BSI values.

[0258] D. Device Description of an Instantaneous aBSI Platform

[0259] In some embodiments, the systems and methods described herein can be implemented as cloud-based platforms for automated and semi-automated image analysis to detect and assess a patient's cancer status. The exemplary apparatus described herein is an automated BSI device (aBSI), a cloud-based software platform with a web interface, where users can upload bone scan image data in the form of specific image files (e.g., DICOM files). The software conforms to the Digital Medical Imaging and Communications (DICOM 3) standard.

[0260] In some embodiments, the apparatus (e.g., computer-aided image analysis tools) according to the systems and methods described herein is programmed for a specific user (typically a healthcare professional who uses the software to view patient images and analyze the results). The user operates the service in a web browser (e.g., Google Chrome) on a computer running an operating system (e.g., 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. The software can be operated using keyboard and mouse controls.

[0261] Multiple scans for each patient can be uploaded, and the system provides individual image-based automated analysis for each patient. Physicians review the automated analyses, guided by quality control and reporting workflows. If the quality control of the automated assessment is approved, a report can be created and signed. The service can be configured to comply with HIPAA and 21 CFR Part 11.

[0262] i. Service Access

[0263] In some embodiments, access to software tools according to the systems and methods described herein is limited and protected by security measures. For example, access to cloud-based implementations of the systems and methods described herein, aBSI, is protected by multi-factor authentication in the form of a username, password, and verification code, which are sent in text message form to a telephone number associated with the account.

[0264] ii. System Requirements

[0265] In some embodiments, the software mandates one or more of the following:

[0266] • Computers with Windows or OS X and internet access,

[0267] Chrome browser,

[0268] • A personal mobile phone is acceptable (for multi-factor authentication only).

[0269] In some embodiments, the user mandates the inclusion of one or more of the following:

[0270] Chrome browser

[0271] a. At least version 54

[0272] b. JavaScript must be allowed.

[0273] c. HTML5 is required

[0274] d. Requires writing to local storage and session storage

[0275] • The monitor resolution must be at least 1280×960

[0276] iii. Image Requirements

[0277] In some embodiments, the software mandates one or more of the following:

[0278] • The image must be uncompressed and in DICOM 3 format.

[0279] • Modality (0008, 0060) must be "NM"

[0280] • Image type (0008, 0008) must be "ORIGINAL\PRIMARY\WHOLE BODY\EMISSION"

[0281] • The research date (0008, 0020) must include a valid date.

[0282] • The number of frames (0028, 0008) must be 1 or 2.

[0283] • The number of slices (0054, 0081) must be 1 or 2.

[0284] • The pixel spacing (0028, 0030) must be ≥1.8mm / pixel and ≤2.8mm / pixel.

[0285] • The shape of the image must be such that the number of rows is greater than or equal to the number of columns.

[0286] • The patient's gender (0010, 0040) must be M.

[0287] • Front and back images can be stored in the form of two different series (two files with different series instance UIDs) or a multi-frame series (one file) containing two frames.

[0288] • Image pixel data should be in the 16-bit range. Images with a pixel range of 0 to 255 (8 bits) are insufficient.

[0289] In some embodiments, the user mandates the inclusion of one or more of the following:

[0290] • The anterior and posterior images should cover at least one region from the scalp to the upper portion of the tibia and the upper portion of the forearm of each arm.

[0291] • There is no need to apply filtering or other post-processing techniques to the image.

[0292] In some embodiments, whole-body bone scintillation images are acquired in accordance with relevant guidelines (e.g., “EANM Bone Scintillation Examination: A Procedural Guide for Tumor Imaging” and “ACR-SPR Practice Parameters for the Performance of Bone Scintillation Examination (Bone Scan)”).

[0293] iv. Workflow

[0294] Figure 1 This is a block flowchart illustrating a quality control and reporting workflow 100 for generating BSI reports, according to an illustrative embodiment. In some embodiments, the systems and methods described herein include a GUI for guiding users (e.g., healthcare professionals) through the review and analysis of patient image data to calculate automated BSI values ​​and generate reports. Workflow 100 allows a user to select 102 and upload 104 an image file of patient data, which is then analyzed by software and reviewed by the user.

[0295] It can present the first GUI window to the user, for example Figure 2 The window 200 shown in the image allows the user to select a specific patient from a list for review / analysis. Figure 2 After selecting the row corresponding to a specific patient from the list, it will display Figure 3 The window displayed shows the patient information in the new window and / or the updated window 300.

[0296] Turn Figure 4 The user can then access the review page 400 of the guided analysis software. Review page 400 provides a GUI for the user (e.g., a healthcare professional, such as a physician) that allows the user to review image data and automated hotspot identification performed by the software backend to calculate the patient's automated BSI index. The automated BSI calculation technique is described in more detail herein, and prior art (without the improvements of this invention) is described in detail in U.S. Patent Application No. 15 / 282,422, filed September 30, 2016, and U.S. Patent Application No. 8,855,387, published October 7, 2014 (U.S. Patent Application No. 15 / 282,422 is a republication), and PCT Application No. PCT / US17 / 58418, filed October 26, 2017, the entire contents of each of these applications are hereby incorporated by reference.

[0297] The review page allows users to review 106 hotspots representing cancerous lesions that have been automatically identified in the image by the software. Users can use the review page GUI to edit 108 the set of regions identified as hotspots, and must verify image quality, skeletal segmentation (as illustrated in the outline depiction in the screenshot), and confirm that the identified hotspot set has been reviewed and accepted 110 to proceed with report generation. Once the user's review and quality control have been confirmed 112, a report 114 containing the final BSI calculation value (e.g., ...) can be generated. Figure 5 The report shown in the document is 500.

[0298] v. Image Processing

[0299] like Figure 6 As shown in the block flowchart, the system and methods described herein can be used in process 600 to automate the detection and pre-selection of hotspots, receive user verification of the pre-selected hotspot set, and calculate risk index values, such as, more precisely, bone scan index (BSI) values.

[0300] Specifically, in some embodiments, such as Figure 6 As shown, in segmentation step 602, the software tools described herein use an image registration algorithm to bring the patient image into a reference coordinate system. This is done by non-rigidly adjusting the atlas image relative to each patient image. By fitting the atlas image to the patient image, the patient image can be segmented into skeletal and background regions. The skeletal region can be further subdivided into smaller regions of interest, which is also referred to herein as localization. In normalization step 604, the whole-body bone scan image is normalized to provide the user with a normalized intensity range when viewing the image at different contrast and brightness levels. Hotspots are detected in hotspot detection step 606. In some embodiments, threshold determination rules are used to detect hotspots. In another step, hotspot pre-selection 608 is performed. In some embodiments, hotspot pre-selection is based on image analysis and machine learning techniques designed to pre-select significant hotspots to be included in a pre-selected hotspot set, which can be reviewed by the user in hotspot verification step 610. This pre-selection is based on a range of features of the hotspots, such as size, location, orientation, shape, and texture. In some embodiments, pre-selection is a user-friendly tool designed to reduce the number of manual clicks that users must perform (e.g., to select hotspots for calculating a risk index, such as a BSI value). In some embodiments, a mandatory hotspot verification step may follow the pre-selection step, in which the user must review and approve the pre-selected hotspots or, if necessary, manually include and / or exclude hotspots in order to create a report.

[0301] In another step 612, a specific risk index (referred to as the Bone Scan Index (BSI)) is calculated using the validated set of hotspots. In some embodiments, the Bone Scan Index (BSI) is defined as the sum of the bone coverage rates of all included hotspots. The coverage rate is an estimate of the proportion of total bone mass to the volume corresponding to the hotspot and is expressed as a percentage. The coverage rate can be calculated using the following formula, where C is an anatomical region coefficient related to bone density:

[0302]

[0303] Skeletal segmentation

[0304] In some embodiments, bone segmentation is performed using a registration method that aligns atlas images with bone scan images to be segmented. In this method, the apparatus automatically delineates bones into distinct skeletal regions by elastically fitting a manually drawn, annotated set of template images to each analyzed image set. This set of annotated template images is referred to as a skeletal atlas image set. The structure of this atlas image set is similar to any patient image—it looks like a routine bone scan and includes an anterior image and a posterior image. The atlas images provide a fixed reference during scan analysis. The atlas is manually annotated with regions of interest (bone regions) that can be transferred to new scans for accurate BSI calculation. Figure 7 The image set displays 700 instance images featuring 31 manually depicted skeletal regions. (Example image) Figure 7 As shown, similar to the bone scan image set, the image set 700 may include anterior image 702 and posterior image 704.

[0305] When analyzing bone scan images, the atlas image is elastically deformed to resemble the bone scan image. The same transformation is then applied to the atlas contour, resulting in the delineation / segmentation of each bone region of interest from the patient's bone scan image. Additional details regarding the construction of the suitable atlas are described at the end of this section.

[0306] The skeletal atlas is deformed to fit a patient scan. In some embodiments, deforming the skeletal atlas to fit a patient's bone scan image follows an iterative approach. A piecewise algorithm continues iteratively, wherein in each iteration, a vector is estimated for each pixel, describing how the pixel should be displaced to its corresponding position in the target image. In the case of individual displacement of each pixel, the displacement vectors may intersect or share the target position, which will result in holes and / or tears in the deformed image. To avoid this, a filtering method is used to smooth the vector field. The displacement is estimated by applying complex-valued filters to the atlas and the target image. The composite filter response can be represented by the amplitude and phase of each pixel. It can be shown that the local phase difference, i.e., the phase difference between pixels within short distances from each other, is proportional to the magnitude of the displacement required to align them. To obtain an estimate of the displacement direction, this process is repeated several times for different filter angles. Knowing the angle of each filter and the resulting displacement magnitude makes it possible to infer the direction of the observable maximum displacement. While this method is suitable for small displacements, it must also be able to be applied when the atlas and the target image are far apart. To achieve this, a quadratic sampling method is employed, where the algorithm is first applied to a quadratic (resized) version of the image. This method treats large displacements as local differences. The algorithm then continues to increasingly detailed (fewer quadratic samples) images to add more detail and variability to the resulting displacement field. The algorithm operates on a fixed pyramid of the quadratic image, performing a fixed number of iterations at each pyramid level with a predetermined smoothness.

[0307] Construction of the skeletal atlas. The exemplary aBSI device described in this paper relies on a single atlas image. The contour rendering algorithm is driven by structural information in both the atlas and the target image and attempts to deform the atlas image to minimize the distance between similar structures in the two images. Structures are defined by edges and ridges (lines) in the image. Therefore, the algorithm ignores global intensity differences and texture patterns. Thus, a suitable atlas image exhibits two important properties:

[0308] • It displays the same pattern of edges and ridges; and

[0309] • The (elastic) transformations required to align the images are typically minimized over the expected set of anatomical changes in the analyzed images.

[0310] In some embodiments, to meet these requirements, an atlas of images based on a database of real, normal (e.g., without metastases or other visible medical conditions) bone scan images is used. Contour rendering algorithms are used to align all the images in the database with each other. An average transformation is then calculated from all the resulting transformations. Subsequently, all images are transformed to this anatomical average to represent the average anatomy in the database. During this process, the intensity is also normalized, thereby creating a typical bone scan image suitable as an anatomical reference.Figure 8 The schematic diagram shown illustrates this concept—how to infer average anatomical structure and intensity (center image, 802) from multiple bone scan images 804a, 804b, 804c, and 804d in the database.

[0311] In some embodiments, as the number of scans included increases, the average anatomical structure for the atlas of images converges rapidly to a stable estimate. A relatively small number of scans (e.g., 30) is sufficient to create representative reference images. Furthermore, because the algorithm is driven by the main structures in the images and is insensitive to differences in shape and / or size, a single atlas can be applied to any bone scan image for bone segmentation.

[0312] Intensity normalization and hotspot detection

[0313] A challenge in reading scintillation examination images (e.g., bone scans) is that intensity levels between scans can vary due to various parameters, such as injection dose, time from injection to scan, scan time, body type, camera hardware and configuration, etc. In some embodiments, to facilitate user reading and as part of a quantization pipeline (e.g., as...), Figure 6 As shown in the diagram, the input bone scan is normalized so that the average intensity of healthy bone tissue is scaled to a predetermined reference level. At this stage of the quantization pipeline, the pixels that have been segmented into bones and belong to the bones are known. However, in order to measure the average intensity of healthy tissue and thus normalize the scan, high-intensity areas must be identified and excluded. If the image has been normalized, then this detection of hotspots in the bone is straightforward. This is a chicken-and-egg problem where hotspot detection is needed to normalize the image, and hotspot detection depends on the normalized image. Therefore, iterative methods (such as the steps listed below) can be used to address this challenge:

[0314] (1) Estimating regularization under the assumption that all bone tissue is healthy (without hot spots);

[0315] (2) In view of the current focus on standardized testing;

[0316] (3) Given the current hotspot set estimation regularization; and

[0317] (4) Iterate through steps (2) and (3) until convergence.

[0318] This iterative process converges to a stable value within 3 or 4 iterations to achieve both normalization and hotspot detection. A simple thresholding method is used to detect hotspots, where the image is filtered using a Gaussian difference bandpass filter that emphasizes cells that are high in intensity relative to their surroundings. This filtered image is then thresholded to a constant level based on the region.

[0319] In some embodiments, different thresholds are used for different skeletal regions of interest. For example, the threshold levels used in the cloud-based aBSI exemplary embodiment are 650 for the cervical spine, clavicle, ribs, lumbar spine, pelvis, sacrum, scapula, skull, thoracic spine, and sternum, and 500 for the femur and humerus.

[0320] The output of hotspot detection is a set of ROIs (Regions of Interest) representing hotspots in an image, along with normalization factors used to set initial maximum and minimum thresholds for image windowing.

[0321] Hotspot Pre-selection

[0322] A data-driven learning method based on artificial neural networks (ANNs) can be used to classify hotspots as either included or excluded from a pre-selection. The ANN can be tuned / trained based on a training database of patients, ranging from normal bone scans to bone scans with many extensive hotspots.

[0323] Each hotspot in the training database is characterized using a set of features (measurements) related to its size, location, orientation, shape, and texture. These features are first fed into the ANN during the training phase, where the parameters of the ANN are set to maximize classification performance in cross-validation studies, and then the hotspots are classified as included or excluded in the actual software.

[0324] Pre-selection training analyzes hotspots and their immediate neighbors in the image, making the classifier robust to large-scale variations in the training material. Therefore, the classifier is suitable for a wide variety of input data. However, small performance differences between cohorts are expected. To avoid the influence of localization on ANN parameters, separate networks can be constructed for different skeletal localizations. Each of these ANNs has a different set of input features. For example, symmetry features are only applicable to localizations with naturally symmetrical counterparts.

[0325] Additionally, hotspots in the training set are typically manually marked as included or excluded by medical experts trained in reading bone scans. Target markers can be validated by a second medical expert. In one exemplary approach, US and European guidelines for bone scintillation examination procedures are consistent, and the equipment used to obtain bone scans is identical in both the US and Europe. Furthermore, the guidelines used to interpret bone scans from clinical trials (e.g., the Prostate II Working Group guidelines) are global. This is based on common sense in nuclear medicine that changes in bone scan appearance caused by cancer are far more significant than, for example, the minute variations in normal bone mineral density measurements across different ethnicities. The parameters of the ANN are optimized so that the final classifier mimics the choices of medical experts. To avoid bias towards the training set, cross-validation can be used.

[0326] In some implementations, while pre-selection may save readers time, all hotspots need to be reviewed and approved by readers before a report is created.

[0327] The ANN described herein can be implemented via one or more machine learning modules. As used herein, the term "machine learning module" refers to a computer-implemented process (e.g., function) that implements one or more specific machine learning algorithms to determine one or more output values ​​for a given input (e.g., an image (e.g., a 2D image; e.g., a 3D image), a dataset, etc.). For example, a machine learning module may receive a 3D image of a subject as input (e.g., a CT image; e.g., an MRI) and, for each stereo pixel of the image, determine a value representing the probability that the stereo pixel lies within a region of the 3D image corresponding to a representation of a specific organ or tissue of the subject. In some embodiments, two or more machine learning modules may be combined and implemented as a single module and / or a single software application. In some embodiments, two or more machine learning modules may also be implemented individually, for example, as a single software application. A machine learning module can be software and / or hardware. For example, a machine learning module may be implemented entirely as software, or some functions of a CNN module may be implemented via dedicated hardware (e.g., via an application-specific integrated circuit (ASIC)).

[0328] E. Graphical user interface and image display

[0329] In some embodiments, the systems and methods described herein include a graphical user interface (GUI) for reviewing patient data and images. The GUI allows a user to review a list of patients and select patients to review and analyze their images. Figure 9A and Figure 9B Displays an instanced GUI window that provides a list of patients from which a specific patient can be selected. Once a specific patient has been selected, the user can use the GUI (e.g., ...). Figure 10A and Figure 10B The GUI shown in the image allows you to view the patient's bone scan images and review automated analyses (e.g., hotspot detection and pre-selection). Various color maps can be used to colorize the bone scan images. Figure 10C and Figure 10D Show examples of color maps that can be used.

[0330] In some embodiments, the GUI system according to the methods described herein facilitates viewing images. For example, automatic resizing of the image to the screen size may be provided. Figure 11A A GUI implementation without this functionality is shown, in which the user clicks a zoom icon to change the size of the image. Figure 11B Displays a screenshot of the GUI, where the image is automatically sized to fill the screen vertically. Figure 11C and Figure 11DThis demonstrates two methods that provide zoom functionality for viewing images in detail. Figure 11C In the embodiment shown, when the scroll wheel button is clicked and held, the area around the mouse pointer is magnified. Figure 11D The embodiments shown in the document provide the functionality of zooming and panning using a mouse wheel as well as click and drag operations.

[0331] Figure 12A to 12F This diagram illustrates the intensity windowing method used to display images. More precisely, Figure 12B and 12D This document demonstrates a GUI implementation of a custom intensity window slider that allows users to adjust the intensity window threshold.

[0332] Figure 13A and Figure 13B A screenshot of an instanced GUI window displaying local intensity is shown. Specifically, in some embodiments, local intensity is displayed when the user hovers the mouse pointer over the image. Figure 13A In the embodiment shown, a local intensity value is displayed at the location of the mouse pointer. Figure 13B In the embodiment shown, the intensity is displayed in the lower left corner of the image rather than next to the mouse pointer, where it hides a portion of the image (e.g., thus providing improved security).

[0333] Figure 14A to 14D An exemplary GUI implementation for displaying multiple bone scan images is shown. Figure 14A and Figure 14C In the embodiments shown, the tabular approach used allows users to bimorphically transform different sets of bone scan images for different studies, or to bimorphically transform anterior and posterior images. Figure 14B and Figure 14C In the embodiments shown, checkboxes are used for bi-state throttling of the visibility of various images, thereby allowing multiple images from different studies to be displayed side by side.

[0334] In some embodiments, examples of GUI tools used for performing the image analysis described herein may provide information for quality control. For example, the total image intensity may be displayed. Figure 15A This is a screenshot of a GUI according to an illustrative embodiment, showing the display of total image intensity and total bone intensity. In some embodiments, for example... Figure 15B The GUI shown in the image displays only the total image intensity (not the total bone intensity) to provide a simpler user interface (e.g., to avoid clutter).

[0335] The GUI tool for reviewing bone scan images can display graphical indicators of detected (e.g., and pre-selected) hotspots overlaid on the bone scan image, such as... Figure 16A andFigure 16B As shown in the figure. In some embodiments, it is also shown Figure 17A The hotspot table is shown in the image. In some embodiments, the GUI tool described herein allows the user to select or deselect pre-selected hotspots for inclusion in the final selected hotspot set used to calculate the BSI value. The GUI can display the resulting BSI value 1602 above the image, allowing the user to observe its changes (compare 1704a and 1704b) while selecting and / or deselecting various hotspots (1702a and 1702b), as shown in the image. Figure 17B and 17C As shown in the image, various graphical control methods can be used to allow users to select pre-selected hotspots to include or exclude them. For example, Figure 18A Displays a pop-up control that appears after the user right-clicks. Figure 18B A dual-state tactile switch 1802 (“Edit Hotspot”) is shown, capable of being turned on and off. Once the dual-state tactile switch is on, the user clicks on various hotspots to select or deselect them. In some embodiments, the GUI tool may include safety / quality control features, such as… Figure 18C The pop-up window shown prompts users to verify quality control requirements before the report is generated.

[0336] In some embodiments, once BSI values ​​are calculated for various studies, they are displayed for user review. For example, the calculated BSI values ​​can be displayed in tabular form, such as... Figure 19A As shown in the figure. In some embodiments, BSI values ​​calculated for different studies are displayed above the displayed bone scan image for use in their respective studies, such as... Figure 19B As shown in the image. Figure 20A and 20B The display also includes a graph showing the time evolution of the calculated BSI value. Additional information related to the calculated BSI value can also be shown. Figure 21 Screenshots of a GUI according to an illustrative embodiment are shown, providing a table listing the number of hotspots in specific anatomical regions used to calculate BSI values ​​in different studies. In some embodiments, the systems and methods described herein provide automated report generation. Figure 22A and Figure 22B Displays automatically generated, instance-based reports.

[0337] F. Improved image processing methods

[0338] In some embodiments, the systems and methods described herein include... Figure 6 The improvement of one or more of the image processing steps shown in the figure.

[0339] i. Skeletal Segmentation

[0340] As described herein, a skeletal atlas image set can be used for image segmentation. The skeletal atlas image set contains a pair of template bone scan images (anterior and posterior) representing a typical normal bone scan, along with artificial contour depictions of 31 skeletal regions. These regions are raised to fit the current patient image to be analyzed. In some embodiments, a limited atlas is used, containing skeletal identifications covering only three-quarters of the femur and humerus. In some embodiments, an improved full-length atlas is used, containing skeletal region identifications covering the entire femur and humerus. Figure 23 A collection of bone scan images overlaid with atlas-based bone segmentation is shown, thus comparing the use of a finite atlas 2300a (left) and a full-length atlas 2300b (right). Figure 23 As shown, the limited atlas 2300a includes only three-quarters or less of the length 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. Using the full-length atlas improves the stability of bone segmentation. Specifically, the increased atlas allows the knee and elbow to be used as references during registration. Such reference points with clear contrast are beneficial for image analysis and provide improved stability. The increased atlas also allows for the detection of hotspots at more distant points in the limb.

[0341] ii. Hotspot Detection Thresholds

[0342] As described herein, an initial set of candidate hotspots is found using image intensity thresholding. In some embodiments, global and fixed thresholds are used, ensuring the same value is used across all bone regions of interest and across all images. Another improved approach sets regional thresholds that vary within different bone regions of interest. For example, this approach allows for the use of reduced thresholds (e.g., from 650 to 500) for the femoral and humeral regions to improve detection sensitivity. The femoral and humeral regions exhibit less uptake in bone scan images than other bone regions. Therefore, lower thresholds can be used for these regions to achieve a similar level of sensitivity as the rest of the body. Setting individual threshold values ​​for different bone regions allows this functionality and increases the detection of lower-intensity hotspots in these bone regions.

[0343] Figure 24An exemplary procedure 2400 is presented for lesion labeling and analysis using the improved segmentation and region-dependent thresholding method described above. In the exemplary procedure 2400, a set of bone scan images of a subject is accessed 2410. Each member image in the set of bone scan images is automatically segmented 2420 to identify the bone region of interest, including a femoral region corresponding to three-quarters of the length of the femur of a subject 2422 and / or a humeral region corresponding to three-quarters of the length of the humerus of a subject 2424. An initial set 2430 of identified bone regions of interest is then analyzed to automatically detect hotspots. As described herein, this step of automatically detecting hotspots may include applying a thresholding operation using region-dependent thresholding to each bone region of interest—that is, the values ​​are non-uniform and vary between different bone regions. Specifically, lower threshold values ​​for the femoral and / or humeral regions (2432 and 2434, respectively) can be used to enhance detection sensitivity in the femoral and / or humeral regions, thereby taking into account reduced drug (e.g., radiopharmaceutical) uptake in the femoral and / or humeral regions.

[0344] Once the initial set of hotspots is detected, a set of hotspot features 2440 is extracted for each hotspot, and a transfer probability value 2450 is calculated for each hotspot using the hotspot feature set. The detected hotspots can be rendered for graphical display to the user 2460, and in some embodiments, displayed to the user along with additional information (e.g., calculated probability values) 2460. Hotspots can be filtered (pre-selected) based on the calculated probability values ​​to be included in a first subset for presentation to the user and / or for calculating a risk index, such as a BSI value. The user can review the detected hotspots—e.g., the initial set of hotspots or the first subset including the filtered hotspots—via a graphical display and confirm or reject hotspots to be included in a second subset. The risk index value can then be calculated using this final subset, thereby incorporating the user's expertise into the decision-making process.

[0345] In some embodiments, global dynamic threshold adjustment is used. This method examines the obtained BSI values ​​and fine-tunes the global threshold scaling factor to accommodate high-burden disease. Scaling is calculated according to the following formula:

[0346]

[0347] Where t i This is the original preliminary threshold. Fine-tuning of the global threshold scaling improves linearity in high-burden disease cases while keeping low-burden diseases unaffected. Therefore, this global scaling factor method increases the useful range of BSI calculations.

[0348] Specifically, in some embodiments, the global scaling factor method is a data-driven approach that takes into account errors that may underestimate BSI values ​​at higher disease levels—i.e., at high metastasis levels. These errors are identified using a simulation system that allows for the simulation of bone scan images of patients with any (e.g., selected) degree of disease. This simulation system generates realistic bone scan images, thus also taking into account specific camera and review parameters. Therefore, realistic bone scan images can be generated based on known specific input parameters, allowing for the understanding of fundamental facts regarding lesion volume, bone volume, and therefore BSI values. This method thus allows BSI values ​​to be calculated via the image analysis methods described herein, and compared and checked against known fundamental facts determined based on the image simulation input parameters. Running a large number of simulations at different degrees of disease burden has demonstrated that no previous system using the global threshold scaling method described herein underestimates BSI values ​​in a non-linear manner at higher disease burdens. The non-linear functional form of the global threshold scaling factor in Equation 1 is based on error patterns observed in simulation studies to correct for observed non-linear underestimations of the calculated BSI values.

[0349] Figure 25A This is a graph showing how the scaling factor changes with b. As shown in the graph, multiplying the initial threshold by the global threshold scaling factor decreases the threshold as the disease burden (measured in hotspot area fractions) b increases. The reduced (adjusted) threshold causes larger areas to be identified as hotspots, thereby increasing the calculated BSI value, which measures the total fraction of a patient's bone occupied by hotspots. In this way, adjusting the threshold used for hotspot detection using the global threshold scaling factor corrects for underestimations observed in BSI values ​​calculated at high disease levels.

[0350] Figure 25B An example procedure 2500 for detecting hotspots using a global threshold scaling method is demonstrated. In a first step 2510, a set of bone scan images of the subject is accessed. The set of bone scan images is automatically segmented 2520 to produce an annotated set of images including the identification of the bone region of interest. In some embodiments, segmentation may include identifying the full length of the femur and / or humerus, as described above regarding... Figure 24As described. In another step 2530, an initial set of hotspots is automatically detected. In this hotspot detection step 2530, a set of potential hotspots is first detected using a preliminary threshold 2532. Then, a global threshold scaling factor 2534 is calculated using the set of potential hotspots, as described herein, and then the preliminary threshold 2536 is adjusted using the global threshold scaling factor 2534. Hotspots are then automatically detected to be included in the initial set of hotspots using the adjusted threshold 2538. Similar to process 2400, once the initial set of hotspots is detected, a set of hotspot features is extracted for each hotspot 2540 and the hotspot feature set is used to calculate a transition probability value 2550 for each hotspot. The detected hotspots can be rendered for graphical display to the user 2560.

[0351] iii. Hotspot Preselection

[0352] As described herein, hotspots are categorized to determine whether they should be pre-selected. In some embodiments, hotspot classification is performed via a two-step process. In some embodiments, for example, if the patient has no other hotspots or has many other hotspots, the first step categorizes the specific hotspot using local features of the specific hotspot rather than the rest of the image. In some embodiments, a second step is included to incorporate global information about the hotspot into the rest of the image.

[0353] Clinical experience suggests that hotspot selection depends on the rest of the image. A hotspot is more likely to be selected if many other hotspots are present, and less likely if it is the only hotspot. Therefore, using a single-step process may result in an underestimation of the BSI value in patients with numerous hotspots. A two-step process can improve performance in patients with numerous hotspots and high BSI. Hotspot selection using global hotspot features can be performed using a machine learning module. For example, in some embodiments, while a first machine learning module calculates the transfer probability value for each hotspot, a second machine learning module (e.g., implementing a different ANN) may receive the calculated probability value along with global hotspot features to determine whether the hotspot should be included in a subset of pre-selected hotspots.

[0354] Figure 26An exemplary process 2600 for hotspot pre-selection using global hotspot features is demonstrated. Process 2600 begins by accessing a set of bone scan images 2610, automatically segmenting the images (members) of the 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 2640 are extracted for each hotspot and used to calculate a metastasis probability value 2650 for each hotspot. In process 2600, a first subset 2662 of the initial hotspots is pre-selected using the metastasis probability value and global hotspot features. This first subset thus filters the hotspots, allowing a smaller set 2664 of target hotspots that have been automatically identified as potentially cancerous lesions to be displayed to the user.

[0355] iv. Atlas Weights

[0356] In some embodiments, correction factors for the sacral, pelvic, and lumbar regions are adjusted so that hot spots of equal area correspond to a more uniform measurement of BSI involvement. In some embodiments, without this adjustment, the sacral region differs significantly from the adjacent pelvic and lumbar regions. Figure 27 These areas – sacral region 2706, pelvic region 2702, and lumbar region 2704 – are shown as depicted in the skeletal atlas.

[0357] To calculate the BSI value, the total bone fraction for each selected hot spot is calculated, and the BSI value is calculated as the sum of all said fractions. For each hot spot, the fraction is calculated as follows: The hot spot size is divided by the size of the corresponding bone region (e.g., skull, ribs, lumbar vertebrae, pelvis) obtained from the bone segmentation, and multiplied by a weight fraction constant of the current bone region relative to the total bone weight. These constants (one for each bone region) can be determined based on International Commission on Radiation Protection (ICRP) Publication 23.

[0358] Use formula The cumulative rate is calculated, where c is a correction factor that incorporates several properties, such as hotspots typically visible in both the front and back images. This constant is 1.0 for all three regions before adjusting the correction factors described in this paper.

[0359] In some embodiments, this basic approach works well in most skeletal regions, but poorly in the sacral region and the surrounding pelvic and lumbar regions. The sacrum is a complex three-dimensional structure, and it is difficult to distinguish hotspots in different regions and locate them correctly using two-dimensional bone scans. Depending on the assigned location of hotspots (e.g., pelvis, lumbar spine, or sacrum), hotspots of similar size can vary significantly in their impact on the calculated BSI score. To reduce these differences, the coefficient c in the above formula is adjusted for the sacrum to make the difference from the pelvic to the lumbar region smoother. Specifically, the correction factor is adjusted to make the ratio... Describe this gradient. The value of c is adjusted accordingly to c. 骶骨,前侧 =0.16 and c 骶骨,后侧 =0.28, which makes the fractional BSI value in the sacrum between the fractional BSI values ​​in the pelvic and lumbar regions.

[0360] Figure 28 An exemplary procedure 2800 is presented, utilizing the correction factors described herein to adjust the skeletal involvement factor and calculate a risk index value from it. Procedure 2800 includes the steps of accessing a bone scan image 2810; automatically segmenting the bone scan image to identify skeletal regions 2820; automatically detecting an initial set of hotspots 2830; and extracting hotspot features for each hotspot 2840 and calculating a transition probability value 2850. The segmentation step 2820 and the hotspot detection step 2830 may utilize the improved segmentation method of procedure 2400 and / or the global threshold scaling method of procedure 2500. A first subset 2860 of hotspots in the initial set is automatically selected, at least based on the transition probability value calculated at step 2850. Global hotspot features (e.g., referring to the global hotspot features described in procedure 2600) may also be used, and user input (e.g., received via interaction with a GUI) may also be used to select the first subset. The hotspots of the first subset are then used to calculate the subject's risk index value 2870. As described in this article, the risk index calculation may include calculating the skeletal involvement factor 2872, adjusting the skeletal involvement factor 2874 using the region dependency correction factor, and summing the adjusted skeletal involvement factor 2876.

[0361] G. Example: BSI Computational Performance

[0362] This example demonstrates the linearity, accuracy, and reproducibility of the calculated BSI value.

[0363] i. Linearity and Accuracy

[0364] Automated bone sample index (BSI) (dependent variable) was determined from two sets of simulated bone scans and measured against a known prosthesis BSI (which was considered a dependent variable). In the first set of 50 simulated bone scans, the Shapiro-Wilk test confirmed that the residuals of the dependent variable were normally distributed (p = 0.850). Furthermore, a mean residual value of 0.00 and a standard deviation of 0.25 confirmed constant homoscedasticity across all values ​​of the dependent variable. Given the normality of the residuals and... Homoscedasticity The model is considered linear. Figure 25 shows the scatter plot with a linear fit and the associated parameters of the linear regression in the range of 0.10 to 13.0 BSI, as presented in Table 1 below.

[0365] Table 1. Parameters of the linear regression model in the first set of 50 prostheses, where the predefined BSI ranges from 0.10 to 13.0.

[0366] Linearity Measure Value 95% CI Sig. R 0.99 (0.99–0.99) <0.0001 Slope 0.98 (0.96–1.00) <0.0001

[0367] ii. Precision of BSI Calculation

[0368] For a second set of 50 simulated bone scans, the coefficient of variation and standard deviation of the automated BSI values ​​for each of the five predefined tumor burdens with different localizations were determined. The coefficient of variation at each of the five predefined prosthetic BSIs was less than or equal to 30%. The results are presented in Table 2 below.

[0369] Table 2. Coefficients of variation and standard deviations of automated BSI values ​​for each of the five predefined tumor burdens.

[0370]

[0371]

[0372] iii. Reproducibility with different cameras

[0373] Table 3 below shows the simulation results of BSI values ​​calculated for five disease burdens and different cameras. Different camera collimator settings have minimal impact on the reproducibility of BSI values. The standard deviation for each disease burden is <10%.

[0374] Table 3. BSI values ​​calculated for simulations using different cameras.

[0375]

[0376] iv. Reproducibility at different image counts

[0377] Figure 26Bland-Altman plots are presented to evaluate the reproducibility of automated BSI reads from 50 simulated prostheses. The difference between the mean and median BSI is 0.0 (solid horizontal line), with a standard deviation of 0.20, and repeatability coefficients (2xSD) of 0.40 and -0.40 (dashed horizontal lines). Illustrative statistics are presented in Table 4 below. Paired t-tests demonstrate a p-value of 0.15, suggesting no statistically significant difference between two aBSI values ​​obtained in repeated scans.

[0378] Table 4. Descriptive statistics showing the reproducibility of automated BSI calculations from 50 simulated prostheses.

[0379]

[0380] v. Reproducibility for patient repeat scans

[0381] Figure 27 Brand-Atman plots are presented to evaluate the reproducibility of automated BSI reads from repeated bone scans of 35 metastatic patients. The difference between the mean BSI and the median BSI is 0.0 (solid horizontal line), with a standard deviation of 0.18, where the repeatability coefficients (2xSD) are 0.36 and -0.36 (dashed horizontal line). Descriptive statistics are presented in Table 5 below. Paired t-tests demonstrated a p-value of 0.09, suggesting no statistically significant difference between the two aBSI values ​​obtained from repeated scans.

[0382] Table 5. Descriptive statistics showing the reproducibility of automated BSI calculations from repeated bone scans of 35 metastatic patients.

[0383]

[0384] vi. Comparison to approved market devices

[0385] Figure 28 A and Figure 28 B compares the Brand-Atman diagrams of BSI calculations using the improved automated BSI apparatus of the present invention (software aBSI version 3.4) with those using an approved market device implementing an earlier version of the software that does not utilize the improvements of the present invention (software EXINI version 1.7). Each diagram shows the difference between the known BSI value of the simulated prosthesis and the calculated value of the prosthesis using one of the two BSI software versions. Figure 28 A shows the difference between the known prosthesis value and the calculated value of aBSI 3.4, while Figure 28 B shows the difference between the known prosthesis value and the value calculated using EXINI 1.7. Figure 28 A and Figure 28The horizontal solid line in B indicates that for aBSI 3.4 and -0.89, the mean BSI difference (between the spur and the calculated BSI value) is 0.14 and -0.89, respectively. The standard deviation (SD = 0.30) of aBSI 3.4 is observed to be significantly lower than the standard deviation (SD = 0.88) of EXINI 1.7. Descriptive statistics are presented in Tables 6A and 6B below.

[0386] Table 6A. Descriptive statistics of BSI values ​​calculated using the automated BSI (software version 3.4) according to the method described in this paper.

[0387]

[0388]

[0389] Table 6B. Descriptive statistics of BSI values ​​calculated using the approved software (EXINI 1.7)

[0390]

[0391] H. Computer systems and network environment

[0392] In some embodiments, the systems and methods described herein are implemented using a cloud-based microservices architecture. Figure 33A Demonstrates an example cloud platform architecture, and Figure 33B Showcases a diagram illustrating an example of microservice communication design.

[0393] Figure 34 An illustrative network environment 3400 is shown, illustrating the methods and systems described herein. In brief, reference is now made to... Figure 34 This document illustrates and describes a block diagram of an exemplary cloud computing environment 3400. The cloud computing environment 3400 may include one or more resource providers 3402a, 3402b, 3402c (collectively referred to as 3402). Each resource provider 3402 may include computing resources. In some embodiments, the computing resources may include any hardware and / or software for processing data. For example, the computing resources may include hardware and / or software capable of executing algorithms, computer programs, and / or computer applications. In some embodiments, the exemplary computing resources may include application servers and / or databases with storage and capture capabilities. Each resource provider 3402 may be connected to any other resource provider 3402 in the cloud computing environment 3400. In some embodiments, the resource providers 3402 may be connected via a computer network 3408. Each resource provider 3402 may be connected via the computer network 3408 to one or more computing devices 3404a, 3404b, 3404c (collectively referred to as 3404).

[0394] The cloud computing environment 3400 may include a resource manager 3406. The resource manager 3406 may be connected to resource providers 3404 and computing devices 3404 via a computer network 3408. In some embodiments, the resource manager 3406 may facilitate one or more resource providers 3402 to provide computing resources to one or more computing devices 3404. The resource manager 3406 may receive requests for computing resources from a specific computing device 3404. The resource manager 3406 may identify one or more resource providers 3402 capable of providing the computing resources requested by the computing device 3404. The resource manager 3406 may select resource providers 3402 to provide computing resources. The resource manager 3406 may facilitate connections between resource providers 3402 and specific computing devices 3404. In some embodiments, the resource manager 3406 may establish connections between specific resource providers 3402 and specific computing devices 3404. In some implementations, resource manager 3406 may redirect a particular computing device 3404 to a particular resource provider 3402 that has the requested computing resources.

[0395] Figure 35 Examples of computing devices 3500 and mobile computing devices 3550 that can be used in the methods and systems described herein are shown. Computing device 3500 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. Mobile computing device 3550 is intended to represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, and other similar computing devices. The components, their connections and relationships, and their functions shown herein are intended to be illustrative only and are not intended to be limiting.

[0396] The computing device 3500 includes a processor 3502, a memory 3504, a storage device 3506, a high-speed interface 3508 connected to the memory 3504, and multiple high-speed expansion ports 3510, as well as a low-speed interface 3512 connected to low-speed expansion ports 3514 and the 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 can be mounted on a common motherboard or otherwise suitably mounted. The processor 3502 can process instructions for execution within the computing device 3500, including instructions stored in the memory 3504 or on the storage device 3506 for displaying graphical information for a GUI on an external input / output device (e.g., a display 3516 coupled to the high-speed interface 3508). In other embodiments, multiple processors and / or multiple buses may be suitably used in conjunction with multiple memories and several types of memory. Furthermore, multiple computing devices can be connected, each providing several parts of the necessary operation (e.g., as a server group, blade server cluster, or multiprocessor system). Therefore, as a term used herein, the description of multiple functions as being performed by a “processor” encompasses embodiments in which any number of computing devices (one or more) and any number of processors (one or more) perform said multiple functions. Furthermore, when a function is described as being performed by a “processor,” this encompasses embodiments in which any number of computing devices (one or more) and any number of processors (one or more) perform the function (e.g., in a distributed computing system).

[0397] Memory 3504 stores information within computing device 3500. In some embodiments, memory 3504 is one or more volatile memory cells. In some embodiments, memory 3504 is one or more non-volatile memory cells. Memory 3504 may also be another form of computer-readable media, such as a magnetic disk or optical disk.

[0398] Storage device 3506 provides a large-capacity storage device for computing device 3500. In some embodiments, storage device 3506 may be or contain computer-readable media, such as floppy disk devices, hard disk devices, optical disk devices, magnetic tape devices, flash memory or other similar solid-state storage devices or device arrays, including devices in a storage area network or other configuration. Instructions may be stored in an information carrier. When executed by one or more processing devices (for example, processor 3502), the instructions perform one or more methods, such as those described above. Instructions may also be stored by one or more storage devices (e.g., computer or machine-readable media (for example, memory 3504, storage device 3506, or memory on processor 3502)).

[0399] High-speed interface 3508 manages bandwidth-intensive operations of computing device 3500, while low-speed interface 3512 manages lower bandwidth-intensive operations. This functional allocation is merely an example. In some embodiments, high-speed interface 3508 is coupled to memory 3504, display 3516 (e.g., via a graphics processor or accelerator), and high-speed expansion port 3510 that accepts various expansion cards (not shown). In embodiments, low-speed interface 3512 is coupled to storage device 3506 and low-speed expansion port 3514. Various communication ports (e.g., USB, ...) may be included. The low-speed expansion port 3514 (Ethernet, Wireless Ethernet) can be coupled to one or more input / output devices, such as keyboards, pointing devices, scanners, or networking devices (e.g., switches or routers), via a network adapter.

[0400] As shown in the figure, the computing device 3500 can be implemented in several different forms. For example, it can be implemented as a standard server 3520 or multiple times in a group of such servers. Alternatively, it can be implemented in a personal computer (e.g., a laptop computer 3522). It can also be implemented as part of a rack server system 3524. Alternatively, components from the computing device 3500 can be combined with other components in a mobile device (not shown) (e.g., a mobile computing device 3550). Each of these devices may contain one or more of the computing device 3500 and the mobile computing device 3550, and the entire system may consist of multiple computing devices communicating with each other.

[0401] Among other components, the mobile computing device 3550 also includes a processor 3552, a memory 3564, an input / output device (e.g., a display 3554), a communication interface 3566, and a transceiver 3568. The mobile computing device 3550 may also include a storage device (e.g., a microdrive or other device) 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 the components may be mounted on a common motherboard or otherwise suitably mounted.

[0402] Processor 3552 executes instructions within mobile computing device 3550, including instructions stored in memory 3564. Processor 3552 may be implemented as a chipset comprising individual and multiple analog and digital processors. For example, processor 3552 may provide coordination of other components of mobile computing device 3550, such as control of the user interface, applications running by mobile computing device 3550, and wireless communications via mobile computing device 3550.

[0403] Processor 3552 can communicate with the user via control interface 3558 and display interface 3556 coupled to display 3554. For example, display 3554 may be a TFT (Thin Film Transistor Liquid Crystal Display) or OLED (Organic Light Emitting Diode) display or other suitable display technology. Display interface 3556 may include suitable circuitry for driving display 3554 to present graphics and other information to the user. Control interface 3558 can receive commands from the user and translate them for submission to processor 3552. Additionally, external interface 3562 can provide communication with processor 3552 to enable near-field communication between mobile computing device 3550 and other devices. For example, external interface 3562 may provide wired communication in some embodiments or wireless communication in others, and multiple interfaces may be used.

[0404] 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 cells, or one or more non-volatile memory cells. Extended memory 3574 may also be provided and connected to the mobile computing device 3550 via an extended interface 3572, for example, which may include a SIMM (Single In-line Memory Module) card interface. Extended memory 3574 may provide additional storage space for the mobile computing device 3550 or may also store applications or other information for the mobile computing device 3550. Specifically, extended memory 3574 may contain instructions for performing or supplementing the processes described above, and may also contain security information. Thus, for example, extended memory 3574 may be programmed as a security module for the mobile computing device 3550 and may allow secure use of the mobile computing device 3550. Additionally, secure applications may be provided via the SIMM card along with additional information, such as placing identification information on the SIMM card in a non-hacking manner.

[0405] For example, the memory may include flash memory and / or NVRAM (non-volatile random access memory), as discussed below. In some embodiments, instructions are stored in an information carrier and, when executed by one or more processing devices (for example, processor 3552), perform one or more methods, such as those described above. Instructions may also be stored by one or more storage devices (e.g., one or more computer or machine-readable media (for example, memory 3564, extended memory 3574, or memory on processor 3552)). In some embodiments, instructions may be received (for example) via transceiver 3568 or external interface 3562 in the form of a transmitted signal.

[0406] The mobile computing device 3550 can communicate wirelessly via a communication interface 3566, which may include a digital signal processing circuitry system if needed. The communication interface 3566 can provide communication under various modes or protocols, such as GSM voice calls (Global System for Mobile Communications), SMS (Short Message Service), EMS (Enhanced Messaging Service) or MMS messages (Multimedia Messaging Service), CDMA (Code Division Multiple Access), TDMA (Time Division Multiple Access), PDC (Personal Digital Cellular System), WCDMA (Wideband Code Division Multiple Access), CDMA2000 or GPRS (General Packet Radio Service), and other modes or protocols. For example, this communication can be initiated using radio frequency via transceiver 3568. Additionally, short-range communication may (for example) use... This occurs via Wi-Fi™ or other such transceivers (not shown). Additionally, the GPS (Global Positioning System) receiver module 3570 can provide additional navigation and location-related wireless data to the mobile computing device 3550, which can be used by applications running on the mobile computing device 3550.

[0407] The mobile computing device 3550 can also communicate audio using an audio codec 3560, which receives information from a user and converts it into usable digital information. The audio codec 3560 can also, for example, produce audible sound to the user via a speaker in the mobile phone of the mobile computing device 3550. This sound may include sounds from voice phone calls, recorded sounds (e.g., voice messages, music files, etc.), and sounds generated by applications operating on the mobile computing device 3550.

[0408] As shown in the figure, the mobile computing device 3550 can be implemented in several different forms. For example, it can be implemented as a cellular phone 3580. It can also be implemented as part of a smartphone 3582, a personal digital assistant, or other similar mobile device.

[0409] Various implementations of the systems and techniques described herein can be implemented using 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 in one or more computer programs executable and / or interpretable on a programmable system, said programmable system comprising at least one programmable processor (coupled to receive data and instructions from and transfer data and instructions to a storage system, whether dedicated or general purpose), at least one input device, and at least one output device.

[0410] These computer programs (also referred to as programs, software, software applications, or code) contain machine instructions for a programmable processor and can be implemented in high-level programming and / or object-oriented programming languages ​​and / or in assembly / machine language. As used herein, the terms machine-readable media and computer-readable media refer to any computer program product, device, and / or apparatus (e.g., disk, optical disk, memory, programmable logic device (PLD)) for providing machine instructions and / or data to a programmable processor, which includes a machine-readable medium for receiving machine instructions as machine-readable signals. The term machine-readable signal refers to any signal for providing machine instructions and / or data to a programmable processor.

[0411] To provide interaction with the user, the systems and techniques described herein can be implemented on a computer having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor for displaying information to the user) and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound, voice, or tactile input).

[0412] The systems and techniques described herein can be implemented in computing systems (including back-end components (e.g., as data servers); or middleware components (e.g., application servers); or front-end components (e.g., client computers having a graphical user interface or web browser through which users can interact with implementations of the systems and techniques described herein) or any combination of such back-end, middleware, or front-end components. The components of the system can be interconnected via any form of digital data communication or media (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0413] The computing system may include clients and servers. Clients and servers are typically geographically separated and usually interact via a communication network. The client-server relationship is established by means of computer programs running on respective computers that have a client-server relationship with each other. In some embodiments, the modules and / or services described herein may be separated, combined, or incorporated into single or combined modules and / or services. The modules and / or services illustrated in the figures are not intended to limit the system described herein to the software architecture shown herein.

[0414] Although the invention has been specifically shown and described with reference to particular preferred embodiments, those skilled in the art will understand that various changes in form and detail may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.

Claims

1. A method for marking and quantifying lesions in nuclear medicine images of subjects, the method comprising: (a) Accessing a set of bone scan images of the subject via a processor of a computing device, the set of bone scan images being obtained after the subject has been given the drug; (b) The processor automatically segments each image in the bone scan image set to identify one or more skeletal regions of interest, thereby obtaining an annotated image set, wherein each of the one or more skeletal regions of interest corresponds to a specific anatomical region of the subject's skeleton, wherein the one or more skeletal regions of interest include at least one of (i) and (ii): (i) Femoral region, which corresponds to a portion of the femur of the subject, the femoral portion encompassing at least three-quarters of the femur along its length; and (ii) Humeral region, which corresponds to a portion of the humerus of the subject, the humeral portion encompassing at least three-quarters of the humerus along its length; (c) Automatically detecting an initial set of one or more hotspots by the processor, each hotspot corresponding to a high-intensity region in the annotated image set, the automatic detection comprising using the intensity of pixels in the annotated image set and using one or more region-dependent thresholds to identify the one or more hotspots, wherein the one or more region-dependent thresholds include one or more values ​​associated with the femoral region and / or the humeral region, the one or more values ​​providing enhanced hotspot detection sensitivity in the femoral region and / or the humeral region to compensate for reduced drug uptake in the femoral region and the humeral region; (d) For each hotspot in the initial hotspot set, the processor extracts a set of hotspot features associated with the hotspot; (e) For each hotspot in the initial hotspot set, the processor calculates a transition probability value corresponding to the probability of the hotspot representing a transition based on the hotspot feature set associated with the hotspot; and (f) The processor causes a graphical representation of at least a portion of the initial hotspot set to be rendered for display within a graphical user interface (GUI). Step (b) includes: Each member in the bone scan image set is compared with a corresponding image in the image set, each image set including one or more identifications of one or more bone regions of interest, the bone regions of interest including (i) the femoral region including at least a portion of the subject's knee region and / or (ii) the humeral region including at least a portion of the subject's elbow region; and For each image in the bone scan image set, the corresponding atlas image is registered with the bone scan image using the knee region and / or the elbow region identified in the image in the bone scan image set as landmarks, such that the identification of the one or more bone regions of interest in the atlas image is applied to the image in the bone scan image set.

2. The method of claim 1, wherein the location of at least one detected hot spot in the initial hot spot set corresponds to a body location in or on the femur, the body location being located at more than three-quarters of the distance along the femur from one end of the femur oriented toward the subject's hip to the other end of the femur oriented toward the subject's knee.

3. The method of claim 1, wherein the location of at least one detected hot spot in the initial hot spot set corresponds to a body location in or on the humerus, the body location being located at more than three-quarters of the distance along the humerus from one end of the humerus oriented toward the subject's shoulder to the other end of the humerus oriented toward the subject's elbow.

4. The method of claim 1, wherein step (c) comprises: The processor identifies healthy tissue areas in the bone scan image set that are determined not to contain any hot spots. The processor calculates a normalization factor such that the product of the normalization factor and the average intensity of the identified healthy tissue region is a predefined intensity level. and The processor normalizes the images in the bone scan image set according to the normalization factor.

5. The method of claim 1, further comprising: (g) The processor calculates one or more risk index values ​​for the subject based at least in part on a calculated score of the subject’s skeleton being occupied by the initial hotspot set.

6. The method according to claim 1, comprising: (h) The processor selects a first subset of the initial hotspot set based at least in part on the transition probability value; and (i) The processor causes the rendering of a graphical representation of the first subset for display within a graphical user interface (GUI).

7. The method of claim 6, further comprising: (j) The processor calculates one or more risk index values ​​for the subject based at least in part on a calculated score of the subject’s skeleton being occupied by the first subset of hotspots.

8. The method according to claim 1, comprising: (k) The processor receives a user's selection of a second subset of the initial hotspot set via the GUI; and (l) The processor calculates one or more risk index values ​​for the subject based at least in part on a calculated score of the subject’s skeleton being occupied by a second subset of hotspots.

9. The method of claim 5, wherein at least one of the risk index values ​​indicates the risk of the subject developing and / or acquiring metastatic cancer.

10. The method of claim 9, wherein the metastatic cancer is metastatic prostate cancer.

11. The method of claim 5, wherein at least one of the risk index values ​​indicates that the subject has a specific state of metastatic cancer.

12. The method of claim 1, wherein the processor is a processor of a cloud-based system.

13. The method of claim 1, wherein the GUI is part of a common picture archiving and communication system (PACS).

14. The method according to claim 1, wherein the reagent comprises technetium 99m methylene diphosphonate ( 99m Tc-MDP).

15. A system for marking and quantifying lesions in nuclear medicine images of subjects, the system comprising: processor; and A memory having instructions, wherein the instructions, when executed by the processor, cause the processor to: (a) Accessing a set of bone scan images of the subject, the set of bone scan images being obtained after the subject has been given the drug; (b) Each image in the bone scan image set is automatically segmented to identify one or more bone regions of interest, thereby obtaining an annotated image set, wherein each of the one or more bone regions of interest corresponds to a specific anatomical region of the subject's skeleton, and wherein the one or more bone regions of interest include at least one of (i) and (ii): (i) Femoral region, which corresponds to a portion of the femur of the subject, the femoral portion encompassing at least three-quarters of the femur along its length; and (ii) Humeral region, which corresponds to a portion of the humerus of the subject, the humeral portion encompassing at least three-quarters of the humerus along its length; (c) Automatically detect an initial set of one or more hot spots, each hot spot corresponding to a high-intensity region in the annotated image set, the automatic detection comprising using the intensity of pixels in the annotated image set and using one or more region-dependent thresholds to identify the one or more hot spots, and wherein the one or more region-dependent thresholds include one or more values ​​associated with the femoral region and / or the humeral region, the one or more values ​​providing enhanced hot spot detection sensitivity in the femoral region and / or the humeral region to compensate for reduced drug uptake in the femoral region and the humeral region; (d) For each hotspot in the initial hotspot set, extract the hotspot feature set associated with the hotspot; (e) For each hotspot in the initial hotspot set, calculate a transfer probability value corresponding to the likelihood of the hotspot representing a transfer, based on the hotspot feature set associated with the hotspot; and (f) Causes the rendering of a graphical representation of at least a portion of the initial hotspot set for display within a graphical user interface (GUI), and wherein, in step (b), the instruction causes the processor to: Each member in the bone scan image set is compared with a corresponding image in the image set, each image set including one or more identifications of one or more bone regions of interest, the bone regions of interest including (i) the femoral region including at least a portion of the subject's knee region and / or (ii) the humeral region including at least a portion of the subject's elbow region; and For each image in the bone scan image set, the corresponding atlas image is registered with the bone scan image using the knee region and / or the elbow region identified in the image in the bone scan image set as landmarks, such that the identification of the one or more bone regions of interest in the atlas image is applied to the image in the bone scan image set.

16. A computer-aided image analysis apparatus comprising the system according to claim 15.

17. The device of claim 16, wherein the device is programmed for use by trained healthcare professionals and / or researchers.

18. The apparatus of claim 17, wherein the apparatus is programmed to analyze bone scan images to evaluate and / or detect metastatic cancer.

19. The apparatus of claim 17 or 18, wherein the apparatus is programmed to analyze bone scan images to evaluate and / or detect prostate cancer.

20. The device according to any one of claims 16 to 18, comprising a sign indicating that the device is intended for use by trained healthcare professionals and / or researchers.

21. The apparatus of claim 20, wherein the mark further specifies that the apparatus is intended for analyzing bone scan images to assess and / or detect metastatic cancer.

22. The apparatus of claim 20, wherein the mark further specifies that the apparatus is intended for analyzing bone scan images to assess and / or detect prostate cancer.

Images